Datasets:

---

bigcode

/

jupyter-parsed

like 4

Follow

BigCode 1.4k

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<!--HEADER-->\n[*NBJoint test on a collection of notebooks about some thermodynamic properperties of water*](https://github.com/rmsrosa/nbjoint)", "_____no_output_____" ], [ "<!--BADGES-->\n<a href=\"https://nbviewer.jupyter.org/github/rmsrosa/nbjoint/blob/master/tests/nb_export_builds/nb_water_md/BA.00-References.md\"><img align=\"left\" src=\"https://img.shields.io/badge/view-markdown-orange\" alt=\"View Markdown\" title=\"View Markdown\"></a><a href=\"https://nbviewer.jupyter.org/github/rmsrosa/nbjoint/blob/master/tests/nb_export_builds/nb_water_pdf/BA.00-References.pdf\"><img align=\"left\" src=\"https://img.shields.io/badge/view-pdf-blueviolet\" alt=\"View PDF\" title=\"View PDF\"></a>&nbsp;", "_____no_output_____" ], [ "<!--NAVIGATOR-->\n[<- Choosing the Best Fit with AIC](05.00-Best_AIC_Fitting.ipynb) | [Water Contents](00.00-Water_Contents.ipynb) | [References](BA.00-References.ipynb) \n\n---\n", "_____no_output_____" ], [ "# References", "_____no_output_____" ], [ "- G. K. Batchelor (2000); \"An Introduction to Fluid Dynamics\"; Cambridge University Press, UK.\n- E. A. Bender (2000), \"An Introduction to Mathematical Modeling\"; (Dover Books on Computer Science) Dover Publications; 1st edition.\n- K. P. Burnham, D. R. Anderson (2002), \"Model Selection and Multimodel Inference: A practical information-theoretic approach\"; Springer-Verlag; 2nd edition.\n- G. H. Golub and C. F. Van Loan (1996), \"Matrix Computations\", Johns Hopkins University Press, 3rd edition.\n- L. N. Trefethen and D. Bau III (1997); \"Numerical Linear Algebra\"; SIAM: Society for Industrial and Applied Mathematics; 1st edition.", "_____no_output_____" ], [ "<!--NAVIGATOR-->\n\n---\n[<- Choosing the Best Fit with AIC](05.00-Best_AIC_Fitting.ipynb) | [Water Contents](00.00-Water_Contents.ipynb) | [References](BA.00-References.ipynb) ", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# All those moments will be los(s)t in time, like tears in rain.\n> I am completely operational, and all my circuits are functioning perfectly.\n\n- toc: true \n- badges: true\n- comments: true\n- categories: [jupyter]\n- image: images/posts/2020-12-10-Tears-In-Rain/Tears-In-Rain.jpg", "_____no_output_____" ] ], [ [ "#hide\n!pip install -Uqq fastbook\nimport fastbook\nfastbook.setup_book()", "_____no_output_____" ], [ "#hide\nfrom fastai.vision.all import *\nfrom fastbook import *\n\nmatplotlib.rc('image', cmap='Greys')", "_____no_output_____" ] ], [ [ "# Under the Hood: Training a Digit Classifier", "_____no_output_____" ], [ "This one is the big one. Now we have made it past the introduction of the course it is time to get under the hood and start implementing some of the functionality for ourselves. I am going to be taking quite thorough notes as I want to make sure I understand everything before I move onto Chapter 5.", "_____no_output_____" ], [ "So far all the heavy lifting has been done for us, the fastai library is a nice high level API written on top of PyTorch and abstracted in such a way that the \"magic\" / \"obfuscation\" is a little hard to determine what is actually happening. So in this chapter we will be making a hand written digit classifier, a simple classifier that can determine whether or not a 28px * 28px image of a hand written digit is either a **3** or a **7**. We will try to figure out a good baseline from which we can assess our model and then proceed to write out each element of the model in simple python, before seeing it all wrapped up in the nice and tidy fastai API.", "_____no_output_____" ], [ "## Pixels: The Foundations of Computer Vision", "_____no_output_____" ], [ "Here we are going to take a look at what we are actually dealing with when it comes to our hand written digits. Thankfully there is a great training set put together by Yann LeCun called the [MNIST](https://en.wikipedia.org/wiki/MNIST_database) or Modified National Institute of Standards and Technology database. It contains thousands of individual hand written digits that have been collated as 28px * 28px grayscale images. This is the data we will be using to build our classifier. ", "_____no_output_____" ], [ "The team at fastai have made it simple to download and unzip the data we are going to use for this lesson. Instead of having to manually go to the [fastai datasets](https://course.fast.ai/datasets) documentation page, download the tar file, unzip it in a location accessible to your notebooks they have a handy [utility](https://github.com/fastai/fastai/blob/715c027b0ad8586feda09e29fb2b483dfe30c910/fastai/data/external.py) that will do all of that in single line of code:", "_____no_output_____" ] ], [ [ "path = untar_data(URLs.MNIST_SAMPLE)", "_____no_output_____" ], [ "#hide\nPath.BASE_PATH = path", "_____no_output_____" ] ], [ [ "The have also added the UNIX [ls()](https://en.wikipedia.org/wiki/Ls) function to Python [path](https://docs.python.org/3/library/pathlib.html) to give a handy way to see what is in our directory. Here you can see we have our training and validation directory along with a label.csv file. ", "_____no_output_____" ] ], [ [ "path.ls()", "_____no_output_____" ] ], [ [ "Inside each of the training and validation directories we have a folder which contains all of our **3's** and **7's** respectively", "_____no_output_____" ] ], [ [ "(path/'train').ls()", "_____no_output_____" ] ], [ [ "Now we can create a list of the file paths and store them in variables", "_____no_output_____" ] ], [ [ "threes = (path/'train'/'3').ls().sorted()\nsevens = (path/'train'/'7').ls().sorted()\nthrees", "_____no_output_____" ] ], [ [ "Let's take a look at what we are working with. By indexing into the above **threes** list we can retrive the first path in the list. Using [PIL](https://pillow.readthedocs.io/en/stable/) the Python Imaging Library, we can display the data at that path.", "_____no_output_____" ] ], [ [ "im3_path = threes[1]\nim3 = Image.open(im3_path)\nim3", "_____no_output_____" ] ], [ [ "At this point it may not be entirely intuative what image information is. It is just an array of values for 0 to 255 for an 8-bit grayscale image like we have above. By casting into a Numpy array and taking a slice we can se what that might look like.", "_____no_output_____" ] ], [ [ "array(im3)[4:10,4:10]", "_____no_output_____" ] ], [ [ "As we have seen earlier in the course the PyTorch tensors have very similar functionality to Numpy arrays, but have the added benefit of being pushed to the GPU, this is a optimization and can be used in replacement of standard Numpy arrays. As a huge fan of Numpy my natural propensity is to use them all the time. But I think I need to reconsider . . .", "_____no_output_____" ] ], [ [ "tensor(im3)[4:10,4:10]", "_____no_output_____" ] ], [ [ "By loading the tensor into a Pandas Dataframe we are able to see what the pixel values look like by using the .background_gradients method", "_____no_output_____" ] ], [ [ "im3_t = tensor(im3)\ndf = pd.DataFrame(im3_t[4:15,4:22])\ndf.style.set_properties(**{'font-size':'6pt'}).background_gradient('Greys')", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ], [ "## First Try: Pixel Similarity", "_____no_output_____" ], [ "Before we start building our model we should consider if there are feasible alternatives to our problem. Working with Deep Learning is like weilding a very large hammer and every problem can appear to us as a nail. But for small self contained problem that do not require scale, simple alternatives are preferable. Here we are exploring an alternative to determine a baseline. What results can we achieve with a naive approach, is this sufficient to solve our problem and can they be improved by the use of a deep learning model. This is an important step, making sure that you are making headway on your problem is more important than having a shiny new model.", "_____no_output_____" ], [ "### Problem Statement", "_____no_output_____" ], [ "Can we create a simple baseline application that can classify an unseen 28px * 28px image of a **3** or a **7**. Theoretically, this is a simpler task than differentiating between a **3** and an **8** for example because both have similar curves at the top of the number and the bottom which might be difficult to disambiguate. If we try making a `mean` image from all the data in the **threes** folder and another for the **sevens** and compare the `pixel similarity` of an unseen image with this `mean` image we might be able to classify it with some manner of accuracy ", "_____no_output_____" ], [ "### Process", "_____no_output_____" ], [ "Let's create two more arrays that store a tensor of each image in our **threes** and **sevens** folders.", "_____no_output_____" ] ], [ [ "seven_tensors = [tensor(Image.open(o)) for o in sevens]\nthree_tensors = [tensor(Image.open(o)) for o in threes]\nlen(three_tensors),len(seven_tensors)", "_____no_output_____" ] ], [ [ "In a similar way to the method above to cast our image to an array or tensor, we can use the fastai `show_image` method to take a tensor and display it as an image, this will be useful for debugging", "_____no_output_____" ] ], [ [ "show_image(three_tensors[1]);", "_____no_output_____" ] ], [ [ "Our next operation is to normalize the values of the images between 0 and 1, first we stack the images on top of one another using `torch.stack`, convert the integer values in the tensor stack to a float to ensure that values aren't rounded after our division operation, and then we divide the image by 255. ", "_____no_output_____" ], [ "By looking at the shape of the torch tensor stack we can see that we have 6131 image tensors that have 28 rows and 28 columns", "_____no_output_____" ] ], [ [ "stacked_sevens = torch.stack(seven_tensors).float()/255\nstacked_threes = torch.stack(three_tensors).float()/255\nstacked_threes.shape", "_____no_output_____" ] ], [ [ "The length of the tensor.shape is the same as the tensor rank or number of dimensions, we can see that below by using the `ndim` method. This is a fast way of checking that your tensors have the correct rank before moving forward building your model or baseline ", "_____no_output_____" ] ], [ [ "len(stacked_threes.shape)", "_____no_output_____" ], [ "stacked_threes.ndim", "_____no_output_____" ] ], [ [ "Now that we have all of our image tensors stack on top of one another and normalized, we can use the `mean` method to find the average. By passing in the argument 0, we are telling the operation to take the `mean` across the first axis. In this example this is the `mean` of the 6131 images.", "_____no_output_____" ] ], [ [ "mean3 = stacked_threes.mean(0)\nshow_image(mean3);", "_____no_output_____" ], [ "mean7 = stacked_sevens.mean(0)\nshow_image(mean7);", "_____no_output_____" ] ], [ [ "Now that we have our `mean` images, we can compare one of the images against them to see what the `pixel similarity` is and determine if the image is either a **3** or a **7**", "_____no_output_____" ] ], [ [ "a_3 = stacked_threes[1]\nshow_image(a_3);", "_____no_output_____" ] ], [ [ "### L1 and L2 norm", "_____no_output_____" ], [ "- Take the mean of the absolute value of differences (absolute value is the function that replaces negative values with positive values). This is called the mean absolute difference or L1 norm\n- Take the mean of the square of differences (which makes everything positive) and then take the square root (which undoes the squaring). This is called the root mean squared error (RMSE) or L2 norm.", "_____no_output_____" ], [ "To determine the `pixel similarity` of our test image and our mean image we are going to use the `mean absolte difference` or the `L1 Norm` and the `root mean squared error` or the `L2 Norm`. ", "_____no_output_____" ], [ "If we were simply going to subtract one for the other we could end up with negative values, which once averaged would negate positive values and give us inconcluse information. By comparing the test image againt the mean of the **3's** and the mean of the **7's** we can see that error is lower when comparing to the mean of the **3's** giving us a classification that this image is of a **3**", "_____no_output_____" ] ], [ [ "dist_3_abs = (a_3 - mean3).abs().mean()\ndist_3_sqr = ((a_3 - mean3)**2).mean().sqrt()\ndist_3_abs,dist_3_sqr", "_____no_output_____" ], [ "dist_7_abs = (a_3 - mean7).abs().mean()\ndist_7_sqr = ((a_3 - mean7)**2).mean().sqrt()\ndist_7_abs,dist_7_sqr", "_____no_output_____" ] ], [ [ "Here we can see by using the `l1_loss` and the `mse_loss` methods we are getting the same answers our implementations above", "_____no_output_____" ] ], [ [ "F.l1_loss(a_3.float(),mean7), F.mse_loss(a_3,mean7).sqrt()", "_____no_output_____" ] ], [ [ "## Computing Metrics Using Broadcasting", "_____no_output_____" ], [ "At this point we could simply use a loop and iterate through each of the images in the validation set, calculate the `L1` loss for each image and determine whether or not our baseline application determines the number to be a **3** or a **7**, check the prediction against and determine an accuracy. This would take a very long time and wouldn't make use of the GPU accelleration needed for Deep Learning.", "_____no_output_____" ], [ "Here we are going to look at the secret sauce that makes Python (an interpreted language) powerful enough to be used in Deep Learning applications. What is that secret sauce you ask? `Broadcasting`", "_____no_output_____" ], [ "So lets start by stacking the images from our validation directories in the same way we did with our training images. We will normalize them and check the shape of the tensor stack", "_____no_output_____" ] ], [ [ "valid_3_tens = torch.stack([tensor(Image.open(o)) \n for o in (path/'valid'/'3').ls()])\nvalid_3_tens = valid_3_tens.float()/255\nvalid_7_tens = torch.stack([tensor(Image.open(o)) \n for o in (path/'valid'/'7').ls()])\nvalid_7_tens = valid_7_tens.float()/255\nvalid_3_tens.shape,valid_7_tens.shape", "_____no_output_____" ] ], [ [ "Here we are going to write our `L1 Norm` in the form of a function, here taking the mean across the rows and columns of the absolute difference of the tow images.", "_____no_output_____" ] ], [ [ "def mnist_distance(a,b): return (a-b).abs().mean((-1,-2))\nmnist_distance(a_3, mean3)", "_____no_output_____" ] ], [ [ "But by using `Broadcasting`, we can instead use our entire tensor stack and compare it against the `mean` image of the **3** all at once. What is happening under the hood is that PyTorch is making a \"virtual\" copy of the mean image tensor for every image tensor in the validation stack so it can determine the `pixel similarity` for all of them at once. What it returns is a tensor of all of the results. ", "_____no_output_____" ] ], [ [ "valid_3_dist = mnist_distance(valid_3_tens, mean3)\nvalid_3_dist, valid_3_dist.shape", "_____no_output_____" ] ], [ [ "Now lets create a simple definition to determine if the prediction is a **3** if the result is `False` then we assume that the image is classified as a **7**", "_____no_output_____" ], [ "If the `mnist_distance` measured againt the `mean` **3** is lower than the `mean` **7** then the function will return `True`", "_____no_output_____" ] ], [ [ "def is_3(x): return mnist_distance(x,mean3) < mnist_distance(x,mean7)", "_____no_output_____" ] ], [ [ "Here by passing in our test image we can see that it returns `True`, but by casting it to a float we can get a value instead", "_____no_output_____" ] ], [ [ "is_3(a_3), is_3(a_3).float()", "_____no_output_____" ] ], [ [ "Taking advantage of `Broadcasting` we can pass the entire tensor stack to the function and we get back an array of predictions", "_____no_output_____" ] ], [ [ "is_3(valid_3_tens)", "_____no_output_____" ] ], [ [ "Lets check the accuracy of our classification application. We are expecting every image tensor in the `valid_3_tens` tensor to return true and all the image tensors in the `valid_7_tens` tensor to return false. We can convert them to a floating point value and take the `mean` to determine the accuracy", "_____no_output_____" ] ], [ [ "accuracy_3s = is_3(valid_3_tens).float() .mean()\naccuracy_7s = (1 - is_3(valid_7_tens).float()).mean()\n\naccuracy_3s,accuracy_7s,(accuracy_3s+accuracy_7s)/2", "_____no_output_____" ] ], [ [ "Wow! It looks like our naive `pixel similarity` application gives us a **95%** accuracy on this task! ", "_____no_output_____" ], [ "We have only used PyTorch for its power tensor operations in this task and proven that with a simple baseline we can determine a highly accurate classifier. Let's see if making a Deep Learning model we can improve the accuracy even further!", "_____no_output_____" ], [ "## Stochastic Gradient Descent (SGD)", "_____no_output_____" ], [ "Up until this point we have a classifier but it really does follow the description by Arthur Samuel ", "_____no_output_____" ], [ "**Suppose we arrange for some automatic means of testing the effectiveness of any current weight assignment in terms of actual performance and provide a mechanism for altering the weight assignment so as to maximize the performance. We need not go into the details of such a procedure to see that it could be made entirely automatic and to see that a machine so programmed would \"learn\" from its experience.**", "_____no_output_____" ], [ "To turn our function into a machine learning classifier we will need:", "_____no_output_____" ], [ "- Initialize the weights.\n- For each image, use these weights to predict whether it appears to be a 3 or a 7.\n- Based on these predictions, calculate how good the model is (its loss).\n- Calculate the gradient, which measures for each weight, how changing that weight would change the loss\n- Step (that is, change) all the weights based on that calculation.\n- Go back to the step 2, and repeat the process.\n- Iterate until you decide to stop the training process (for instance, because the model is good enough or you don't want to wait any longer).", "_____no_output_____" ] ], [ [ "gv('''\ninit->predict->loss->gradient->step->stop\nstep->predict[label=repeat]\n''')", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ], [ "To understand `SGD` a little better lets start with a simpler example using this quadratic function:", "_____no_output_____" ] ], [ [ "def f(x): return x**2", "_____no_output_____" ] ], [ [ "Let's plot what that function looks like:", "_____no_output_____" ] ], [ [ "plot_function(f, 'x', 'x**2')", "/opt/conda/envs/fastai/lib/python3.8/site-packages/fastbook/__init__.py:73: UserWarning: Not providing a value for linspace's steps is deprecated and will throw a runtime error in a future release. This warning will appear only once per process. (Triggered internally at /opt/conda/conda-bld/pytorch_1603729096996/work/aten/src/ATen/native/RangeFactories.cpp:23.)\n x = torch.linspace(min,max)\n" ] ], [ [ "The sequence of steps we described earlier starts by picking some random value for a parameter, and calculating the value of the loss:", "_____no_output_____" ] ], [ [ "plot_function(f, 'x', 'x**2')\nplt.scatter(-1.5, f(-1.5), color='red');", "_____no_output_____" ] ], [ [ "Now we look to see what would happen if we increased or decreased our parameter by a little bit—the adjustment. This is simply the slope at a particular point:", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "We can change our weight by a little in the direction of the slope, calculate our loss and adjustment again, and repeat this a few times. Eventually, we will get to the lowest point on our curve:", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "This basic idea goes all the way back to Isaac Newton, who pointed out that we can optimize arbitrary functions in this way", "_____no_output_____" ], [ "### Calculating Gradients", "_____no_output_____" ], [ "*This is a lot of text from this portion of the book, its **really** important and I getting it right is paramount*", "_____no_output_____" ], [ "\"The one magic step is the bit where we calculate the gradients. As we mentioned, we use calculus as a performance optimization; it allows us to more quickly calculate whether our loss will go up or down when we adjust our parameters up or down. In other words, the gradients will tell us how much we have to change each weight to make our model better.\"", "_____no_output_____" ], [ "Thankfully PyTorch is a very powerful auto-differential library. It utilises the [Chain Rule](https://medium.com/machine-learning-and-math/deep-learning-and-chain-rule-of-calculus-80896a1e91f9) to [calculate the derivative](https://pytorch.org/docs/stable/notes/autograd.html) of our functions.", "_____no_output_____" ], [ "First, let's pick a tensor value which we want gradients at:", "_____no_output_____" ] ], [ [ "xt = tensor(3.).requires_grad_()", "_____no_output_____" ] ], [ [ "Notice the special method requires_grad_? That's the magical incantation we use to tell PyTorch that we want to calculate gradients with respect to that variable at that value. It is essentially tagging the variable, so PyTorch will remember to keep track of how to compute gradients of the other, direct calculations on it that you will ask for.", "_____no_output_____" ], [ "This API might throw you off if you're coming from math or physics. In those contexts the \"gradient\" of a function is just another function (i.e., its derivative), so you might expect gradient-related APIs to give you a new function. **But in deep learning, \"gradients\" usually means the value of a function's derivative at a particular argument value**. The PyTorch API also puts the focus on the argument, not the function you're actually computing the gradients of. It may feel backwards at first, but it's just a different perspective.", "_____no_output_____" ], [ "Now we calculate our function with that value. Notice how PyTorch prints not just the value calculated, but also a note that it has a gradient function it'll be using to calculate our gradients when needed:", "_____no_output_____" ] ], [ [ "yt = f(xt)\nyt", "_____no_output_____" ] ], [ [ "Calculating the derivative value of our function at this input tensor is simple, we just call the `backward` method.", "_____no_output_____" ] ], [ [ "yt.backward()", "_____no_output_____" ] ], [ [ "The \"backward\" here refers to backpropagation, which is the name given to the process of calculating the derivative of each layer.", "_____no_output_____" ], [ "We can now view the gradients by checking the grad attribute of our tensor:", "_____no_output_____" ] ], [ [ "xt.grad", "_____no_output_____" ] ], [ [ "If you remember your high school calculus rules, the derivative of x***2 is 2*x, and we have x=3, so the gradients should be 2*3=6, which is what PyTorch calculated for us!\n\nNow we'll repeat the preceding steps, but with a vector argument for our function:", "_____no_output_____" ] ], [ [ "xt = tensor([3.,4.,10.]).requires_grad_()\nxt", "_____no_output_____" ] ], [ [ "And we'll add sum to our function so it can take a vector (i.e., a rank-1 tensor), and return a scalar (i.e., a rank-0 tensor):", "_____no_output_____" ] ], [ [ "def f(x): return (x**2).sum()\n\nyt = f(xt)\nyt", "_____no_output_____" ] ], [ [ "Our gradients are 2*xt, as we'd expect!", "_____no_output_____" ] ], [ [ "yt.backward()\nxt.grad", "_____no_output_____" ] ], [ [ "The gradients only tell us the slope of our function, they don't actually tell us exactly how far to adjust the parameters. But it gives us some idea of how far; if the slope is very large, then that may suggest that we have more adjustments to do, whereas if the slope is very small, that may suggest that we are close to the optimal value.", "_____no_output_____" ], [ "### Stepping With a Learning Rate", "_____no_output_____" ], [ "Because our gradient only shows the slope, or the direction, in which we need to change our parameters. We will need to figure out how much we move in this direction. This is where we introduce the learning rate. The learning rate is often a number between 0.001 and 0.1, although it could be anything. ", "_____no_output_____" ], [ "Once you've picked a learning rate, you can adjust your parameters using this simple function:", "_____no_output_____" ] ], [ [ "w -= gradient(w) * lr", "_____no_output_____" ] ], [ [ "If you pick a learning rate that's too low, it can mean having to do a lot of steps", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "But picking a learning rate that's too high is even worse—it can actually result in the loss getting worse", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "If the learning rate is too high, it may also \"bounce\" around, rather than actually diverging", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "### An End-to-End SGD Example", "_____no_output_____" ], [ "Let's work through a practical example from end-to-end. First we will make a float tensor that represents our time variable. Each index is a time in seconds", "_____no_output_____" ] ], [ [ "time = torch.arange(0,20).float(); time", "_____no_output_____" ] ], [ [ "Here we create a quadratic function, a function of the form a*(time***2)+(b*time)+c, and add some noiose to simulate realworld measurments", "_____no_output_____" ] ], [ [ "speed = torch.randn(20)*3 + 0.75*(time-9.5)**2 + 1\nplt.scatter(time,speed);", "_____no_output_____" ] ], [ [ "We want to distinguish clearly between the function's input (the time when we are measuring the coaster's speed) and its parameters (the values that define which quadratic we're trying). So, let's collect the parameters in one argument and thus separate the input, t, and the parameters, params, in the function's signature:", "_____no_output_____" ] ], [ [ "def f(t, params):\n a,b,c = params\n return a*(t**2) + (b*t) + c", "_____no_output_____" ] ], [ [ "We need to determine which loss function we would like to use. For continuous data, it's common to use mean squared error:", "_____no_output_____" ] ], [ [ "def mse(preds, targets): return ((preds-targets)**2).mean().sqrt()", "_____no_output_____" ] ], [ [ "#### Step 1: Initialize the parameters", "_____no_output_____" ], [ "First thing we want to do is create a tensor for our parameters and to tell PyTorch that this tensor `requires_grad_()`.", "_____no_output_____" ] ], [ [ "params = torch.randn(3).requires_grad_()", "_____no_output_____" ], [ "#hide\norig_params = params.clone()", "_____no_output_____" ] ], [ [ "#### Step 2: Calculate the predictions", "_____no_output_____" ], [ "Lets pass our time tensor and our params into our function", "_____no_output_____" ] ], [ [ "preds = f(time, params)", "_____no_output_____" ] ], [ [ "Here we are creating a small matplotlib function to show a comparison between our predictions and our observations", "_____no_output_____" ] ], [ [ "def show_preds(preds, ax=None):\n if ax is None: ax=plt.subplots()[1]\n ax.scatter(time, speed)\n ax.scatter(time, to_np(preds), color='red')\n ax.set_ylim(-300,100)", "_____no_output_____" ], [ "show_preds(preds)", "_____no_output_____" ] ], [ [ "#### Step 3: Calculate the loss", "_____no_output_____" ], [ "Now we can use our `L2` or `RMSE` function to determine our loss", "_____no_output_____" ] ], [ [ "loss = mse(preds, speed)\nloss", "_____no_output_____" ] ], [ [ "#### Step 4: Calculate the gradients", "_____no_output_____" ], [ "By performing `Back Propogation` with our `backward()` method on the loss, we can calculate the slope of our gradient and which direction the we need to move in to improve our loss", "_____no_output_____" ] ], [ [ "loss.backward()\nparams.grad", "_____no_output_____" ] ], [ [ "Here we use a small learning rate and multiply that to our parameter's gradient to make sure we move towards the optimal solution over each iteration", "_____no_output_____" ] ], [ [ "params.grad * 1e-5", "_____no_output_____" ], [ "params", "_____no_output_____" ] ], [ [ "#### Step 5: Step the weights. ", "_____no_output_____" ] ], [ [ "lr = 1e-5\nparams.data -= lr * params.grad.data\nparams.grad = None", "_____no_output_____" ] ], [ [ "**Understanding this bit depends on remembering recent history. To calculate the gradients we call backward on the loss. But this loss was itself calculated by mse, which in turn took preds as an input, which was calculated using f taking as an input params, which was the object on which we originally called required_grads_—which is the original call that now allows us to call backward on loss. This chain of function calls represents the mathematical composition of functions, which enables PyTorch to use calculus's chain rule under the hood to calculate these gradients.**", "_____no_output_____" ] ], [ [ "preds = f(time,params)\nmse(preds, speed)", "_____no_output_____" ], [ "show_preds(preds)", "_____no_output_____" ] ], [ [ "Here we simply wrap our previous steps into a function", "_____no_output_____" ], [ "- Make a prediction\n- Calculate the loss\n- Perform back propogation on the loss, see above for details\n- Apply the learning rate \n- Zero our gradients\n- Return our predictions", "_____no_output_____" ] ], [ [ "def apply_step(params, prn=True):\n preds = f(time, params)\n loss = mse(preds, speed)\n loss.backward()\n params.data -= lr * params.grad.data\n params.grad = None\n if prn: print(loss.item())\n return preds", "_____no_output_____" ] ], [ [ "#### Step 6: Repeat the process ", "_____no_output_____" ], [ "Now we simply call our `apply_step` function a number of times, as we can see our loss improves each iteration", "_____no_output_____" ] ], [ [ "for i in range(10): apply_step(params)", "155.75035095214844\n155.4757537841797\n155.20118713378906\n154.92662048339844\n154.65211486816406\n154.37762451171875\n154.1031494140625\n153.82872009277344\n153.55430603027344\n153.27992248535156\n" ], [ "#hide\nparams = orig_params.detach().requires_grad_()", "_____no_output_____" ], [ "_,axs = plt.subplots(1,4,figsize=(12,3))\nfor ax in axs: show_preds(apply_step(params, False), ax)\nplt.tight_layout()", "_____no_output_____" ] ], [ [ "#### Step 7: stop", "_____no_output_____" ], [ "Either we can stop after a certain accuracy or simply after a number of iterations", "_____no_output_____" ], [ "### Summarizing Gradient Descent", "_____no_output_____" ] ], [ [ "gv('''\ninit->predict->loss->gradient->step->stop\nstep->predict[label=repeat]\n''')", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ], [ "- Initialize the weights.\n- For each image, use these weights to predict whether it appears to be a 3 or a 7.\n- Based on these predictions, calculate how good the model is (its loss).\n- Calculate the gradient, which measures for each weight, how changing that weight would change the loss\n- Step (that is, change) all the weights based on that calculation.\n- Go back to the step 2, and repeat the process.\n- Iterate until you decide to stop the training process (for instance, because the model is good enough or you don't want to wait any longer).", "_____no_output_____" ], [ "To summarize, at the beginning, the weights of our model can be random (training from scratch) or come from a pretrained model (transfer learning). In the first case, the output we will get from our inputs won't have anything to do with what we want, and even in the second case, it's very likely the pretrained model won't be very good at the specific task we are targeting. So the model will need to learn better weights.\n\nWe begin by comparing the outputs the model gives us with our targets (we have labeled data, so we know what result the model should give) using a loss function, which returns a number that we want to make as low as possible by improving our weights. To do this, we take a few data items (such as images) from the training set and feed them to our model. We compare the corresponding targets using our loss function, and the score we get tells us how wrong our predictions were. We then change the weights a little bit to make it slightly better.\n\nTo find how to change the weights to make the loss a bit better, we use calculus to calculate the gradients. (Actually, we let PyTorch do it for us!) Let's consider an analogy. Imagine you are lost in the mountains with your car parked at the lowest point. To find your way back to it, you might wander in a random direction, but that probably wouldn't help much. Since you know your vehicle is at the lowest point, you would be better off going downhill. By always taking a step in the direction of the steepest downward slope, you should eventually arrive at your destination. We use the magnitude of the gradient (i.e., the steepness of the slope) to tell us how big a step to take; specifically, we multiply the gradient by a number we choose called the learning rate to decide on the step size. We then iterate until we have reached the lowest point, which will be our parking lot, then we can stop.", "_____no_output_____" ], [ "## The MNIST Loss Function", "_____no_output_____" ], [ "Now we have seen how the mothod works on a simple function, we can now put this into practice on our MNIST **3's** and **7's** problem. As we have our dependent variable x:", "_____no_output_____" ], [ "Remember:\n- **Dependent Variable** == input variables\n- **Independent Varibale** == output variables", "_____no_output_____" ], [ "Our dependent variable in this is example are the images themselves. Here we concatinate the stacked image tensors of the **3's** and the **7's**. Having the image as Matrix is irrelevant we can use the Pytorch `view` method to reshape every tensor to rank 1,", "_____no_output_____" ] ], [ [ "train_x = torch.cat([stacked_threes, stacked_sevens]).view(-1, 28*28)", "_____no_output_____" ] ], [ [ "We also need a label for each of the images, here we can simply create a tensor by combining an array of 1's of length of the number of **3's** and an array of 0's the length of the number of **7's**. We use the PyTorch function `unsqueeze` to transpose the tensor from a vector with 123396 elements into a tensor with 12396 rows and a single column", "_____no_output_____" ] ], [ [ "train_y = tensor([1]*len(threes) + [0]*len(sevens)).unsqueeze(1)\ntrain_x.shape,train_y.shape", "_____no_output_____" ] ], [ [ "Now we need to create a dataset, a dataset needs to be able to be indexable, and at that index we expect a tuple (data, label). Here we are using the python `zip` function to take the `train_x` and `train_y` varibles and combine them into a tuple at each index as described", "_____no_output_____" ] ], [ [ "dset = list(zip(train_x,train_y))\nx,y = dset[0]\nx.shape,y", "_____no_output_____" ] ], [ [ "We need to do the same operations above for our validation set as well\n- Concatinate the images and reshape\n- Create labels and unsqueeze\n- Zip data and label into dataset", "_____no_output_____" ] ], [ [ "valid_x = torch.cat([valid_3_tens, valid_7_tens]).view(-1, 28*28)\nvalid_y = tensor([1]*len(valid_3_tens) + [0]*len(valid_7_tens)).unsqueeze(1)\nvalid_dset = list(zip(valid_x,valid_y))", "_____no_output_____" ] ], [ [ "Here is a simple function that will initialize our parameters with random values. PyTorch `randn` returns a tensor the shape and size of its argument with normalized values between 0 and 1. We can use a variance argument here to scale the random values if necessary, but not in this example. We want to be able to calculate the gradient of this tensor, so we use the `requires_grad` method ", "_____no_output_____" ] ], [ [ "def init_params(size, std=1.0): return (torch.randn(size)*std).requires_grad_()", "_____no_output_____" ] ], [ [ "We create and initialize our weights and bias variables", "_____no_output_____" ] ], [ [ "weights = init_params((28*28,1))", "_____no_output_____" ], [ "bias = init_params(1)", "_____no_output_____" ] ], [ [ "In neural networks, the w in the equation y=w*x+b is called the weights, and the b is called the bias. Together, the weights and bias make up the parameters.", "_____no_output_____" ], [ "We can now calculate a prediction for one image:", "_____no_output_____" ] ], [ [ "(train_x[0]*weights.T).sum() + bias", "_____no_output_____" ] ], [ [ "While we could use a Python for loop to calculate the prediction for each image, that would be very slow. Because Python loops don't run on the GPU, and because Python is a slow language for loops in general, we need to represent as much of the computation in a model as possible using higher-level functions.\n\nIn this case, there's an extremely convenient mathematical operation that calculates w*x for every row of a matrix—it's called matrix multiplication.", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "In Python, matrix multiplication is represented with the @ operator. Let's try it:", "_____no_output_____" ] ], [ [ "def linear1(xb): return xb@weights + bias\npreds = linear1(train_x)\npreds", "_____no_output_____" ] ], [ [ "Lets check how good our random initialization is. Here is make a determination that if a prediction is over 0 it's a **3** else its a **7** and we compare it against our labels", "_____no_output_____" ] ], [ [ "corrects = (preds>0.0).float() == train_y\ncorrects", "_____no_output_____" ] ], [ [ "As you can see, and could have predicted, a random initilization is correct roughly 50% of the time.", "_____no_output_____" ] ], [ [ "corrects.float().mean().item()", "_____no_output_____" ] ], [ [ "So lets change one of our parameters a little bit and see if that changes our result", "_____no_output_____" ] ], [ [ "weights[0] *= 1.0001", "_____no_output_____" ], [ "preds = linear1(train_x)\n((preds>0.0).float() == train_y).float().mean().item()", "_____no_output_____" ] ], [ [ "As we can see above changing one parameter a little has absolutely no affect on the results of our network. What does this mean practically?\n\nBy changing this pixel by some small amount is not sufficient in itself to change the prediction of an image from a **3** to a **7**\n\nBecaue there is no change we cannot calculate a gradient and we cannot make a step to improve our predictions. This is because of the thresholding in our determining our correctness.", "_____no_output_____" ], [ "**In mathematical terms, accuracy is a function that is constant almost everywhere (except at the threshold, 0.5), so its derivative is nil almost everywhere (and infinity at the threshold). This then gives gradients that are 0 or infinite, which are useless for updating the model.**", "_____no_output_____" ], [ "So how do we fix this?", "_____no_output_____" ] ], [ [ "trgts = tensor([1,0,1])\nprds = tensor([0.9, 0.4, 0.2])", "_____no_output_____" ] ], [ [ "Here's a first try at a loss function that measures the distance between predictions and targets:", "_____no_output_____" ] ], [ [ "def mnist_loss(predictions, targets):\n return torch.where(targets==1, 1-predictions, predictions).mean()", "_____no_output_____" ] ], [ [ "**Read the Docs: It's important to learn about PyTorch functions like this, because looping over tensors in Python performs at Python speed, not C/CUDA speed! Try running help(torch.where) now to read the docs for this function, or, better still, look it up on the PyTorch documentation site.**", "_____no_output_____" ] ], [ [ "torch.where(trgts==1, 1-prds, prds)", "_____no_output_____" ] ], [ [ "You can see that this function returns a lower number when predictions are more accurate, when accurate predictions are more confident (higher absolute values), and when inaccurate predictions are less confident. In PyTorch, we always assume that a lower value of a loss function is better. Since we need a scalar for the final loss, mnist_loss takes the mean of the previous tensor:", "_____no_output_____" ] ], [ [ "mnist_loss(prds,trgts)", "_____no_output_____" ] ], [ [ "For instance, if we change our prediction for the one \"false\" target from 0.2 to 0.8 the loss will go down, indicating that this is a better prediction:", "_____no_output_____" ] ], [ [ "mnist_loss(tensor([0.9, 0.4, 0.8]),trgts)", "_____no_output_____" ] ], [ [ "One problem with mnist_loss as currently defined is that it assumes that predictions are always between 0 and 1. We need to ensure, then, that this is actually the case! As it happens, there is a function that does exactly that—let's take a look.", "_____no_output_____" ], [ "### Sigmoid", "_____no_output_____" ], [ "The sigmoid function always outputs a number between 0 and 1. It's defined as follows:", "_____no_output_____" ] ], [ [ "def sigmoid(x): return 1/(1+torch.exp(-x))", "_____no_output_____" ], [ "plot_function(torch.sigmoid, title='Sigmoid', min=-4, max=4)", "_____no_output_____" ] ], [ [ "As you can see, it takes any input value, positive or negative, and smooshes it onto an output value between 0 and 1. It's also a smooth curve that only goes up, which makes it easier for SGD to find meaningful gradients.\n\nLet's update mnist_loss to first apply sigmoid to the inputs:", "_____no_output_____" ] ], [ [ "def mnist_loss(predictions, targets):\n predictions = predictions.sigmoid()\n return torch.where(targets==1, 1-predictions, predictions).mean()", "_____no_output_____" ] ], [ [ "### SGD and Mini-Batches", "_____no_output_____" ], [ "In the context of SGD, \"Minibatch\" means that the gradient is calculated across the entire batch before updating weights. If you are not using a \"minibatch\", every training example in a \"batch\" updates the learning algorithm's parameters independently.", "_____no_output_____" ] ], [ [ "coll = range(15)\ndl = DataLoader(coll, batch_size=5, shuffle=True)\nlist(dl)", "_____no_output_____" ], [ "ds = L(enumerate(string.ascii_lowercase))\nds", "_____no_output_____" ], [ "dl = DataLoader(ds, batch_size=6, shuffle=True)\nlist(dl)", "_____no_output_____" ] ], [ [ "## Putting It All Together", "_____no_output_____" ], [ "First, let's re-initialize our parameters:", "_____no_output_____" ] ], [ [ "weights = init_params((28*28,1))\nbias = init_params(1)", "_____no_output_____" ] ], [ [ "A DataLoader can be created from a Dataset and we can set our bacth size here:", "_____no_output_____" ] ], [ [ "dl = DataLoader(dset, batch_size=256)\nxb,yb = first(dl)\nxb.shape,yb.shape", "_____no_output_____" ] ], [ [ "We'll do the same for the validation set:", "_____no_output_____" ] ], [ [ "valid_dl = DataLoader(valid_dset, batch_size=256)", "_____no_output_____" ] ], [ [ "Let's create a mini-batch of size 4 for testing:", "_____no_output_____" ] ], [ [ "batch = train_x[:4]\nbatch.shape", "_____no_output_____" ], [ "preds = linear1(batch)\npreds", "_____no_output_____" ], [ "loss = mnist_loss(preds, train_y[:4])\nloss", "_____no_output_____" ] ], [ [ "Now we can calculate the gradients:", "_____no_output_____" ] ], [ [ "loss.backward()\nweights.grad.shape,weights.grad.mean(),bias.grad", "_____no_output_____" ] ], [ [ "Let's put this into a function:", "_____no_output_____" ] ], [ [ "def calc_grad(xb, yb, model):\n preds = model(xb)\n loss = mnist_loss(preds, yb)\n loss.backward()", "_____no_output_____" ], [ "calc_grad(batch, train_y[:4], linear1)\nweights.grad.mean(),bias.grad", "_____no_output_____" ] ], [ [ "But look what happens if we call it twice:", "_____no_output_____" ] ], [ [ "calc_grad(batch, train_y[:4], linear1)\nweights.grad.mean(),bias.grad", "_____no_output_____" ] ], [ [ "The gradients have changed! The reason for this is that loss.backward actually adds the gradients of loss to any gradients that are currently stored. So, we have to set the current gradients to 0 first:", "_____no_output_____" ] ], [ [ "weights.grad.zero_()\nbias.grad.zero_();", "_____no_output_____" ] ], [ [ "**Inplace Operations: Methods in PyTorch whose names end in an underscore modify their objects in place. For instance, bias.zero_() sets all elements of the tensor bias to 0.**", "_____no_output_____" ], [ "Our only remaining step is to update the weights and biases based on the gradient and learning rate. When we do so, we have to tell PyTorch not to take the gradient of this step too—otherwise things will get very confusing when we try to compute the derivative at the next batch! \n\n**If we assign to the data attribute of a tensor then PyTorch will not take the gradient of that step. Here's our basic training loop for an epoch:**", "_____no_output_____" ] ], [ [ "def train_epoch(model, lr, params):\n for xb,yb in dl:\n calc_grad(xb, yb, model)\n for p in params:\n p.data -= p.grad*lr\n p.grad.zero_()", "_____no_output_____" ] ], [ [ "We also want to check how we're doing, by looking at the accuracy of the validation set. To decide if an output represents a 3 or a 7, we can just check whether it's greater than 0. So our accuracy for each item can be calculated (using broadcasting, so no loops!) with:", "_____no_output_____" ] ], [ [ "(preds>0.0).float() == train_y[:4]", "_____no_output_____" ] ], [ [ "That gives us this function to calculate our validation accuracy:", "_____no_output_____" ] ], [ [ "def batch_accuracy(xb, yb):\n preds = xb.sigmoid()\n correct = (preds>0.5) == yb\n return correct.float().mean()", "_____no_output_____" ], [ "batch_accuracy(linear1(batch), train_y[:4])", "_____no_output_____" ] ], [ [ "and then put the batches together:", "_____no_output_____" ] ], [ [ "def validate_epoch(model):\n accs = [batch_accuracy(model(xb), yb) for xb,yb in valid_dl]\n return round(torch.stack(accs).mean().item(), 4)", "_____no_output_____" ], [ "validate_epoch(linear1)", "_____no_output_____" ] ], [ [ "That's our starting point. Let's train for one epoch, and see if the accuracy improves:", "_____no_output_____" ] ], [ [ "lr = 1.\nparams = weights,bias\ntrain_epoch(linear1, lr, params)\nvalidate_epoch(linear1)", "_____no_output_____" ] ], [ [ "Then do a few more:", "_____no_output_____" ] ], [ [ "for i in range(20):\n train_epoch(linear1, lr, params)\n print(validate_epoch(linear1), end=' ')", "0.8265 0.89 0.9183 0.9276 0.9398 0.9467 0.9506 0.9525 0.9559 0.9579 0.9598 0.9608 0.9613 0.9618 0.9633 0.9637 0.9647 0.9657 0.9672 0.9677 " ] ], [ [ "Looking good! We're already about at the same accuracy as our \"pixel similarity\" approach, and we've created a general-purpose foundation we can build on. Our next step will be to create an object that will handle the SGD step for us. In PyTorch, it's called an optimizer.", "_____no_output_____" ], [ "### Creating an Optimizer", "_____no_output_____" ], [ "PyTorch's `nn.Linear` does the same thing as our init_params and linear together. It contains both the weights and biases in a single class. Here's how we replicate our model from the previous section:", "_____no_output_____" ] ], [ [ "linear_model = nn.Linear(28*28,1)", "_____no_output_____" ] ], [ [ "Every PyTorch module knows what parameters it has that can be trained; they are available through the parameters method:", "_____no_output_____" ] ], [ [ "w,b = linear_model.parameters()\nw.shape,b.shape", "_____no_output_____" ] ], [ [ "Now, let's create an optimizer:", "_____no_output_____" ] ], [ [ "class BasicOptim:\n def __init__(self,params,lr): self.params,self.lr = list(params),lr\n\n def step(self, *args, **kwargs):\n for p in self.params: p.data -= p.grad.data * self.lr\n\n def zero_grad(self, *args, **kwargs):\n for p in self.params: p.grad = None", "_____no_output_____" ] ], [ [ "We can create our optimizer by passing in the model's parameters:", "_____no_output_____" ] ], [ [ "opt = BasicOptim(linear_model.parameters(), lr)", "_____no_output_____" ], [ "def train_epoch(model):\n for xb,yb in dl:\n calc_grad(xb, yb, model)\n opt.step()\n opt.zero_grad()", "_____no_output_____" ], [ "validate_epoch(linear_model)", "_____no_output_____" ], [ "def train_model(model, epochs):\n for i in range(epochs):\n train_epoch(model)\n print(validate_epoch(model), end=' ')", "_____no_output_____" ], [ "train_model(linear_model, 20)", "0.4932 0.7685 0.8554 0.9136 0.9346 0.9482 0.957 0.9634 0.9658 0.9678 0.9697 0.9717 0.9736 0.9746 0.9761 0.977 0.9775 0.9775 0.978 0.9785 " ] ], [ [ "fastai provides the SGD class which, by default, does the same thing as our BasicOptim:", "_____no_output_____" ] ], [ [ "linear_model = nn.Linear(28*28,1)\nopt = SGD(linear_model.parameters(), lr)\ntrain_model(linear_model, 20)", "0.4932 0.8179 0.8496 0.914 0.9346 0.9482 0.957 0.9619 0.9658 0.9673 0.9692 0.9712 0.9741 0.9751 0.9761 0.9775 0.9775 0.978 0.9785 0.979 " ] ], [ [ "fastai also provides Learner.fit, which we can use instead of train_model. To create a Learner we first need to create a DataLoaders, by passing in our training and validation DataLoaders:", "_____no_output_____" ] ], [ [ "dls = DataLoaders(dl, valid_dl)", "_____no_output_____" ] ], [ [ "To create a Learner without using an application (such as cnn_learner) we need to pass in all the elements that we've created in this chapter: the DataLoaders, the model, the optimization function (which will be passed the parameters), the loss function, and optionally any metrics to print:", "_____no_output_____" ] ], [ [ "learn = Learner(dls, nn.Linear(28*28,1), opt_func=SGD,\n loss_func=mnist_loss, metrics=batch_accuracy)", "_____no_output_____" ], [ "learn.fit(10, lr=lr)", "_____no_output_____" ] ], [ [ "## Adding a Nonlinearity", "_____no_output_____" ], [ "So far we have a general procedure for optimizing the parameters of a function, and we have tried it out on a very boring function: a simple linear classifier. A linear classifier is very constrained in terms of what it can do. To make it a bit more complex (and able to handle more tasks), we need to add something nonlinear between two linear classifiers—this is what gives us a neural network.\n\nHere is the entire definition of a basic neural network:", "_____no_output_____" ] ], [ [ "def simple_net(xb): \n res = xb@w1 + b1\n res = res.max(tensor(0.0))\n res = res@w2 + b2\n return res", "_____no_output_____" ] ], [ [ "That's it! All we have in simple_net is two linear classifiers with a max function between them.\n\nHere, w1 and w2 are weight tensors, and b1 and b2 are bias tensors; that is, parameters that are initially randomly initialized, just like we did in the previous section:", "_____no_output_____" ] ], [ [ "w1 = init_params((28*28,30))\nb1 = init_params(30)\nw2 = init_params((30,1))\nb2 = init_params(1)", "_____no_output_____" ] ], [ [ "The key point about this is that w1 has 30 output activations (which means that w2 must have 30 input activations, so they match). That means that the first layer can construct 30 different features, each representing some different mix of pixels. You can change that 30 to anything you like, to make the model more or less complex.\n\nThat little function res.max(tensor(0.0)) is called a rectified linear unit, also known as ReLU. We think we can all agree that rectified linear unit sounds pretty fancy and complicated... But actually, there's nothing more to it than res.max(tensor(0.0))—in other words, replace every negative number with a zero. This tiny function is also available in PyTorch as F.relu:", "_____no_output_____" ] ], [ [ "plot_function(F.relu)", "_____no_output_____" ] ], [ [ "**Mathematically, we say the composition of two linear functions is another linear function. So, we can stack as many linear classifiers as we want on top of each other, and without nonlinear functions between them, it will just be the same as one linear classifier.**", "_____no_output_____" ] ], [ [ "simple_net = nn.Sequential(\n nn.Linear(28*28,30),\n nn.ReLU(),\n nn.Linear(30,1)\n)", "_____no_output_____" ] ], [ [ "nn.Sequential creates a module that will call each of the listed layers or functions in turn.\n\nnn.ReLU is a PyTorch module that does exactly the same thing as the F.relu function. Most functions that can appear in a model also have identical forms that are modules. Generally, it's just a case of replacing F with nn and changing the capitalization. When using nn.Sequential, PyTorch requires us to use the module version. Since modules are classes, we have to instantiate them, which is why you see nn.ReLU() in this example.\n\nBecause nn.Sequential is a module, we can get its parameters, which will return a list of all the parameters of all the modules it contains. Let's try it out! As this is a deeper model, we'll use a lower learning rate and a few more epochs.", "_____no_output_____" ] ], [ [ "learn = Learner(dls, simple_net, opt_func=SGD,\n loss_func=mnist_loss, metrics=batch_accuracy)", "_____no_output_____" ], [ "learn.fit(40, 0.1)", "_____no_output_____" ], [ "plt.plot(L(learn.recorder.values).itemgot(2));", "_____no_output_____" ] ], [ [ "And we can view the final accuracy:", "_____no_output_____" ] ], [ [ "learn.recorder.values[-1][2]", "_____no_output_____" ] ], [ [ "At this point we have something that is rather magical:\n\nA function that can solve any problem to any level of accuracy (the neural network) given the correct set of parameters\nA way to find the best set of parameters for any function (stochastic gradient descent)", "_____no_output_____" ], [ "### Going Deeper", "_____no_output_____" ], [ "Here what happens when we train an 18-layer model . . .", "_____no_output_____" ] ], [ [ "dls = ImageDataLoaders.from_folder(path)\nlearn = cnn_learner(dls, resnet18, pretrained=False,\n loss_func=F.cross_entropy, metrics=accuracy)\nlearn.fit_one_cycle(1, 0.1)", "_____no_output_____" ] ], [ [ "Almost 100% accuracy!", "_____no_output_____" ], [ "## Jargon Recap", "_____no_output_____" ], [ "- **ReLU :** Function that returns 0 for negative numbers and doesn't change positive numbers.\n- **Mini-batch :** A small group of inputs and labels gathered together in two arrays. A gradient descent step is updated on this batch (rather than a whole epoch).\n- **Forward pass :** Applying the model to some input and computing the predictions.\n- **Loss :** A value that represents how well (or badly) our model is doing.\n- **Gradient :** The derivative of the loss with respect to some parameter of the model.\n- **Backward pass :** Computing the gradients of the loss with respect to all model parameters.\n- **Gradient descent :** Taking a step in the directions opposite to the gradients to make the model parameters a little bit better.\n- **Learning rate :** The size of the step we take when applying SGD to update the parameters of the model.", "_____no_output_____" ], [ "## Questionnaire", "_____no_output_____" ], [ "**How is a grayscale image represented on a computer? How about a color image?**", "_____no_output_____" ], [ "Images are represented by arrays with pixel values representing the content of the image. For greyscale images, a 2-dimensional array is used with the pixels representing the greyscale values, with a range of 256 integers. A value of 0 would represent white, and a value of 255 represents black, and different shades of greyscale in between. For color images, three color channels (red, green, blue) are typicall used, with a separate 256-range 2D array used for each channel. A pixel value of 0 again represents white, with 255 representing solid red, green, or blue. The three 2-D arrays form a final 3-D array (rank 3 tensor) representing the color image.", "_____no_output_____" ], [ "**How are the files and folders in the `MNIST_SAMPLE` dataset structured? Why?**", "_____no_output_____" ], [ "There are two subfolders, train and valid, the former contains the data for model training, the latter contains the data for validating model performance after each training step. Evaluating the model on the validation set serves two purposes: a) to report a human-interpretable metric such as accuracy (in contrast to the often abstract loss functions used for training), b) to facilitate the detection of overfitting by evaluating the model on a dataset it hasn’t been trained on (in short, an overfitting model performs increasingly well on the training set but decreasingly so on the validation set). Of course, every practicioner could generate their own train/validation-split of the data. Public datasets are usually pre-split to simplifiy comparing results between implementations/publications.\n\nEach subfolder has two subsubfolders 3 and 7 which contain the .jpg files for the respective class of images. This is a common way of organizing datasets comprised of pictures. For the full MNIST dataset there are 10 subsubfolders, one for the images for each digit.", "_____no_output_____" ], [ "**Explain how the \"pixel similarity\" approach to classifying digits works.**", "_____no_output_____" ], [ "In the “pixel similarity” approach, we generate an archetype for each class we want to identify. In our case, we want to distinguish images of 3’s from images of 7’s. We define the archetypical 3 as the pixel-wise mean value of all 3’s in the training set. Analoguously for the 7’s. You can visualize the two archetypes and see that they are in fact blurred versions of the numbers they represent.\nIn order to tell if a previously unseen image is a 3 or a 7, we calculate its distance to the two archetypes (here: mean pixel-wise absolute difference). We say the new image is a 3 if its distance to the archetypical 3 is lower than two the archetypical 7.", "_____no_output_____" ], [ "**What is a list comprehension? Create one now that selects odd numbers from a list and doubles them.**", "_____no_output_____" ], [ "Lists (arrays in other programming languages) are often generated using a for-loop. A list comprehension is a Pythonic way of condensing the creation of a list using a for-loop into a single expression. List comprehensions will also often include if clauses for filtering.", "_____no_output_____" ] ], [ [ "lst_in = range(10)\nlst_out = [2*el for el in lst_in if el%2==1]\n# is equivalent to:\nlst_out = []\nfor el in lst_in:\n if el%2==1:\n lst_out.append(2*el)", "_____no_output_____" ] ], [ [ "**What is a \"rank-3 tensor\"?**", "_____no_output_____" ], [ "The rank of a tensor is the number of dimensions it has. An easy way to identify the rank is the number of indices you would need to reference a number within a tensor. A scalar can be represented as a tensor of rank 0 (no index), a vector can be represented as a tensor of rank 1 (one index, e.g., v[i]), a matrix can be represented as a tensor of rank 2 (two indices, e.g., a[i,j]), and a tensor of rank 3 is a cuboid or a “stack of matrices” (three indices, e.g., b[i,j,k]). In particular, the rank of a tensor is independent of its shape or dimensionality, e.g., a tensor of shape 2x2x2 and a tensor of shape 3x5x7 both have rank 3.\nNote that the term “rank” has different meanings in the context of tensors and matrices (where it refers to the number of linearly independent column vectors).", "_____no_output_____" ], [ "**What is the difference between tensor rank and shape? How do you get the rank from the shape?**", "_____no_output_____" ], [ "Rank is the number of axes or dimensions in a tensor; shape is the size of each axis of a tensor.", "_____no_output_____" ], [ "**How do you get the rank from the shape?**", "_____no_output_____" ], [ "The length of a tensor’s shape is its rank.\n\nSo if we have the images of the 3 folder from the MINST_SAMPLE dataset in a tensor called stacked_threes and we find its shape like this.", "_____no_output_____" ] ], [ [ "stacked_threes.shape", "_____no_output_____" ] ], [ [ "torch.Size([6131, 28, 28])\n\nWe just need to find its length to know its rank. This is done as follows.", "_____no_output_____" ] ], [ [ "len(stacked_threes.shape)", "_____no_output_____" ] ], [ [ "3\n\nYou can also get a tensor’s rank directly with ndim .", "_____no_output_____" ] ], [ [ "stacked_threes.ndim", "_____no_output_____" ] ], [ [ "3 ", "_____no_output_____" ], [ "**What are RMSE and L1 norm?**", "_____no_output_____" ], [ "Root mean square error (RMSE), also called the L2 norm, and mean absolute difference (MAE), also called the L1 norm, are two commonly used methods of measuring “distance”. Simple differences do not work because some difference are positive and others are negative, canceling each other out. Therefore, a function that focuses on the magnitudes of the differences is needed to properly measure distances. The simplest would be to add the absolute values of the differences, which is what MAE is. RMSE takes the mean of the square (makes everything positive) and then takes the square root (undoes squaring).", "_____no_output_____" ], [ "**How can you apply a calculation on thousands of numbers at once, many thousands of times faster than a Python loop?**", "_____no_output_____" ], [ "As loops are very slow in Python, it is best to represent the operations as array operations rather than looping through individual elements. If this can be done, then using NumPy or PyTorch will be thousands of times faster, as they use underlying C code which is much faster than pure Python. Even better, PyTorch allows you to run operations on GPU, which will have significant speedup if there are parallel operations that can be done.", "_____no_output_____" ], [ "**Create a 3×3 tensor or array containing the numbers from 1 to 9. Double it. Select the bottom-right four numbers.**", "_____no_output_____" ] ], [ [ "a = torch.Tensor(list(range(1,10))).view(3,3); print(a)", "_____no_output_____" ] ], [ [ "tensor([[1., 2., 3.],\n [4., 5., 6.],\n [7., 8., 9.]])", "_____no_output_____" ] ], [ [ "b = 2*a; \nprint(b)", "_____no_output_____" ] ], [ [ "tensor([[ 2., 4., 6.],\n [ 8., 10., 12.],\n [14., 16., 18.]])", "_____no_output_____" ] ], [ [ "b[1:,1:]", "_____no_output_____" ] ], [ [ "tensor([[10., 12.],\n [16., 18.]])", "_____no_output_____" ], [ "**What is broadcasting?**", "_____no_output_____" ], [ "Scientific/numerical Python packages like NumPy and PyTorch will often implement broadcasting that often makes code easier to write. In the case of PyTorch, tensors with smaller rank are expanded to have the same size as the larger rank tensor. In this way, operations can be performed between tensors with different rank.", "_____no_output_____" ], [ "**Are metrics generally calculated using the training set, or the validation set? Why?**", "_____no_output_____" ], [ "Metrics are generally calculated on a validation set. As the validation set is unseen data for the model, evaluating the metrics on the validation set is better in order to determine if there is any overfitting and how well the model might generalize if given similar data.", "_____no_output_____" ], [ "**What is SGD?**", "_____no_output_____" ], [ "SGD, or stochastic gradient descent, is an optimization algorithm. Specifically, SGD is an algorithm that will update the parameters of a model in order to minimize a given loss function that was evaluated on the predictions and target. The key idea behind SGD (and many optimization algorithms, for that matter) is that the gradient of the loss function provides an indication of how that loss function changes in the parameter space, which we can use to determine how best to update the parameters in order to minimize the loss function. This is what SGD does.", "_____no_output_____" ], [ "**Why does SGD use mini-batches?**", "_____no_output_____" ], [ "We need to calculate our loss function (and our gradient) on one or more data points. We cannot calculate on the whole datasets due to compute limitations and time constraints. If we iterated through each data point, however, the gradient will be unstable and imprecise, and is not suitable for training. As a compromise, we calculate the average loss for a small subset of the dataset at a time. This subset is called a mini-batch. Using mini-batches are also more computationally efficient than single items on a GPU.", "_____no_output_____" ], [ "**What are the seven steps in SGD for machine learning?**", "_____no_output_____" ], [ "- Initialize the weights.\n- For each image, use these weights to predict whether it appears to be a 3 or a 7.\n- Based on these predictions, calculate how good the model is (its loss).\n- Calculate the gradient, which measures for each weight, how changing that weight would change the loss\n- Step (that is, change) all the weights based on that calculation.\n- Go back to the step 2, and repeat the process.\n- Iterate until you decide to stop the training process (for instance, because the model is good enough or you don't want to wait any longer).", "_____no_output_____" ], [ "**How do we initialize the weights in a model?**", "_____no_output_____" ], [ "Random weights work pretty well.", "_____no_output_____" ], [ "**What is \"loss\"?**", "_____no_output_____" ], [ "The loss function will return a value based on the given predictions and targets, where lower values correspond to better model predictions.", "_____no_output_____" ], [ "**Why can't we always use a high learning rate?**", "_____no_output_____" ], [ "The loss may “bounce” around (oscillate) or even diverge, as the optimizer is taking steps that are too large, and updating the parameters faster than it should be.", "_____no_output_____" ], [ "**What is a \"gradient\"?**", "_____no_output_____" ], [ "The gradients tell us how much we have to change each weight to make our model better. It is essentially a measure of how the loss function changes with changes of the weights of the model (the derivative).", "_____no_output_____" ], [ "**Do you need to know how to calculate gradients yourself?**", "_____no_output_____" ], [ "Manual calculation of the gradients are not required, as deep learning libraries will automatically calculate the gradients for you. This feature is known as automatic differentiation. In PyTorch, if requires_grad=True, the gradients can be returned by calling the backward method: a.backward()", "_____no_output_____" ], [ "**Why can't we use accuracy as a loss function?**", "_____no_output_____" ], [ "A loss function needs to change as the weights are being adjusted. Accuracy only changes if the predictions of the model change. So if there are slight changes to the model that, say, improves confidence in a prediction, but does not change the prediction, the accuracy will still not change. Therefore, the gradients will be zero everywhere except when the actual predictions change. The model therefore cannot learn from the gradients equal to zero, and the model’s weights will not update and will not train. A good loss function gives a slightly better loss when the model gives slightly better predictions. Slightly better predictions mean if the model is more confident about the correct prediction. For example, predicting 0.9 vs 0.7 for probability that a MNIST image is a 3 would be slightly better prediction. The loss function needs to reflect that.", "_____no_output_____" ], [ "**Draw the sigmoid function. What is special about its shape?**", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "Sigmoid function is a smooth curve that squishes all values into values between 0 and 1. Most loss functions assume that the model is outputting some form of a probability or confidence level between 0 and 1 so we use a sigmoid function at the end of the model in order to do this.", "_____no_output_____" ], [ "**What is the difference between a loss function and a metric?**", "_____no_output_____" ], [ "The key difference is that metrics drive human understanding and losses drive automated learning. In order for loss to be useful for training, it needs to have a meaningful derivative. Many metrics, like accuracy are not like that. Metrics instead are the numbers that humans care about, that reflect the performance of the model.", "_____no_output_____" ], [ "**What is the function to calculate new weights using a learning rate?**", "_____no_output_____" ], [ "The optimizer step function", "_____no_output_____" ], [ "**What does the `DataLoader` class do?**", "_____no_output_____" ], [ "The DataLoader class can take any Python collection and turn it into an iterator over many batches.", "_____no_output_____" ], [ "**Write pseudocode showing the basic steps taken in each epoch for SGD.**", "_____no_output_____" ] ], [ [ "for x,y in dl:\n pred = model(x)\n loss = loss_func(pred, y)\n loss.backward()\n parameters -= parameters.grad * lr", "_____no_output_____" ] ], [ [ "**Create a function that, if passed two arguments `[1,2,3,4]` and `'abcd'`, returns `[(1, 'a'), (2, 'b'), (3, 'c'), (4, 'd')]`. What is special about that output data structure?**", "_____no_output_____" ] ], [ [ "def func(a,b): return list(zip(a,b))", "_____no_output_____" ] ], [ [ "This data structure is useful for machine learning models when you need lists of tuples where each tuple would contain input data and a label.", "_____no_output_____" ], [ "**What does `view` do in PyTorch?**", "_____no_output_____" ], [ "It changes the shape of a Tensor without changing its contents.", "_____no_output_____" ], [ "**What are the \"bias\" parameters in a neural network? Why do we need them?**", "_____no_output_____" ], [ "Without the bias parameters, if the input is zero, the output will always be zero. Therefore, using bias parameters adds additional flexibility to the model.", "_____no_output_____" ], [ "**What does the `@` operator do in Python?**", "_____no_output_____" ], [ "This is the matrix multiplication operator.", "_____no_output_____" ], [ "**What does the `backward` method do?**", "_____no_output_____" ], [ "This method returns the current gradients.", "_____no_output_____" ], [ "**Why do we have to zero the gradients?**", "_____no_output_____" ], [ "PyTorch will add the gradients of a variable to any previously stored gradients. If the training loop function is called multiple times, without zeroing the gradients, the gradient of current loss would be added to the previously stored gradient value.", "_____no_output_____" ], [ "**What information do we have to pass to `Learner`?**", "_____no_output_____" ], [ "We need to pass in the DataLoaders, the model, the optimization function, the loss function, and optionally any metrics to print.", "_____no_output_____" ], [ "**Show Python or pseudocode for the basic steps of a training loop.**", "_____no_output_____" ] ], [ [ "def train_epoch(model, lr, params):\n for xb,yb in dl:\n calc_grad(xb, yb, model)\n for p in params:\n p.data -= p.grad*lr\n p.grad.zero_()\n\nfor i in range(20):\n train_epoch(model, lr, params)", "_____no_output_____" ] ], [ [ "**What is \"ReLU\"? Draw a plot of it for values from `-2` to `+2`.**", "_____no_output_____" ], [ "ReLU just means “replace any negative numbers with zero”. It is a commonly used activation function.", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "**What is an \"activation function\"?**", "_____no_output_____" ], [ "The activation function is another function that is part of the neural network, which has the purpose of providing non-linearity to the model. The idea is that without an activation function, we just have multiple linear functions of the form y=mx+b. However, a series of linear layers is equivalent to a single linear layer, so our model can only fit a line to the data. By introducing a non-linearity in between the linear layers, this is no longer true. Each layer is somewhat decoupled from the rest of the layers, and the model can now fit much more complex functions. In fact, it can be mathematically proven that such a model can solve any computable problem to an arbitrarily high accuracy, if the model is large enough with the correct weights. This is known as the universal approximation theorem.", "_____no_output_____" ], [ "**What's the difference between `F.relu` and `nn.ReLU`?**", "_____no_output_____" ], [ "F.relu is a Python function for the relu activation function. On the other hand, nn.ReLU is a PyTorch module. This means that it is a Python class that can be called as a function in the same way as F.relu.", "_____no_output_____" ], [ "**The universal approximation theorem shows that any function can be approximated as closely as needed using just one nonlinearity. So why do we normally use more?**", "_____no_output_____" ], [ "There are practical performance benefits to using more than one nonlinearity. We can use a deeper model with less number of parameters, better performance, faster training, and less compute/memory requirements.", "_____no_output_____" ], [ "### Further Research", "_____no_output_____" ], [ "1. Create your own implementation of `Learner` from scratch, based on the training loop shown in this chapter.\n1. Complete all the steps in this chapter using the full MNIST datasets (that is, for all digits, not just 3s and 7s). This is a significant project and will take you quite a bit of time to complete! You'll need to do some of your own research to figure out how to overcome some obstacles you'll meet on the way.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ [ [ "# 08 - Common problems & bad data situations\n\n<a rel=\"license\" href=\"http://creativecommons.org/licenses/by/4.0/\"><img alt=\"Creative Commons Licence\" style=\"border-width:0\" src=\"https://i.creativecommons.org/l/by/4.0/88x31.png\" title='This work is licensed under a Creative Commons Attribution 4.0 International License.' align=\"right\"/></a>\n\nIn this notebook, we will revise common problems that might come up when dealing with real-world data.\n\nMaintainers: [@thempel](https://github.com/thempel), [@cwehmeyer](https://github.com/cwehmeyer), [@marscher](https://github.com/marscher), [@psolsson](https://github.com/psolsson)\n\n**Remember**:\n- to run the currently highlighted cell, hold <kbd>&#x21E7; Shift</kbd> and press <kbd>&#x23ce; Enter</kbd>;\n- to get help for a specific function, place the cursor within the function's brackets, hold <kbd>&#x21E7; Shift</kbd>, and press <kbd>&#x21E5; Tab</kbd>;\n- you can find the full documentation at [PyEMMA.org](http://www.pyemma.org).\n\n---\n\nMost problems in Markov modeling of MD data arise from bad sampling combined with a poor discretization.\nFor estimating a Markov model, it is required to have a connected data set,\ni.e., we must have observed each process we want to describe in both directions.\nPyEMMA checks if this requirement is fulfilled but, however, in certain situations this might be less obvious.", "_____no_output_____" ] ], [ [ "%matplotlib inline\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport mdshare\nimport pyemma", "_____no_output_____" ] ], [ [ "## Case 1: preprocessed, two-dimensional data (toy model)\n\n### well-sampled double-well potential\n\nLet's again have a look at the double-well potential.\nSince we are only interested in the problematic situations here,\nwe will simplify our data a bit and work with a 1D projection.", "_____no_output_____" ] ], [ [ "file = mdshare.fetch('hmm-doublewell-2d-100k.npz', working_directory='data')\nwith np.load(file) as fh:\n data = [fh['trajectory'][:, 1]]", "_____no_output_____" ] ], [ [ "Since this particular example is simple enough, we can define a plotting function that combines histograms with trajectory data:", "_____no_output_____" ] ], [ [ "def plot_1D_histogram_trajectories(data, cluster=None, max_traj_length=200, ax=None):\n if ax is None:\n fig, ax = plt.subplots()\n for n, _traj in enumerate(data):\n ax.hist(_traj, bins=30, alpha=.33, density=True, color='C{}'.format(n));\n ylims = ax.get_ylim()\n xlims = ax.get_xlim()\n for n, _traj in enumerate(data):\n ax.plot(\n _traj[:min(len(_traj), max_traj_length)], \n np.linspace(*ylims, min(len(_traj), max_traj_length)), \n alpha=0.6, color='C{}'.format(n), label='traj {}'.format(n))\n if cluster is not None:\n ax.plot(\n cluster.clustercenters[cluster.dtrajs[n][:min(len(_traj), max_traj_length)], 0], \n np.linspace(*ylims, min(len(_traj), max_traj_length)), \n '.-', alpha=.6, label='dtraj {}'.format(n), linewidth=.3)\n ax.annotate(\n '', xy=(0.8500001 * xlims[1], 0.7 * ylims[1]), xytext=(0.85 * xlims[1], 0.3 * ylims[1]),\n arrowprops=dict(fc='C0', ec='None', alpha=0.6, width=2))\n ax.text(0.86 * xlims[1], 0.5 * ylims[1], '$x(time)$', ha='left', va='center', rotation=90)\n ax.set_xlabel('TICA coordinate')\n ax.set_ylabel('histogram counts & trajectory time')\n ax.legend(loc=2)", "_____no_output_____" ] ], [ [ "As a reference, we visualize the histogram of this well-sampled trajectory along with the first $200$ steps (left panel) and the MSM implied timescales (right panel):", "_____no_output_____" ] ], [ [ "fig, axes = plt.subplots(1, 2, figsize=(10, 4))\n\ncluster = pyemma.coordinates.cluster_regspace(data, dmin=0.05)\n\nplot_1D_histogram_trajectories(data, cluster=cluster, ax=axes[0])\n\nlags = [i + 1 for i in range(10)]\nits = pyemma.msm.its(cluster.dtrajs, lags=lags)\npyemma.plots.plot_implied_timescales(its, marker='o', ax=axes[1], nits=4)\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "We see a nice, reversibly connected trajectory.\nThat means we have sampled transitions between the basins in both directions that are correctly resolved by the discretization.\nAs we see from the almost perfect overlay of discrete and continuous trajectory, nearly no discretization error is made. \n\n### irreversibly connected double-well trajectories\n\nIn MD simulations, we often face the problem that a process is sampled only in one direction.\nFor example, consider protein-protein binding.\nThe unbinding might take on the order of seconds to minutes and is thus difficult to sample.\nWe will have a look what happens with the MSM in this case. \n\nOur example are two trajectories sampled from a double-well potential, each started in a different basin.\nThey will be color coded.", "_____no_output_____" ] ], [ [ "file = mdshare.fetch('doublewell_oneway.npy', working_directory='data')\ndata = [trj for trj in np.load(file)]\n\nplot_1D_histogram_trajectories(data, max_traj_length=data[0].shape[0])", "_____no_output_____" ] ], [ [ "We note that the orange trajectory does not leave its potential well while the blue trajectory does overcome the barrier exactly once.\n\n⚠️ Even though we have sampled one direction of the process,\nwe do not sample the way out of one of the potential wells, thus effectively finding a sink state in our data. \n\nLet's have a look at the MSM.\nSince in higher dimensions, we often face the problem of poor discretization,\nwe will simulate this situation by using too few cluster centers.", "_____no_output_____" ] ], [ [ "cluster_fine = pyemma.coordinates.cluster_regspace(data, dmin=0.1)\ncluster_poor = pyemma.coordinates.cluster_regspace(data, dmin=0.7)\nprint(cluster_fine.n_clusters, cluster_poor.n_clusters)", "_____no_output_____" ], [ "fig, axes = plt.subplots(2, 2, figsize=(10, 8), sharey='col')\nfor cluster, ax in zip([cluster_poor, cluster_fine], axes):\n plot_1D_histogram_trajectories(data, cluster=cluster, max_traj_length=data[0].shape[0], ax=ax[0])\n its = pyemma.msm.its(cluster.dtrajs, lags=[1, 10, 100, 200, 300, 500, 800, 1000])\n pyemma.plots.plot_implied_timescales(its, marker='o', ax=ax[1], nits=4)\naxes[0, 0].set_title('poor discretization')\naxes[1, 0].set_title('fine discretization')\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "#### What do we see? \n\n1) We observe implied timescales that even look converged in the fine discretization case. \n\n2) With poor clustering, the process cannot be resolved any more, i.e., the ITS does not convergence before the lag time exceeds the implied time scale. \n\nThe obvious question is, what is the process that can be observed in the fine discretization case?\nPyEMMA checks for disconnectivity and thus should not find the process between the two wells.\nWe follow this question by taking a look at the first eigenvector, which corresponds to that process.", "_____no_output_____" ] ], [ [ "msm = pyemma.msm.estimate_markov_model(cluster_fine.dtrajs, 200)\nfig, ax = plt.subplots()\nax.plot(\n cluster_fine.clustercenters[msm.active_set, 0],\n msm.eigenvectors_right()[:, 1],\n 'o:',\n label='first eigvec')\ntx = ax.twinx()\ntx.hist(np.concatenate(data), bins=30, alpha=0.33)\ntx.set_yticklabels([])\ntx.set_yticks([])\nfig.legend()\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "We observe a process which is entirely taking place in the left potential well.\nHow come?\nPyEMMA estimates MSMs only on the largest connected set because they are only defined on connected sets.\nIn this particular example, the largest connected set is the microstates in the left potential well.\nThat means that we find a transition between the right and the left side of this well.\nThis is not wrong, it might just be non-informative or even irrelevant. \n\nThe set of microstates which is used for the MSM estimation is stored in the MSM object `msm` and can be retrieved via `.active_set`.", "_____no_output_____" ] ], [ [ "print('Active set: {}'.format(msm.active_set))\nprint('Active state fraction: {:.2}'.format(msm.active_state_fraction))", "_____no_output_____" ] ], [ [ "In this example we clearly see that some states are missing.\n\n### disconnected double-well trajectories with cross-overs\n\nThis example covers the worst-case scenario.\nWe have two trajectories that live in two separated wells and never transition to the other one.\nDue to a very bad clustering, we believe that the data is connected.\nThis can happen if we cluster a large dataset in very high dimensions where it is especially difficult to debug. ", "_____no_output_____" ] ], [ [ "file = mdshare.fetch('doublewell_disconnected.npy', working_directory='data')\ndata = [trj for trj in np.load(file)]\n\nplot_1D_histogram_trajectories(data, max_traj_length=data[0].shape[0])", "_____no_output_____" ] ], [ [ "We, again, compare a reasonable to a deliberately poor discretization:", "_____no_output_____" ] ], [ [ "cluster_fine = pyemma.coordinates.cluster_regspace(data, dmin=0.1)\ncluster_poor = pyemma.coordinates.cluster_regspace(data, dmin=0.7)\nprint(cluster_fine.n_clusters, cluster_poor.n_clusters)", "_____no_output_____" ], [ "fig, axes = plt.subplots(2, 2, figsize=(10, 8), sharey='col')\nfor cluster, ax in zip([cluster_poor, cluster_fine], axes):\n plot_1D_histogram_trajectories(data, cluster=cluster, max_traj_length=data[0].shape[0], ax=ax[0])\n its = pyemma.msm.its(cluster.dtrajs, lags=[1, 10, 100, 200, 300, 500, 800, 1000])\n pyemma.plots.plot_implied_timescales(its, marker='o', ax=ax[1], nits=4)\naxes[0, 0].set_title('poor discretization')\naxes[1, 0].set_title('fine discretization')\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "#### What do we see?\n\n1) With the fine discretization, we observe some timescales that are converged. These are most probably processes within one of the wells, similar to the ones we saw before.\n\n2) The poor discretization induces a large error and describes artificial short visits to the other basin.\n\n3) The timescales in the poor discretization are much higher but not converged. \n\nThe reason for the high timescales in 3) are in fact the artificial cross-over events created by the poor discretization.\nThis process was not actually sampled and is an artifact of bad clustering.\nLet's look at it in more detail and see what happens if we estimate an MSM and even compute metastable states with PCCA++.", "_____no_output_____" ] ], [ [ "msm = pyemma.msm.estimate_markov_model(cluster_poor.dtrajs, 200)\n\nnstates = 2\nmsm.pcca(nstates)\n\nindex_order = np.argsort(cluster_poor.clustercenters[:, 0])\n\nfig, axes = plt.subplots(1, 3, figsize=(12, 3))\naxes[0].plot(\n cluster_poor.clustercenters[index_order, 0],\n msm.eigenvectors_right()[index_order, 1],\n 'o:',\n label='1st eigvec')\naxes[0].set_title('first eigenvector')\nfor n, metastable_distribution in enumerate(msm.metastable_distributions):\n axes[1].step(\n cluster_poor.clustercenters[index_order, 0],\n metastable_distribution[index_order],\n ':', \n label='md state {}'.format(n + 1),\n where='mid')\naxes[1].set_title('metastable distributions (md)')\naxes[2].step(\n cluster_poor.clustercenters[index_order, 0],\n msm.pi[index_order],\n 'k--',\n label='$\\pi$',\n where='mid')\naxes[2].set_title('stationary distribution $\\pi$')\nfor ax in axes:\n tx = ax.twinx()\n tx.hist(np.concatenate(data), bins=30, alpha=0.33)\n tx.set_yticklabels([])\n tx.set_yticks([])\nfig.legend(loc=7)\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "We observe that the first eigenvector represents a process that does not exist, i.e., is an artifact.\nNevertheless, the PCCA++ algorithm can separate metastable states in a way we would expect.\nIt finds the two disconnected states. However, the stationary distribution yields arbitrary results. \n\n#### How to detect disconnectivity?\n\nGenerally, hidden Markov models (HMMs) are much more reliable because they come with an additional layer of hidden states.\nCross-over events are thus unlikely to be counted as \"real\" transitions.\nThus, it is a good idea to estimate an HMM.\nWhat happens if we try to estimate a two state HMM on the same, poorly discretized data? \n\n⚠️ It is important to note that the HMM estimation is initialized from the PCCA++ metastable states that we already analyzed.", "_____no_output_____" ] ], [ [ "hmm = pyemma.msm.estimate_hidden_markov_model(cluster_poor.dtrajs, nstates, msm.lag)", "_____no_output_____" ] ], [ [ "We are getting an error message which already explains what is going wrong, i.e.,\nthat the (macro-) states are not connected and thus no unique stationary distribution can be estimated.\nThis is equivalent to having two eigenvalues of magnitude 1 or an implied timescale of infinity which is what we observe in the implied timescales plot.", "_____no_output_____" ] ], [ [ "its = pyemma.msm.timescales_hmsm(cluster_poor.dtrajs, nstates, lags=[1, 3, 4, 10, 100])\npyemma.plots.plot_implied_timescales(its, marker='o', ylog=True);", "_____no_output_____" ] ], [ [ "As we see, the requested timescales above $4$ steps could not be computed because the underlying HMM is disconnected,\ni.e., the corresponding timescales are infinity.\nThe implied timescales that could be computed are most likely the same process that we observed from the fine clustering before, i.e., jumps within one basin.\n\nIn general, it is a non-trivial problem to show that processes were not sampled reversibly.\nIn our experience, HMMs are a good choice here, even though situations can occur where they might not detect the problem as easily as in this example. \n\n<a id=\"poorly_sampled_dw\"></a>\n### poorly sampled double-well trajectories\n\nLet's now assume that everything worked out fine but our sampling is somewhat poor.\nThis is a realistic scenario when dealing with large systems that were well-sampled but still contain only few events of interest.\nWe expect that our trajectories are just long enough to sample a certain process but are too short to capture them with a large lag time.\nTo rule out discretization issues and to make the example clear, we use the full data set for discretization.", "_____no_output_____" ] ], [ [ "file = mdshare.fetch('hmm-doublewell-2d-100k.npz', working_directory='data')\nwith np.load(file) as fh:\n data = [fh['trajectory'][:, 1]]\ncluster = pyemma.coordinates.cluster_regspace(data, dmin=0.05)", "_____no_output_____" ] ], [ [ "We want to simulate a process that happens on a timescale that is on the order of magnitude of the trajectory length.\nTo do so, we choose `n_trajs` chunks from the full data set that contain `traj_length` steps by splitting the original trajectory:", "_____no_output_____" ] ], [ [ "traj_length = 10\nn_trajs = 50\n\ndata_short_trajs = list(data[0].reshape((data[0].shape[0] // traj_length, traj_length)))[:n_trajs]\ndtrajs_short = list(cluster.dtrajs[0].reshape((data[0].shape[0] // traj_length, traj_length)))[:n_trajs]", "_____no_output_____" ] ], [ [ "Now, let's plot the trajectories (left panel) and estimate implied timescales (right panel) as above.\nSince we know the true ITS of this process, we visualize it as a dotted line.", "_____no_output_____" ] ], [ [ "fig, axes = plt.subplots(1, 2, figsize=(10, 4))\n\nfor n, _traj in enumerate(data_short_trajs):\n axes[0].plot(_traj, np.linspace(0, 1, _traj.shape[0]) + n)\n\nlags = [i + 1 for i in range(9)]\n\nits = pyemma.msm.its(dtrajs_short, lags=lags)\npyemma.plots.plot_implied_timescales(its, marker='o', ax=axes[1], nits=1)\nits_reference = pyemma.msm.its(cluster.dtrajs, lags=lags)\npyemma.plots.plot_implied_timescales(its_reference, linestyle=':', ax=axes[1], nits=1)\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "We note that the slowest process is clearly contained in the data chunks and is reversibly sampled (left panel, short trajectory pieces color coded and stacked).\nDue to very short trajectories, we find that this process can only be captured at a very short MSM lag time (right panel).\nAbove that interval, the slowest timescale diverges.\nLuckily, here we know that it is already converged at $\\tau = 1$, so we estimate an MSM:", "_____no_output_____" ] ], [ [ "msm_short_trajectories = pyemma.msm.estimate_markov_model(dtrajs_short, 1)", "_____no_output_____" ] ], [ [ "Let's now have a look at the CK-test:", "_____no_output_____" ] ], [ [ "pyemma.plots.plot_cktest(msm_short_trajectories.cktest(2), marker='.');", "_____no_output_____" ] ], [ [ "As already discussed, we cannot expect new estimates above a certain lag time to agree with the model prediction due to too short trajectories.\nIndeed, we find that new estimates and model predictions diverge at very high lag times.\nThis does not necessarily mean that the model at $\\tau=1$ is wrong and in this particular case,\nwe can even explain the divergence and find that it fits to the implied timescales divergence. \n\nThis example mirrors another incarnation of the sampling problem: Working with large systems,\nwe often have comparably short trajectories with few rare events.\nThus, implied timescales convergence can often be achieved only in a certain interval and CK-tests will not converge up to arbitrary multiples of the lag time.\nIt is the responsibility of the modeler to interpret these results and to ensure that a valid model can be obtained from the data.\n\nPlease note that this is only a special case of a failed CK test.\nMore general information about CK tests and what it means if it fails are explained in\n[Notebook 03 ➜ 📓](03-msm-estimation-and-validation.ipynb).\n\n## Case 2: low-dimensional molecular dynamics data (alanine dipeptide)\n\nIn this example, we will show how an ill-conducted TICA analysis can yield results that look metastable in the 2D histogram,\nbut in fact are not describing the slow dynamics.\nPlease note that this was deliberately broken with a nonsensical TICA-lagtime of almost trajectory length, which is 250 ns.\n\nWe start off with adding all atom coordinates.\nThat is a non-optimal choice because it artificially blows up the dimensionality,\nbut might still be a reasonable choice depending on the problem.\nA well-conducted TICA projection can extract the slow coordinates, as we will see at the end of this example.", "_____no_output_____" ] ], [ [ "pdb = mdshare.fetch('alanine-dipeptide-nowater.pdb', working_directory='data')\nfiles = mdshare.fetch('alanine-dipeptide-*-250ns-nowater.xtc', working_directory='data')\nfeat = pyemma.coordinates.featurizer(pdb)\n\nfeat.add_all()\ndata = pyemma.coordinates.load(files, features=feat)", "_____no_output_____" ] ], [ [ "TICA analysis is conducted with an extremely high lag time of almost $249.9$ ns. We map down to two dimensions.", "_____no_output_____" ] ], [ [ "tica = pyemma.coordinates.tica(data, lag=data[0].shape[0] - 100, dim=2)\ntica_output = tica.get_output()\n\npyemma.plots.plot_free_energy(*np.concatenate(tica_output).T, legacy=False);", "_____no_output_____" ] ], [ [ "In the free energy plot, we recognize two defined basins that are nicely separated by the first TICA component. We thus continue with a discretization of this space and estimate MSM implied timescales.", "_____no_output_____" ] ], [ [ "cluster = pyemma.coordinates.cluster_kmeans(tica_output, k=200, max_iter=30, stride=100)", "_____no_output_____" ], [ "its = pyemma.msm.its(cluster.dtrajs, lags=[1, 5, 10, 20, 30, 50])\npyemma.plots.plot_implied_timescales(its, marker='o', units='ps', nits=3);", "_____no_output_____" ] ], [ [ "Indeed, we observe a converged implied timescale.\nIn this example we already know that it is way lower than expected,\nbut in the general case we are unaware of the real dynamics of the system. \n\nThus, we estimate an MSM at lag time $20 $ ps.\nCoarse graining and validation will be done with $2$ metastable states since we found $2$ basins in the free energy landscape and have one slow process in the ITS plot.", "_____no_output_____" ] ], [ [ "msm = pyemma.msm.estimate_markov_model(cluster.dtrajs, 20)\n\nnstates = 2\nmsm.pcca(nstates);", "_____no_output_____" ], [ "stride = 10\nmetastable_trajs_strided = [msm.metastable_assignments[dtrj[::stride]] for dtrj in cluster.dtrajs]\ntica_output_strided = [i[::stride] for i in tica_output]\n_, _, misc = pyemma.plots.plot_state_map(*np.concatenate(tica_output_strided).T, \n np.concatenate(metastable_trajs_strided));\nmisc['cbar'].set_ticklabels(range(1, nstates + 1)) # set state numbers 1 ... nstates", "_____no_output_____" ] ], [ [ "As we see, the PCCA++ algorithm is perfectly able to separate the two basins.\nLet's go on with a Chapman-Kolmogorow validation.", "_____no_output_____" ] ], [ [ "pyemma.plots.plot_cktest(msm.cktest(nstates), units='ps');", "_____no_output_____" ] ], [ [ "Congratulations, we have estimated a well-validated MSM.\nThe only question remaining is: What does it actually describe?\nFor this, we usually extract representative structures as described in [Notebook 00 ➜ 📓](00-pentapeptide-showcase.ipynb).\nWe will not do this here but look at the metastable trajectories instead.\n\n#### What could be wrong about it?\n\nLet's have a look at the trajectories as assigned to PCCA++ metastable states.\nWe have already computed them before but not looked at their time dependence.", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots(1, 1, figsize=(15, 6), sharey=True, sharex=True)\nax_yticks_labels = []\nfor n, pcca_traj in enumerate(metastable_trajs_strided):\n ax.plot(range(len(pcca_traj)), msm.n_metastable * n + pcca_traj, color='k', linewidth=0.3)\n ax.scatter(range(len(pcca_traj)), msm.n_metastable * n + pcca_traj, c=pcca_traj, s=0.1)\n ax_yticks_labels.append(((msm.n_metastable * (2 * n + 1) - 1) / 2, n + 1))\nax.set_yticks([l[0] for l in ax_yticks_labels])\nax.set_yticklabels([str(l[1]) for l in ax_yticks_labels])\nax.set_ylabel('Trajectory #')\nax.set_xlabel('time / {} ps'.format(stride))\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "#### What do we see?\nThe above figure shows the metastable states visited by the trajectory over time.\nEach metastable state is color-coded, the trajectory is shown by the black line.\nThis is clearly not a metastable trajectory as we would have expected. \n\nWhat did we do wrong?\nLet's have a look at the TICA trajectories, not only the histogram!", "_____no_output_____" ] ], [ [ "fig, axes = plt.subplots(2, 3, figsize=(12, 6), sharex=True, sharey='row')\n\nfor n, trj in enumerate(tica_output):\n for dim, traj1d in enumerate(trj.T):\n axes[dim, n].plot(traj1d[::stride], linewidth=.5)\nfor ax in axes[1]:\n ax.set_xlabel('time / {} ps'.format(stride))\nfor dim, ax in enumerate(axes[:, 0]):\n ax.set_ylabel('IC {}'.format(dim + 1))\nfor n, ax in enumerate(axes[0]):\n ax.set_title('Trajectory # {}'.format(n + 1))\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "This is essentially noise, so it is not surprising that the metastable trajectories do not show significant metastability.\nThe MSM nevertheless found a process in the above TICA components which, however,\ndoes not seem to describe any of the slow dynamics.\nThus, the model is not wrong, it is just not informative. \n\nAs we see in this example, it can be instructive to keep the trajectories in mind and not to rely on the histograms alone.\n\n⚠️ Histograms are no proof of metastability,\nthey can only give us a hint towards defined states in a multi-dimensional state space which can be metastable.\n\n#### How to fix it?\n\nIn this particular example, we already know the issue:\nthe TICA lag time was deliberately chosen way too high.\nThat's easy to fix.\n\nLet's now have a look at how the metastable trajectories should look for a decent model such as the one estimated in [Notebook 05 ➜ 📓](05-pcca-tpt.ipynb).\nWe will take the same input data,\ndo a TICA transform with a realistic lag time of $10$ ps,\nand coarse grain into $2$ metastable states in order to compare with the example above.", "_____no_output_____" ] ], [ [ "tica = pyemma.coordinates.tica(data, lag=10, dim=2)\ntica_output = tica.get_output()\ncluster = pyemma.coordinates.cluster_kmeans(tica_output, k=200, max_iter=30, stride=100)\n\npyemma.plots.plot_free_energy(*np.concatenate(tica_output).T, legacy=False);", "_____no_output_____" ] ], [ [ "As wee see, TICA yields a very nice state separation.\nWe will see that these states are in fact metastable.", "_____no_output_____" ] ], [ [ "msm = pyemma.msm.estimate_markov_model(cluster.dtrajs, lag=20)\nmsm.pcca(nstates);", "_____no_output_____" ], [ "metastable_trajs_strided = [msm.metastable_assignments[dtrj[::stride]] for dtrj in cluster.dtrajs]", "_____no_output_____" ], [ "stride = 10\ntica_output_strided = [i[::stride] for i in tica_output]\n_, _, misc = pyemma.plots.plot_state_map(*np.concatenate(tica_output_strided).T, \n np.concatenate(metastable_trajs_strided));\nmisc['cbar'].set_ticklabels(range(1, nstates + 1)) # set state numbers 1 ... nstates", "_____no_output_____" ] ], [ [ "We note that PCCA++ separates the two basins of the free energy plot.\nLet's have a look at the metastable trajectories:", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots(1, 1, figsize=(12, 6), sharey=True, sharex=True)\nax_yticks_labels = []\nfor n, pcca_traj in enumerate(metastable_trajs_strided):\n ax.plot(range(len(pcca_traj)), msm.n_metastable * n + pcca_traj, color='k', linewidth=0.3)\n ax.scatter(range(len(pcca_traj)), msm.n_metastable * n + pcca_traj, c=pcca_traj, s=0.1)\n ax_yticks_labels.append(((msm.n_metastable * (2 * n + 1) - 1) / 2, n + 1))\nax.set_yticks([l[0] for l in ax_yticks_labels])\nax.set_yticklabels([str(l[1]) for l in ax_yticks_labels])\nax.set_ylabel('Trajectory #')\nax.set_xlabel('time / {} ps'.format(stride))\nfig.tight_layout()", "_____no_output_____" ] ], [ [ "These trajectories show the expected behavior of a metastable trajectory,\ni.e., it does not quickly jump back and forth between the states.\n\n## Wrapping up\n\nIn this notebook, we have learned about some problems that can arise when estimating MSMs with \"real world\" data at simple examples.\nIn detail, we have seen\n- irreversibly connected dynamics and what it means for MSM estimation,\n- fully disconnected trajectories and how to identify them,\n- connected but poorly sampled trajectories and how convergence looks in this case,\n- ill-conducted TICA analysis and what it yields.\n\nThe most important lesson from this tutorial is that histograms, which are usually calculated in a projected space, are not a sufficient means of identifying metastability or connectedness.\nIt is crucial to remember that the underlying trajectories play the role of ground truth for the model. \nUltimately, histograms only help us to understand this ground truth but cannot provide a complete picture.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# pip install joblib\n", "_____no_output_____" ], [ "import pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport joblib\nplt.style.use('fivethirtyeight')", "_____no_output_____" ], [ "class Perceptron:\n def __init__(self,eta: float=None, epochs: int=None):\n self.weights = np.random.randn(3)*1e-4\n self.eta=eta #learning rate\n self.epochs = epochs # number of iteration\n def _z_outcome(self, inputs, weights):\n return np.dot(inputs,weights)\n def activation_function(self,z):\n return np.where(z>0,1,0)\n def fit(self, X, y):\n self.X = X\n self.y = y\n \n X_with_bias=np.c_[self.X,-np.ones(((len(self.X)),1))]\n print(X_with_bias)\n# self.z=_z_cal()\n for epoch in range(self.epochs):\n print(\"__\"*10)\n print(f'for epoch: {epoch+1}')\n z = self._z_outcome(X_with_bias,self.weights)\n y_hat = self.activation_function(z)\n \n print(f'predicted value after forward pass: \\n{y_hat}')\n \n self.error = self.y-y_hat\n print(f'error : \\n{self.error}')\n \n self.weights = self.weights + self.eta*np.dot(X_with_bias.T,self.error)\n print(f'Updated weights after epoch : {epoch +1}/{self.epochs} \\n{self.weights}')\n def predict(self,X):\n X_with_bais = np.c_[X,-np.ones(((len(X)),1))]\n z=self._z_outcome(X_with_bais,self.weights)\n return self.activation_function(z)\n ", "_____no_output_____" ], [ "obj= Perceptron(eta=.2,epochs=10)", "_____no_output_____" ], [ "OR ={\n \"x1\":[0,0,1,1],\n \"x2\":[0,1,0,1],\n \"y\":[0,1,1,1]\n}\ndf_or = pd.DataFrame(OR)\ndf_or", "_____no_output_____" ], [ "AND ={\n \"x1\":[0,0,1,1],\n \"x2\":[0,1,0,1],\n \"y\":[0,0,0,1]\n}\ndf_and = pd.DataFrame(AND)\ndf_and", "_____no_output_____" ], [ "def prepare_data(df, target_col='y'):\n X = df.drop(target_col,axis=1)\n y = df[target_col]\n return X, y", "_____no_output_____" ], [ "X,y = prepare_data(df_and)\nETA= 0.1\nEPOCHS = 10\nmodel_and = Perceptron(eta=ETA, epochs=EPOCHS)\nmodel_and.fit(X,y)", "[[ 0. 0. -1.]\n [ 0. 1. -1.]\n [ 1. 0. -1.]\n [ 1. 1. -1.]]\n____________________\nfor epoch: 1\npredicted value after forward pass: \n[1 1 0 0]\nerror : \n0 -1\n1 -1\n2 0\n3 1\nName: y, dtype: int64\nUpdated weights after epoch : 1/10 \n[9.98616217e-02 4.88378852e-05 9.99127134e-02]\n____________________\nfor epoch: 2\npredicted value after forward pass: \n[0 0 0 0]\nerror : \n0 0\n1 0\n2 0\n3 1\nName: y, dtype: int64\nUpdated weights after epoch : 2/10 \n[ 1.99861622e-01 1.00048838e-01 -8.72865745e-05]\n____________________\nfor epoch: 3\npredicted value after forward pass: \n[1 1 1 1]\nerror : \n0 -1\n1 -1\n2 -1\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 3/10 \n[9.98616217e-02 4.88378852e-05 2.99912713e-01]\n____________________\nfor epoch: 4\npredicted value after forward pass: \n[0 0 0 0]\nerror : \n0 0\n1 0\n2 0\n3 1\nName: y, dtype: int64\nUpdated weights after epoch : 4/10 \n[0.19986162 0.10004884 0.19991271]\n____________________\nfor epoch: 5\npredicted value after forward pass: \n[0 0 0 1]\nerror : \n0 0\n1 0\n2 0\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 5/10 \n[0.19986162 0.10004884 0.19991271]\n____________________\nfor epoch: 6\npredicted value after forward pass: \n[0 0 0 1]\nerror : \n0 0\n1 0\n2 0\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 6/10 \n[0.19986162 0.10004884 0.19991271]\n____________________\nfor epoch: 7\npredicted value after forward pass: \n[0 0 0 1]\nerror : \n0 0\n1 0\n2 0\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 7/10 \n[0.19986162 0.10004884 0.19991271]\n____________________\nfor epoch: 8\npredicted value after forward pass: \n[0 0 0 1]\nerror : \n0 0\n1 0\n2 0\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 8/10 \n[0.19986162 0.10004884 0.19991271]\n____________________\nfor epoch: 9\npredicted value after forward pass: \n[0 0 0 1]\nerror : \n0 0\n1 0\n2 0\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 9/10 \n[0.19986162 0.10004884 0.19991271]\n____________________\nfor epoch: 10\npredicted value after forward pass: \n[0 0 0 1]\nerror : \n0 0\n1 0\n2 0\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 10/10 \n[0.19986162 0.10004884 0.19991271]\n" ], [ "xor ={\n \"x1\":[0,0,1,1],\n \"x2\":[0,1,0,1],\n \"y\":[0,1,1,0]\n}\ndf_xor = pd.DataFrame(xor)\ndf_xor", "_____no_output_____" ], [ "X,y = prepare_data(df_xor)\nETA= 0.1\nEPOCHS = 10\nmodel_xor = Perceptron(eta=ETA, epochs=EPOCHS)\nmodel_xor.fit(X,y)", "[[ 0. 0. -1.]\n [ 0. 1. -1.]\n [ 1. 0. -1.]\n [ 1. 1. -1.]]\n____________________\nfor epoch: 1\npredicted value after forward pass: \n[0 1 0 0]\nerror : \n0 0\n1 0\n2 1\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 1/10 \n[ 0.0999219 0.000151 -0.09989023]\n____________________\nfor epoch: 2\npredicted value after forward pass: \n[1 1 1 1]\nerror : \n0 -1\n1 0\n2 0\n3 -1\nName: y, dtype: int64\nUpdated weights after epoch : 2/10 \n[-7.81023823e-05 -9.98489975e-02 1.00109772e-01]\n____________________\nfor epoch: 3\npredicted value after forward pass: \n[0 0 0 0]\nerror : \n0 0\n1 1\n2 1\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 3/10 \n[ 0.0999219 0.000151 -0.09989023]\n____________________\nfor epoch: 4\npredicted value after forward pass: \n[1 1 1 1]\nerror : \n0 -1\n1 0\n2 0\n3 -1\nName: y, dtype: int64\nUpdated weights after epoch : 4/10 \n[-7.81023823e-05 -9.98489975e-02 1.00109772e-01]\n____________________\nfor epoch: 5\npredicted value after forward pass: \n[0 0 0 0]\nerror : \n0 0\n1 1\n2 1\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 5/10 \n[ 0.0999219 0.000151 -0.09989023]\n____________________\nfor epoch: 6\npredicted value after forward pass: \n[1 1 1 1]\nerror : \n0 -1\n1 0\n2 0\n3 -1\nName: y, dtype: int64\nUpdated weights after epoch : 6/10 \n[-7.81023823e-05 -9.98489975e-02 1.00109772e-01]\n____________________\nfor epoch: 7\npredicted value after forward pass: \n[0 0 0 0]\nerror : \n0 0\n1 1\n2 1\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 7/10 \n[ 0.0999219 0.000151 -0.09989023]\n____________________\nfor epoch: 8\npredicted value after forward pass: \n[1 1 1 1]\nerror : \n0 -1\n1 0\n2 0\n3 -1\nName: y, dtype: int64\nUpdated weights after epoch : 8/10 \n[-7.81023823e-05 -9.98489975e-02 1.00109772e-01]\n____________________\nfor epoch: 9\npredicted value after forward pass: \n[0 0 0 0]\nerror : \n0 0\n1 1\n2 1\n3 0\nName: y, dtype: int64\nUpdated weights after epoch : 9/10 \n[ 0.0999219 0.000151 -0.09989023]\n____________________\nfor epoch: 10\npredicted value after forward pass: \n[1 1 1 1]\nerror : \n0 -1\n1 0\n2 0\n3 -1\nName: y, dtype: int64\nUpdated weights after epoch : 10/10 \n[-7.81023823e-05 -9.98489975e-02 1.00109772e-01]\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Define the Convolutional Neural Network\n\nAfter you've looked at the data you're working with and, in this case, know the shapes of the images and of the keypoints, you are ready to define a convolutional neural network that can *learn* from this data.\n\nIn this notebook and in `models.py`, you will:\n1. Define a CNN with images as input and keypoints as output\n2. Construct the transformed FaceKeypointsDataset, just as before\n3. Train the CNN on the training data, tracking loss\n4. See how the trained model performs on test data\n5. If necessary, modify the CNN structure and model hyperparameters, so that it performs *well* **\\***\n\n**\\*** What does *well* mean?\n\n\"Well\" means that the model's loss decreases during training **and**, when applied to test image data, the model produces keypoints that closely match the true keypoints of each face. And you'll see examples of this later in the notebook.\n\n---\n", "_____no_output_____" ], [ "## CNN Architecture\n\nRecall that CNN's are defined by a few types of layers:\n* Convolutional layers\n* Maxpooling layers\n* Fully-connected layers\n\nYou are required to use the above layers and encouraged to add multiple convolutional layers and things like dropout layers that may prevent overfitting. You are also encouraged to look at literature on keypoint detection, such as [this paper](https://arxiv.org/pdf/1710.00977.pdf), to help you determine the structure of your network.\n\n\n### TODO: Define your model in the provided file `models.py` file\n\nThis file is mostly empty but contains the expected name and some TODO's for creating your model.\n\n---", "_____no_output_____" ], [ "## PyTorch Neural Nets\n\nTo define a neural network in PyTorch, you define the layers of a model in the function `__init__` and define the feedforward behavior of a network that employs those initialized layers in the function `forward`, which takes in an input image tensor, `x`. The structure of this Net class is shown below and left for you to fill in.\n\nNote: During training, PyTorch will be able to perform backpropagation by keeping track of the network's feedforward behavior and using autograd to calculate the update to the weights in the network.\n\n#### Define the Layers in ` __init__`\nAs a reminder, a conv/pool layer may be defined like this (in `__init__`):\n```\n# 1 input image channel (for grayscale images), 32 output channels/feature maps, 3x3 square convolution kernel\nself.conv1 = nn.Conv2d(1, 32, 3)\n\n# maxpool that uses a square window of kernel_size=2, stride=2\nself.pool = nn.MaxPool2d(2, 2) \n```\n\n#### Refer to Layers in `forward`\nThen referred to in the `forward` function like this, in which the conv1 layer has a ReLu activation applied to it before maxpooling is applied:\n```\nx = self.pool(F.relu(self.conv1(x)))\n```\n\nBest practice is to place any layers whose weights will change during the training process in `__init__` and refer to them in the `forward` function; any layers or functions that always behave in the same way, such as a pre-defined activation function, should appear *only* in the `forward` function.", "_____no_output_____" ], [ "#### Why models.py\n\nYou are tasked with defining the network in the `models.py` file so that any models you define can be saved and loaded by name in different notebooks in this project directory. For example, by defining a CNN class called `Net` in `models.py`, you can then create that same architecture in this and other notebooks by simply importing the class and instantiating a model:\n```\n from models import Net\n net = Net()\n```", "_____no_output_____" ] ], [ [ "# import the usual resources\nimport matplotlib.pyplot as plt\nimport numpy as np\n\n# watch for any changes in model.py, if itchanges, re-load it automatically\n%load_ext autoreload\n%autoreload 2", "_____no_output_____" ], [ "## TODO: Define the Net in models.py\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\n\n## TODO: Once you've define the network, you can instantiate it\n# one example conv layer has been provided for you\nfrom models import Net\n\nnet = Net()\nprint(net)", "Net(\n (conv1): Conv2d(1, 32, kernel_size=(4, 4), stride=(1, 1))\n (conv2): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1))\n (conv3): Conv2d(64, 128, kernel_size=(2, 2), stride=(1, 1))\n (conv4): Conv2d(128, 256, kernel_size=(2, 2), stride=(1, 1))\n (drop_out1): Dropout(p=0.1)\n (drop_out2): Dropout(p=0.1)\n (drop_out3): Dropout(p=0.2)\n (drop_out4): Dropout(p=0.2)\n (drop_out5): Dropout(p=0.5)\n (drop_out6): Dropout(p=0.5)\n (fc1): Linear(in_features=36864, out_features=3200, bias=True)\n (fc2): Linear(in_features=3200, out_features=1600, bias=True)\n (fc3): Linear(in_features=1600, out_features=136, bias=True)\n (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n)\n" ] ], [ [ "## Transform the dataset \n\nTo prepare for training, create a transformed dataset of images and keypoints.\n\n### TODO: Define a data transform\n\nIn PyTorch, a convolutional neural network expects a torch image of a consistent size as input. For efficient training, and so your model's loss does not blow up during training, it is also suggested that you normalize the input images and keypoints. The necessary transforms have been defined in `data_load.py` and you **do not** need to modify these; take a look at this file (you'll see the same transforms that were defined and applied in Notebook 1).\n\nTo define the data transform below, use a [composition](http://pytorch.org/tutorials/beginner/data_loading_tutorial.html#compose-transforms) of:\n1. Rescaling and/or cropping the data, such that you are left with a square image (the suggested size is 224x224px)\n2. Normalizing the images and keypoints; turning each RGB image into a grayscale image with a color range of [0, 1] and transforming the given keypoints into a range of [-1, 1]\n3. Turning these images and keypoints into Tensors\n\nThese transformations have been defined in `data_load.py`, but it's up to you to call them and create a `data_transform` below. **This transform will be applied to the training data and, later, the test data**. It will change how you go about displaying these images and keypoints, but these steps are essential for efficient training.\n\nAs a note, should you want to perform data augmentation (which is optional in this project), and randomly rotate or shift these images, a square image size will be useful; rotating a 224x224 image by 90 degrees will result in the same shape of output.", "_____no_output_____" ] ], [ [ "from torch.utils.data import Dataset, DataLoader\nfrom torchvision import transforms, utils\n\n# the dataset we created in Notebook 1 is copied in the helper file `data_load.py`\nfrom data_load import FacialKeypointsDataset\n# the transforms we defined in Notebook 1 are in the helper file `data_load.py`\nfrom data_load import Rescale, RandomCrop, Normalize, ToTensor\n\n\n## TODO: define the data_transform using transforms.Compose([all tx's, . , .])\n# order matters! i.e. rescaling should come before a smaller crop\ndata_transform = transforms.Compose([Rescale(256), RandomCrop(224),Normalize(), ToTensor()])\n#data_transform = transforms.Compose([Rescale((224,224)),Normalize(), ToTensor()])\n\n# testing that you've defined a transform\nassert(data_transform is not None), 'Define a data_transform'", "_____no_output_____" ], [ "# create the transformed dataset\ntransformed_dataset = FacialKeypointsDataset(csv_file='data/training_frames_keypoints.csv',\n root_dir='data/training/',\n transform=data_transform)\n\n\nprint('Number of images: ', len(transformed_dataset))\n\n# iterate through the transformed dataset and print some stats about the first few samples\nfor i in range(4):\n sample = transformed_dataset[i]\n print(i, sample['image'].size(), sample['keypoints'].size())\n \n\n", "Number of images: 3462\n0 torch.Size([1, 224, 224]) torch.Size([68, 2])\n1 torch.Size([1, 224, 224]) torch.Size([68, 2])\n2 torch.Size([1, 224, 224]) torch.Size([68, 2])\n3 torch.Size([1, 224, 224]) torch.Size([68, 2])\n" ] ], [ [ "## Batching and loading data\n\nNext, having defined the transformed dataset, we can use PyTorch's DataLoader class to load the training data in batches of whatever size as well as to shuffle the data for training the model. You can read more about the parameters of the DataLoader, in [this documentation](http://pytorch.org/docs/master/data.html).\n\n#### Batch size\nDecide on a good batch size for training your model. Try both small and large batch sizes and note how the loss decreases as the model trains.\n\n**Note for Windows users**: Please change the `num_workers` to 0 or you may face some issues with your DataLoader failing.", "_____no_output_____" ] ], [ [ "# load training data in batches\nbatch_size = 128\n\ntrain_loader = DataLoader(transformed_dataset, \n batch_size=batch_size,\n shuffle=True, \n num_workers=4)\n", "_____no_output_____" ] ], [ [ "## Before training\n\nTake a look at how this model performs before it trains. You should see that the keypoints it predicts start off in one spot and don't match the keypoints on a face at all! It's interesting to visualize this behavior so that you can compare it to the model after training and see how the model has improved.\n\n#### Load in the test dataset\n\nThe test dataset is one that this model has *not* seen before, meaning it has not trained with these images. We'll load in this test data and before and after training, see how your model performs on this set!\n\nTo visualize this test data, we have to go through some un-transformation steps to turn our images into python images from tensors and to turn our keypoints back into a recognizable range. ", "_____no_output_____" ] ], [ [ "# load in the test data, using the dataset class\n# AND apply the data_transform you defined above\n\n# create the test dataset\ntest_dataset = FacialKeypointsDataset(csv_file='data/test_frames_keypoints.csv',\n root_dir='data/test/',\n transform=data_transform)\n\n", "_____no_output_____" ], [ "# load test data in batches\nbatch_size = 128\n\ntest_loader = DataLoader(test_dataset, \n batch_size=batch_size,\n shuffle=True, \n num_workers=4)", "_____no_output_____" ] ], [ [ "## Apply the model on a test sample\n\nTo test the model on a test sample of data, you have to follow these steps:\n1. Extract the image and ground truth keypoints from a sample\n2. Make sure the image is a FloatTensor, which the model expects.\n3. Forward pass the image through the net to get the predicted, output keypoints.\n\nThis function test how the network performs on the first batch of test data. It returns the images, the transformed images, the predicted keypoints (produced by the model), and the ground truth keypoints.", "_____no_output_____" ] ], [ [ "# test the model on a batch of test images\n\ndef net_sample_output():\n \n # iterate through the test dataset\n for i, sample in enumerate(test_loader):\n \n # get sample data: images and ground truth keypoints\n images = sample['image']\n key_pts = sample['keypoints']\n\n # convert images to FloatTensors\n images = images.type(torch.FloatTensor)\n\n # forward pass to get net output\n output_pts = net(images)\n \n # reshape to batch_size x 68 x 2 pts\n output_pts = output_pts.view(output_pts.size()[0], 68, -1)\n \n # break after first image is tested\n if i == 0:\n return images, output_pts, key_pts\n ", "_____no_output_____" ] ], [ [ "#### Debugging tips\n\nIf you get a size or dimension error here, make sure that your network outputs the expected number of keypoints! Or if you get a Tensor type error, look into changing the above code that casts the data into float types: `images = images.type(torch.FloatTensor)`.", "_____no_output_____" ] ], [ [ "# call the above function\n# returns: test images, test predicted keypoints, test ground truth keypoints\ntest_images, test_outputs, gt_pts = net_sample_output()\n\n# print out the dimensions of the data to see if they make sense\nprint(test_images.data.size())\nprint(test_outputs.data.size())\nprint(gt_pts.size())", "_____no_output_____" ] ], [ [ "## Visualize the predicted keypoints\n\nOnce we've had the model produce some predicted output keypoints, we can visualize these points in a way that's similar to how we've displayed this data before, only this time, we have to \"un-transform\" the image/keypoint data to display it.\n\nNote that I've defined a *new* function, `show_all_keypoints` that displays a grayscale image, its predicted keypoints and its ground truth keypoints (if provided).", "_____no_output_____" ] ], [ [ "def show_all_keypoints(image, predicted_key_pts, gt_pts=None):\n \"\"\"Show image with predicted keypoints\"\"\"\n # image is grayscale\n plt.imshow(image, cmap='gray')\n plt.scatter(predicted_key_pts[:, 0], predicted_key_pts[:, 1], s=20, marker='.', c='m')\n # plot ground truth points as green pts\n if gt_pts is not None:\n plt.scatter(gt_pts[:, 0], gt_pts[:, 1], s=20, marker='.', c='g')\n", "_____no_output_____" ] ], [ [ "#### Un-transformation\n\nNext, you'll see a helper function. `visualize_output` that takes in a batch of images, predicted keypoints, and ground truth keypoints and displays a set of those images and their true/predicted keypoints.\n\nThis function's main role is to take batches of image and keypoint data (the input and output of your CNN), and transform them into numpy images and un-normalized keypoints (x, y) for normal display. The un-transformation process turns keypoints and images into numpy arrays from Tensors *and* it undoes the keypoint normalization done in the Normalize() transform; it's assumed that you applied these transformations when you loaded your test data.", "_____no_output_____" ] ], [ [ "# visualize the output\n# by default this shows a batch of 10 images\ndef visualize_output(test_images, test_outputs, gt_pts=None, batch_size=10):\n\n for i in range(batch_size):\n plt.figure(figsize=(20,10))\n ax = plt.subplot(1, batch_size, i+1)\n\n # un-transform the image data\n image = test_images[i].data # get the image from it's wrapper\n image = image.numpy() # convert to numpy array from a Tensor\n image = np.transpose(image, (1, 2, 0)) # transpose to go from torch to numpy image\n\n # un-transform the predicted key_pts data\n predicted_key_pts = test_outputs[i].data\n predicted_key_pts = predicted_key_pts.numpy()\n # undo normalization of keypoints \n predicted_key_pts = predicted_key_pts*50.0+100\n \n # plot ground truth points for comparison, if they exist\n ground_truth_pts = None\n if gt_pts is not None:\n ground_truth_pts = gt_pts[i] \n ground_truth_pts = ground_truth_pts*50.0+100\n \n # call show_all_keypoints\n show_all_keypoints(np.squeeze(image), predicted_key_pts, ground_truth_pts)\n \n plt.axis('off')\n\n plt.show()\n \n# call it\n#visualize_output(test_images, test_outputs, gt_pts)", "_____no_output_____" ] ], [ [ "## Training\n\n#### Loss function\nTraining a network to predict keypoints is different than training a network to predict a class; instead of outputting a distribution of classes and using cross entropy loss, you may want to choose a loss function that is suited for regression, which directly compares a predicted value and target value. Read about the various kinds of loss functions (like MSE or L1/SmoothL1 loss) in [this documentation](http://pytorch.org/docs/master/_modules/torch/nn/modules/loss.html).\n\n### TODO: Define the loss and optimization\n\nNext, you'll define how the model will train by deciding on the loss function and optimizer.\n\n---", "_____no_output_____" ] ], [ [ "## TODO: Define the loss and optimization\nimport torch.optim as optim\n\n#criterion = nn.MSELoss()\ncriterion = nn.SmoothL1Loss()\n\noptimizer = optim.Adam(net.parameters(), lr=1e-3, betas=(0.9, 0.999))\n#optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)", "_____no_output_____" ] ], [ [ "## Training and Initial Observation\n\nNow, you'll train on your batched training data from `train_loader` for a number of epochs. \n\nTo quickly observe how your model is training and decide on whether or not you should modify it's structure or hyperparameters, you're encouraged to start off with just one or two epochs at first. As you train, note how your the model's loss behaves over time: does it decrease quickly at first and then slow down? Does it take a while to decrease in the first place? What happens if you change the batch size of your training data or modify your loss function? etc. \n\nUse these initial observations to make changes to your model and decide on the best architecture before you train for many epochs and create a final model.", "_____no_output_____" ] ], [ [ "from livelossplot import PlotLosses\n\ndef train_net(n_epochs):\n \n # prepare the net for training\n net.train()\n liveloss = PlotLosses()\n\n\n for epoch in range(n_epochs): # loop over the dataset multiple times\n epoch_loss = 0.0\n epoch_correct = 0\n epoch_loss_val = 0.0\n epoch_correct_val = 0\n running_loss = 0.0\n\n # train on batches of data, assumes you already have train_loader\n for batch_i, data in enumerate(train_loader):\n # get the input images and their corresponding labels\n images = data['image']\n key_pts = data['keypoints']\n\n # flatten pts\n key_pts = key_pts.view(key_pts.size(0), -1)\n\n # convert variables to floats for regression loss\n key_pts = key_pts.type(torch.FloatTensor)\n images = images.type(torch.FloatTensor)\n\n # forward pass to get outputs\n output_pts = net(images)\n\n # calculate the loss between predicted and target keypoints\n loss = criterion(output_pts, key_pts)\n\n # zero the parameter (weight) gradients\n optimizer.zero_grad()\n \n # backward pass to calculate the weight gradients\n loss.backward()\n\n # update the weights\n optimizer.step()\n \n epoch_loss += loss.data[0]\n epoch_correct += (output_pts.max(1)[1] == key_pts).sum().data[0]\n\n # print loss statistics\n # to convert loss into a scalar and add it to the running_loss, use .item()\n running_loss += loss.item()\n if batch_i % 10 == 9: # print every 10 batches\n print('Epoch: {}, Batch: {}, Avg. Loss: {}'.format(epoch + 1, batch_i+1, running_loss/1000))\n running_loss = 0.0\n avg_loss = epoch_loss / len(train_loader.dataset)\n avg_accuracy = epoch_correct / len(train_loader.dataset)\n liveloss.update({\n 'log loss': avg_loss,\n 'val_log loss': avg_loss_val,\n 'accuracy': avg_accuracy,\n 'val_accuracy': avg_accuracy_val})\n liveloss.draw()\n \n print('Finished Training')\n", "_____no_output_____" ], [ "def train_net (n_epochs):\n\n # prepare the net for training\n net.train()\n\n for epoch in range(n_epochs): # loop over the dataset multiple times\n \n running_loss = 0.0\n\n # train on batches of data, assumes you already have train_loader\n for batch_i, data in enumerate(train_loader):\n # get the input images and their corresponding labels\n images = data['image']\n key_pts = data['keypoints']\n\n # flatten pts\n key_pts = key_pts.view(key_pts.size(0), -1)\n\n # convert variables to floats for regression loss\n key_pts = key_pts.type(torch.FloatTensor)\n images = images.type(torch.FloatTensor)\n\n # forward pass to get outputs\n output_pts = net(images)\n\n # calculate the loss between predicted and target keypoints\n loss = criterion(output_pts, key_pts)\n\n # zero the parameter (weight) gradients\n optimizer.zero_grad()\n \n # backward pass to calculate the weight gradients\n loss.backward()\n\n # update the weights\n optimizer.step()\n\n # print loss statistics\n # to convert loss into a scalar and add it to the running_loss, use .item()\n running_loss += loss.item()\n if batch_i % 10 == 9: # print every 10 batches\n print('Epoch: {}, Batch: {}, Avg. Loss: {}'.format(epoch + 1, batch_i+1, running_loss/1000))\n running_loss = 0.0\n model_dir = 'saved_models/'\n model_name = 'model_3200_1600_smoothL1.pt'\n\n # after training, save your model parameters in the dir 'saved_models'\n #torch.save(net.state_dict(), model_dir+model_name)\n\n print('Finished Training')", "_____no_output_____" ], [ "net.load_state_dict(torch.load('saved_models/model_3200_1600_smoothL1.pt'))\n", "_____no_output_____" ], [ "# train your network\nn_epochs = 5 # start small, and increase when you've decided on your model structure and hyperparams\n\ntrain_net(n_epochs)", "Epoch: 1, Batch: 10, Avg. Loss: 0.0005024574361741543\nEpoch: 1, Batch: 20, Avg. Loss: 0.00045902186259627343\nEpoch: 2, Batch: 10, Avg. Loss: 0.0004713563397526741\nEpoch: 2, Batch: 20, Avg. Loss: 0.000502167847007513\nEpoch: 3, Batch: 10, Avg. Loss: 0.0005260687507688999\nEpoch: 3, Batch: 20, Avg. Loss: 0.00044130625203251837\n" ] ], [ [ "## Test data\n\nSee how your model performs on previously unseen, test data. We've already loaded and transformed this data, similar to the training data. Next, run your trained model on these images to see what kind of keypoints are produced. You should be able to see if your model is fitting each new face it sees, if the points are distributed randomly, or if the points have actually overfitted the training data and do not generalize.", "_____no_output_____" ] ], [ [ "# get a sample of test data again\ntest_images, test_outputs, gt_pts = net_sample_output()\ntorch.cuda.get_device_name(0)\nprint(test_images.data.size())\nprint(test_outputs.data.size())\nprint(gt_pts.size())", "torch.Size([128, 1, 224, 224])\ntorch.Size([128, 68, 2])\ntorch.Size([128, 68, 2])\n" ], [ "visualize_output(test_images, test_outputs, gt_pts)\n", "_____no_output_____" ], [ "## TODO: visualize your test output\n# you can use the same function as before, by un-commenting the line below:\n\nvisualize_output(test_images, test_outputs, gt_pts)\n", "_____no_output_____" ] ], [ [ "Once you've found a good model (or two), save your model so you can load it and use it later!", "_____no_output_____" ] ], [ [ "## TODO: change the name to something uniqe for each new model\nmodel_dir = 'saved_models/'\nmodel_name = 'model_3200_1600_smoothL1.pt'\n\n\n# after training, save your model parameters in the dir 'saved_models'\ntorch.save(net.state_dict(), model_dir+model_name)", "_____no_output_____" ] ], [ [ "After you've trained a well-performing model, answer the following questions so that we have some insight into your training and architecture selection process. Answering all questions is required to pass this project.", "_____no_output_____" ], [ "### Question 1: What optimization and loss functions did you choose and why?\n", "_____no_output_____" ], [ "**Answer**: write your answer here (double click to edit this cell)\nI use the ADAM optimization and MSE loss functions. I choose it because the network's output error should be as close as the desire value ( not classification). Therefore the ADAM and MSE works well. ", "_____no_output_____" ], [ "### Question 2: What kind of network architecture did you start with and how did it change as you tried different architectures? Did you decide to add more convolutional layers or any layers to avoid overfitting the data?", "_____no_output_____" ], [ "**Answer**: write your answer here\n\nFirstly, i try to apply LeNet architecture because it simple and fast. However, the result looks not good. Therefore, i change to the NaimishNet. This CNN apply many dropout layers to avoid overfitting the data.", "_____no_output_____" ], [ "### Question 3: How did you decide on the number of epochs and batch_size to train your model?", "_____no_output_____" ], [ "**Answer**: write your answer here\n•\tEpoch: 30. The bigger the epoch, the better model is. Because there mare many dropout() layer, the network does not overfit the data with 30 epoch.\n•\tBatch size: 10 . The bigger the batch size, the faster model is.\n", "_____no_output_____" ], [ "## Feature Visualization\n\nSometimes, neural networks are thought of as a black box, given some input, they learn to produce some output. CNN's are actually learning to recognize a variety of spatial patterns and you can visualize what each convolutional layer has been trained to recognize by looking at the weights that make up each convolutional kernel and applying those one at a time to a sample image. This technique is called feature visualization and it's useful for understanding the inner workings of a CNN.", "_____no_output_____" ], [ "In the cell below, you can see how to extract a single filter (by index) from your first convolutional layer. The filter should appear as a grayscale grid.", "_____no_output_____" ] ], [ [ "# Get the weights in the first conv layer, \"conv1\"\n# if necessary, change this to reflect the name of your first conv layer\nweights1 = net.conv1.weight.data\n\nw = weights1.numpy()\n\nfilter_index = 0\n\nprint(w[filter_index][0])\nprint(w[filter_index][0].shape)\n\n# display the filter weights\nplt.imshow(w[filter_index][0], cmap='gray')\n", "_____no_output_____" ] ], [ [ "## Feature maps\n\nEach CNN has at least one convolutional layer that is composed of stacked filters (also known as convolutional kernels). As a CNN trains, it learns what weights to include in it's convolutional kernels and when these kernels are applied to some input image, they produce a set of **feature maps**. So, feature maps are just sets of filtered images; they are the images produced by applying a convolutional kernel to an input image. These maps show us the features that the different layers of the neural network learn to extract. For example, you might imagine a convolutional kernel that detects the vertical edges of a face or another one that detects the corners of eyes. You can see what kind of features each of these kernels detects by applying them to an image. One such example is shown below; from the way it brings out the lines in an the image, you might characterize this as an edge detection filter.\n\n<img src='images/feature_map_ex.png' width=50% height=50%/>\n\n\nNext, choose a test image and filter it with one of the convolutional kernels in your trained CNN; look at the filtered output to get an idea what that particular kernel detects.\n\n### TODO: Filter an image to see the effect of a convolutional kernel\n---", "_____no_output_____" ] ], [ [ "##TODO: load in and display any image from the transformed test dataset\n\n## TODO: Using cv's filter2D function,\n## apply a specific set of filter weights (like the one displayed above) to the test image\n\n\nfig = plt.figure(figsize=(120, 4))\nfor idx in np.arange(4):\n ax = fig.add_subplot(2, 128/2, idx+1, xticks=[], yticks=[])\n image = test_dataset[idx]['image'].numpy()\n ax.imshow(np.squeeze(image), cmap='gray')\n\n ", "_____no_output_____" ], [ "import cv2\nnum = 3\nimg = np.squeeze(test_dataset[num]['image'].numpy())\nplt.imshow(img, cmap='gray')\nweights = net.conv1.weight.data\nw = weights.numpy()\nfig=plt.figure(figsize=(30, 10))\ncolumns = 5*2\nrows = 2\nfor i in range(0, columns*rows):\n fig.add_subplot(rows, columns, i+1)\n if ((i%2)==0):\n plt.imshow(w[int(i/2)][0], cmap='gray')\n else:\n c = cv2.filter2D(img, -1, w[int((i-1)/2)][0])\n plt.imshow(c, cmap='gray')\nplt.show()", "_____no_output_____" ] ], [ [ "### Question 4: Choose one filter from your trained CNN and apply it to a test image; what purpose do you think it plays? What kind of feature do you think it detects?\n", "_____no_output_____" ], [ "**Answer**: (does it detect vertical lines or does it blur out noise, etc.) write your answer here", "_____no_output_____" ], [ "---\n## Moving on!\n\nNow that you've defined and trained your model (and saved the best model), you are ready to move on to the last notebook, which combines a face detector with your saved model to create a facial keypoint detection system that can predict the keypoints on *any* face in an image!", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "import tensorflow as tf", "_____no_output_____" ], [ "import tensorflow as tf\nfrom tensorflow.python.keras.applications.vgg19 import VGG19", "_____no_output_____" ], [ "model=VGG19(\n include_top=False,\n weights='imagenet'\n)\nmodel.trainable=False\nmodel.summary()", "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg19/vgg19_weights_tf_dim_ordering_tf_kernels_notop.h5\n80142336/80134624 [==============================] - 36s 0us/step\nModel: \"vgg19\"\n_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\ninput_1 (InputLayer) [(None, None, None, 3)] 0 \n_________________________________________________________________\nblock1_conv1 (Conv2D) (None, None, None, 64) 1792 \n_________________________________________________________________\nblock1_conv2 (Conv2D) (None, None, None, 64) 36928 \n_________________________________________________________________\nblock1_pool (MaxPooling2D) (None, None, None, 64) 0 \n_________________________________________________________________\nblock2_conv1 (Conv2D) (None, None, None, 128) 73856 \n_________________________________________________________________\nblock2_conv2 (Conv2D) (None, None, None, 128) 147584 \n_________________________________________________________________\nblock2_pool (MaxPooling2D) (None, None, None, 128) 0 \n_________________________________________________________________\nblock3_conv1 (Conv2D) (None, None, None, 256) 295168 \n_________________________________________________________________\nblock3_conv2 (Conv2D) (None, None, None, 256) 590080 \n_________________________________________________________________\nblock3_conv3 (Conv2D) (None, None, None, 256) 590080 \n_________________________________________________________________\nblock3_conv4 (Conv2D) (None, None, None, 256) 590080 \n_________________________________________________________________\nblock3_pool (MaxPooling2D) (None, None, None, 256) 0 \n_________________________________________________________________\nblock4_conv1 (Conv2D) (None, None, None, 512) 1180160 \n_________________________________________________________________\nblock4_conv2 (Conv2D) (None, None, None, 512) 2359808 \n_________________________________________________________________\nblock4_conv3 (Conv2D) (None, None, None, 512) 2359808 \n_________________________________________________________________\nblock4_conv4 (Conv2D) (None, None, None, 512) 2359808 \n_________________________________________________________________\nblock4_pool (MaxPooling2D) (None, None, None, 512) 0 \n_________________________________________________________________\nblock5_conv1 (Conv2D) (None, None, None, 512) 2359808 \n_________________________________________________________________\nblock5_conv2 (Conv2D) (None, None, None, 512) 2359808 \n_________________________________________________________________\nblock5_conv3 (Conv2D) (None, None, None, 512) 2359808 \n_________________________________________________________________\nblock5_conv4 (Conv2D) (None, None, None, 512) 2359808 \n_________________________________________________________________\nblock5_pool (MaxPooling2D) (None, None, None, 512) 0 \n=================================================================\nTotal params: 20,024,384\nTrainable params: 0\nNon-trainable params: 20,024,384\n_________________________________________________________________\n" ], [ "from tensorflow.python.keras.preprocessing.image import load_img, img_to_array\nfrom tensorflow.python.keras.applications.vgg19 import preprocess_input\nfrom tensorflow.python.keras.models import Model\nimport numpy as np\nimport matplotlib.pyplot as plt\n%matplotlib inline", "_____no_output_____" ], [ "def load_and_process_image(image_path):\n img=load_img(image_path)\n img=img_to_array(img)\n img=preprocess_input(img)\n img=np.expand_dims(img,axis=0)\n return img", "_____no_output_____" ], [ "def deprocess(x):\n x[:,:,0]+=103.939\n x[:,:,1]+=116.779\n x[:,:,2]+=123.68\n x=x[:,:,::-1]\n \n x=np.clip(x,0,255).astype('uint8')\n return x\ndef display_image(image):\n if len(image.shape)==4:\n img=np.squeeze(image,axis=0)\n img=deprocess(img)\n \n plt.grid(False)\n plt.xticks([])\n plt.yticks([])\n plt.imshow(img)\n return", "_____no_output_____" ], [ "display_image(load_and_process_image('style.jpg'))", "_____no_output_____" ], [ "style_layers = [\n 'block1_conv1', \n 'block3_conv1', \n 'block5_conv1'\n]\n\ncontent_layer = 'block5_conv2'\n\n# intermediate models\ncontent_model = Model(\n inputs = model.input, \n outputs = model.get_layer(content_layer).output\n)\n\nstyle_models = [Model(inputs = model.input, \n outputs = model.get_layer(layer).output) for layer in style_layers]", "_____no_output_____" ], [ "# Content Cost\ndef content_cost(content, generated):\n a_C = content_model(content)\n a_G = content_model(generated)\n cost = tf.reduce_mean(tf.square(a_C - a_G))\n return cost", "_____no_output_____" ], [ "def gram_matrix(A):\n channels = int(A.shape[-1])\n a = tf.reshape(A, [-1, channels])\n n = tf.shape(a)[0]\n gram = tf.matmul(a, a, transpose_a = True)\n return gram / tf.cast(n, tf.float32)", "_____no_output_____" ], [ "lam = 1. / len(style_models)\n\ndef style_cost(style, generated):\n J_style = 0\n \n for style_model in style_models:\n a_S = style_model(style)\n a_G = style_model(generated)\n GS = gram_matrix(a_S)\n GG = gram_matrix(a_G)\n current_cost = tf.reduce_mean(tf.square(GS - GG))\n J_style += current_cost * lam\n \n return J_style", "_____no_output_____" ], [ "import time\n\ngenerated_images = []\n\ndef training_loop(content_path, style_path, iterations = 20, a = 10., b = 20.):\n # initialise\n content = load_and_process_image(content_path)\n style = load_and_process_image(style_path)\n generated = tf.Variable(content, dtype = tf.float32)\n \n opt = tf.optimizers.Adam(learning_rate = 7.)\n \n best_cost = 1e12+0.1\n best_image = None\n \n start_time = time.time()\n \n for i in range(iterations):\n \n with tf.GradientTape() as tape:\n J_content = content_cost(content, generated)\n J_style = style_cost(style, generated)\n J_total = a * J_content + b * J_style\n \n grads = tape.gradient(J_total, generated)\n opt.apply_gradients([(grads, generated)])\n \n if J_total < best_cost:\n best_cost = J_total\n best_image = generated.numpy()\n \n if i % int(iterations/10) == 0:\n time_taken = time.time() - start_time\n print('Cost at {}: {}. Time elapsed: {}'.format(i, J_total, time_taken))\n generated_images.append(generated.numpy())\n \n return best_image", "_____no_output_____" ], [ "final = training_loop('content.jpg','style.jpg')", "Cost at 0: 6672084992.0. Time elapsed: 3.4547882080078125\nCost at 2: 1479381760.0. Time elapsed: 10.296310186386108\nCost at 4: 863365824.0. Time elapsed: 17.23086667060852\nCost at 6: 594945280.0. Time elapsed: 24.286461353302002\nCost at 8: 454304224.0. Time elapsed: 31.091986417770386\nCost at 10: 369753824.0. Time elapsed: 37.903504848480225\nCost at 12: 308917152.0. Time elapsed: 44.70805072784424\nCost at 14: 260217664.0. Time elapsed: 51.643970251083374\nCost at 16: 220326704.0. Time elapsed: 58.485512256622314\nCost at 18: 188584880.0. Time elapsed: 65.35505986213684\n" ], [ "plt.figure(figsize = (12, 12))\n\nfor i in range(10):\n plt.subplot(5, 2, i + 1)\n display_image(generated_images[i])\nplt.show()", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "from sklearn import datasets\nwine = datasets.load_wine()\nprint(wine.DESCR)", ".. _wine_dataset:\n\nWine recognition dataset\n------------------------\n\n**Data Set Characteristics:**\n\n :Number of Instances: 178 (50 in each of three classes)\n :Number of Attributes: 13 numeric, predictive attributes and the class\n :Attribute Information:\n \t\t- Alcohol\n \t\t- Malic acid\n \t\t- Ash\n\t\t- Alcalinity of ash \n \t\t- Magnesium\n\t\t- Total phenols\n \t\t- Flavanoids\n \t\t- Nonflavanoid phenols\n \t\t- Proanthocyanins\n\t\t- Color intensity\n \t\t- Hue\n \t\t- OD280/OD315 of diluted wines\n \t\t- Proline\n\n - class:\n - class_0\n - class_1\n - class_2\n\t\t\n :Summary Statistics:\n \n ============================= ==== ===== ======= =====\n Min Max Mean SD\n ============================= ==== ===== ======= =====\n Alcohol: 11.0 14.8 13.0 0.8\n Malic Acid: 0.74 5.80 2.34 1.12\n Ash: 1.36 3.23 2.36 0.27\n Alcalinity of Ash: 10.6 30.0 19.5 3.3\n Magnesium: 70.0 162.0 99.7 14.3\n Total Phenols: 0.98 3.88 2.29 0.63\n Flavanoids: 0.34 5.08 2.03 1.00\n Nonflavanoid Phenols: 0.13 0.66 0.36 0.12\n Proanthocyanins: 0.41 3.58 1.59 0.57\n Colour Intensity: 1.3 13.0 5.1 2.3\n Hue: 0.48 1.71 0.96 0.23\n OD280/OD315 of diluted wines: 1.27 4.00 2.61 0.71\n Proline: 278 1680 746 315\n ============================= ==== ===== ======= =====\n\n :Missing Attribute Values: None\n :Class Distribution: class_0 (59), class_1 (71), class_2 (48)\n :Creator: R.A. Fisher\n :Donor: Michael Marshall (MARSHALL%[email protected])\n :Date: July, 1988\n\nThis is a copy of UCI ML Wine recognition datasets.\nhttps://archive.ics.uci.edu/ml/machine-learning-databases/wine/wine.data\n\nThe data is the results of a chemical analysis of wines grown in the same\nregion in Italy by three different cultivators. There are thirteen different\nmeasurements taken for different constituents found in the three types of\nwine.\n\nOriginal Owners: \n\nForina, M. et al, PARVUS - \nAn Extendible Package for Data Exploration, Classification and Correlation. \nInstitute of Pharmaceutical and Food Analysis and Technologies,\nVia Brigata Salerno, 16147 Genoa, Italy.\n\nCitation:\n\nLichman, M. (2013). UCI Machine Learning Repository\n[https://archive.ics.uci.edu/ml]. Irvine, CA: University of California,\nSchool of Information and Computer Science. \n\n.. topic:: References\n\n (1) S. Aeberhard, D. Coomans and O. de Vel, \n Comparison of Classifiers in High Dimensional Settings, \n Tech. Rep. no. 92-02, (1992), Dept. of Computer Science and Dept. of \n Mathematics and Statistics, James Cook University of North Queensland. \n (Also submitted to Technometrics). \n\n The data was used with many others for comparing various \n classifiers. The classes are separable, though only RDA \n has achieved 100% correct classification. \n (RDA : 100%, QDA 99.4%, LDA 98.9%, 1NN 96.1% (z-transformed data)) \n (All results using the leave-one-out technique) \n\n (2) S. Aeberhard, D. Coomans and O. de Vel, \n \"THE CLASSIFICATION PERFORMANCE OF RDA\" \n Tech. Rep. no. 92-01, (1992), Dept. of Computer Science and Dept. of \n Mathematics and Statistics, James Cook University of North Queensland. \n (Also submitted to Journal of Chemometrics).\n\n" ], [ "print('Features: ', wine.feature_names)\nprint('Labels: ', wine.target_names)", "Features: ['alcohol', 'malic_acid', 'ash', 'alcalinity_of_ash', 'magnesium', 'total_phenols', 'flavanoids', 'nonflavanoid_phenols', 'proanthocyanins', 'color_intensity', 'hue', 'od280/od315_of_diluted_wines', 'proline']\nLabels: ['class_0' 'class_1' 'class_2']\n" ], [ "#wine=wine.sample(frac=1)", "_____no_output_____" ], [ "data = wine.data\ntarget = wine.target", "_____no_output_____" ], [ "from sklearn.model_selection import train_test_split\nX_train, X_test, y_train, y_test = train_test_split(data, target, test_size=.3, random_state=109)", "_____no_output_____" ], [ "print('Xtrain : ',X_train.shape)\nprint('Xtest : ',X_test.shape)\nprint('Ytrain : ',y_train.shape)\nprint('Ytest : ',y_test.shape)", "Xtrain : (124, 13)\nXtest : (54, 13)\nYtrain : (124,)\nYtest : (54,)\n" ], [ "print('Xtrain : ',X_train[:5])\nprint('Xtest : ',X_test[:5])\nprint('Ytrain : ',y_train[:5])\nprint('Ytest : ',y_test[:5])", "Xtrain : [[1.323e+01 3.300e+00 2.280e+00 1.850e+01 9.800e+01 1.800e+00 8.300e-01\n 6.100e-01 1.870e+00 1.052e+01 5.600e-01 1.510e+00 6.750e+02]\n [1.384e+01 4.120e+00 2.380e+00 1.950e+01 8.900e+01 1.800e+00 8.300e-01\n 4.800e-01 1.560e+00 9.010e+00 5.700e-01 1.640e+00 4.800e+02]\n [1.220e+01 3.030e+00 2.320e+00 1.900e+01 9.600e+01 1.250e+00 4.900e-01\n 4.000e-01 7.300e-01 5.500e+00 6.600e-01 1.830e+00 5.100e+02]\n [1.288e+01 2.990e+00 2.400e+00 2.000e+01 1.040e+02 1.300e+00 1.220e+00\n 2.400e-01 8.300e-01 5.400e+00 7.400e-01 1.420e+00 5.300e+02]\n [1.305e+01 3.860e+00 2.320e+00 2.250e+01 8.500e+01 1.650e+00 1.590e+00\n 6.100e-01 1.620e+00 4.800e+00 8.400e-01 2.010e+00 5.150e+02]]\nXtest : [[1.330e+01 1.720e+00 2.140e+00 1.700e+01 9.400e+01 2.400e+00 2.190e+00\n 2.700e-01 1.350e+00 3.950e+00 1.020e+00 2.770e+00 1.285e+03]\n [1.293e+01 3.800e+00 2.650e+00 1.860e+01 1.020e+02 2.410e+00 2.410e+00\n 2.500e-01 1.980e+00 4.500e+00 1.030e+00 3.520e+00 7.700e+02]\n [1.221e+01 1.190e+00 1.750e+00 1.680e+01 1.510e+02 1.850e+00 1.280e+00\n 1.400e-01 2.500e+00 2.850e+00 1.280e+00 3.070e+00 7.180e+02]\n [1.253e+01 5.510e+00 2.640e+00 2.500e+01 9.600e+01 1.790e+00 6.000e-01\n 6.300e-01 1.100e+00 5.000e+00 8.200e-01 1.690e+00 5.150e+02]\n [1.421e+01 4.040e+00 2.440e+00 1.890e+01 1.110e+02 2.850e+00 2.650e+00\n 3.000e-01 1.250e+00 5.240e+00 8.700e-01 3.330e+00 1.080e+03]]\nYtrain : [2 2 2 2 1]\nYtest : [0 0 1 2 0]\n" ], [ "from sklearn.naive_bayes import GaussianNB\nnb = GaussianNB()\nnb.fit(X_train, y_train)", "_____no_output_____" ], [ "y_pred = nb.predict(X_test)", "_____no_output_____" ], [ "from sklearn import metrics\nscores = metrics.accuracy_score(y_test, y_pred)\nprint('Accuracy: ','{:2.2%}'.format(scores))", "Accuracy: 90.74%\n" ], [ "cm = metrics.confusion_matrix(y_test, y_pred)\nprint(cm)", "[[20 1 0]\n [ 2 15 2]\n [ 0 0 14]]\n" ], [ "from sklearn.metrics import classification_report\nprint(classification_report(y_test, y_pred))", " precision recall f1-score support\n\n 0 0.91 0.95 0.93 21\n 1 0.94 0.79 0.86 19\n 2 0.88 1.00 0.93 14\n\n accuracy 0.91 54\n macro avg 0.91 0.91 0.91 54\nweighted avg 0.91 0.91 0.91 54\n\n" ], [ "import numpy as np\nprint(np.sum(np.diag(cm)/np.sum(cm)))", "0.9074074074074074\n" ], [ "import itertools\nimport matplotlib.pyplot as plt\n%matplotlib inline\ndef plot_confusion_matrix(cm,\n target_names,\n title='Confusion matrix',\n cmap=None,\n normalize=True):\n \"\"\"\n given a sklearn confusion matrix (cm), make a nice plot\n\n Arguments\n ---------\n cm: confusion matrix from sklearn.metrics.confusion_matrix\n\n target_names: given classification classes such as [0, 1, 2]\n the class names, for example: ['high', 'medium', 'low']\n\n title: the text to display at the top of the matrix\n\n cmap: the gradient of the values displayed from matplotlib.pyplot.cm\n see http://matplotlib.org/examples/color/colormaps_reference.html\n plt.get_cmap('jet') or plt.cm.Blues\n\n normalize: If False, plot the raw numbers\n If True, plot the proportions\n\n Usage\n -----\n plot_confusion_matrix(cm = cm, # confusion matrix created by\n # sklearn.metrics.confusion_matrix\n normalize = True, # show proportions\n target_names = y_labels_vals, # list of names of the classes\n title = best_estimator_name) # title of graph\n\n Citiation\n ---------\n https://www.kaggle.com/grfiv4/plot-a-confusion-matrix\n \n \"\"\"\n\n\n accuracy = np.trace(cm) / float(np.sum(cm))\n misclass = 1 - accuracy\n\n if cmap is None:\n cmap = plt.get_cmap('Blues')\n\n plt.figure(figsize=(8, 6))\n plt.imshow(cm, interpolation='nearest', cmap=cmap)\n plt.title(title)\n plt.colorbar()\n\n if target_names is not None:\n tick_marks = np.arange(len(target_names))\n plt.xticks(tick_marks, target_names, rotation=45)\n plt.yticks(tick_marks, target_names)\n\n if normalize:\n cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n\n\n thresh = cm.max() / 1.5 if normalize else cm.max() / 2\n for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n if normalize:\n plt.text(j, i, \"{:0.4f}\".format(cm[i, j]),\n horizontalalignment=\"center\",\n color=\"white\" if cm[i, j] > thresh else \"black\")\n else:\n plt.text(j, i, \"{:,}\".format(cm[i, j]),\n horizontalalignment=\"center\",\n color=\"white\" if cm[i, j] > thresh else \"black\")\n\n\n plt.tight_layout()\n plt.ylabel('True label')\n plt.xlabel('Predicted label\\naccuracy={:0.4f}; misclass={:0.4f}'.format(accuracy, misclass))\n plt.show()", "_____no_output_____" ], [ "plot_confusion_matrix(cm,wine.target_names)", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%reload_ext watermark\n%matplotlib inline\nfrom os.path import exists\n\nfrom metapool.metapool import *\nfrom metapool import (validate_plate_metadata, assign_emp_index, make_sample_sheet, KLSampleSheet, parse_prep, validate_and_scrub_sample_sheet, generate_qiita_prep_file)\n%watermark -i -v -iv -m -h -p metapool,sample_sheet,openpyxl -u", "Last updated: 2021-12-15T17:15:09.841175-06:00\n\nPython implementation: CPython\nPython version : 3.9.7\nIPython version : 7.30.1\n\nmetapool : 0+untagged.112.g8fed443.dirty\nsample_sheet: 0.12.0\nopenpyxl : 3.0.9\n\nCompiler : Clang 10.0.0 \nOS : Darwin\nRelease : 20.6.0\nMachine : x86_64\nProcessor : i386\nCPU cores : 12\nArchitecture: 64bit\n\nHostname: Kelly-Fogelsons-MacBook-Pro.local\n\nseaborn : 0.11.2\nmatplotlib: 3.5.0\nre : 2.2.1\npandas : 1.3.4\nnumpy : 1.21.2\n\n" ] ], [ [ "# Knight Lab Amplicon Sample Sheet and Mapping (preparation) File Generator \n\n### What is it?\n\nThis Jupyter Notebook allows you to automatically generate sample sheets for amplicon sequencing. \n\n\n### Here's how it should work.\n\nYou'll start out with a **basic plate map** (platemap.tsv) , which just links each sample to it's approprite row and column.\n\nYou can use this google sheet template to generate your plate map:\n\nhttps://docs.google.com/spreadsheets/d/1xPjB6iR3brGeG4bm2un4ISSsTDxFw5yME09bKqz0XNk/edit?usp=sharing\n\nNext you'll automatically assign EMP barcodes in order to produce a **sample sheet** (samplesheet.csv) that can be used in combination with the rest of the sequence processing pipeline. \n\n**Please designate what kind of amplicon sequencing you want to perform:**", "_____no_output_____" ] ], [ [ "seq_type = '16S'\n#options are ['16S', '18S', 'ITS']", "_____no_output_____" ] ], [ [ "## Step 1: read in plate map\n\n**Enter the correct path to the plate map file**. This will serve as the plate map for relating all subsequent information.", "_____no_output_____" ] ], [ [ "plate_map_fp = './test_data/amplicon/compressed-map.tsv'\n\nif not exists(plate_map_fp):\n print(\"Error: %s is not a path to a valid file\" % plate_map_fp)", "_____no_output_____" ] ], [ [ "**Read in the plate map**. It should look something like this:\n\n```\nSample\tRow\tCol\tBlank\nGLY_01_012\tA\t1\tFalse\nGLY_14_034\tB\t1\tFalse\nGLY_11_007\tC\t1\tFalse\nGLY_28_018\tD\t1\tFalse\nGLY_25_003\tE\t1\tFalse\nGLY_06_106\tF\t1\tFalse\nGLY_07_011\tG\t1\tFalse\nGLY_18_043\tH\t1\tFalse\nGLY_28_004\tI\t1\tFalse\n```\n\n**Make sure there a no duplicate IDs.** If each sample doesn't have a different name, an error will be thrown and you won't be able to generate a sample sheet.", "_____no_output_____" ] ], [ [ "plate_df = read_plate_map_csv(open(plate_map_fp,'r'))\n\nplate_df.head()", "_____no_output_____" ] ], [ [ "# Assign barcodes according to primer plate\n\nThis portion of the notebook will assign a barcode to each sample according to the primer plate number.\n\nAs inputs, it requires:\n1. A plate map dataframe (from previous step)\n2. Preparation metadata for the plates, importantly we need the Primer Plate # so we know what **EMP barcodes** to assign to each plate.\n\nThe workflow then:\n1. Joins the preparation metadata with the plate metadata.\n2. Assigns indices per sample", "_____no_output_____" ], [ "## Enter and validate the plating metadata\n\n- In general you will want to update all the fields, but the most important ones are the `Primer Plate #` and the `Plate Position`. `Primer Plate #` determines which EMP barcodes will be used for this plate. `Plate Position` determines the physical location of the plate.\n- If you are plating less than four plates, then remove the metadata for that plate by deleting the text between the curly braces.\n- For missing fields, write NA between the single quotes for example `'NA'`.\n- To enter a plate copy and paste the contents from the plates below.", "_____no_output_____" ] ], [ [ "_metadata = [\n {\n # top left plate\n 'Plate Position': '1',\n 'Primer Plate #': '1',\n \n 'Sample Plate': 'THDMI_UK_Plate_2',\n 'Project_Name': 'THDMI UK',\n\n 'Plating': 'SF',\n 'Extraction Kit Lot': '166032128',\n 'Extraction Robot': 'Carmen_HOWE_KF3',\n 'TM1000 8 Tool': '109379Z',\n 'Primer Date': '2021-08-17', # yyyy-mm-dd\n 'MasterMix Lot': '978215',\n 'Water Lot': 'RNBJ0628',\n 'Processing Robot': 'Echo550',\n 'Original Name': ''\n },\n {\n # top right plate\n 'Plate Position': '2',\n 'Primer Plate #': '2',\n \n 'Sample Plate': 'THDMI_UK_Plate_3',\n 'Project_Name': 'THDMI UK',\n\n 'Plating':'AS',\n 'Extraction Kit Lot': '166032128',\n 'Extraction Robot': 'Carmen_HOWE_KF4',\n 'TM1000 8 Tool': '109379Z',\n 'Primer Date': '2021-08-17', # yyyy-mm-dd\n 'MasterMix Lot': '978215',\n 'Water Lot': 'RNBJ0628',\n 'Processing Robot': 'Echo550',\n 'Original Name': ''\n },\n {\n # bottom left plate\n 'Plate Position': '3',\n 'Primer Plate #': '3',\n \n 'Sample Plate': 'THDMI_UK_Plate_4',\n 'Project_Name': 'THDMI UK',\n\n 'Plating':'MB_SF',\n 'Extraction Kit Lot': '166032128',\n 'Extraction Robot': 'Carmen_HOWE_KF3',\n 'TM1000 8 Tool': '109379Z',\n 'Primer Date': '2021-08-17', # yyyy-mm-dd\n 'MasterMix Lot': '978215',\n 'Water Lot': 'RNBJ0628',\n 'Processing Robot': 'Echo550',\n 'Original Name': ''\n },\n {\n # bottom right plate\n 'Plate Position': '4',\n 'Primer Plate #': '4',\n \n 'Sample Plate': 'THDMI_US_Plate_6',\n 'Project_Name': 'THDMI US',\n\n 'Plating':'AS',\n 'Extraction Kit Lot': '166032128',\n 'Extraction Robot': 'Carmen_HOWE_KF4',\n 'TM1000 8 Tool': '109379Z',\n 'Primer Date': '2021-08-17', # yyyy-mm-dd\n 'MasterMix Lot': '978215',\n 'Water Lot': 'RNBJ0628',\n 'Processing Robot': 'Echo550',\n 'Original Name': ''\n },\n]\n\nplate_metadata = validate_plate_metadata(_metadata)\nplate_metadata", "_____no_output_____" ] ], [ [ "The `Plate Position` and `Primer Plate #` allow us to figure out which wells are associated with each of the EMP barcodes.", "_____no_output_____" ] ], [ [ "if plate_metadata is not None:\n plate_df = assign_emp_index(plate_df, plate_metadata, seq_type).reset_index()\n\n plate_df.head()\nelse:\n print('Error: Please fix the errors in the previous cell')", "_____no_output_____" ] ], [ [ "As you can see in the table above, the resulting table is now associated with the corresponding EMP barcodes (`Golay Barcode`, `Forward Primer Linker`, etc), and the plating metadata (`Primer Plate #`, `Primer Date`, `Water Lot`, etc).", "_____no_output_____" ] ], [ [ "plate_df.head()", "_____no_output_____" ] ], [ [ "# Combine plates (optional)\n\nIf you would like to combine existing plates with these samples, enter the path to their corresponding sample sheets and mapping (preparation) files below. Otherwise you can skip to the next section.\n\n- sample sheet and mapping (preparation)", "_____no_output_____" ] ], [ [ "files = [\n # uncomment the line below and point to the correct filepaths to combine with previous plates\n # ['test_output/amplicon/2021_08_17_THDMI-4-6_samplesheet.csv', 'test_output/amplicon/2021-08-01-515f806r_prep.tsv'],\n]\nsheets, preps = [], []\n\nfor sheet, prep in files:\n sheets.append(KLSampleSheet(sheet))\n preps.append(parse_prep(prep))\n \nif len(files):\n print('%d pair of files loaded' % len(files))", "_____no_output_____" ] ], [ [ "# Make Sample Sheet\n\nThis workflow takes the pooled sample information and writes an Illumina sample sheet that can be given directly to the sequencing center or processing pipeline. Note that as of writing `bcl2fastq` does not support error-correction in Golay barcodes so the sample sheet is used to generate a mapping (preparation) file but not to demultiplex sequences. Demultiplexing takes place in [Qiita](https://qiita.ucsd.edu).\n\nAs inputs, this notebook requires:\n1. A plate map DataFrame (from previous step)\n\nThe workflow:\n1. formats sample names as bcl2fastq-compatible\n2. formats sample data\n3. sets values for sample sheet fields and formats sample sheet.\n4. writes the sample sheet to a file", "_____no_output_____" ], [ "## Step 1: Format sample names to be bcl2fastq-compatible\n\nbcl2fastq requires *only* alphanumeric, hyphens, and underscore characters. We'll replace all non-those characters\nwith underscores and add the bcl2fastq-compatible names to the DataFrame.", "_____no_output_____" ] ], [ [ "plate_df['sample sheet Sample_ID'] = plate_df['Sample'].map(bcl_scrub_name)\n\nplate_df.head()", "_____no_output_____" ] ], [ [ "## Format the sample sheet data\n\nThis step formats the data columns appropriately for the sample sheet, using the values we've calculated previously.\n\nThe newly-created `bcl2fastq`-compatible names will be in the `Sample ID` and `Sample Name` columns. The original sample names will be in the Description column.\n\nModify lanes to indicate which lanes this pool will be sequenced on.\n\nThe `Project Name` and `Project Plate` columns will be placed in the `Sample_Project` and `Sample_Name` columns, respectively.\n\nsequencer is important for making sure the i5 index is in the correct orientation for demultiplexing. `HiSeq4000`, `HiSeq3000`, `NextSeq`, and `MiniSeq` all require reverse-complemented i5 index sequences. If you enter one of these exact strings in for sequencer, it will revcomp the i5 sequence for you.\n\n`HiSeq2500`, `MiSeq`, and `NovaSeq` will not revcomp the i5 sequence.", "_____no_output_____" ] ], [ [ "sequencer = 'HiSeq4000'\nlanes = [1]\n\nmetadata = {\n 'Bioinformatics': [\n {\n 'Sample_Project': 'THDMI_10317',\n 'QiitaID': '10317',\n 'BarcodesAreRC': 'False',\n 'ForwardAdapter': '',\n 'ReverseAdapter': '',\n 'HumanFiltering': 'True',\n 'library_construction_protocol': 'Illumina EMP protocol 515fbc, 806r amplification of 16S rRNA V4',\n 'experiment_design_description': 'Equipment',\n },\n ],\n 'Contact': [\n {\n 'Sample_Project': 'THDMI_10317',\n # non-admin contacts who want to know when the sequences\n # are available in Qiita\n 'Email': '[email protected],[email protected]'\n },\n ],\n 'Chemistry': 'Amplicon',\n 'Assay': 'TruSeq HT',\n}\n\nsheet = make_sample_sheet(metadata, plate_df, sequencer, lanes)\n\n\nsheet.Settings['Adapter'] = 'AGATCGGAAGAGCACACGTCTGAACTCCAGTCA'\nsheet.Settings['AdapterRead2'] = 'AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT'", "/Users/kfolelso/Documents/metagenomics_pooling_notebook_troubleshooting/metapool/sample_sheet.py:473: UserWarning: The column I5_Index_ID in the sample sheet is empty\n warnings.warn('The column %s in the sample sheet is empty' %\n/Users/kfolelso/Documents/metagenomics_pooling_notebook_troubleshooting/metapool/sample_sheet.py:473: UserWarning: The column index2 in the sample sheet is empty\n warnings.warn('The column %s in the sample sheet is empty' %\n" ] ], [ [ "Check for any possible errors in the sample sheet", "_____no_output_____" ] ], [ [ "sheet = validate_and_scrub_sample_sheet(sheet)", "_____no_output_____" ] ], [ [ "Add the other sample sheets", "_____no_output_____" ] ], [ [ "if len(sheets):\n sheet.merge(sheets)", "_____no_output_____" ] ], [ [ "## Step 3: Write the sample sheet to file", "_____no_output_____" ] ], [ [ "# write sample sheet as .csv\nsample_sheet_fp = './test_output/amplicon/2021_08_17_THDMI-4-6_samplesheet16S.csv'\n\nif exists(sample_sheet_fp):\n print(\"Warning! This file exists already.\")", "_____no_output_____" ], [ "with open(sample_sheet_fp,'w') as f:\n sheet.write(f)\n \n!head -n 30 {sample_sheet_fp}\n!echo ...\n!tail -n 15 {sample_sheet_fp}", "[Header],,,,,,,,,,\nIEMFileVersion,4,,,,,,,,,\nDate,2021-12-15,,,,,,,,,\nWorkflow,GenerateFASTQ,,,,,,,,,\nApplication,FASTQ Only,,,,,,,,,\nAssay,TruSeq HT,,,,,,,,,\nDescription,,,,,,,,,,\nChemistry,Amplicon,,,,,,,,,\n,,,,,,,,,,\n[Reads],,,,,,,,,,\n151,,,,,,,,,,\n151,,,,,,,,,,\n,,,,,,,,,,\n[Settings],,,,,,,,,,\nReverseComplement,0,,,,,,,,,\nAdapter,AGATCGGAAGAGCACACGTCTGAACTCCAGTCA,,,,,,,,,\nAdapterRead2,AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT,,,,,,,,,\n,,,,,,,,,,\n[Data],,,,,,,,,,\nSample_ID,Sample_Name,Sample_Plate,Sample_Well,I7_Index_ID,index,Sample_Project,Well_description,I5_Index_ID,index2,Lane\nX00180471,X00180471,THDMI_10317_PUK2,A1,515rcbc0,AGCCTTCGTCGC,THDMI_10317,X00180471,,,1\nX00180199,X00180199,THDMI_10317_PUK2,C1,515rcbc12,CGTATAAATGCG,THDMI_10317,X00180199,,,1\nX00179789,X00179789,THDMI_10317_PUK2,E1,515rcbc24,TGACTAATGGCC,THDMI_10317,X00179789,,,1\nX00180201,X00180201,THDMI_10317_PUK2,G1,515rcbc36,GTGGAGTCTCAT,THDMI_10317,X00180201,,,1\nX00180464,X00180464,THDMI_10317_PUK2,I1,515rcbc48,TGATGTGCTAAG,THDMI_10317,X00180464,,,1\nX00179796,X00179796,THDMI_10317_PUK2,K1,515rcbc60,TGTGCACGCCAT,THDMI_10317,X00179796,,,1\nX00179888,X00179888,THDMI_10317_PUK2,M1,515rcbc72,GGTGAGCAAGCA,THDMI_10317,X00179888,,,1\nX00179969,X00179969,THDMI_10317_PUK2,O1,515rcbc84,CTATGTATTAGT,THDMI_10317,X00179969,,,1\nBLANK2_2A,BLANK2.2A,THDMI_10317_PUK2,A3,515rcbc1,TCCATACCGGAA,THDMI_10317,BLANK2.2A,,,1\nBLANK2_2B,BLANK2.2B,THDMI_10317_PUK2,C3,515rcbc13,ATGCTGCAACAC,THDMI_10317,BLANK2.2B,,,1\n...\nX00179670,X00179670,THDMI_10317_PUS6,F24,515rcbc323,GTCAGTATGGCT,THDMI_10317,X00179670,,,1\nX00179548,X00179548,THDMI_10317_PUS6,H24,515rcbc335,GTCCTCGCGACT,THDMI_10317,X00179548,,,1\nX00179326,X00179326,THDMI_10317_PUS6,J24,515rcbc347,CGTTCGCTAGCC,THDMI_10317,X00179326,,,1\nX00179165,X00179165,THDMI_10317_PUS6,L24,515rcbc359,TGCCTGCTCGAC,THDMI_10317,X00179165,,,1\nX00179035,X00179035,THDMI_10317_PUS6,N24,515rcbc371,TCTTACCCATAA,THDMI_10317,X00179035,,,1\nX00179260,X00179260,THDMI_10317_PUS6,P24,515rcbc383,TGTGCTTGTAGG,THDMI_10317,X00179260,,,1\n,,,,,,,,,,\n[Bioinformatics],,,,,,,,,,\nSample_Project,QiitaID,BarcodesAreRC,ForwardAdapter,ReverseAdapter,HumanFiltering,library_construction_protocol,experiment_design_description,,,\nTHDMI_10317,10317,False,,,True,\"Illumina EMP protocol 515fbc, 806r amplification of 16S rRNA V4\",Equipment,,,\n,,,,,,,,,,\n[Contact],,,,,,,,,,\nSample_Project,Email,,,,,,,,,\nTHDMI_10317,\"[email protected],[email protected]\",,,,,,,,,\n,,,,,,,,,,\n" ] ], [ [ "# Create a mapping (preparation) file for Qiita", "_____no_output_____" ] ], [ [ "output_filename = 'test_output/amplicon/2021-08-01-515f806r_prep.tsv'", "_____no_output_____" ], [ "qiita_df = generate_qiita_prep_file(plate_df, seq_type)\n\nqiita_df.head()", "_____no_output_____" ], [ "qiita_df.set_index('sample_name', verify_integrity=True).to_csv(output_filename, sep='\\t')", "_____no_output_____" ] ], [ [ "Add the previous sample sheets", "_____no_output_____" ] ], [ [ "if len(preps):\n prep = prep.append(preps, ignore_index=True)", "_____no_output_____" ], [ "!head -n 5 {output_filename}", "sample_name\tbarcode\tprimer\tprimer_plate\twell_id\tplating\textractionkit_lot\textraction_robot\ttm1000_8_tool\tprimer_date\tmastermix_lot\twater_lot\tprocessing_robot\ttm300_8_tool\ttm50_8_tool\tsample_plate\tproject_name\torig_name\twell_description\texperiment_design_description\tlibrary_construction_protocol\tlinker\tplatform\trun_center\trun_date\trun_prefix\tpcr_primers\tsequencing_meth\ttarget_gene\ttarget_subfragment\tcenter_name\tcenter_project_name\tinstrument_model\trunid\r\nX00180471\tAGCCTTCGTCGC\tGTGYCAGCMGCCGCGGTAA\t1\tA1\tSF\t166032128\tCarmen_HOWE_KF3\t109379Z\t2021-08-17\t978215\tRNBJ0628\tEcho550\t\t\tTHDMI_UK_Plate_2\tTHDMI_10317\tX00180471\tTHDMI_UK_Plate_2.X00180471.A1\t\tIllumina EMP protocol 515fbc, 806r amplification of 16S rRNA V4\tGT\tIllumina\tUCSDMI\t\t\tFWD:GTGYCAGCMGCCGCGGTAA; REV:GGACTACNVGGGTWTCTAAT\tSequencing by synthesis\t16S rRNA\tV4\tUCSDMI\t\t\t\r\nX00180199\tCGTATAAATGCG\tGTGYCAGCMGCCGCGGTAA\t1\tC1\tSF\t166032128\tCarmen_HOWE_KF3\t109379Z\t2021-08-17\t978215\tRNBJ0628\tEcho550\t\t\tTHDMI_UK_Plate_2\tTHDMI_10317\tX00180199\tTHDMI_UK_Plate_2.X00180199.C1\t\tIllumina EMP protocol 515fbc, 806r amplification of 16S rRNA V4\tGT\tIllumina\tUCSDMI\t\t\tFWD:GTGYCAGCMGCCGCGGTAA; REV:GGACTACNVGGGTWTCTAAT\tSequencing by synthesis\t16S rRNA\tV4\tUCSDMI\t\t\t\r\nX00179789\tTGACTAATGGCC\tGTGYCAGCMGCCGCGGTAA\t1\tE1\tSF\t166032128\tCarmen_HOWE_KF3\t109379Z\t2021-08-17\t978215\tRNBJ0628\tEcho550\t\t\tTHDMI_UK_Plate_2\tTHDMI_10317\tX00179789\tTHDMI_UK_Plate_2.X00179789.E1\t\tIllumina EMP protocol 515fbc, 806r amplification of 16S rRNA V4\tGT\tIllumina\tUCSDMI\t\t\tFWD:GTGYCAGCMGCCGCGGTAA; REV:GGACTACNVGGGTWTCTAAT\tSequencing by synthesis\t16S rRNA\tV4\tUCSDMI\t\t\t\r\nX00180201\tGTGGAGTCTCAT\tGTGYCAGCMGCCGCGGTAA\t1\tG1\tSF\t166032128\tCarmen_HOWE_KF3\t109379Z\t2021-08-17\t978215\tRNBJ0628\tEcho550\t\t\tTHDMI_UK_Plate_2\tTHDMI_10317\tX00180201\tTHDMI_UK_Plate_2.X00180201.G1\t\tIllumina EMP protocol 515fbc, 806r amplification of 16S rRNA V4\tGT\tIllumina\tUCSDMI\t\t\tFWD:GTGYCAGCMGCCGCGGTAA; REV:GGACTACNVGGGTWTCTAAT\tSequencing by synthesis\t16S rRNA\tV4\tUCSDMI\t\t\t\r\n" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "<p align=\"center\">\n <h1 align=\"center\">Machine Learning and Statistics Tasks 2020</h1>\n <h1 align=\"center\"> Task 1: Python function sqrt2</h1>\n <h2 align=\"center\"> Author: Ezekiel Onaloye</h2>\n <h2 align=\"center\"> Created: November 2020 </h2> \n</p>", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "### Task 1\nWrite a Python function called sqrt2 that calculates and prints to the screen the square root of 2 to 100 decimal places. ", "_____no_output_____" ], [ "### Introduction\n\nA square root of a number is a value that gives the original number when multiplied by itself. For example, 2 x 2 = 4, so a square root of 4 is 2. -2 x -2 is 4 too, so -2 is also a square root of 4.\n\nA square root is like asking ourselves, \"what value can we multiply by itself to get this outcome?\". That multiplication by itself, is also called squaring. So, in other words, 3 squared is 9, and so the square root of 9 is 3.", "_____no_output_____" ], [ "### Simple Python function sqrt2", "_____no_output_____" ], [ "<h4 align=\"left\"> Algorithm </h4>\n\n1. Take the input from the user\n \n2. Create a function that takes one argument\n \n3. Then, if the number is negative, return nothing \n \n4. As a square root of a number is simply the number raised to the power 0.5, raise the given number to the power of 0.5\n\n5. This will give us the square root of the number; return it\n \n6. Print out the result to the user", "_____no_output_____" ] ], [ [ "# Python function called sqrt2 \n# Adapted from https://www.educative.io/edpresso/how-to-take-the-square-root-of-a-number-in-python\n# Function takes input,calculates and prints to the screen the square root of 2 to 100 decimal places\n\nn = int(input(\"Please input the base value: \"))\n \ndef sqrt(n):\n if n < 0:\n return\n else:\n return n**0.5\n \nprint(\"The square root of\", n, \"is\", format(sqrt(n),'.100f'))", "Please input the base value: 2\nThe square root of 2 is 1.4142135623730951454746218587388284504413604736328125000000000000000000000000000000000000000000000000\n" ] ], [ [ "### Python function sqrt2", "_____no_output_____" ], [ "<h4 align=\"left\"> Algorithm </h4> \n\n1. Let n be the given number\n2. create function sqrt2(takes a parameter)\n3. Given number stored as variable x\n4. y equate given number + 1 divided by 2\n5. while y is less than x\n6. then x = y\n7. y store the value of x+n divided by itself and overall by 2\n8. return y as the square root", "_____no_output_____" ] ], [ [ "# Python function called sqrt2 \n# Adapted from \n# Function takes input,calculates and prints to the screen the square root of 2 to 100 decimal places\n\n# user asked to enter a number \nprint(\"\")\nn = int(input(\"Please enter any number to find square root: \"))\n\n# function declared\ndef sqrt2(n):\n # number entered stored as x, x is a variable carrying entered number\n x = n\n # y store x + 1 divided by 2 based on \n y = (x + 1) / 2\n while y < x:\n x = y\n y = (x + n / x) / 2\n return y;\nprint(\"Square root of\", n, \"the given number is %.100f\" % sqrt2(n))", "\nPlease enter any number to find square root: 2\nSquare root of 2 the given number is 1.4142135623730949234300169337075203657150268554687500000000000000000000000000000000000000000000000000\n" ] ], [ [ "### References", "_____no_output_____" ], [ "[1] https://www.educative.io/edpresso/how-to-take-the-square-root-of-a-number-in-python\n\n[2] https://www.codegrepper.com/code-examples/python/python+print+upto+two+decimal+places \n\n[3] https://kodify.net/python/math/square-root/\n\n[4] https://www.educba.com/square-root-in-python/ ", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%load_ext autoreload\n%autoreload 2", "The autoreload extension is already loaded. To reload it, use:\n %reload_ext autoreload\n" ], [ "import numpy as np\nimport matplotlib.pyplot as plt\n%matplotlib inline\nfrom tqdm.autonotebook import tqdm\nfrom joblib import Parallel, delayed\nimport umap\nimport pandas as pd", "_____no_output_____" ], [ "from avgn.utils.paths import DATA_DIR, most_recent_subdirectory, ensure_dir", "_____no_output_____" ], [ "DATASET_ID = 'swamp_sparrow'", "_____no_output_____" ], [ "from avgn.utils.hparams import HParams\nfrom avgn.dataset import DataSet", "_____no_output_____" ], [ "from avgn.signalprocessing.create_spectrogram_dataset import prepare_wav, create_label_df, get_row_audio", "_____no_output_____" ] ], [ [ "### create dataset", "_____no_output_____" ] ], [ [ "hparams = HParams(\n num_mel_bins = 32,\n mel_lower_edge_hertz=100,\n mel_upper_edge_hertz=22000,\n butter_lowcut = 100,\n butter_highcut = 22000,\n ref_level_db = 25,\n min_level_db = -50,\n mask_spec = True,\n win_length_ms = 5,\n hop_length_ms = .5,\n mask_spec_kwargs = {\"spec_thresh\": 0.9, \"offset\": 1e-10}\n)", "_____no_output_____" ], [ "# create a dataset object\ndataset = DataSet(DATASET_ID, hparams = hparams)", "_____no_output_____" ], [ "dataset.sample_json", "_____no_output_____" ] ], [ [ "#### Create dataset based upon JSON", "_____no_output_____" ] ], [ [ "from joblib import Parallel, delayed\nn_jobs = -1; verbosity = 10", "_____no_output_____" ], [ "with Parallel(n_jobs=n_jobs, verbose=verbosity) as parallel:\n syllable_dfs = parallel(\n delayed(create_label_df)(\n dataset.data_files[key].data,\n hparams=dataset.hparams,\n labels_to_retain=[\"syllable\", \"pos_in_syllable\"],\n unit=\"elements\",\n key = key,\n )\n for key in tqdm(dataset.data_files.keys())\n )\nsyllable_df = pd.concat(syllable_dfs)\nlen(syllable_df)", "_____no_output_____" ], [ "syllable_df[:3]", "_____no_output_____" ], [ "syllable_df", "_____no_output_____" ] ], [ [ "### get audio for dataset", "_____no_output_____" ] ], [ [ "with Parallel(n_jobs=n_jobs, verbose=verbosity) as parallel:\n syllable_dfs = parallel(\n delayed(get_row_audio)(\n syllable_df[syllable_df.key == key], \n dataset.data_files[key].data['wav_loc'], \n dataset.hparams\n )\n for key in tqdm(syllable_df.key.unique())\n )\nsyllable_df = pd.concat(syllable_dfs)\nlen(syllable_df)", "_____no_output_____" ], [ "syllable_df[:3]", "_____no_output_____" ], [ "syllable_df.indvi.values[:100]", "_____no_output_____" ], [ "sylls = syllable_df.audio.values", "_____no_output_____" ], [ "nrows = 5\nncols = 10\nzoom = 2\nfig, axs = plt.subplots(ncols=ncols, nrows = nrows,figsize = (ncols*zoom, nrows+zoom/1.5))\nfor i, syll in tqdm(enumerate(sylls), total = nrows*ncols):\n ax = axs.flatten()[i]\n ax.plot(syll)\n if i == nrows*ncols -1:\n break", "_____no_output_____" ] ], [ [ "### Create spectrograms", "_____no_output_____" ] ], [ [ "from avgn.visualization.spectrogram import draw_spec_set\nfrom avgn.signalprocessing.create_spectrogram_dataset import make_spec, mask_spec, log_resize_spec, pad_spectrogram", "_____no_output_____" ], [ "syllables_wav = syllable_df.audio.values\nsyllables_rate = syllable_df.rate.values", "_____no_output_____" ], [ "with Parallel(n_jobs=n_jobs, verbose=verbosity) as parallel:\n # create spectrograms\n syllables_spec = parallel(\n delayed(make_spec)(\n syllable,\n rate,\n hparams=dataset.hparams,\n mel_matrix=dataset.mel_matrix,\n use_mel=True,\n use_tensorflow=False,\n )\n for syllable, rate in tqdm(\n zip(syllables_wav, syllables_rate),\n total=len(syllables_rate),\n desc=\"getting syllable spectrograms\",\n leave=False,\n )\n )", "_____no_output_____" ] ], [ [ "### Rescale spectrogram\n- using log rescaling", "_____no_output_____" ] ], [ [ "log_scaling_factor = 4", "_____no_output_____" ], [ "with Parallel(n_jobs=n_jobs, verbose=verbosity) as parallel:\n syllables_spec = parallel(\n delayed(log_resize_spec)(spec, scaling_factor=log_scaling_factor)\n for spec in tqdm(syllables_spec, desc=\"scaling spectrograms\", leave=False)\n )", "_____no_output_____" ], [ "draw_spec_set(syllables_spec, zoom=1, maxrows=10, colsize=25)", "(25.0, 10) (320, 800) 25.0 32 800\n" ] ], [ [ "### Pad spectrograms", "_____no_output_____" ] ], [ [ "syll_lens = [np.shape(i)[1] for i in syllables_spec]\npad_length = np.max(syll_lens)", "_____no_output_____" ], [ "plt.hist(syll_lens)", "_____no_output_____" ], [ "with Parallel(n_jobs=n_jobs, verbose=verbosity) as parallel:\n\n syllables_spec = parallel(\n delayed(pad_spectrogram)(spec, pad_length)\n for spec in tqdm(\n syllables_spec, desc=\"padding spectrograms\", leave=False\n )\n )", "_____no_output_____" ], [ "draw_spec_set(syllables_spec, zoom=1, maxrows=10, colsize=25)", "(25.0, 10) (320, 800) 25.0 32 800\n" ] ], [ [ "### save dataset", "_____no_output_____" ] ], [ [ "np.shape(syllables_spec)", "_____no_output_____" ], [ "syllable_df[:3]", "_____no_output_____" ], [ "syllable_df['spectrogram'] = syllables_spec", "_____no_output_____" ], [ "save_loc = DATA_DIR / 'syllable_dfs' / DATASET_ID / 'swampsparrow.pickle'\nensure_dir(save_loc)\nsyllable_df.to_pickle(save_loc)", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Model structure\n<img src=\"../freaminsg_integrated_model/notebooks/meetings/figures/model_structure.jpeg\" alt=\"Alt text\" width=\"800\" title=\"Title text\" />", "_____no_output_____" ], [ "# Project structure\n\n```\nDREAMINSG_INTEGRATED_MODEL/\n|-- dreaminsg_integrated_model/\n|\t|-- data/\n|\t|\t|-- disruptive_scenarios/\n|\t|\t|\t|-- disrupt_generator.py\n|\t|\t|\n|\t|\t|-- networks/\n|\t|\t\t|-- examples/\n|\t|\t\t|-- power/\n|\t|\t\t|-- water/\n|\t|\t\t|-- transportation/\n|\t|\n|\t|-- network_sim_models/\n|\t|\t|-- power/power_system_model.py\n|\t|\t|-- transportation/\n|\t|\t|-- water/water_network_model.py\n|\t|\t|-- interdependencies.py\n|\t|\n|\t|-- results/\n|\t|\t|-- figures/plots.py\n|\t|\n|\t|-- main.py\n|\n|-- notebooks\n|\t|-- demo.ipynb\n|\t|-- model_structure.ipynb\n|\t|-- plots.ipynb\n|\n|-- setup.py\n|-- README.md\n|-- LICENSE\n```\n\nModel available on [Github](https://github.com/srijithbalakrishnan/dreaminsg-integrated-model.git)", "_____no_output_____" ], [ "# Model Demonstration\n\n[demo.ipynb](./demo.ipynb)", "_____no_output_____" ], [ "# _To do list_\n\n## Ongoing (April - May)\n1. ### Scaling the model to larger networks.\n \n (a) Some issues related to wntr package identified and are being fixed.\n \n (b) Discrete event simulation is being tested. Saves a lot of time, but there are some issues with it while implementing the wntr simulation. Simulates only the network conditions during:\n \n - the start of the simulation;\n - when some component fails;\n - when the crew reaches a failed component for repair; and\n - when a component is completely repaired and the end of the simulation. \n \n2. ### Changes to the way recovery is modeled -- recovery times instead of recovery rates.\n\n3. ### Incorporate pipe leaks and failures in the model. (duration of pipe leaks + repair time)\n4. ### Identification of relevant metrics for resilience quantification.\n\n (a) Currently using ratio of consumption to pre-disaster consumption (water and power)\n \n (b) Additional suggestions provided by Dr. Giovini Sansavinni to be incorporated.\n\n5. ### Add more interdependencies and direct impact scenarios.\n\n## Mid-term (May - July)\n1. ### Integrating module for optimizing the recovery strategy. \n\n2. ### Network generation and automation.", "_____no_output_____" ], [ "## Comments and Feedback\n\n### 1. What data to be collected from the model?\n - Network\n - Disruptive scenario\n\n### 2. Weighting\n- Fuzzy Ordered Weighting Methods\n- ", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Construction of pYPK0_PDC1_HIS3_RPS19b\n\n[pYPKa_Z_PDC1](pYPKa_Z_PDC1.ipynb)\n\n[pYPKa_E_RPS19b](pYPKa_E_RPS19b.ipynb) ", "_____no_output_____" ] ], [ [ "from pydna.all import *", "_____no_output_____" ], [ "p567,p577,p468,p467,p568,p578,p775,p778,p167,p166 = parse(\"yeast_pahtway_kit_standard_primers.txt\")", "_____no_output_____" ] ], [ [ "[Yeast Pathway Kit Standard Primers](ypk_std_primers.ipynb)", "_____no_output_____" ] ], [ [ "from Bio.Restriction import ZraI, AjiI, EcoRV", "_____no_output_____" ], [ "pYPK0 =read(\"pYPK0.gb\")", "_____no_output_____" ], [ "promoter_clone = pYPKa_Z_PDC1 =read(\"pYPKa_Z_PDC1.gb\")", "_____no_output_____" ], [ "gene_clone =read(\"pYPKa_A_ScHIS3.gb\")", "_____no_output_____" ], [ "terminator_clone = pYPKa_E_RPS19b =read(\"pYPKa_E_RPS19b.gb\")", "_____no_output_____" ], [ "p =pcr( p167, p567, promoter_clone)\ng =pcr( p468, p467, gene_clone)\nt =pcr( p568, p166, terminator_clone)", "_____no_output_____" ], [ "pYPK0_E_Z, stuffer = pYPK0.cut((EcoRV, ZraI))", "_____no_output_____" ], [ "(pYPK0_E_Z, p, g, t)", "_____no_output_____" ], [ "asm =Assembly((pYPK0_E_Z, p, g, t), limit=31)", "_____no_output_____" ], [ "asm", "_____no_output_____" ], [ "candidate = asm.assemble_circular()[0]\ncandidate.figure()", "_____no_output_____" ], [ "result = candidate.synced(pYPK0)", "_____no_output_____" ] ], [ [ "The new construct should have cseguid ```m-7lOIJ60aZl4b7rN4vanlD7OWk``` and 9404 bp.", "_____no_output_____" ] ], [ [ "result.write(\"pYPK0_PDC1_HIS3_RPS19b.gb\")", "_____no_output_____" ] ], [ [ "###[Download](pYPK0_PDC1_HIS3_RPS19b.gb)", "_____no_output_____" ] ], [ [ "from pydna.all import *\nreloaded =read(\"pYPK0_PDC1_HIS3_RPS19b.gb\")", "_____no_output_____" ], [ "reloaded =read(\"pYPK0_PDC1_HIS3_RPS19b.gb\")", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import caffe\nimport numpy as np\nimport matplotlib.pyplot as plt\n%matplotlib inline", "_____no_output_____" ], [ "net = caffe.Net('deploy.prototxt',caffe.TEST)", "_____no_output_____" ], [ "# net.layers[0].blobs>0\n\nfor i ,(name,layer) in enumerate(zip(net._layer_names,net.layers)):\n \n# display all layers\n print(name)\n for bottom_name in net.bottom_names[name]:\n \n print('layer_%s bottom name: '%name,bottom_name)\n if(len(net.bottom_names[name])>1):\n print('True')\n# parameters layers\n# if(len(layer.blobs)>0):\n# print(name)\n ", "data\nconv1\n('layer_conv1 bottom name: ', 'data')\nrelu1\n('layer_relu1 bottom name: ', 'conv1')\nnorm1\n('layer_norm1 bottom name: ', 'conv1')\npool1\n('layer_pool1 bottom name: ', 'norm1')\nconv2\n('layer_conv2 bottom name: ', 'pool1')\nrelu2\n('layer_relu2 bottom name: ', 'conv2')\nnorm2\n('layer_norm2 bottom name: ', 'conv2')\npool2\n('layer_pool2 bottom name: ', 'norm2')\nconv3\n('layer_conv3 bottom name: ', 'pool2')\nrelu3\n('layer_relu3 bottom name: ', 'conv3')\nconv4\n('layer_conv4 bottom name: ', 'conv3')\nrelu4\n('layer_relu4 bottom name: ', 'conv4')\nconv5\n('layer_conv5 bottom name: ', 'conv4')\nrelu5\n('layer_relu5 bottom name: ', 'conv5')\npool5\n('layer_pool5 bottom name: ', 'conv5')\nfc6\n('layer_fc6 bottom name: ', 'pool5')\nrelu6\n('layer_relu6 bottom name: ', 'fc6')\ndrop6\n('layer_drop6 bottom name: ', 'fc6')\nfc7\n('layer_fc7 bottom name: ', 'fc6')\nrelu7\n('layer_relu7 bottom name: ', 'fc7')\ndrop7\n('layer_drop7 bottom name: ', 'fc7')\nfc8\n('layer_fc8 bottom name: ', 'fc7')\nprob\n('layer_prob bottom name: ', 'fc8')\n" ], [ "arr = np.array([[1,2,3],[2,3,4]])\narr.swapaxes(0,1)", "_____no_output_____" ], [ "arr2 = np.array([[2,6,7],[8,9,5]])\nnp.concatenate([arr,arr2],axis=1)", "_____no_output_____" ], [ "# np.concatenate([[[1,2]],[[2,3]],[[3,3]]],axis=1)\nprint(arr2[:])\nprint(arr2[None])\nprint(arr2[:,None])\nprint(arr2[:,:,None])", "[[2 6 7]\n [8 9 5]]\n[[[2 6 7]\n [8 9 5]]]\n[[[2 6 7]]\n\n [[8 9 5]]]\n[[[2]\n [6]\n [7]]\n\n [[8]\n [9]\n [5]]]\n" ], [ "def transform(D,W):\n W, D = np.concatenate([w[:,None] for w in W], axis=1), np.concatenate([d[:,:,None] for d in D], axis=2)\n return W.reshape((-1,)+W.shape[2:]), D.reshape((D.shape[0], -1)+D.shape[3:])", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code" ] ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "<div class=\"alert alert-block alert-info\">\n <b><h1>ENGR 1330 Computational Thinking with Data Science </h1></b> \n</div> \n\nCopyright © 2021 Theodore G. Cleveland and Farhang Forghanparast\n\nLast GitHub Commit Date: 5 Mar 2022\n \n# 19: Simulation\n- Simulating random values\n- Draw with/without replacement\n- Histograms to check behavior\n- Concept of an interval", "_____no_output_____" ], [ "## References", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<img src=\"../../images/banners/python-basics.png\" width=\"600\"/>", "_____no_output_____" ], [ "# <img src=\"../../images/logos/python.png\" width=\"23\"/> Conda Environments \n", "_____no_output_____" ], [ "## <img src=\"../../images/logos/toc.png\" width=\"20\"/> Table of Contents \n* [Understanding Conda Environments](#understanding_conda_environments)\n* [Understanding Basic Package Management With Conda](#understanding_basic_package_management_with_conda)\n * [Searching and Installing Packages](#searching_and_installing_packages)\n * [Updating and Removing Packages](#updating_and_removing_packages)\n* [Cheat Sheet](#cheat_sheet)\n* [<img src=\"../../images/logos/web.png\" width=\"20\"/> Read More](#<img_src=\"../../images/logos/web.png\"_width=\"20\"/>_read_more)\n\n---", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"understanding_conda_environments\"></a>\n## Understanding Conda Environments", "_____no_output_____" ], [ "When you start developing a project from scratch, it’s recommended that you use the latest versions of the libraries you need. However, when working with someone else’s project, such as when running an example from [Kaggle](https://www.kaggle.com/) or [Github](https://github.com/), you may need to install specific versions of packages or even another version of Python due to compatibility issues.", "_____no_output_____" ], [ "This problem may also occur when you try to run an application you’ve developed long ago, which uses a particular library version that does not work with your application anymore due to updates.", "_____no_output_____" ], [ "Virtual environments are a solution to this kind of problem. By using them, it is possible to create multiple environments, each one with different versions of packages. A typical Python set up includes [Virtualenv](https://virtualenv.pypa.io/en/stable/#), a tool to create isolated Python virtual environments, widely used in the Python community.", "_____no_output_____" ], [ "Conda includes its own environment manager and presents some advantages over Virtualenv, especially concerning numerical applications, such as the ability to manage non-Python dependencies and the ability to manage different versions of Python, which is not possible with Virtualenv. Besides that, Conda environments are entirely compatible with default [Python packages](https://realpython.com/python-modules-packages/) that may be installed using pip.", "_____no_output_____" ], [ "Miniconda installation provides Conda and a root environment with a version of Python and some basic packages installed. Besides this root environment, it is possible to set up additional environments including different versions of Python and packages.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"conda_environments:\"></a>\nUsing the Anaconda prompt, it is possible to check the available Conda environments by running `conda env list`:\n\n```bash\n\n$ (base) ~ % conda env list\n\n# conda environments:\n#\nbase * /home/ali/anaconda3\n```", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"package_plan_##\"></a>\nThis base environment is the root environment, created by the Miniconda installer. It is possible to create another environment, named `otherenv`, by running `conda create --name otherenv`:\n\n\n```bash\n$ (base) ~ % conda create --name otherenv\nSolving environment: done\n\n## Package Plan ##\n\n environment location: C:\\Users\\IEUser\\Miniconda3\\envs\\otherenv\n\n\nProceed ([y]/n)? y\n\nPreparing transaction: done\nVerifying transaction: done\nExecuting transaction: done\n#\n# To activate this environment, use\n#\n# $ conda activate otherenv\n#\n# To deactivate an active environment, use\n#\n# $ conda deactivate\n```", "_____no_output_____" ], [ "As notified after the environment creation process is finished, it is possible to activate the otherenv environment by running `conda activate otherenv`. You’ll notice the environment has changed by the indication between parentheses in the beginning of the prompt:\n\n```bash\n$ (base) ~ % conda activate otherenv\n$ (otherenv) ~ %\n```", "_____no_output_____" ], [ "You can open the Python interpreter within this environment by running `python`:\n\n```bash\n$ (otherenv) ~ % python\n\nPython 3.7.0 (default, Jun 28 2018, 08:04:48) [MSC v.1912 64 bit (AMD64)] :: Anaconda, Inc. on win32\nType \"help\", \"copyright\", \"credits\" or \"license\" for more information.\n>>>\n```", "_____no_output_____" ], [ "The environment includes Python 3.7.0, the same version included in the root base environment. To exit the Python interpreter, just run `quit()`:\n\n```bash\n>>> quit()\n\n(otherenv) ~ %\n```", "_____no_output_____" ], [ "To deactivate the otherenv environment and go back to the root base environment, you should run `deactivate`:\n\n```bash\n(otherenv) ~ % conda deactivate\n\n(base) ~ %\n```", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"package_plan_##\"></a>\nAs mentioned earlier, Conda allows you to easily create environments with different versions of Python, which is not straightforward with Virtualenv. To include a different Python version within an environment, you have to specify it by using `python=<version>` when running conda create. For example, to create an environment named `py2` with `Python 2.7`, you have to run `conda create --name py2 python=2.7`:\n\n\n```bash\n(base) ~ % create --name py2 python=2.7\nSolving environment: done\n\n## Package Plan ##\n\n environment location: C:\\Users\\IEUser\\Miniconda3\\envs\\py2\n\n added / updated specs:\n - python=2.7\n\n\nThe following NEW packages will be INSTALLED:\n\n certifi: 2018.8.24-py27_1\n pip: 10.0.1-py27_0\n python: 2.7.15-he216670_0\n setuptools: 40.2.0-py27_0\n vc: 9-h7299396_1\n vs2008_runtime: 9.00.30729.1-hfaea7d5_1\n wheel: 0.31.1-py27_0\n wincertstore: 0.2-py27hf04cefb_0\n\nProceed ([y]/n)? y\n\nPreparing transaction: done\nVerifying transaction: done\nExecuting transaction: done\n#\n# To activate this environment, use\n#\n# $ conda activate py2\n#\n# To deactivate an active environment, use\n#\n# $ conda deactivate\n\n(base) /mnt/c/Users/username%\n```", "_____no_output_____" ], [ "As shown by the output of `conda create`, this time some new packages were installed, since the new environment uses Python 2. You can check the new environment indeed uses Python 2 by activating it and running the Python interpreter:\n\n```\n(base) ~ % conda activate py2\n```", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"conda_environments:\"></a>\nNow, if you run `conda env list`, you should see the two environments that were created, besides the root base environment:\n\n```bash\n(py2) ~ % conda env list\n# conda environments:\n#\nbase C:\\Users\\IEUser\\Miniconda3\notherenv C:\\Users\\IEUser\\Miniconda3\\envs\\otherenv\npy2 * C:\\Users\\IEUser\\Miniconda3\\envs\\py2\n\n\n(py2) ~ %\n```", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"package_plan_##\"></a>\nIn the list, the asterisk indicates the activated environment. It is possible to remove an environment by running `conda remove --name <environment name> --all`. Since it is not possible to remove an activated environment, you should first deactivate the `py2` environment, to remove it:\n\n```bash\n(py2) ~ % conda deactivate\n(base) ~ % conda remove --name py2 --all\n\nRemove all packages in environment C:\\Users\\IEUser\\Miniconda3\\envs\\py2:\n\n\n## Package Plan ##\n\n environment location: C:\\Users\\IEUser\\Miniconda3\\envs\\py2\n\n\nThe following packages will be REMOVED:\n\n certifi: 2018.8.24-py27_1\n pip: 10.0.1-py27_0\n python: 2.7.15-he216670_0\n setuptools: 40.2.0-py27_0\n vc: 9-h7299396_1\n vs2008_runtime: 9.00.30729.1-hfaea7d5_1\n wheel: 0.31.1-py27_0\n wincertstore: 0.2-py27hf04cefb_0\n\nProceed ([y]/n)? y\n\n\n(base) /mnt/c/Users/username%\n```", "_____no_output_____" ], [ "Now that you’ve covered the basics of managing environments with Conda, let’s see how to manage packages within the environments.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"understanding_basic_package_management_with_conda\"></a>\n## Understanding Basic Package Management With Conda\n\nWithin each environment, packages of software can be installed using the Conda package manager. The root base environment created by the Miniconda installer includes some packages by default that are not part of Python standard library.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"packages_in_environment_at_c:\\users\\ieuser\\miniconda3:\"></a>\nThe default installation includes the minimum packages necessary to use Conda. To check the list of installed packages in an environment, you just have to make sure it is activated and run `conda list`. In the root environment, the following packages are installed by default:\n\n```bash\n(base) ~ % conda list\n# packages in environment at C:\\Users\\IEUser\\Miniconda3:\n#\n# Name Version Build Channel\nasn1crypto 0.24.0 py37_0\nca-certificates 2018.03.07 0\ncertifi 2018.8.24 py37_1\ncffi 1.11.5 py37h74b6da3_1\nchardet 3.0.4 py37_1\nconda 4.5.11 py37_0\nconda-env 2.6.0 1\nconsole_shortcut 0.1.1 3\ncryptography 2.3.1 py37h74b6da3_0\nidna 2.7 py37_0\nmenuinst 1.4.14 py37hfa6e2cd_0\nopenssl 1.0.2p hfa6e2cd_0\npip 10.0.1 py37_0\npycosat 0.6.3 py37hfa6e2cd_0\npycparser 2.18 py37_1\npyopenssl 18.0.0 py37_0\npysocks 1.6.8 py37_0\npython 3.7.0 hea74fb7_0\npywin32 223 py37hfa6e2cd_1\nrequests 2.19.1 py37_0\nruamel_yaml 0.15.46 py37hfa6e2cd_0\nsetuptools 40.2.0 py37_0\nsix 1.11.0 py37_1\nurllib3 1.23 py37_0\nvc 14 h0510ff6_3\nvs2015_runtime 14.0.25123 3\nwheel 0.31.1 py37_0\nwin_inet_pton 1.0.1 py37_1\nwincertstore 0.2 py37_0\nyaml 0.1.7 hc54c509_2\n```", "_____no_output_____" ], [ "To manage the packages, you should also use Conda. Next, let’s see how to search, install, update, and remove packages using Conda.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"searching_and_installing_packages\"></a>\n### Searching and Installing Packages\n\nPackages are installed from repositories called **channels** by Conda, and some default channels are configured by the installer. To search for a specific package, you can run `conda search <package name>`. For example, this is how you search for the `keras` package (a machine learning library):\n\n```bash\n(base) ~ % conda search keras\nLoading channels: done\n# Name Version Build Channel\nkeras 2.0.8 py35h15001cb_0 pkgs/main\nkeras 2.0.8 py36h65e7a35_0 pkgs/main\nkeras 2.1.2 py35_0 pkgs/main\nkeras 2.1.2 py36_0 pkgs/main\nkeras 2.1.3 py35_0 pkgs/main\nkeras 2.1.3 py36_0 pkgs/main\n\n... (more)\n```", "_____no_output_____" ], [ "According to the previous output, there are different versions of the package and different builds for each version, such as for Python 3.5 and 3.6.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"name_version_build_channel\"></a>\nThe previous search shows only exact matches for packages named `keras`. To perform a broader search, including all packages containing `keras` in their names, you should use the wildcard `*`. For example, when you run conda search `*keras*`, you get the following:\n\n```bash\n(base) ~ % conda search \"*keras*\"\nLoading channels: done\n# Name Version Build Channel\nkeras 2.0.8 py35h15001cb_0 pkgs/main\nkeras 2.0.8 py36h65e7a35_0 pkgs/main\nkeras 2.1.2 py35_0 pkgs/main\nkeras 2.1.2 py36_0 pkgs/main\nkeras 2.1.3 py35_0 pkgs/main\nkeras 2.1.3 py36_0 pkgs/main\n\n... (more)\n\nkeras-applications 1.0.2 py35_0 pkgs/main\nkeras-applications 1.0.2 py36_0 pkgs/main\nkeras-applications 1.0.4 py35_0 pkgs/main\n\n... (more)\n\nkeras-base 2.2.0 py35_0 pkgs/main\nkeras-base 2.2.0 py36_0 pkgs/main\n\n... (more)\n```", "_____no_output_____" ], [ "As the previous output shows, there are some other keras related packages in the default channels.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"package_plan_##\"></a>\nTo install a package, you should run `conda install <package name>`. By default, the newest version of the package will be installed in the active environment. So, let’s install the package `keras` in the environment `otherenv` that you’ve already created:\n\n```bash\n(base) ~ % conda activate otherenv\n\n(otherenv) ~ % conda install keras\nSolving environment: done\n\n## Package Plan ##\n\n environment location: C:\\Users\\IEUser\\Miniconda3\\envs\\otherenv\n\n added / updated specs:\n - keras\n\n\nThe following NEW packages will be INSTALLED:\n\n _tflow_1100_select: 0.0.3-mkl\n absl-py: 0.4.1-py36_0\n astor: 0.7.1-py36_0\n blas: 1.0-mkl\n certifi: 2018.8.24-py36_1\n gast: 0.2.0-py36_0\n grpcio: 1.12.1-py36h1a1b453_0\n h5py: 2.8.0-py36h3bdd7fb_2\n hdf5: 1.10.2-hac2f561_1\n icc_rt: 2017.0.4-h97af966_0\n intel-openmp: 2018.0.3-0\n keras: 2.2.2-0\n keras-applications: 1.0.4-py36_1\n keras-base: 2.2.2-py36_0\n keras-preprocessing: 1.0.2-py36_1\n libmklml: 2018.0.3-1\n libprotobuf: 3.6.0-h1a1b453_0\n markdown: 2.6.11-py36_0\n mkl: 2019.0-117\n mkl_fft: 1.0.4-py36h1e22a9b_1\n mkl_random: 1.0.1-py36h77b88f5_1\n numpy: 1.15.1-py36ha559c80_0\n numpy-base: 1.15.1-py36h8128ebf_0\n pip: 10.0.1-py36_0\n protobuf: 3.6.0-py36he025d50_0\n python: 3.6.6-hea74fb7_0\n pyyaml: 3.13-py36hfa6e2cd_0\n scipy: 1.1.0-py36h4f6bf74_1\n setuptools: 40.2.0-py36_0\n six: 1.11.0-py36_1\n tensorboard: 1.10.0-py36he025d50_0\n tensorflow: 1.10.0-mkl_py36hb361250_0\n tensorflow-base: 1.10.0-mkl_py36h81393da_0\n termcolor: 1.1.0-py36_1\n vc: 14-h0510ff6_3\n vs2013_runtime: 12.0.21005-1\n vs2015_runtime: 14.0.25123-3\n werkzeug: 0.14.1-py36_0\n wheel: 0.31.1-py36_0\n wincertstore: 0.2-py36h7fe50ca_0\n yaml: 0.1.7-hc54c509_2\n zlib: 1.2.11-h8395fce_2\n\nProceed ([y]/n)?\n```", "_____no_output_____" ], [ "Conda manages the necessary dependencies for a package when it is installed. Since the package keras has a lot of dependencies, when you install it, Conda manages to install this big list of packages.", "_____no_output_____" ], [ "> **Note:** The paragraph below may not happen when you run it as newer versions of `keras` may be available that use python 3.7.\n\nIt’s worth noticing that, since the keras package’s newest build uses Python 3.6 and the otherenv environment was created using Python 3.7, the package python version 3.6.6 was included as a dependency. After confirming the installation, you can check that the Python version for the otherenv environment is downgraded to the 3.6.6 version.", "_____no_output_____" ], [ "Sometimes, you don’t want packages to be downgraded, and it would be better to just create a new environment with the necessary version of Python. To check the list of new packages, updates, and downgrades necessary for a package without installing it, you should use the parameter `--dry-run`. For example, to check the packages that will be changed by the installation of the package keras, you should run the following:\n\n```\n(base) ~ % conda install keras --dry-run\n```", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"package_plan_##\"></a>\nHowever, if necessary, it is possible to change the default Python of a Conda environment by installing a specific version of the package python. To demonstrate that, let’s create a new environment called envpython:\n\n```bash\n(otherenv) ~ % conda create --name envpython\nSolving environment: done\n\n## Package Plan ##\n\n environment location: C:\\Users\\IEUser\\Miniconda3\\envs\\envpython\n\n\nProceed ([y]/n)? y\n\nPreparing transaction: done\nVerifying transaction: done\nExecuting transaction: done\n#\n# To activate this environment, use\n#\n# $ conda activate envpython\n#\n# To deactivate an active environment, use\n#\n# $ conda deactivate\n```", "_____no_output_____" ], [ "As you saw before, since the root base environment uses Python 3.7, envpython is created including this same version of Python:\n\n```bash\n(base) ~ % conda activate envpython\n(envpython) ~ % python\nPython 3.7.0 (default, Jun 28 2018, 08:04:48) [MSC v.1912 64 bit (AMD64)] :: Anaconda, Inc. on win32\nType \"help\", \"copyright\", \"credits\" or \"license\" for more information.\n>>> quit()\n```", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"package_plan_##\"></a>\nTo install a specific version of a package, you can run `conda install <package name>=<version>`. For example, this is how you install Python 3.6 in the envpython environment:\n\n```bash\n(envpython) ~ % conda install python=3.6\nSolving environment: done\n\n## Package Plan ##\n\n environment location: C:\\Users\\IEUser\\Miniconda3\\envs\\envpython\n\n added / updated specs:\n - python=3.6\n\n\nThe following NEW packages will be INSTALLED:\n\n certifi: 2018.8.24-py36_1\n pip: 10.0.1-py36_0\n python: 3.6.6-hea74fb7_0\n setuptools: 40.2.0-py36_0\n vc: 14-h0510ff6_3\n vs2015_runtime: 14.0.25123-3\n wheel: 0.31.1-py36_0\n wincertstore: 0.2-py36h7fe50ca_0\n\nProceed ([y]/n)?\n```", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"package_plan_##\"></a>\nIn case you need to install more than one package in an environment, it is possible to run conda install only once, passing the names of the packages. To illustrate that, let’s install `numpy`, `scipy`, and `matplotlib`, basic packages for numerical computation:\n\n```bash\n(envpython) ~ % conda install numpy scipy matplotlib\n\nSolving environment: done\n\n## Package Plan ##\n\n environment location: C:\\Users\\IEUser\\Miniconda3\n\n added / updated specs:\n - matplotlib\n - numpy\n - scipy\n\n\nThe following packages will be downloaded:\n\n package | build\n ---------------------------|-----------------\n libpng-1.6.34 | h79bbb47_0 1.3 MB\n mkl_random-1.0.1 | py37h77b88f5_1 267 KB\n intel-openmp-2019.0 | 117 1.7 MB\n qt-5.9.6 | vc14h62aca36_0 92.5 MB\n matplotlib-2.2.3 | py37hd159220_0 6.5 MB\n tornado-5.1 | py37hfa6e2cd_0 668 KB\n pyqt-5.9.2 | py37ha878b3d_0 4.6 MB\n pytz-2018.5 | py37_0 232 KB\n scipy-1.1.0 | py37h4f6bf74_1 13.5 MB\n jpeg-9b | hb83a4c4_2 313 KB\n python-dateutil-2.7.3 | py37_0 260 KB\n numpy-base-1.15.1 | py37h8128ebf_0 3.9 MB\n numpy-1.15.1 | py37ha559c80_0 37 KB\n mkl_fft-1.0.4 | py37h1e22a9b_1 120 KB\n kiwisolver-1.0.1 | py37h6538335_0 61 KB\n pyparsing-2.2.0 | py37_1 96 KB\n cycler-0.10.0 | py37_0 13 KB\n freetype-2.9.1 | ha9979f8_1 470 KB\n icu-58.2 | ha66f8fd_1 21.9 MB\n sqlite-3.24.0 | h7602738_0 899 KB\n sip-4.19.12 | py37h6538335_0 283 KB\n ------------------------------------------------------------\n Total: 149.5 MB\n\nThe following NEW packages will be INSTALLED:\n\n blas: 1.0-mkl\n cycler: 0.10.0-py37_0\n freetype: 2.9.1-ha9979f8_1\n icc_rt: 2017.0.4-h97af966_0\n icu: 58.2-ha66f8fd_1\n intel-openmp: 2019.0-117\n jpeg: 9b-hb83a4c4_2\n kiwisolver: 1.0.1-py37h6538335_0\n libpng: 1.6.34-h79bbb47_0\n matplotlib: 2.2.3-py37hd159220_0\n mkl: 2019.0-117\n mkl_fft: 1.0.4-py37h1e22a9b_1\n mkl_random: 1.0.1-py37h77b88f5_1\n numpy: 1.15.1-py37ha559c80_0\n numpy-base: 1.15.1-py37h8128ebf_0\n pyparsing: 2.2.0-py37_1\n pyqt: 5.9.2-py37ha878b3d_0\n python-dateutil: 2.7.3-py37_0\n pytz: 2018.5-py37_0\n qt: 5.9.6-vc14h62aca36_0\n scipy: 1.1.0-py37h4f6bf74_1\n sip: 4.19.12-py37h6538335_0\n sqlite: 3.24.0-h7602738_0\n tornado: 5.1-py37hfa6e2cd_0\n zlib: 1.2.11-h8395fce_2\n\nProceed ([y]/n)?\n```", "_____no_output_____" ], [ "Now that you’ve covered how to search and install packages, let’s see how to update and remove them using Conda.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"updating_and_removing_packages\"></a>\n### Updating and Removing Packages\n\nSometimes, when new packages are released, you need to update them. To do so, you may run `conda update <package name>`. In case you wish to update all the packages within one environment, you should activate the environment and run `conda update --all`.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"package_plan_##\"></a>\nTo remove a package, you can run `conda remove <package name>`. For example, this is how you remove numpy from the root base environment:\n\n```bash\n(envpython) ~ % conda remove numpy\nSolving environment: done\n\n## Package Plan ##\n\n environment location: C:\\Users\\IEUser\\Miniconda3\n\n removed specs:\n - numpy\n\n\nThe following packages will be REMOVED:\n\n matplotlib: 2.2.3-py37hd159220_0\n mkl_fft: 1.0.4-py37h1e22a9b_1\n mkl_random: 1.0.1-py37h77b88f5_1\n numpy: 1.15.1-py37ha559c80_0\n scipy: 1.1.0-py37h4f6bf74_1\n\nProceed ([y]/n)?\n```", "_____no_output_____" ], [ "> **Note:** It’s worth noting that when you remove a package, all packages that depend on it are also removed.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"cheat_sheet\"></a>\n## Cheat Sheet\n\n[Click here to get access to a Conda cheat sheet](https://static.realpython.com/conda-cheatsheet.pdf) with handy usage examples for managing your Python environment and packages.", "_____no_output_____" ], [ "<a class=\"anchor\" id=\"read_more\"></a>\n## <img src=\"../../images/logos/web.png\" width=\"20\"/> Read More \n", "_____no_output_____" ], [ "Also, if you’d like a deeper understanding of Anaconda and Conda, check out the following links:\n\n- [Why you need Python environments and how to manage them with Conda](https://medium.freecodecamp.org/why-you-need-python-environments-and-how-to-manage-them-with-conda-85f155f4353c)\n- [Conda: Myths and Misconceptions](http://jakevdp.github.io/blog/2016/08/25/conda-myths-and-misconceptions/)", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Configuraciones para el Grupo de Estudio\n\n<img src=\"./img/f_mail.png\" style=\"width: 700px;\"/>\n\n## Contenidos\n- ¿Por qué jupyter notebooks?\n- Bash\n- ¿Que es un *kernel*?\n- Instalación\n- Deberes", "_____no_output_____" ], [ "## Python y proyecto Jupyter\n\n<img src=\"./img/py.jpg\" style=\"width: 500px;\"/>\n<img src=\"./img/jp.png\" style=\"width: 100px;\"/>\n\n- Necesitamos llevar un registro del avance de cada integrante. \n- Lenguaje de programación interpretado de alto nivel.\n- Jupyter notebooks: son fáciles de usar\n- `Necesitamos que todos tengan una versión de Python con jupyter lab`", "_____no_output_____" ], [ "## ¿Cómo funciona Jupyter?\n\n- Es un derivado del proyecto `iPython`, que ofrece una interfaz interactiva para programadores.\n- Tiene formato `.ipynb`\n- Es posible usar otros lenguajes de programación diferentes a Python.\n- Permite al usuario configurar cómo se visualiza su código mediante `Markdown`.\n- Ahora una demostración\n\n<img src=\"./img/jupex.png\" style=\"width: 500px;\"/>", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt\nimport numpy as np\nimport math\n\n# constantes\npi = math.pi; h = 6.626e-34; kB = 1.380e-23; c = 3.0e+8;\nTemps = [9940.00, 8500.00, 7500.00, 6627.00, 5810.93, 4231.15, 3000.00, 2973.15, 288.15]\nlabels = ['Sirius', 'White star', 'Yellow-white star', 'Polaris', 'Sol', 'HfC', 'Bombilla', 'TaN', 'Atmósfera ']\ncolors = ['r','g','#FF9633','c','m','#eeefff','y','b','k']\n\n# arreglo de frecuencias\nfreq = np.arange(0.25e14,3e15,0.25e14)\n\n# funcion spectral energy density (SED)\ndef SED(f, T):\n energyDensity = ( 8*pi*h*(np.power(f, 3.0))/(c**3) ) / (np.exp((h/kB)*f/T) - 1)\n return energyDensity\n\n# Calculo de SED para temperaturas\nfor i in range(len(Temps)):\n r = SED(freq,Temps[i])\n plt.plot(freq*1e-12,r,color=colors[i],label=labels[i])\n\nplt.legend(); plt.xlabel('frequency ( THz )'); plt.ylabel('SED_frequency ( J $m^{-3}$ $Hz^{-1}$ )')\nplt.xlim(0.25e2,2.5e3); plt.show()", "_____no_output_____" ] ], [ [ "### Permite escribir expresiones matemáticas complejas\n\nEs posible escribir código en $\\LaTeX$ si es necesario", "_____no_output_____" ], [ "\\begin{align}\n \\frac{\\partial u(\\lambda, T)}{\\partial \\lambda} &= \\frac{\\partial}{\\partial \\lambda} \\left( \\frac{C_{1}}{\\lambda^{5}}\\left(\\frac{1}{e^{C_{2}/T\\lambda} -1}\\right) \\right) \\\\\n 0 &= \\left(\\frac{-5}{e^{C_{2}/T\\lambda} -1}\\frac{1}{\\lambda^{6}}\\right) + \\left( \\frac{C_{2}e^{C_{2}/T\\lambda}}{T\\lambda^{7}} \\right)\\left(\\frac{1}{e^{C_{2}/T\\lambda} -1}\\right)^{2} \\\\\n 0 &= \\frac{-\\lambda T5}{C_{2}} + \\frac{e^{C_{2}/T\\lambda}}{e^{C_{2}/T\\lambda} -1} \\\\\n 0 &= -5 + \\left(\\frac{C_{2}}{\\lambda T}\\right) \\left(\\frac{e^{C_{2}/T\\lambda}}{e^{C_{2}/T\\lambda} -1}\\right)\n\\end{align}", "_____no_output_____" ], [ "## ¿Cómo es que usa un lenguaje diferente a Python?\n\n- Un kernel es una especie de `motor computacional` que ejecuta el código dentro de un archivo `.ipynb`. \n- Los kernels hay para varios lengajes de programación, como R, Bash, C++, julia.\n\n<img src=\"./img/ker.png\" style=\"width: 250px;\"/>\n\n## ¿Por qué Bash?\n\n- Bash es un lenguaje de scripting que se comunica con la shell e históricamente ha ayudado a científicos a llevarse mejor con la bioinformática.", "_____no_output_____" ], [ "## ¿Dónde encontramos las instrucciones para instalar Python?\n\n- Es posible hacerlo de varias maneras: `Anaconda` y el `intérprete oficial` desde https://www.python.org/downloads/\n- Usaremos el intérprete de `Anaconda`: es más fácil la instalación si no te acostumbras a usar la línea de comandos.\n- Si ustedes ya están familiarizados con Python y no desean instalar el intérprete de `Anaconda` pueden usar `pip` desde https://pypi.org/project/bash_kernel/\n\n<img src=\"./img/qrgit.png\" style=\"width: 250px;\"/>", "_____no_output_____" ], [ "## Deberes\n- Creamos una carpeta en `google Drive` donde harán subirán los archivos `.ipynb` y una conversión a HTML, u otro tipo de archivo dependiendo de la sesión.\n- Vamos a tener un quiz cada semana, que les enviaremos por el servidor de Discord del grupo de estudio.\n- El deber para la siguiente semana: \n 1. Instalar Ubuntu si aún no lo poseen usando cualquiera de las alternativas presentadas.\n 2. Instalar Anaconda, jupyter lab y el kernel de bash. \n\nSe deben enviar un documento word o pdf con capturas de pantalla que compruebe esto.\nSi tienen algún problema, usen por favor los foros de `Discord` y nos ayudamos entre todos.\n\n<img src=\"./img/deberes.png\" style=\"width: 500px;\"/>", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Generación de observaciones aleatorias a partir de una distribución de probabilidad", "_____no_output_____" ], [ "La primera etapa de la simulación es la **generación de números aleatorios**. Los números aleatorios sirven como el bloque de construcción de la simulación. La segunda etapa de la simulación es la **generación de variables aleatorias basadas en números aleatorios**. Esto incluye generar variables aleatorias <font color ='red'> discretas y continuas de distribuciones conocidas </font>. En esta clase, estudiaremos técnicas para generar variables aleatorias.\n\nIntentaremos dar respuesta a el siguiente interrogante:\n>Dada una secuencia de números aleatorios, ¿cómo se puede generar una secuencia de observaciones aleatorias a partir de una distribución de probabilidad dada? Varios enfoques diferentes están disponibles, dependiendo de la naturaleza de la distribución", "_____no_output_____" ], [ "Considerando la generación de números alestorios estudiados previamente, asumiremos que tenemos disponble una secuencia $U_1,U_2,\\cdots$ variables aleatorias independientes, para las cuales se satisface que:\n$$\nP(U_i\\leq u) = \\begin{cases}0,& u<0\\\\ u,&0\\leq u \\leq 1\\\\ 1,& u>1 \\end{cases}\n$$\nes decir, cada variable se distribuye uniformemente entre 0 y 1.\n\n**Recordar:** En clases pasadas, observamos como transformar un número p-seudoaletorio distribuido uniformemte entre 0 y 1, en una distribución normalmente distribuida con media $(\\mu,\\sigma^2)\\longrightarrow$ <font color='red'> [Médoto de Box Muller](http://www.lmpt.univ-tours.fr/~nicolis/Licence_NEW/08-09/boxmuller.pdf) </font> como un caso particular.\n\nEn esta sesión, se presentarán dos de los técnicas más ampliamente utilizados para generar variables aletorias, a partir de una distribución de probabilidad.", "_____no_output_____" ], [ "## 1. Método de la transformada inversa", "_____no_output_____" ], [ "Este método puede ser usado en ocasiones para generar una observación aleatoria. Tomando $X$ como la variable aletoria involucrada, denotaremos la función de distribución de probabilidad acumulada por\n$$F(x)=P(X\\leq x),\\quad \\forall x$$\n<font color ='blue'> Dibujar graficamente esta situación en el tablero</font>\n\nEl método de la transformada inversa establece\n$$X = F^{-1}(U),\\quad U \\sim \\text{Uniforme[0,1]}$$\ndonde $F^{-1}$ es la transformada inversa de $F$.\n\nRecordar que $F^{-1}$ está bien definida si $F$ es estrictamente creciente, de otro modo necesitamos una regla para solucionar los casos donde esta situación no se satisface. Por ejemplo, podríamos tomar\n$$F^{-1}(u)=\\inf\\{x:F(x)\\geq u\\}$$ \nSi hay muchos valores de $x$ para los cuales $F(x)=u$, esta regla escoje el valor mas pequeño. Observar esta situación en el siguiente ejemplo:\n\n\nObserve que en el intervalo $(a,b]$ si $X$ tiene distribución $F$, entonces\n$$P(a<X\\leq b)=F(b)-F(a)=0\\longrightarrow \\text{secciones planas}$$\n\nPor lo tanto si $F$ tienen una densidad continua, entonces $F$ es estrictamente creciente y su inversa está bien definida. \n", "_____no_output_____" ], [ "Ahora observemos cuando se tienen las siguientes funciones:\n\nObservemos que sucede en $x_0$\n$$\\lim_{x \\to x_0^-} F(x)\\equiv F(x^-)<F(x^+)\\equiv \\lim_{x\\to x_0^+}F(x)$$\nBajo esta distribución el resultado $x_0$ tiene probabilidad $F(x^+)-F(x^-)$. Por otro lado todos los valores de $u$ entre $[u_2,u_1]$ serán mapeados a $x_0$.\n\nLos siguientes ejemplos mostrarán una implementación directa de este método.", "_____no_output_____" ], [ "### Ejemplo 1: Distribución exponencial\nLa distribución exponencial con media $\\theta$ tiene distribución \n$$F(x)=1-e^{-x/\\theta}, \\quad x\\geq 0$$\n> Distrubución exponencial python: https://en.wikipedia.org/wiki/Exponential_distribution", "_____no_output_____" ], [ ">### <font color= blue> Mostrar en el tablero la demostración ", "_____no_output_____" ] ], [ [ "# Importamos las librerías principales\nimport numpy as np\nimport matplotlib.pyplot as plt", "_____no_output_____" ], [ "# Creamos la función que crea muestras distribuidas exponencialmente\ndef D_exponential(theta,N):\n return -np.log(np.random.random(N))*theta", "_____no_output_____" ], [ "theta = 4 # Media\nN = 10**6 # Número de muestras\n# creamos muestras exponenciales con la función que esta en numpy\nx = np.random.exponential(theta,N) \n# creamos muestras exponenciales con la función creada\nx2 = D_exponential(theta,N)\n# Graficamos el historial para x\nplt.hist(x,100,density=True)\nplt.xlabel('valores aleatorios')\nplt.ylabel('probabilidad')\nplt.title('histograma función de numpy')\nprint(np.mean(x))\nplt.show()", "4.003380413190708\n" ], [ "plt.hist(x2,100,density=True)\nplt.xlabel('valores aleatorios')\nplt.ylabel('probabilidad')\nplt.title('histograma función creada')\nprint(np.mean(x2))\nplt.show()", "4.001007346372537\n" ] ], [ [ "### Ejemplo 2\nSe sabe que la distribución Erlang resulta de la suma de $k$ variables distribuidas exponencialmente cada una con media $\\theta$, y por lo tanto esta variable resultante tiene distribución Erlang de tamaño $k$ y media $theta$.\n\n> Enlace distribución Erlang: https://en.wikipedia.org/wiki/Erlang_distribution", "_____no_output_____" ] ], [ [ "N = 10**4\n# Variables exponenciales\nx1 = np.random.exponential(4,N)\nx2 = np.random.exponential(4,N)\nx3 = np.random.exponential(4,N)\nx4 = np.random.exponential(4,N)\nx5 = np.random.exponential(4,N)\n\n# Variables erlang\ne0 = x1\ne1 = (x1+x2)\ne2 = (x3+x4+x5)\ne3 = (x1+x2+x3+x4)\ne4 = x1+x2+x3+x4+x5\nplt.hist(e0,100,density=True,label='1 exponencial')\nplt.hist(e1,100,density=True,label='suma de 2 exp')\nplt.hist(e2,100,density=True,label='suma de 3 exp')\nplt.hist(e3,100,density=True,label='suma de 4 exp')\nplt.hist(e4,100,density=True,label='suma de 5 exp')\nplt.legend()\n\nplt.show()", "_____no_output_____" ], [ "# Función para crear variables aleatorias Erlang\ndef D_erlang(theta:'media distrubucion',k,N):\n f = np.random.rand(N,k) # Matriz de variables aleatorias de dim N*k mejora la velocidad del algoritmo\n y =list(map(lambda i:-(theta)*np.log(np.prod(f[i,:])),range(N)))\n return y", "_____no_output_____" ], [ "# Prueba de la función creada\nN = 10**4\nks = [1,2,3,4,5]\ntheta = 4\ny = list(map(lambda k:D_erlang(theta,k,N),ks))\n\n[plt.hist(y[i],bins=100,density=True,label='suma de %i exp'%(i+1)) for i in range(len(y))]\nplt.legend()\nplt.show()", "_____no_output_____" ] ], [ [ "### Función de densidad variables Erlang\n\n$$p(x)=x^{k-1}\\frac{e^{-x/\\theta}}{\\theta^k\\Gamma(k)}\\equiv x^{k-1}\\frac{e^{-x/\\theta}}{\\theta^k(k-1)!}$$", "_____no_output_____" ] ], [ [ "#Librería que tiene la función gamma y factorial \n# Para mostrar la equivalencia entre el factorial y la función gamma\nimport scipy.special as sps \nfrom math import factorial as fac\nk = 4\ntheta = 4\n\nx = np.arange(0,60,0.01)\nplt.show() \ny= x**(k-1)*(np.exp(-x/theta) /(sps.gamma(k)*theta**k))\ny2 = x**(k-1)*(np.exp(-x/theta) /(fac(k-1)*theta**k))\nplt.plot(x,y,'r')\nplt.plot(x,y2,'b--')\n# plt.show()\n\n# Creo variables aleatorias erlang y obtengo su histograma en la misma gráfica anterior\nN = 10**4\nr1 = D_erlang(theta,k,N)\nplt.hist(r1,bins=50,density=True)\nplt.show()", "_____no_output_____" ] ], [ [ "Para mejorar la eficiencia, creemos una función que grafique la misma gráfica anterior pero este caso que le podamos variar los parámetros `k` y $\\theta$ de la distribución", "_____no_output_____" ] ], [ [ "# Función que grafica subplots para cada señal de distribución Erlang\ndef histograma_erlang(signal:'señal que desea graficar',\n k:'Parámetro de la función Erlang'):\n\n plt.figure(figsize=(8,3))\n count, x, _ = plt.hist(signal,100,density=True,label='k=%d'%k)\n y = x**(k-1)*(np.exp(-x/theta) /(sps.gamma(k)*theta**k))\n plt.plot(x, y, linewidth=2,color='k')\n plt.ylabel('Probabilidad')\n plt.xlabel('Muestras')\n plt.legend()\n plt.show()", "_____no_output_____" ] ], [ [ "Con la función anterior, graficar la función de distribución de una Erlang con parámetros $\\theta = 4$ y `Ks = [1,8,3,6] `", "_____no_output_____" ] ], [ [ "theta = 4 # media \nN = 10**5 # Número de muestras\nKs = [1,8,3,6] # Diferentes valores de k para la distribución Erlang\n# Obtengo\nY = list(map(lambda k:D_erlang(theta,k,N),Ks))\nlist(map(histograma_erlang,Y,Ks));", "_____no_output_____" ] ], [ [ "### Ejemplo 4\nDistribución de Rayleigh\n$$F(x)=1-e^{-2x(x-b)},\\quad x\\geq b $$", "_____no_output_____" ], [ "> Fuente: https://en.wikipedia.org/wiki/Rayleigh_distribution", "_____no_output_____" ] ], [ [ "# Función del ejemplo 4\ndef D_rayleigh(b,N):\n return (b/2)+np.sqrt(b**2-2*np.log(np.random.rand(N)))/2\n\nnp.random.rayleigh?\n\n# Función de Raylegh que contiene numpy\ndef D_rayleigh2(sigma,N):\n return np.sqrt(-2*sigma**2*np.log(np.random.rand(N)))", "_____no_output_____" ], [ "b = 0.5; N =10**6;sigma = 2\nr = D_rayleigh(b,N) # Función del ejemplo \nr2 = np.random.rayleigh(sigma,N) # Función que contiene python\nr3 = D_rayleigh2(sigma,N) # Función creada de acuerdo a la función de python\n\nplt.figure(1,figsize=(10,8))\nplt.subplot(311)\nplt.hist(r3,100,density=True)\nplt.xlabel('valores aleatorios')\nplt.ylabel('probabilidad')\nplt.title('histograma función D_rayleigh2')\n\nplt.subplot(312)\nplt.hist(r2,100,density=True)\nplt.xlabel('valores aleatorios')\nplt.ylabel('probabilidad')\nplt.title('histograma función numpy')\n\nplt.subplot(313)\nplt.hist(r,100,density=True)\nplt.xlabel('valores aleatorios')\nplt.ylabel('probabilidad')\nplt.title('histograma función D_rayleigh')\nplt.subplots_adjust(top=0.92, bottom=0.08, left=0.10, right=0.95,\n hspace=.5,wspace=0)\nplt.show()", "_____no_output_____" ] ], [ [ "## Distribuciones discretas\n\nPara una variable dicreta, evaluar $F^{-1}$ se reduce a buscar en una tabla. Considere por ejemplo una variable aleatoria discreta, cuyos posibles valores son $c_1<c_2<\\cdots<c_n$. Tome $p_i$ la probabilidad alcanzada por $c_i$, $i=1,\\cdots,n$ y tome $q_0=0$, en donde $q_i$ representa las **probabilidades acumuladas asociadas con $c_i$** y está definido como:\n$$q_i=\\sum_{j=1}^{i}p_j,\\quad i=1,\\cdots,n \\longrightarrow q_i=F(c_i)$$\nEntonces, para tomar muestras de esta distribución se deben de realizar los siguientes pasos:\n 1. Generar un número uniforme $U$ entre (0,1).\n 2. Encontrar $k\\in\\{1,\\cdots,n\\}$ tal que $q_{k-1}<U\\leq q_k$\n 3. Tomar $X=c_k$.", "_____no_output_____" ], [ "### Ejemplo numérico", "_____no_output_____" ] ], [ [ "# Librería para crear tablas\nimport pandas as pd", "_____no_output_____" ], [ "val = [1,2,3,4,5]\np_ocur = [.1,.2,.4,.2,.1]\np_acum = np.cumsum(p_ocur)\n\ndf = pd.DataFrame(index=val,columns=['Probabilidades','Probabilidad acumulada'], dtype='float')\ndf.index.name = \"Valores (índices)\"\ndf.loc[val,'Probabilidades'] = p_ocur\ndf.loc[val,'Probabilidad acumulada'] = p_acum\ndf", "_____no_output_____" ], [ "u = .5\nprint(sum(1 for i in p_acum if i<u) + 1)", "3\n" ], [ "def Gen_distr_discreta(U:'vector de números aleatorios',\n p_acum: 'P.Acumulada de la distribución a generar'):\n '''Tener en cuenta que este arreglo cuenta números empezando del 0'''\n v = np.array(list(map(lambda j:sum(1 for i in p_acum if i<U[j]),range(len(U)))))\n return v", "_____no_output_____" ] ], [ [ "# Lo que no se debe de hacer, cuando queremos graficar el histograma de una distribución discreta", "_____no_output_____" ] ], [ [ "N = 10**4\nu =np.random.rand(N)\nv = Gen_distr_discreta(u,p_acum)+1\nplt.hist(v,bins = 6)\nplt.show()", "_____no_output_____" ], [ "N = 10**4\nu =np.random.rand(N)\nv = Gen_distr_discreta(u,p_acum)+1 #+1 porque los índices comienzan en 1 \n# print(u,v)\n\n# Método 1 (Correcto)\nhist,bins = np.histogram(v,bins=len(val))\n# print(hist,bins)\nplt.bar(val,hist)\nplt.title('METODO CORRECTO')\nplt.xlabel('valores (índices)')\nplt.ylabel('frecuencias')\nplt.show()\n \n# Método 2 (incorrecto)\ny,x,_ = plt.hist(v,bins=len(val))\nplt.title('METODO INCORRECTO')\nplt.xlabel('valores (índices)')\nplt.ylabel('frecuencias')\nplt.legend(['incorrecto'])\nplt.show()\n", "_____no_output_____" ], [ "def plot_histogram_discrete(distribucion:'distribución a graficar histograma',\n label:'label del legend'):\n # len(set(distribucion)) cuenta la cantidad de elementos distintos de la variable 'distribucion'\n plt.figure(figsize=[8,4])\n y,x = np.histogram(distribucion,bins = len(set(distribucion))) \n plt.bar(list(set(distribucion)),y,label=label)\n plt.legend()\n plt.show()", "_____no_output_____" ] ], [ [ ">### <font color ='red'> **Tarea 4** \n> 1. Generación variable aleatoria continua\n>El tiempo en el cual un movimiento browniano se mantiene sobre su punto máximo en el intervalo [0,1] tiene una distribución\n>$$F(x)=\\frac{2}{\\pi}\\sin^{-1}(\\sqrt x),\\quad 0\\leq x\\leq 1$$ </font>\n\n> 2. Generación variable aleatoria Discreta\n> La distribución binomial modela el número de éxitos de n ensayos independientes donde hay una probabilidad p de éxito en cada ensayo.\n> Generar una variable aletoria binomial con parámetros $n=10$ y $p=0.7$. Recordar que $$X\\sim binomial(n,p) \\longrightarrow p_i=P(X=i)=\\frac{n!}{i!(n-i)!}p^i(1-p)^{n-i},\\quad i=0,1,\\cdots,n$$\n> Por propiedades de la operación factorial la anterior $p_i$ se puede escribir como:\n> $$p_{i+1}=\\frac{n-i}{i+1}\\frac{p}{1-p} p_i $$\n\n> **Nota:** Por notación recuerde que para el caso continuo $f(x)$ es la distribución de probabilidad (PDF), mientras $F(x)$ corresponde a la distribución de probabilidad acumulada (CDF). Para el caso discreto, $P(X=i)$ corresponde a su distribución de probabilidad (PMF) y $ F_{X}(x)=\\operatorname {P} (X\\leq x)=\\sum _{x_{i}\\leq x}\\operatorname {P} (X=x_{i})=\\sum _{x_{i}\\leq x}p(x_{i})$, corresponde a su distribución de probabilidad acumulada (CDF).\n\nGenere muestres aleatorias que distribuyan según la función dada usando el método de la transformada inversa y grafique el histograma de 100 muestras generadas con el método y compárela con el función $f(x)$ dada, esto con el fín de validar que el procedimiento fue realizado de manera correcta", "_____no_output_____" ], [ "<script>\n $(document).ready(function(){\n $('div.prompt').hide();\n $('div.back-to-top').hide();\n $('nav#menubar').hide();\n $('.breadcrumb').hide();\n $('.hidden-print').hide();\n });\n</script>\n\n<footer id=\"attribution\" style=\"float:right; color:#808080; background:#fff;\">\nCreated with Jupyter by Oscar David Jaramillo Z.\n</footer>", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "print(\"Hello Github\")", "Hello Github\n" ], [ "", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# MHKiT Quality Control Module\nThe following example runs a simple quality control analysis on wave elevation data using the [MHKiT QC module](https://mhkit-software.github.io/MHKiT/mhkit-python/api.qc.html). The data file used in this example is stored in the [\\\\\\\\MHKiT\\\\\\\\examples\\\\\\\\data](https://github.com/MHKiT-Software/MHKiT-Python/tree/master/examples/data) directory.\n\nStart by importing the necessary Python packages and MHKiT modules.", "_____no_output_____" ] ], [ [ "import pandas as pd\nfrom mhkit import qc, utils", "_____no_output_____" ] ], [ [ "## Load Data", "_____no_output_____" ], [ "The wave elevation data used in this example includes several issues, including timestamps that are out of order, corrupt data with values of -999, data outside the expected range, and stagnant data. \n\nThe data is loaded into a pandas DataFrame using the pandas method `read_csv`. The first 5 rows of data are shown below, along with a plot.", "_____no_output_____" ] ], [ [ "# Load data from the csv file into a DataFrame\ndata = pd.read_csv('data/qc/wave_elevation_data.csv', index_col='Time') \n\n# Plot the data\ndata.plot(figsize=(15,5), ylim=(-60,60)) \n\n# Print the first 5 rows of data\nprint(data.head()) ", " probe1 probe2 probe3\nTime \n10.000 24.48 28.27 1.3\n10.002 34.48 40.27 -8.7\n10.004 30.48 38.27 -13.7\n10.006 12.48 24.27 -32.7\n10.008 13.48 22.27 -21.7\n" ] ], [ [ "The data is indexed by time in seconds. To use the quality control functions, the data must be indexed by datetime. The index can be converted to datetime using the following utility function.", "_____no_output_____" ] ], [ [ "# Convert the index to datetime\ndata.index = utils.index_to_datetime(data.index, origin='2019-05-20') \n\n# Print the first 5 rows of data\nprint(data.head())", " probe1 probe2 probe3\nTime \n2019-05-20 00:00:10.000 24.48 28.27 1.3\n2019-05-20 00:00:10.002 34.48 40.27 -8.7\n2019-05-20 00:00:10.004 30.48 38.27 -13.7\n2019-05-20 00:00:10.006 12.48 24.27 -32.7\n2019-05-20 00:00:10.008 13.48 22.27 -21.7\n" ] ], [ [ "## Quality control tests\nThe following quality control tests are used to identify timestamp issues, corrupt data, data outside the expected range, and stagnant data.\n\nEach quality control tests results in the following information:\n\n* Cleaned data, which is a DataFrame that has *NaN* in place of data that did not pass the quality control test\n* Boolean mask, which is a DataFrame with True/False that indicates if each data point passed the quality control test\n* Summary of the quality control test results, the summary includes the variable name (which is blank for timestamp issues), the start and end time of the test failure, and an error flag for each test failure", "_____no_output_____" ], [ "### Check timestamp\nQuality control analysis generally starts by checking the timestamp index of the data. \n\nThe following test checks to see if 1) the data contains duplicate timestamps, 2) timestamps are not monotonically increasing, and 3) timestamps occur at irregular intervals (an interval of 0.002s is expected for this data). \n\nIf duplicate timestamps are found, the resulting DataFrames (cleaned data and mask) keep the first occurrence. If timestamps are not monotonic, the timestamps in the resulting DataFrames are reordered.", "_____no_output_____" ] ], [ [ "# Define expected frequency of the data, in seconds\nfrequency = 0.002 \n\n# Run the timestamp quality control test\nresults = qc.check_timestamp(data, frequency) ", "_____no_output_____" ] ], [ [ "The cleaned data, boolean mask, and test results summary are shown below. The summary is transposed (using .T) so that it is easier to read.", "_____no_output_____" ] ], [ [ "# Plot cleaned data\nresults['cleaned_data'].plot(figsize=(15,5), ylim=(-60,60)) \n\n# Print the first 5 rows of the cleaned data\nprint(results['cleaned_data'].head()) ", " probe1 probe2 probe3\n2019-05-20 00:00:10.000 24.48 28.27 1.3\n2019-05-20 00:00:10.002 34.48 40.27 -8.7\n2019-05-20 00:00:10.004 30.48 38.27 -13.7\n2019-05-20 00:00:10.006 12.48 24.27 -32.7\n2019-05-20 00:00:10.008 13.48 22.27 -21.7\n" ], [ "# Print the first 5 rows of the mask\nprint(results['mask'].head()) ", " probe1 probe2 probe3\n2019-05-20 00:00:10.000 True True True\n2019-05-20 00:00:10.002 True True True\n2019-05-20 00:00:10.004 True True True\n2019-05-20 00:00:10.006 True True True\n2019-05-20 00:00:10.008 True True True\n" ], [ "# Print the test results summary\n# The summary is transposed (using .T) so that it is easier to read.\nprint(results['test_results'].T) ", " 0 1 \\\nVariable Name \nStart Time 2019-05-20 00:00:10.230000 2019-05-20 00:00:10.340000 \nEnd Time 2019-05-20 00:00:10.230000 2019-05-20 00:00:10.340000 \nTimesteps 1 1 \nError Flag Nonmonotonic timestamp Duplicate timestamp \n\n 2 \nVariable Name \nStart Time 2019-05-20 00:00:10.042000 \nEnd Time 2019-05-20 00:00:10.044000 \nTimesteps 2 \nError Flag Missing timestamp \n" ] ], [ [ "### Check for corrupt data\nIn the following quality control tests, the cleaned data from the previous test is used as input to the subsequent test. For each quality control test, a plot of the cleaned data is shown along with the test results summary.\n\nNote, that if you want to run a series of quality control tests before extracting the cumulative cleaned data, boolean mask, and summary, we recommend using Pecos directly with the object-oriented approach, see https://pecos.readthedocs.io/ for more details.\n\nThe quality control test below checks for corrupt data, indicated by a value of -999.", "_____no_output_____" ] ], [ [ "# Define corrupt values\ncorrupt_values = [-999] \n\n# Run the corrupt data quality control test\nresults = qc.check_corrupt(results['cleaned_data'], corrupt_values) \n\n# Plot cleaned data\nresults['cleaned_data'].plot(figsize=(15,5), ylim=(-60,60)) \n\n# Print test results summary\nprint(results['test_results'].T)", " 0 1\nVariable Name probe1 probe3\nStart Time 2019-05-20 00:00:10.110000 2019-05-20 00:00:10.834000\nEnd Time 2019-05-20 00:00:10.134000 2019-05-20 00:00:10.848000\nTimesteps 13 8\nError Flag Corrupt data Corrupt data\n" ] ], [ [ "### Check for data outside the expected range\nThe next quality control test checks for data that is greater than 50 or less than -50. Note that expected range tests can also be used to compare measured values to a model, or analyze the expected relationships between data columns.", "_____no_output_____" ] ], [ [ "# Define expected lower and upper bound ([lower bound, upper bound])\nexpected_bounds = [-50, 50] \n\n# Run expected range quality control test\nresults = qc.check_range(results['cleaned_data'], expected_bounds) \n\n# Plot cleaned data\nresults['cleaned_data'].plot(figsize=(15,5), ylim=(-60,60)) \n\n# Print test results summary\nprint(results['test_results'].T) ", " 0 1 \\\nVariable Name probe3 probe3 \nStart Time 2019-05-20 00:00:10.240000 2019-05-20 00:00:10.468000 \nEnd Time 2019-05-20 00:00:10.240000 2019-05-20 00:00:10.468000 \nTimesteps 1 1 \nError Flag Data < lower bound, -50 Data < lower bound, -50 \n\n 2 \nVariable Name probe3 \nStart Time 2019-05-20 00:00:10.716000 \nEnd Time 2019-05-20 00:00:10.716000 \nTimesteps 1 \nError Flag Data > upper bound, 50 \n" ] ], [ [ "### Check for stagnant data\nThe final quality control test checks for stagnant data by looking for data that changes by less than 0.001 within a 0.02 second moving window. ", "_____no_output_____" ] ], [ [ "# Define expected lower bound (no upper bound is specified in this example)\nexpected_bound = [0.001, None] \n\n# Define the moving window, in seconds\nwindow = 0.02 \n\n# Run the delta quality control test\nresults = qc.check_delta(results['cleaned_data'], expected_bound, window) \n\n# Plot cleaned data\nresults['cleaned_data'].plot(figsize=(15,5), ylim=(-60,60))\n\n# Print test results summary\nprint(results['test_results'].T) ", " 0\nVariable Name probe2\nStart Time 2019-05-20 00:00:10.400000\nEnd Time 2019-05-20 00:00:10.544000\nTimesteps 73\nError Flag Delta < lower bound, 0.001\n" ] ], [ [ "## Cleaned Data\nThe cleaned data can be used directly in MHKiT analysis, or the missing values can be replaced using various methods before analysis is run. \nData replacement strategies are generally defined on a case by case basis. Pandas includes methods to interpolate, replace, and fill missing values.", "_____no_output_____" ] ], [ [ "# Extract final cleaned data for MHKiT analysis\ncleaned_data = results['cleaned_data'] ", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import numpy as np\nimport pandas as pd\nimport scipy as sp\nfrom scipy.stats import mode\nfrom sklearn import linear_model\nimport matplotlib\nimport matplotlib.pyplot as plt\nfrom sklearn.decomposition import PCA\nfrom sklearn import preprocessing\nimport sklearn as sk\nimport sklearn.discriminant_analysis as da\nimport sklearn.neighbors as knn\nfrom IPython.display import Markdown, display\nfrom sklearn.cross_validation import train_test_split\nfrom sklearn.preprocessing import Imputer\nfrom sklearn.linear_model import LinearRegression\nfrom sklearn import cross_validation, ensemble, preprocessing, metrics\nfrom sklearn.externals import joblib\n\n%matplotlib inline", "C:\\Users\\cdrgv\\Anaconda3\\lib\\site-packages\\sklearn\\cross_validation.py:41: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n \"This module will be removed in 0.20.\", DeprecationWarning)\n" ], [ "df = pd.read_csv('listings_new_york_2018.csv')", "C:\\Users\\cdrgv\\Anaconda3\\lib\\site-packages\\IPython\\core\\interactiveshell.py:2785: DtypeWarning: Columns (43,87,88) have mixed types. Specify dtype option on import or set low_memory=False.\n interactivity=interactivity, compiler=compiler, result=result)\n" ] ], [ [ "## 全部的columns名稱", "_____no_output_____" ] ], [ [ "df.columns.values", "_____no_output_____" ] ], [ [ "## 把全部的分數相關的平均起來成新的column", "_____no_output_____" ] ], [ [ "col = df.loc[: , \"review_scores_accuracy\":\"review_scores_value\"]\ndf['review_scores_mean'] = col.mean(axis=1)\n", "_____no_output_____" ] ], [ [ "## 留下來的attributes", "_____no_output_____" ] ], [ [ "cols_to_keep = [\n 'latitude',\n 'longitude',\n 'property_type',\n 'room_type',\n 'accommodates',\n 'bathrooms',\n 'bedrooms',\n 'beds',\n 'price','review_scores_mean'\n]\ndf = df[cols_to_keep]", "_____no_output_____" ] ], [ [ "## 清理price的符號 只剩下數字和小數點", "_____no_output_____" ] ], [ [ "df['price'] = df['price'].replace('[^(0-9).]','', regex=True).replace('[(]','-', regex=True).astype(float)\n\ndisplay(df.head())\n\n", "_____no_output_____" ] ], [ [ "## 要預測的欄位 price換成log price", "_____no_output_____" ] ], [ [ "df['log_price'] = np.log(df['price'].values)\n", "_____no_output_____" ] ], [ [ "## drop price", "_____no_output_____" ] ], [ [ "data = df.drop('price', axis=1)\nlist(data.columns.values)\ndata.head()", "_____no_output_____" ] ], [ [ "## Missing Values \n找出各欄位有多少NAN", "_____no_output_____" ] ], [ [ "#Replace blanks with NaNs\ndata = data.replace('_', np.nan)\ndata = data.replace(' ', np.nan)\ndata = data.replace([np.inf, -np.inf], np.nan)\ncol_analysis = []\nfor column in data.columns:\n numNulls = len(data[column][data[column].isnull()])\n totalLength = len(data[column])\n dict1 = {'Name':column,'DataType':data[column].dtype, 'NumberOfNulls':numNulls, 'PercentageNulls':numNulls*100.0/totalLength}\n col_analysis.append(dict1)\n \ncol_anal_df = pd.DataFrame(col_analysis)[['Name', 'DataType','NumberOfNulls','PercentageNulls']].sort_values(by='PercentageNulls', ascending=False)\n\nuseful_cols = col_anal_df[col_anal_df.PercentageNulls < 50.0]\n\nprint('List of Predictors and their respective percentages of missing values')\ndisplay(useful_cols.head(28))\n\nfor cols in data.columns.values:\n if (np.any(useful_cols.Name.values == cols) == False):\n data.drop(cols, axis=1, inplace=True)\n \ndata.head(5)\n\n", "List of Predictors and their respective percentages of missing values\n" ] ], [ [ "## Impute Missing Values\n把空白的填補成整個欄位的平均值", "_____no_output_____" ] ], [ [ "#Use Mean for Real values Columns\nreal_value_cols = useful_cols[useful_cols.DataType == 'float64']\n\nimp = Imputer(missing_values='NaN', strategy='mean', axis=0)\n\ndata[real_value_cols.Name.values] = imp.fit_transform(data[real_value_cols.Name.values])\n#Use Highest frequency for categorical columns\ncategorical_value_cols = useful_cols[useful_cols.DataType == 'object'].Name.values\ndata[categorical_value_cols] = data[categorical_value_cols].apply(lambda x:x.fillna(x.value_counts().index[0]))\n\ndata.head()\ndata.dtypes", "_____no_output_____" ], [ "\ndata = data.dropna()\n", "_____no_output_____" ] ], [ [ "## log_price的直方圖 (可以看出試常態分佈)", "_____no_output_____" ] ], [ [ "data.log_price.hist()", "_____no_output_____" ] ], [ [ "## Convert Categorical Variables to dummy integer values here\n- We convert the categorical variables to numeric here such that we can run models that work only with numbers\n\n## One-Hot Encoding\n把nomial的欄位換成數字", "_____no_output_____" ] ], [ [ "\ndata_ohe = data.copy(deep= True)\n\n#Encode categorical variables\ndef encode_categorical(array):\n return preprocessing.LabelEncoder().fit_transform(array) \n\n\ncategorical_value_cols = useful_cols[useful_cols.DataType == 'object'].Name.values\n\n#print(categorical_value_cols)\n#Convert Categories to numbers here\n\n\ndata_ohe[categorical_value_cols] = data_ohe[categorical_value_cols].apply(encode_categorical)\n\n\n#data_ohe['property_type'] = data_ohe['property_type'].apply(encode_categorical)\n\n# Apply one hot endcoing\n# Leads to inferior performance and hence we disable for now\n#data_ohe = pd.get_dummies(data_ohe.ix[:,:-1], columns=categorical_value_cols)\n\nprint ('Final Dataset ready for modelling after filling in missing values, and encoding categorical variables')\ndata_ohe.head()\n", "Final Dataset ready for modelling after filling in missing values, and encoding categorical variables\n" ] ], [ [ "## Separate response from predictors", "_____no_output_____" ] ], [ [ "\nx = data_ohe.values[:, :-1]\ny = data_ohe.values[:, -1]\n\n#response = df_filtered[['log_price']]\n#predictors = df_filtered.drop(['log_price'], axis=1)", "_____no_output_____" ] ], [ [ "## Split into train/test", "_____no_output_____" ] ], [ [ "\nx_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.4, random_state=42)", "_____no_output_____" ] ], [ [ "## Simple Regression Model", "_____no_output_____" ] ], [ [ "#OLS regression\nclf = LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False)\nclf.fit(x_train, y_train)\npredicted = clf.predict(x_test)\n\nscore = sk.metrics.r2_score(y_test, predicted)\nprint('sklearn: R2 score for Linear Regression is: {}'.format(score))", "sklearn: R2 score for Linear Regression is: 0.5703453525474875\n" ], [ "from sklearn import cross_validation, ensemble, preprocessing, metrics\n# 建立 random forest 模型\nforest = ensemble.RandomForestClassifier(n_estimators = 100)\ny_train = np.array(y_train, dtype=int)\nforest_fit = forest.fit(x_train, y_train)\n\n# 預測\ntest_y_predicted = forest.predict(x_test)\ny_test = np.array(y_test, dtype=int)\n# 績效\nscore = sk.metrics.r2_score(y_test, test_y_predicted)\nprint('sklearn: R2 score for Random Forest is: {}'.format(score))", "sklearn: R2 score for Linear Regression is: 0.4082299522656714\n" ] ], [ [ "## 把model打包", "_____no_output_____" ] ], [ [ "joblib.dump(clf, 'predicted.pkl')\nestimator = joblib.load('predicted.pkl')\nestimator", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import random\nimport copy\nimport os\nimport time\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport warnings\n# warnings.filterwarnings(\"ignore\")\n%matplotlib inline", "_____no_output_____" ], [ "class Dataset:\n def __init__(self,X,y,proportion=0.8,shuffle=True, mini_batch=0):\n \"\"\"\n Dataset class provide tools to manage dataset\n\n :param X: ndarray, features, highly recommand ndarray\n :param y: ndarray, labels\n :param proportion: number between 0 and 1, the proportion of train dataset and test dataset\n :param shuffle: boolean,\n :param mini_batch mini batch size, 0 by default, in this case no mini batch size dataset will be generated\n \"\"\"\n self.X = X\n self.y = y\n self.trainset = None\n self.testset = None\n self.validationset = None\n self.proportion = proportion\n self.shuffle = shuffle\n self.mini_batch = mini_batch\n self.allset = np.concatenate((X,y),axis=1)\n self.minisets = []\n\n if self.shuffle:\n # automatic distribution\n self.distribute()\n\n # @classmethod\n # def imageset(cls, path, proportion = 0.8, shuffle = None):\n # pass\n\n def distribute(self):\n \"\"\"\n This function will automatically distribute train and test dataset\n call this function to reshuffle all the dataset and also generate new train and test set\n \"\"\"\n n = np.shape(self.X)[0]\n samples = np.concatenate((self.X,self.y),axis=1)\n random.shuffle(samples)\n # sample train and test dataset\n self.trainset = samples[0:round(n * self.proportion),:]\n self.testset = samples[round(n * self.proportion) + 1:, :]\n\n def getX(self):\n return self.X\n\n def gety(self):\n return self.y\n\n def getminibatch(self):\n return self.mini_batch\n\n def gettrainset(self):\n \"\"\"\n :return: return train dataset with respect of proportion\n \"\"\"\n return Dataset(self.trainset[:, 0:self.X.shape[1]], self.trainset[:, self.X.shape[1]:], mini_batch=self.mini_batch)\n\n def gettestset(self):\n \"\"\"\n :return: test dataset with respect of proportion\n \"\"\"\n return Dataset(self.testset[:, 0:self.X.shape[1]], self.testset[:, self.X.shape[1]:], mini_batch=self.mini_batch)\n\n def getminiset(self):\n \"\"\"\n get mini sets with mini batch size\n :return: Dataset array\n \"\"\"\n spilit_list = np.arange(self.mini_batch, self.allset.shape[0], self.mini_batch)\n minisets = np.split(self.allset, spilit_list)\n for i in range(len(minisets)):\n self.minisets.append(Dataset(minisets[i][:, 0:self.X.shape[1]], minisets[i][:, self.X.shape[1]:],shuffle =False, mini_batch=self.mini_batch))\n return self.minisets\n", "_____no_output_____" ], [ "class NN:\n import numpy as np\n def __init__(self,dataset):\n \"\"\"\n This class contains Activation function util class, Layer class for construct networks, it contains several extend classes like LinearLayer, Conv2D etc.\n\n Examples:\n\n layer_list = [NN.Layer('Linear',3,10,'sigmoid',BN=True), NN.Layer('Linear',10,100,'sigmoid',BN=True),\n NN.Layer('Linear',100,10,'sigmoid',BN=True),NN.Layer('Linear',10,3,'none') ]\n\n dataset = Dataset(X, y, mini_batch= 64)\n\n nn = NN(dataset)\n\n layer_list is a list has 4 layers all are Layer class. Note that here we don't use LinearLayer,\n to use LinearLayer, replace NN.Layer('Linear',3,10,'sigmoid',BN=True) as NN.LinearLayer(,3,10,'sigmoid',BN=True) or\n NN.LinearLayer(3,10,'sigmoid'), NN.BN()\n\n :param dataset: Dataset class\n \"\"\"\n # self.input = input\n self.dataset = dataset\n self.layer_list = []\n\n def addlayers(self,layers):\n self.layer_list = layers\n\n def getlayers(self):\n return self.layer_list\n\n # activation functions\n class ActivationFunc:\n \"\"\"\n ActivationFunc is an util class with different types of activation function.\n it can\n \"\"\"\n @staticmethod\n def sigmoid(x):\n \"\"\"\n Sigmoid function\n \"\"\"\n return 1.0 / (1.0 + np.exp(-x))\n\n @staticmethod\n def ReLU(x):\n \"\"\"\n :param x: ndarray, \n :return:\n \"\"\"\n return np.maximum(0, x)\n\n @staticmethod\n def LeakyReLU(x):\n return np.where(x > 0, x, x * 0.01)\n\n @staticmethod\n def tanh(x):\n return np.tanh(x)\n\n @staticmethod\n def none(x):\n return x\n\n # Layer class\n class Layer:\n def __init__(self, type, input_dim, output_dim, activation, BN = False):\n \"\"\"\n Define a layer contains activation function or other normalization.\n\n :param type: Layer type, choose 'Linear', 'Conv' etc\n :param input_dim: input dim or previous layer's output\n :param output_dim: output dim of this layer\n :param activation: activation function, it now support \"sigmoid\", \"ReLU\", \"LeakyReLU\", \"tanh\" and \"none\" for no activation function\n :param BN, batch normalization , Default False\n\n Examples:\n\n A linear layer with input dim = 3 and output dim = 10, following batch normalization and a sigmoid activation function\n NN.Layer('Linear',3,10,'sigmoid',BN=True)\n\n \"\"\"\n self.type = type\n self.input_dim = input_dim\n self.output_dim = output_dim\n self.activation = activation\n self.BN = BN\n\n def getinputdim(self):\n return self.input_dim\n\n def getoutputdim(self):\n return self.output_dim\n\n def gettype(self):\n return self.type\n\n def getact(self, x):\n func_name = \"NN.ActivationFunc.\"+self.activation\n func = eval(func_name)\n return func(x)\n\n def getactname(self):\n return self.activation\n\n def getBN(self):\n return self.BN\n\n class LinearLayer(Layer):\n \"\"\"\n Define a linear layer\n\n As same as Layer except no need to clarify type\n \"\"\"\n def __init__(self, input_dim, output_dim):\n self.type = \"Linear\"\n self.input_dim = input_dim\n self.output_dim = output_dim\n\n\n class Conv2DLayer(Layer):\n \"\"\"\n Define a 2D convolutional layer_\n \"\"\"\n def __init__(self, input_size, kernel_size, stride, padding):\n \"\"\"\n initialize 2D conv layer\n\n :param input_size: Union[tuple, ndarray] layer's input size\n :param kernel_size: Union[tuple, ndarray] layer's kernel size\n :param stride: Int\n :param padding: Int\n \"\"\"\n self.type = \"Conv2D\"\n self.input_size = input_size\n self.kernel_size = kernel_size\n self.stride = stride\n self.padding = padding\n\n def getimagesize(self):\n return self.image_size\n\n def getkernelsize(self):\n return self.kernel_size\n\n def getstride(self):\n return self.stride\n\n def getpadding(self):\n return self.padding\n\n class BN(Layer):\n def __init__(self):\n \"\"\"\n Define a batch normalization layer\n \"\"\"\n self.type = \"BN\"\n self.activation =\"none\"\n", "_____no_output_____" ], [ "class Optimizer:\n def __init__(self,nn ,optimizer,loss_function, batch_size=8,epoch=20000,lr=0.0001,decay_rate=0):\n \"\"\"\n :param nn: input an NN class\n :param optimizer: optimizer as \"GD\", \"SGD\" etc\n :param batch_size: batch size for mini batch optimization\n :param epoch: epoch number\n :param lr: learning rate\n :param decay_rate: float, learning rate decay rate by default is 0\n \"\"\"\n\n self.nn = nn\n self.optimizer = optimizer\n self.loss_function = loss_function\n self.batch_size = batch_size\n self.epoch = epoch\n self.lr = lr\n self.weight_list = None\n self.gradient_list = None\n self.loss_list = None\n self.passer_list = None\n self.decay_rate = decay_rate\n\n def getgradientlist(self):\n return self.gradient_list\n\n def getlosslist(self):\n return self.loss_list\n\n def getweightlist(self):\n return self.weight_list\n\n class LossFunc:\n class Logarithmic:\n def __init__(self, y_true=None, y_pred=None, eps=1e-16):\n self.y_true = y_true\n self.y_pred = y_pred\n self.eps = eps\n \"\"\"\n Loss function we would like to optimize (minimize)\n We are using Logarithmic Loss\n http://scikit-learn.org/stable/modules/model_evaluation.html#log-loss\n \"\"\"\n def loss(self):\n self.y_pred = np.maximum(self.y_pred, self.eps)\n self.y_pred = np.minimum(self.y_pred, (1 - self.eps))\n return -(np.sum(self.y_true * np.log(self.y_pred)) + np.sum((1 - self.y_true) * np.log(1 - self.y_pred))) / len(self.y_true)\n\n class Quadratic:\n def __init__(self, y_true=None, y_pred=None, norm = 0):\n self.y_true = y_true\n self.y_pred = y_pred\n self.norm = norm\n\n def loss(self):\n return 1 / self.y_true.shape[0] * 0.5 * np.sum((self.y_pred - self.y_true) ** 2)\n\n def diff(self):\n return 2 * (self.y_pred - self.y_true)\n\n class MSE:\n def __init__(self, y_true=None, y_pred=None, x=None):\n self.y_true = y_true\n self.y_pred = y_pred\n self.x = x\n\n def loss(self):\n return 1 / np.shape(self.y_true)[0] * np.sum((self.y_pred - self.y_true) ** 2)\n\n def diff(self):\n return 2 / np.shape(self.y_true)[0] * np.sum(self.x @ (self.y_pred - self.y_true))\n\n\n class Node:\n def __init__(self, data: np.ndarray, type : str):\n \"\"\"\n Node class, is the node of binary tree which has two child node: left and right.\n It can also be presented as weight. Every passer during the back propagation is saved as\n a node class contains data, type, back and cache for calculation\n\n :param data: ndarray, value given during forward propagation\n :param type: str, the type of node, it can be \"weight\", \"data\" or calculation like \"@\", \"+\" etc\n :param back: ndarray, value updated during back propagation\n :param cache: array_like stock forward propagation's detail and middle value for the convenient of back propagation\n \"\"\"\n self.left = None\n self.right = None\n self.data = data\n self.type = type\n self.back = None\n self.cache = None\n self.momentum = None\n\n def getleft(self):\n return self.left\n\n def getright(self):\n return self.right\n\n def gettype(self):\n return self.type\n\n def getdata(self):\n return self.data\n\n def getback(self):\n return self.back\n\n def getmomentum(self):\n return self.momentum\n\n class WeightIni:\n \"\"\"\n Provide weight initial functions. util class\n \"\"\"\n @staticmethod\n def init_linear_weight(input_dim, output_dim):\n return np.random.uniform(-1, 1, (input_dim, output_dim))\n\n @staticmethod\n def init_BN_weight(dim):\n\n return np.ones((1, dim)), np.ones((1, dim), dtype=\"float32\")\n\n @staticmethod\n def init_conv2D_kernel(shape):\n \"\"\"\n :param shape: Union[tuple, int, float] shape of kernel\n :return:\n \"\"\"\n return np.random.random(shape)\n\n @staticmethod\n def initial_weight_list(layer_list):\n \"\"\"\n @Staticmethod. Given layer list and return respected initiall weight list\n\n :param layer_list: list, layer list\n :return: list, list of weight in Node class\n \"\"\"\n weight_list = []\n # initial weights in weight list by their type\n layer_num = len(layer_list)\n for i in range(layer_num):\n # linear weight operation\n if layer_list[i].gettype() == \"Linear\":\n weight_list.append(Optimizer.Node(Optimizer.WeightIni.init_linear_weight(layer_list[i].getinputdim(), layer_list[i].getoutputdim()),\"weight\"))\n elif layer_list[i].gettype() == \"BN\":\n dim = layer_list[i-1].getoutputdim()\n gamma, beta = Optimizer.WeightIni.init_BN_weight(dim)\n weight_list.append(Optimizer.Node(gamma,\"weight\"))\n weight_list.append(Optimizer.Node(beta,\"weight\"))\n layer_list[i].input_dim = dim\n layer_list[i].output_dim = dim\n # kernel parse operation\n elif layer_list[i].gettype() == \"Conv2D\":\n weight_list.append(Optimizer.Node(Optimizer.WeightIni.init_conv2D_kernel(layer_list[i].getkernelsize()),\"weight\"))\n else:\n return NameError\n # check if you need BN init\n if layer_list[i].getBN():\n dim = layer_list[i].getoutputdim()\n gamma, beta = Optimizer.WeightIni.init_BN_weight(dim)\n weight_list.append(Optimizer.Node(gamma,\"weight\"))\n weight_list.append(Optimizer.Node(beta,\"weight\"))\n\n return weight_list\n\n @staticmethod\n def forword(passer, weight_list, layer_list):\n layer_num = len(layer_list)\n passer_list = [Optimizer.Node(passer, \"data\")]\n # Every layer not necessarily has only one weight, like BN has 2 weights in a single layer\n weight_count = 0\n\n for i in range(layer_num):\n if layer_list[i].gettype() =='Linear':\n passer = passer@weight_list[weight_count].getdata()\n # append binary tree after inner product of weight and previous layer\n node = Optimizer.Node(passer,\"@\")\n node.left = passer_list[-1]\n node.right = weight_list[weight_count]\n passer_list.append(node)\n\n weight_count += 1\n\n if layer_list[i].getBN():\n node_cache = [passer, np.var(passer,axis = 0), np.mean(passer, axis=0 )]\n\n passer = (passer - np.mean(passer,axis=0))/np.sqrt(np.var(passer,axis=0))\n node = Optimizer.Node(passer,\"normalization\")\n node.cache = node_cache\n node.left = passer_list[-1]\n passer_list.append(node)\n\n node = Optimizer.Node(passer,\"*scalar\")\n node.left = passer_list[-1]\n node.right = weight_list[weight_count]\n passer_list.append(node)\n\n passer = passer + weight_list[weight_count+1].getdata()\n node = Optimizer.Node(passer,\"+scalar\")\n node.left = passer_list[-1]\n node.right = weight_list[weight_count+1]\n passer_list.append(node)\n\n weight_count += 2\n\n passer = layer_list[i].getact(passer)\n #append binary tree after activation function\n node = Optimizer.Node(passer,layer_list[i].getactname())\n node.left = passer_list[-1]\n passer_list.append(node)\n\n # elif layer_list[j].gettype() == \"Conv2D\":\n else: raise NameError\n\n return passer_list\n\n @staticmethod\n def backpropagation(node):\n epsilon = 1e-8\n if node.getleft() is not None:\n if node.gettype() == \"@\":\n node.getleft().back = node.getback()@node.getright().getdata().T\n node.getright().back = node.getleft().getdata()[email protected]()\n elif node.gettype() == \"sigmoid\":\n node.getleft().back = np.multiply(node.getback(),np.multiply(NN.ActivationFunc.sigmoid(node.getback()),\n 1-NN.ActivationFunc.sigmoid(node.getback())))\n elif node.gettype() == \"ReLU\":\n back = copy.deepcopy(node.getback())\n back[back<=0] = 0\n node.getleft().back = back\n elif node.gettype() == \"LeakyReLU\":\n back = copy.deepcopy(node.getback())\n back[back<0] = 0.01*back[back<0]\n node.getleft().back = back\n elif node.gettype() == \"tanh\":\n node.getleft().back = np.multiply((np.ones(node.getback().shape)-NN.ActivationFunc.tanh(node.getback())**2),\n node.getback())\n elif node.gettype() == \"+\":\n node.getleft().back = node.getback()\n node.getright().back = node.getback()\n elif node.gettype() == \"-\":\n node.getleft().back = node.getback()\n node.getright().back = -node.getback()\n elif node.gettype() == \"+scalar\":\n node.getleft().back = node.getback()\n node.getright().back = np.sum(node.getback(),axis=0)\n elif node.gettype() == \"*scalar\":\n node.getleft().back = node.getright().getdata() * node.getback()\n node.getright().back = np.sum(node.getleft().getdata().T,axis=0)@node.getback()\n elif node.gettype() == \"none\":\n node.getleft().back = node.getback()\n elif node.gettype() == \"normalization\":\n # cache = [x, sigma_beta^2, mu_beta]\n\n # dx = 1/N / std * (N * dx_norm -\n # dx_norm.sum(axis=0) -\n # x_norm * (dx_norm * x_norm).sum(axis=0))\n\n x = node.cache[0]\n sigma2 = node.cache[1]\n mu = node.cache[2]\n\n dl_dx_hat = node.getback()\n dl_dsigma2 = np.sum(dl_dx_hat,axis=0) * (x-mu) * -0.5*(sigma2+epsilon)**-3/2\n dl_dmu = np.sum(dl_dx_hat,axis=0) * -1/np.sqrt(sigma2+epsilon) + dl_dsigma2 * np.sum(-2*(x-mu),axis= 0)/x.shape[0]\n dl_dx = dl_dx_hat * 1/np.sqrt(sigma2+epsilon) + dl_dsigma2*2*(x-mu)/x.shape[0] + dl_dmu /x.shape[0]\n node.getleft().back = dl_dx\n\n Optimizer.backpropagation(node.getleft())\n else:\n return\n\n def lrdecay(self, iter):\n \"\"\"\n Learning rate decay function. Given iteration, modify learning rate\n\n :param iter: int, iteration count\n \"\"\"\n self.lr = 1 / (1 + self.decay_rate * iter) * self.lr\n\n def GD(self, root: Node, weight_list):\n \"\"\"\n Gradient descent, do the back propagation and update weight list\n\n :param root: Node, the root of passer binary tree\n :param weight_list: list, weight list\n :return: list, updated weight list\n \"\"\"\n Optimizer.backpropagation(root)\n gradient_list = []\n\n for node in weight_list:\n node.data = node.data - self.lr * node.back\n gradient_list.append(node.back)\n return weight_list, gradient_list\n\n def SGD(self, weight_list, passer_list):\n # we resume mini-batch equals 1 each time\n \"\"\"\n Stochastic gradient descent. It takes weight list and passer list as inputs, it will\n :param weight_list:\n :param passer_list:\n :return:\n \"\"\"\n def init_random_node(node, random_num_list, mini_weight_list):\n node.data = node.data[random_num_list,:]\n node.back = None\n if node.getright() is not None:\n mini_weight_list.append(node.getright())\n if node.getleft() is not None:\n init_random_node(node.getleft(), random_num_list, mini_weight_list)\n else: return\n\n # obs = observation number = output layer's dim 0\n num_obs = self.nn.dataset.gettrainset().getX().shape[0]\n mini_passer_list = copy.deepcopy(passer_list)\n root = mini_passer_list[-1]\n gradient_list = []\n\n # randomly pick observations from original obs\n random_num_list = np.random.randint(0, num_obs, num_obs)\n\n # initial random node\n mini_weight_list = []\n init_random_node(root, random_num_list, mini_weight_list)\n\n # back propagation\n root.back = 2 * (- self.nn.dataset.gettrainset().gety()[random_num_list] + root.getdata()[random_num_list])\n Optimizer.backpropagation(root)\n\n i = 0\n # update weight list\n for weight in weight_list:\n weight.data = weight.data - self.lr * mini_weight_list[-i-1].back\n gradient_list.append(mini_weight_list[-i-1].back)\n i = i + 1\n\n return weight_list, gradient_list\n\n def momentumgd(self, root: Node, weight_list, beta = 0.2):\n \"\"\"\n\n :param root: Node, the root of passer binary tree\n :param weight_list: list, weight list\n :param beta: momentum conservation rate\n :return: list, updated weight list\n \"\"\"\n Optimizer.backpropagation(root)\n gradient_list = []\n\n for node in weight_list:\n if node.getmomentum() is None:\n node.momentum = (1 - beta) * node.getback()\n else:\n node.momentum = beta * node.getmomentum() + (1 - beta) * node.getback()\n node.data = node.getdata() - self.lr * (1 - beta) * node.getback()\n gradient_list.append(node.back)\n return weight_list, gradient_list\n\n def RMSprop(self, root: Node, weight_list, beta = 0.2, eps =1e-10):\n\n Optimizer.backpropagation(root)\n gradient_list = []\n\n for node in weight_list:\n if node.getmomentum() is None:\n node.momentum = (1 - beta) * node.getback() ** 2\n else:\n node.momentum = beta * node.getmomentum() + (1 - beta) * node.getback() ** 2\n\n node.data = node.getdata() - self.lr * node.getback() / (np.sqrt(node.getmomentum()) + eps)\n gradient_list.append(node.back)\n return weight_list, gradient_list\n\n\n def Adam(self, root: Node, weight_list, beta_mom = 0.2, beta_rms = 0.2, eps = 1e-10):\n \"\"\"\n Adam optimizer\n :param root:\n :param weight_list:\n :param beta_mom:\n :param beta_rms:\n :param eps:\n :return:\n \"\"\"\n Optimizer.backpropagation(root)\n gradient_list = []\n\n for node in weight_list:\n if node.getmomentum() is None:\n node.momentum = [(1 - beta_mom) * node.getback(), (1 - beta_rms) * node.getback() ** 2]\n else:\n node.momentum[0] = (beta_mom * node.getmomentum()[0] + (1 - beta_mom) * node.getback()) / (1 - beta_mom)\n node.momentum[1] = (beta_rms * node.getmomentum()[1] + (1 - beta_rms) * node.getback() ** 2 ) / (1 - beta_rms)\n\n node.data = node.getdata() - self.lr * node.getmomentum()[0] / (np.sqrt(node.getmomentum()[1])+eps)\n gradient_list.append(node.back)\n return weight_list, gradient_list\n\n def train(self):\n \"\"\"\n train process, it will first initial weight, loss, gradient and passer list, then, optimize weights by given optimizer.\n In the end, calculate loss and step to the next epoch.\n\n It will finally stock all the weight, loss, gradient and passer during the training process\n \"\"\"\n layer_list = self.nn.getlayers()\n\n # initial weight, loss and gradient list\n self.weight_list = [[] for i in range(self.epoch+1)]\n self.weight_list[0] = Optimizer.WeightIni.initial_weight_list(layer_list)\n self.loss_list = np.zeros(self.epoch)\n self.gradient_list = [[] for i in range(self.epoch)]\n self.passer_list = [[] for i in range(self.epoch)]\n\n # for GD and SGD, they use full dataset, so need only read X and y once\n if self.optimizer ==\"GD\" or self.optimizer == \"SGD\":\n X = self.nn.dataset.gettrainset().getX()\n X = Optimizer.Node(X, \"data\")\n for i in range(self.epoch):\n # forward propagation\n self.passer_list[i] = Optimizer.forword(X.getdata(), self.weight_list[i],layer_list)\n root = self.passer_list[i][-1]\n\n # calculate loss by using: loss 2 * (-self.nn.dataset.gettrainset().gety() + root.getdata())\n loss_func = self.loss_function(self.nn.dataset.gettrainset().gety(), root.getdata())\n self.loss_list[i] = loss_func.loss()\n\n root.back = loss_func.diff()\n # upgrade gradient by selected optimizer\n if self.optimizer ==\"GD\":\n self.weight_list[i+1], self.gradient_list[i] = Optimizer.GD(self, root, self.weight_list[i])\n\n elif self.optimizer ==\"SGD\":\n self.weight_list[i+1], self.gradient_list[i] = Optimizer.SGD(self, self.weight_list[i], self.passer_list[i])\n\n # mini batch type gradient descent\n else:\n for i in range(self.epoch):\n start_time = time.time()\n # get mini batch\n minisets = self.nn.dataset.gettrainset().getminiset()\n epoch_weight_list = [copy.deepcopy(self.weight_list[i])]\n epoch_loss_list = np.zeros(len(minisets))\n\n # GD for every mini batch\n for j in range(len(minisets)):\n X_bar = minisets[j]\n self.passer_list[i].append(Optimizer.forword(X_bar.getX(), epoch_weight_list[j], layer_list))\n\n root = self.passer_list[i][j][-1]\n loss_func = self.loss_function(X_bar.gety(), root.getdata())\n\n epoch_loss_list[j] = loss_func.loss()\n root.back = loss_func.diff()\n root.momentum = root.getback()\n\n if self.optimizer == \"minibatchgd\":\n weight, gradient = Optimizer.GD(self, root, epoch_weight_list[j])\n elif self.optimizer == \"momentumgd\":\n weight, gradient = Optimizer.momentumgd(self, root, epoch_weight_list[j])\n elif self.optimizer == \"RMSprop\":\n weight, gradient = Optimizer.RMSprop(self, root, epoch_weight_list[j])\n elif self.optimizer == \"Adam\":\n weight, gradient = Optimizer.Adam(self, root, epoch_weight_list[j])\n else: raise NameError\n epoch_weight_list.append(weight)\n\n self.weight_list[i+1]= epoch_weight_list[-1]\n self.gradient_list[i] = gradient\n\n self.loss_list[i] = sum(epoch_loss_list)/len(epoch_loss_list)\n\n # learnign rate decay\n\n self.lrdecay(i)\n # every epoch shuffle the dataset\n self.nn.dataset.distribute()\n\n if (i + 1) % 1 ==0:\n used_time = time.time() - start_time\n print(\"epoch \" + str(i + 1) + ', Training time: %.4f' % used_time + ', Training loss: %.6f' % self.loss_list[i])\n\n def test(self):\n \"\"\"\n Use trained weight on testset for the evaluation of the model\n :return: model prediction and loss on the testset\n \"\"\"\n weight = self.weight_list[-1]\n layer_list = self.nn.getlayers()\n testset = self.nn.dataset.gettestset()\n passer = testset.getX()\n\n passer_list = self.forword(passer,weight,layer_list)\n predicted = passer_list[-1].getdata()\n\n loss = self.loss_function.loss(testset.gety(), predicted)\n return predicted, loss\n\n def predict(self, X):\n \"\"\"\n Use trained weight on X and output prediction\n :param X: ndarray, feature data wish to be predicted\n :return: model's prediction by using trained data\n \"\"\"\n passer = X\n weight = self.weight_list[-1]\n passer_list = self.forword(passer, weight, self.nn.getlayers())\n return passer_list", "_____no_output_____" ], [ "class Visual:\n def __init__(self, optim):\n self.optim = optim\n\n def plotloss(self):\n \"\"\"\n :return: plot loss flow during the training\n \"\"\"\n plt.style.use('seaborn-whitegrid')\n fig = plt.figure()\n ax = plt.axes()\n ax.plot(self.optim.loss_list, label = 'loss')\n ax.legend(loc='upper right')\n ax.set_ylabel('Loss during the training')\n\n def plotgradientnorm(self):\n plt.style.use('seaborn-whitegrid')\n fig, axs = plt.subplots(len(self.optim.getgradientlist()[0]))\n for i in range(len(self.optim.getgradientlist()[0])):\n gradient_norm_list = []\n for j in range(len(self.optim.getgradientlist())):\n gradient_norm_list.append(np.linalg.norm(self.optim.getgradientlist()[j][i]))\n axs[i].plot(gradient_norm_list, label = 'norm 2')\n axs[i].legend(loc='upper right')\n axs[i].set_ylabel('W' + str(i) +\" norm\")\n", "_____no_output_____" ], [ "# total observation number\nn = 300\n# x1, x2 are generated by two\nx1 = np.random.uniform(0,1,n)\nx2 = np.random.uniform(0,1,n)\nconst = np.ones(n)\neps = np.random.normal(0,.05,n)\nb = 1.5\ntheta1 = 2\ntheta2 = 5\nTheta = np.array([b, theta1, theta2])\ny = np.array(b * const+ theta1 * x1 + theta2 * x2 + eps)\ny=np.reshape(y,(-1,1))\nX = np.array([const,x1,x2]).T", "_____no_output_____" ], [ "layer_list = [NN.Layer('Linear',3,100,'LeakyReLU'),NN.Layer('Linear',100,3,'LeakyReLU'),\n NN.Layer('Linear',3,1,'none')]\ndataset = Dataset(X, y)\nnn = NN(dataset)\nnn.addlayers(layer_list)\nloss_func = Optimizer.LossFunc.Quadratic\noptim = Optimizer(nn,\"SGD\",loss_func, epoch = 10000, lr=1e-6)\noptim.train()\nvisual = Visual(optim)\nvisual.plotloss()\nvisual.plotgradientnorm()", "_____no_output_____" ], [ "# total observation number\nn = 10000\n# x1, x2 are generated by two\nx1 = np.random.uniform(0,1,n)\nx2 = np.random.uniform(0,1,n)\nconst = np.ones(n)\neps = np.random.normal(0,.05,n)\nb = 1.5\ntheta1 = 2\ntheta2 = 5\nTheta = np.array([b, theta1, theta2])\ny = np.array(b * const+ theta1 * x1 + theta2 * x2 + eps)\ny=np.reshape(y,(-1,1))\nX = np.array([const,x1,x2]).T", "_____no_output_____" ], [ "layer_list = [NN.Layer('Linear',3,10,'sigmoid',BN=True), NN.Layer('Linear',10,100,'sigmoid',BN=True),\n NN.Layer('Linear',100,10,'sigmoid',BN=True),NN.Layer('Linear',10,3,'none') ]\ndataset = Dataset(X, y, mini_batch= 64)\nnn = NN(dataset)\nnn.addlayers(layer_list)\nloss_func = Optimizer.LossFunc.Quadratic\noptim = Optimizer(nn,\"Adam\", loss_func, epoch = 10, lr=1e-2, decay_rate=0.01)\noptim.train()\nvisual = Visual(optim)\nvisual.plotloss()\nvisual.plotgradientnorm()", "/tmp/ipykernel_15455/216912902.py:43: RuntimeWarning: overflow encountered in exp\n return 1.0 / (1.0 + np.exp(-x))\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "IJG" ]

[ "IJG" ]

[ "IJG" ]

[ [ [ "%pylab inline", "Populating the interactive namespace from numpy and matplotlib\n" ] ], [ [ "** Problem: Fourier transforms of simple functions **\n\nWrite Python programs to calculate the coefficients in the discrete Fourier transforms of the following periodic functions sampled at $N=1000$ evenly spaced points, and make plots of their amplitudes:\n\n1) A single cycle of a square-wave with amplitude 1\n\n2) The sawtooth wave $y_n=n$\n\n3) The modulated sine wave $y_n = \\sin(\\pi n/N) \\sin(20\\pi n/N)$\n", "_____no_output_____" ] ], [ [ "# Import required modules\nfrom numpy import arange\nfrom numpy.fft import rfft\nfrom math import sin,pi\n\n# Square wave\ndef f(x):\n if x<0.5:\n return 1\n else:\n return -1\n \nN=1000\nx=arange(0.0,1.0,1.0/N)\ny=map(f,x)\nc=rfft(y) # Calculate the Fourier coefficients (complex numbers!)\nplot(abs(c)) # plot their magnitudes vs. mode number\nxlim(0,100) # show the first 100 modes\nshow()", "_____no_output_____" ], [ "# Sawtooth function\n\nN=1000\ny=arange(N)\nc=rfft(y)\nplot(abs(c))\nxlim(0,100)\nshow()", "_____no_output_____" ], [ "# sin wave\n\ndef f(x):\n return sin(pi*x)*sin(20*pi*x)\n\nN=1000\nx=arange(0.0,1.0,1.0/N)\ny=map(f,x)\nc=rfft(y) # Calculate the Fourier coefficients (complex numbers!)\nplot(abs(c)) # plot their magnitudes vs. mode number\nxlim(0,100) # show the first 100 modes\nshow()", "_____no_output_____" ] ] ]

[ "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import tushare as ts\nimport sina_data\nimport numpy as np\nimport pandas as pd\nfrom pandas import DataFrame, Series\nfrom datetime import datetime, timedelta\nfrom dateutil.parser import parse\nimport time\nimport common_util\nimport os", "_____no_output_____" ], [ "def get_time(date=False, utc=False, msl=3):\n if date:\n time_fmt = \"%Y-%m-%d %H:%M:%S.%f\"\n else:\n time_fmt = \"%H:%M:%S.%f\"\n\n if utc:\n return datetime.utcnow().strftime(time_fmt)[:(msl-6)]\n else:\n return datetime.now().strftime(time_fmt)[:(msl-6)]\n\n\ndef print_info(status=\"I\"):\n return \"\\033[0;33;1m[{} {}]\\033[0m\".format(status, get_time())", "_____no_output_____" ], [ "def judgement(df, change_rate=0.01, buy1_rate=0.03, buy1_volume=1e5):\n float_share = df['float_share'].to_numpy().astype(np.int)\n open = df['今日开盘价'].to_numpy().astype(np.float)\n pre_close = df['昨日收盘价'].to_numpy().astype(np.float)\n limit_up = limit_up_price(pre_close)\n price = df['当前价'].to_numpy().astype(np.float)\n high = df['今日最高价'].to_numpy().astype(np.float)\n low = df['今日最低价'].to_numpy().astype(np.float)\n volume = df['成交股票数'].to_numpy().astype(np.int)\n buy_1v = df['买一量'].to_numpy().astype(np.int)\n\n judge_list = [\n low < limit_up,\n price < limit_up,\n volume < float_share * change_rate,\n buy_1v < float_share * buy1_rate,\n buy_1v < buy1_volume\n ]\n\n return judge_list\n\n\n# 基于前一交易日收盘价的涨停价计算\ndef limit_up_price(pre_close):\n return np.around(pre_close * 1.1, decimals=2)\n\n\n# 日K数据判断是否开板\ndef is_sold(code, start_date):\n print(code)\n try:\n time.sleep(1)\n pro = ts.pro_api('ba73b3943bdd57c2ff05991f7556ef417f457ac453355972ff5d01ce')\n start_date = (parse(str(start_date))+timedelta(1)).strftime('%Y%m%d')\n end_date = datetime.now().strftime('%Y%m%d')\n daily_k = pro.daily(ts_code=code, start_date=start_date, end_date=end_date)\n if len(daily_k) > 0:\n daily_k['flag'] = daily_k.apply(\n lambda x: x['high'] == x['low'] and x['open'] == x['close'] and x['high'] == x['low'],\n axis=1\n )\n flag = daily_k['flag'].sum()\n result = True\n for each in daily_k['flag'].tolist():\n result = result and each\n return result\n else:\n return True\n except Exception as e:\n print('再次请求ts数据')\n time.sleep(1)\n a = is_sold(code, start_date)\n return a\n\n\n# 获取流通股本\ndef get_float_share(code):\n print(code)\n try:\n time.sleep(1)\n pro = ts.pro_api('ba73b3943bdd57c2ff05991f7556ef417f457ac453355972ff5d01ce')\n # target_date = datetime.now().strftime('%Y%m%d')\n target_data = []\n delta = 0\n count = 1\n while len(target_data) == 0:\n target_date = datetime.now() + timedelta(delta)\n target_data = pro.daily_basic(\n ts_code=code, trade_date=target_date.strftime('%Y%m%d'), fields='free_share'\n )\n delta = delta - 1\n time.sleep(0.5)\n count = count + 1\n if count > 3:\n return 1000000\n return target_data.values[0][0] * 10000\n\n except Exception as e:\n time.sleep(1)\n get_float_share(code)\n print('再次请求ts数据.....')", "_____no_output_____" ], [ "# 新股筛选\n# 获取股票列表\npro = ts.pro_api('ba73b3943bdd57c2ff05991f7556ef417f457ac453355972ff5d01ce')\nbasic_data = pro.stock_basic()\nprint('股票筛选')\n# basic_data.to_excel(r'C:\\Users\\duanp\\Desktop\\test\\stock_basic.xlsx')\n# basic_data = pd.read_excel(r'C:\\Users\\duanp\\Desktop\\test\\stock_basic.xlsx')\n# 筛选上市日期为近一月的股票\nstart_date = datetime.now() + timedelta(-30)\nend_date = datetime.now() + timedelta(1)\nbasic_data['list_date'] = basic_data['list_date'].apply(lambda x: parse(str(x)))\nbasic_data = basic_data[basic_data['list_date'] > start_date]\nbasic_data = basic_data[basic_data['list_date'] < end_date]\n# 剔除科创板股票\nbasic_data = basic_data[basic_data['market'] != '科创板']\n# 筛选未开板的股票\nbasic_data['target_flag'] = basic_data.apply(lambda x: is_sold(x['ts_code'], x['list_date']), axis=1)\n# basic_data = basic_data[basic_data['target_flag']]\nprint('补充流通股本数据')\n# 补充流通股本信息\nbasic_data['float_share'] = basic_data.apply(lambda x: get_float_share(x['ts_code']), axis=1)\nbasic_data['float_share'] = basic_data['float_share'].fillna('100000')\nprint('预警股票如下：')\nprint(basic_data)\n\nchange_rate = 0.01\nbuy1_rate = 0.03\nbuy1_volume = 1e5\n\ntick_list = [\n '股票代码',\n '今日开盘价',\n '昨日收盘价',\n '当前价',\n '今日最高价',\n '今日最低价',\n '成交股票数',\n '买一量'\n]\n\nflag_dict = {\n \"low_flag\": \"当日曾开板！\",\n \"price_flag\": \"已经开板！\",\n \"volume_top_flag\": \"换手率超过 {:.0%}！\".format(change_rate),\n \"buy1_percent_flag\": \"买一量不足总流通市值的 {:.0%}！\".format(buy1_rate),\n \"buy1_volume_flag\": \"买一量不足 {} 股！\".format(buy1_volume),\n}\n\nflag_list = list(flag_dict.keys())\nflag_len = len(flag_list)", "股票筛选\n003001.SZ\n003012.SZ\n003013.SZ\n003015.SZ\n003016.SZ\n003017.SZ\n300898.SZ\n300899.SZ\n300900.SZ\n300901.SZ\n300902.SZ\n300999.SZ\n601187.SH\n601568.SH\n601995.SH\n605058.SH\n605169.SH\n605336.SH\n605338.SH\n补充流通股本数据\n003001.SZ\n003012.SZ\n003013.SZ\n003015.SZ\n003016.SZ\n003017.SZ\n300898.SZ\n300899.SZ\n300900.SZ\n300901.SZ\n300902.SZ\n300999.SZ\n601187.SH\n601568.SH\n601995.SH\n605058.SH\n605169.SH\n605336.SH\n605338.SH\n预警股票如下：\n ts_code symbol name area industry market list_date target_flag \\\n1417 003001.SZ 003001 中岩大地 北京 建筑工程 中小板 2020-10-13 False \n1427 003012.SZ 003012 东鹏控股 广东 陶瓷 中小板 2020-10-19 False \n1428 003013.SZ 003013 地铁设计 广东 建筑工程 中小板 2020-10-22 False \n1429 003015.SZ 003015 日久光电 江苏 元器件 中小板 2020-10-21 False \n1430 003016.SZ 003016 欣贺股份 福建 服饰 中小板 2020-10-26 False \n1431 003017.SZ 003017 大洋生物 浙江 化工原料 中小板 2020-10-26 False \n2296 300898.SZ 300898 熊猫乳品 浙江 乳制品 创业板 2020-10-16 False \n2297 300899.SZ 300899 上海凯鑫 上海 环境保护 创业板 2020-10-16 False \n2298 300900.SZ 300900 C广联 黑龙江 航空 创业板 2020-10-29 False \n2299 300901.SZ 300901 C中胤 浙江 服饰 创业板 2020-10-29 False \n2300 300902.SZ 300902 C国安达 福建 专用机械 创业板 2020-10-29 False \n2301 300999.SZ 300999 金龙鱼 上海 食品 创业板 2020-10-15 False \n3160 601187.SH 601187 厦门银行 福建 银行 主板 2020-10-27 False \n3210 601568.SH 601568 北元集团 陕西 化工原料 主板 2020-10-20 False \n3304 601995.SH 601995 中金公司 北京 证券 主板 2020-11-02 False \n3828 605058.SH 605058 澳弘电子 江苏 元器件 主板 2020-10-21 False \n3843 605169.SH 605169 N洪通 新疆 供气供热 主板 2020-10-30 False \n3854 605336.SH 605336 帅丰电器 浙江 家用电器 主板 2020-10-19 False \n3855 605338.SH 605338 巴比食品 上海 食品 主板 2020-10-12 False \n\n float_share \n1417 24293828.0 \n1427 143000000.0 \n1428 40010000.0 \n1429 70266667.0 \n1430 106666700.0 \n1431 15000000.0 \n2296 26456438.0 \n2297 15950000.0 \n2298 44464076.0 \n2299 56887984.0 \n2300 30346688.0 \n2301 356738334.0 \n3160 263912789.0 \n3210 361111112.0 \n3304 260278568.0 \n3828 35731000.0 \n3843 40000000.0 \n3854 35200000.0 \n3855 62000000.0 \n" ], [ "basic_data['target_code'] = basic_data['ts_code'].apply(lambda x: common_util.get_format_code(x, 'num'))", "_____no_output_____" ], [ "basic_data['ts_code'].to_list()", "_____no_output_____" ], [ "tick_data = sina_data.get_tick_data(basic_data['symbol'].to_list())", "_____no_output_____" ], [ "tick_data['股票代码'] = tick_data['股票代码'].apply(lambda x: common_util.get_format_code(x, 'wind'))", "_____no_output_____" ], [ "tick_data = tick_data[tick_list]", "_____no_output_____" ], [ "temp_data = basic_data.merge(tick_data, left_on='ts_code', right_on='股票代码')", "_____no_output_____" ], [ "judge_list = judgement(temp_data, change_rate, buy1_rate, buy1_volume)", "_____no_output_____" ], [ "judge_list", "_____no_output_____" ], [ "alert_dict = dict()\ncount = 0", "_____no_output_____" ], [ "for idx in range(flag_len):\n temp_data[flag_list[idx]] = judge_list[idx]\n alert_dict[flag_list[idx]] = temp_data[temp_data[flag_list[idx]]][\"name\"].tolist()\n if len(alert_dict[flag_list[idx]]) > 0:\n print(print_info(\"W\"), end=\" \")\n print(flag_dict[flag_list[idx]])\n print(\"，\".join(alert_dict[flag_list[idx]]))\n else:\n count += 1", "\u001b[0;33;1m[W 23:24:45.211]\u001b[0m 当日曾开板！\n中岩大地，东鹏控股，地铁设计，日久光电，欣贺股份，大洋生物，熊猫乳品，上海凯鑫，C广联，C中胤，C国安达，金龙鱼，厦门银行，北元集团，中金公司，澳弘电子，N洪通，帅丰电器，巴比食品\n\u001b[0;33;1m[W 23:24:45.211]\u001b[0m 已经开板！\n中岩大地，东鹏控股，地铁设计，欣贺股份，大洋生物，熊猫乳品，上海凯鑫，C广联，C中胤，C国安达，金龙鱼，厦门银行，北元集团，中金公司，澳弘电子，N洪通，帅丰电器，巴比食品\n\u001b[0;33;1m[W 23:24:45.212]\u001b[0m 买一量不足总流通市值的 3%！\n中岩大地，东鹏控股，地铁设计，日久光电，欣贺股份，大洋生物，熊猫乳品，上海凯鑫，C广联，C中胤，C国安达，金龙鱼，厦门银行，北元集团，中金公司，澳弘电子，N洪通，帅丰电器，巴比食品\n\u001b[0;33;1m[W 23:24:45.213]\u001b[0m 买一量不足 100000.0 股！\n中岩大地，东鹏控股，地铁设计，欣贺股份，大洋生物，熊猫乳品，上海凯鑫，C广联，C国安达，北元集团，澳弘电子，N洪通，帅丰电器，巴比食品\n" ], [ "idx=1\ntemp_data[flag_list[idx]] = judge_list[idx]", "_____no_output_____" ], [ "alert_dict[flag_list[idx]] = temp_data[temp_data[flag_list[idx]]]\nalert_dict", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%matplotlib inline\n%pylab inline\npylab.rcParams['figure.figsize'] = (10, 6)\nimport numpy as np\nfrom numpy.lib import stride_tricks\nimport cv2\nfrom matplotlib.colors import hsv_to_rgb\nimport matplotlib.pyplot as plt\nimport numpy as np\nnp.set_printoptions(precision=3)", "Populating the interactive namespace from numpy and matplotlib\n" ], [ "class PatchMatch(object):\n def __init__(self, a, b, patch_size):\n assert a.shape == b.shape, \"Dimensions were unequal for patch-matching input\"\n self.A = a\n self.B = b\n self.patch_size = patch_size\n self.nnf = np.zeros((2, self.A.shape[0], self.A.shape[1])).astype(np.int)\n self.nnd = np.zeros((self.A.shape[0], self.A.shape[1]))\n self.initialise_nnf()\n \n def initialise_nnf(self):\n self.nnf[0] = np.random.randint(self.B.shape[0], size=(self.A.shape[0], self.A.shape[1]))\n self.nnf[1] = np.random.randint(self.B.shape[1], size=(self.A.shape[0], self.A.shape[1]))\n self.nnf = self.nnf.transpose((1, 2 ,0))\n for i in range(self.A.shape[0]):\n for j in range(self.A.shape[1]):\n pos = self.nnf[i,j]\n self.nnd[i,j] = self.cal_dist(i, j, pos[0], pos[1])\n\n def cal_dist(self, ai ,aj, bi, bj):\n dx0 = dy0 = self.patch_size//2\n dx1 = dy1 = self.patch_size//2 + 1\n dx0 = min(ai, bi, dx0)\n dx1 = min(self.A.shape[0]-ai, self.B.shape[0]-bi, dx1)\n dy0 = min(aj, bj, dy0)\n dy1 = min(self.A.shape[1]-aj, self.B.shape[1]-bj, dy1)\n return np.sum((self.A[ai-dx0:ai+dx1, aj-dy0:aj+dy1]-self.B[bi-dx0:bi+dx1, bj-dy0:bj+dy1])**2) / (dx1+dx0) / (dy1+dy0)\n \n def reconstruct(self):\n ans = np.zeros_like(self.A)\n for i in range(self.A.shape[0]):\n for j in range(self.A.shape[1]):\n pos = self.nnf[i,j]\n ans[i,j] = self.B[pos[0], pos[1]]\n return ans\n \n def reconstruct_img_voting(self, patch_size=3,arr_v=None):\n if patch_size is None:\n patch_size = self.patch_size\n b_prime = np.zeros_like(self.A,dtype=np.uint8)\n\n for i in range(self.A.shape[0]): #traverse down a\n for j in range(self.A.shape[1]): #traverse across a\n \n dx0 = dy0 = patch_size//2\n dx1 = dy1 = patch_size//2 + 1\n dx0 = min(i,dx0)\n dx1 = min(self.A.shape[0]-i, dx1)\n dy0 = min(j, dy0)\n dy1 = min(self.A.shape[1]-j, dy1)\n \n votes = self.nnf[i-dx0:i+dx1, j-dy0:j+dy1] \n b_patch = np.zeros(shape=(votes.shape[0],votes.shape[1],self.A.shape[2]))\n \n for p_i in range(votes.shape[0]):\n for p_j in range(votes.shape[1]):\n \n b_patch[p_i, p_j] = self.B[votes[p_i,p_j][0] , votes[p_i,p_j][1]]\n\n averaged_patch = np.average(b_patch,axis=(0,1))\n b_prime[i, j] = averaged_patch[:]\n plt.imshow(b_prime[:,:,::-1])\n plt.show()\n \n def visualize_nnf(self):\n nnf = self.nnf\n nnd = self.nnd\n def angle_between_alt(p1, p2):\n ang1 = np.arctan2(*p1[::-1])\n ang2 = np.arctan2(*p2[::-1])\n return np.rad2deg((ang1 - ang2) % (2 * np.pi))\n\n def norm_dist(arr):\n return (arr)/(arr.max())\n \n img = np.zeros((nnf.shape[0], nnf.shape[1], 3),dtype=np.uint8)\n for i in range(1, nnf.shape[0]):\n for j in range(1, nnf.shape[1]):\n angle = angle_between_alt([j, i], [nnf[i, j][0], nnf[i, j][1]])\n img[i, j, :] = np.array([angle, nnd[i,j], 250])\n img = hsv_to_rgb(norm_dist(img/255))\n plt.imshow(img)\n plt.show()\n \n def propagate(self):\n compare_value = -1 \n for i in range(self.A.shape[0]):\n for j in range(self.A.shape[1]):\n x,y = self.nnf[i,j]\n bestx, besty, bestd = x, y, self.nnd[i,j]\n \n compare_value *=-1\n \n if (i + compare_value >= 0 and compare_value == -1) or (i + compare_value < self.A.shape[0] and compare_value == 1) :\n rx, ry = self.nnf[i+compare_value, j][0] , self.nnf[i+compare_value, j][1]\n if rx < self.B.shape[0]:\n val = self.cal_dist(i, j, rx, ry)\n if val < bestd:\n bestx, besty, bestd = rx, ry, val\n\n if (j+compare_value >= 0 and compare_value == -1)or (j + compare_value < self.A.shape[1] and compare_value == 1) :\n rx, ry = self.nnf[i, j+compare_value][0], self.nnf[i, j+compare_value][1] \n if ry < self.B.shape[1]:\n val = self.cal_dist(i, j, rx, ry)\n if val < bestd:\n bestx, besty, bestd = rx, ry, val\n \n rand_d = min(self.B.shape[0]//2, self.B.shape[1]//2)\n while rand_d > 0:\n try:\n xmin = max(bestx - rand_d, 0)\n xmax = min(bestx + rand_d, self.B.shape[0])\n ymin = max(besty - rand_d, 0)\n ymax = min(besty + rand_d, self.B.shape[1])\n #print(xmin, xmax)\n rx = np.random.randint(xmin, xmax)\n ry = np.random.randint(ymin, ymax)\n val = self.cal_dist(i, j, rx, ry)\n if val < bestd:\n bestx, besty, bestd = rx, ry, val\n except:\n print(rand_d)\n print(xmin, xmax)\n print(ymin, ymax)\n print(bestx, besty)\n print(self.B.shape)\n rand_d = rand_d // 2\n\n self.nnf[i, j] = [bestx, besty]\n self.nnd[i, j] = bestd\n \n print(\"Done\")", "_____no_output_____" ], [ "x = cv2.imread(\"./blue.jpg\")\ny = cv2.imread(\"./yellow.jpg\")\n\nx = cv2.resize(x,(200,200))\ny = cv2.resize(y,(200,200))", "_____no_output_____" ], [ "pm = PatchMatch(x,y, 3)\npm.visualize_nnf()\ndef do():\n pm.propagate()\n pm.reconstruct_img_voting(patch_size=3)\n# pm.propagate()\n# pm.reconstruct_img_voting(patch_size=3)\n# pm.propagate()\n# pm.reconstruct_img_voting(patch_size=3)\n# pm.propagate()\n# pm.reconstruct_img_voting(patch_size=3)\n\ndo()\npm.visualize_nnf()\n\n", "_____no_output_____" ], [ "do()", "_____no_output_____" ], [ "plt.figure(1)\nplt.subplot(131)\nplt.axis('off')\nplt.imshow(x[:,:,::-1])\n\nplt.subplot(132)\nplt.axis('off')\nplt.imshow(y[:,:,::-1])\n\nplt.subplot(133)\nplt.axis('off')\nplt.imshow(pm.reconstruct()[:,:,::-1])\n\nplt.show()", "_____no_output_____" ], [ "import os\nimport sys\n\n# add the 'src' directory as one where we can import modules\nsrc_dir = os.path.join(os.getcwd(), os.pardir)\nsys.path.append(src_dir)", "_____no_output_____" ], [ "os.path.join(os.getcwd(), os.pardir)", "_____no_output_____" ], [ "from src.PatchMatch import PatchMatchSimple", "_____no_output_____" ], [ "pm = PatchMatchSimple(x,y,patch_size=3)\nfor i in range(15):\n pm.propagate()\n pm.reconstruct_img_voting(patch_size=3)", "Done\n" ], [ "pm.visualize_nnf()", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "This notebook is used to generate data regarding the Goldbach conjecture: rows are of the form $N, a, b$ where $N$ is an even number, $a< b$ are prime and $N=a + b$.\n\nFirst we write a Python function `goldbach` that takes `N` and returns all pairs of numbers $a, b$.", "_____no_output_____" ] ], [ [ "import sympy as sym\nimport pandas as pd\n\ndef goldbach(N):\n \"\"\"Returns all pairs of primes that sum to give N\"\"\"\n primes = list(sym.primerange(1, N))\n sums = []\n for i, p1 in enumerate(primes):\n for p2 in primes[i:]:\n if p1 + p2 == N:\n sums.append((p1, p2))\n return sums", "_____no_output_____" ] ], [ [ "Let us use the above function to create our data:", "_____no_output_____" ] ], [ [ "maxN = 500\ndata = [[N, *pair] for N in range(4, maxN + 1) \n for pair in goldbach(N) if N % 2 == 0 ]", "_____no_output_____" ] ], [ [ "Let us write our data to an excel file:", "_____no_output_____" ] ], [ [ "df = pd.DataFrame(data, columns=[\"N\",\"a\", \"b\"]) # Create a data frame\ndf.to_excel(\"data/goldbach.xlsx\") # Write it to excel", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import matplotlib.pyplot as plt\n%matplotlib inline\nplt.plot([1, 2, 3, 4, 5], [1, 2, 3, 4, 5])", "_____no_output_____" ], [ "plt.plot()", "_____no_output_____" ], [ "plt.plot([1, 7, 3, 5, 11, 1])", "_____no_output_____" ], [ "plt.plot([1, 5, 10, 15, 20], [1, 7, 3, 5, 11])", "_____no_output_____" ], [ "plt.xlabel('Day', fontsize=15, color='blue')", "_____no_output_____" ], [ "plt.title('Chart price', fontsize=17)", "_____no_output_____" ], [ "plt.text(1, 1, 'type: Steel')", "_____no_output_____" ], [ "x = [1, 5, 10, 15, 20]\ny = [1, 7, 3, 5, 11]\nplt.plot(x, y, label='steel price')\nplt.title('Chart price', fontsize=15)\nplt.xlabel('Day', fontsize=12, color='blue')\nplt.ylabel('Price', fontsize=12, color='blue')\nplt.legend()\nplt.grid(True)\nplt.text(15, 4, 'grow up!')", "_____no_output_____" ], [ "plt.plot(x, y, color='red')", "_____no_output_____" ], [ "x = [1, 5, 10, 15, 20]\ny = [1, 7, 3, 5, 11]\nplt.plot(x, y, '--')", "_____no_output_____" ], [ "x = [1, 5, 10, 15, 20]\ny = [1, 7, 3, 5, 11]\nline = plt.plot(x, y)\nplt.setp(line, linestyle='--')", "_____no_output_____" ], [ "x = [1, 5, 10, 15, 20]\ny1 = [1, 7, 3, 5, 11]\ny2 = [i*1.2 + 1 for i in y1]\ny3 = [i*1.2 + 1 for i in y2]\ny4 = [i*1.2 + 1 for i in y3]\nplt.plot(x, y1, '-', x, y2, '--', x, y3, '-.', x, y4, ':')", "_____no_output_____" ], [ "plt.plot(x, y1, '-')\nplt.plot(x, y2, '--')\nplt.plot(x, y3, '-.')\nplt.plot(x, y4, ':')", "_____no_output_____" ], [ "x = [1, 5, 10, 15, 20]\ny = [1, 7, 3, 5, 11]\nplt.plot(x, y, '--r')", "_____no_output_____" ], [ "plt.plot(x, y, 'ro')", "_____no_output_____" ], [ "plt.plot(x, y, 'bx')", "_____no_output_____" ], [ "x = [1, 5, 10, 15, 20]\ny1 = [1, 7, 3, 5, 11]\ny2 = [i*1.2 + 1 for i in y1]\ny3 = [i*1.2 + 1 for i in y2]\ny4 = [i*1.2 + 1 for i in y3]\n\nplt.figure(figsize=(12, 7))\n\nplt.subplot(2, 2, 1)\nplt.plot(x, y1, '-')\n\nplt.subplot(2, 2, 2)\nplt.plot(x, y2, '--')\n\nplt.subplot(2, 2, 3)\nplt.plot(x, y3, '-.')\n\nplt.subplot(2, 2, 4)\nplt.plot(x, y4, ':')", "_____no_output_____" ], [ "plt.figure(figsize=(12, 7))\n\nplt.subplot(221)\nplt.plot(x, y1, '-')\n\nplt.subplot(222)\nplt.plot(x, y2, '--')\n\nplt.subplot(223)\nplt.plot(x, y3, '-.')\n\nplt.subplot(224)\nplt.plot(x, y4, ':')", "_____no_output_____" ], [ "fig, axs = plt.subplots(2, 2, figsize=(12, 7))\n\naxs[0, 0].plot(x, y1, '-')\naxs[0, 1].plot(x, y2, '--')\naxs[1, 0].plot(x, y3, '-.')\naxs[1, 1].plot(x, y4, ':')", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "CC-BY-3.0", "Apache-2.0" ]

[ "CC-BY-3.0", "Apache-2.0" ]

[ "CC-BY-3.0", "Apache-2.0" ]

[ [ [ "# -*- coding: utf-8 -*-\n# เรียกใช้งานโมดูล\nfile_name=\"data\"\nimport codecs\nfrom tqdm import tqdm\nfrom pythainlp.tokenize import word_tokenize\n#import deepcut\nfrom pythainlp.tag import pos_tag\nfrom nltk.tokenize import RegexpTokenizer\nimport glob\nimport nltk\nimport re\n# thai cut\nthaicut=\"newmm\"\nfrom sklearn_crfsuite import scorers,metrics\nfrom sklearn.metrics import make_scorer\nfrom sklearn.model_selection import cross_validate,train_test_split\nimport sklearn_crfsuite\nfrom pythainlp.corpus.common import thai_stopwords\nstopwords = list(thai_stopwords())\n#จัดการประโยคซ้ำ\ndata_not=[]\ndef Unique(p):\n text=re.sub(\"<[^>]*>\",\"\",p)\n text=re.sub(\"\\[(.*?)\\]\",\"\",text)\n text=re.sub(\"\\[\\/(.*?)\\]\",\"\",text)\n if text not in data_not:\n data_not.append(text)\n return True\n else:\n return False\n# เตรียมตัวตัด tag ด้วย re\npattern = r'\\[(.*?)\\](.*?)\\[\\/(.*?)\\]'\ntokenizer = RegexpTokenizer(pattern) # ใช้ nltk.tokenize.RegexpTokenizer เพื่อตัด [TIME]8.00[/TIME] ให้เป็น ('TIME','ไง','TIME')\n# จัดการกับ tag ที่ไม่ได้ tag\ndef toolner_to_tag(text):\n text=text.strip().replace(\"FACILITY\",\"LOCATION\").replace(\"[AGO]\",\"\").replace(\"[/AGO]\",\"\").replace(\"[T]\",\"\").replace(\"[/T]\",\"\")\n text=re.sub(\"<[^>]*>\",\"\",text)\n text=re.sub(\"(\\[\\/(.*?)\\])\",\"\\\\1***\",text)#.replace('(\\[(.*?)\\])','***\\\\1')# text.replace('>','>***') # ตัดการกับพวกไม่มี tag word\n text=re.sub(\"(\\[\\w+\\])\",\"***\\\\1\",text)\n text2=[]\n for i in text.split('***'):\n if \"[\" in i:\n text2.append(i)\n else:\n text2.append(\"[word]\"+i+\"[/word]\")\n text=\"\".join(text2)#re.sub(\"[word][/word]\",\"\",\"\".join(text2))\n return text.replace(\"[word][/word]\",\"\")\n# แปลง text ให้เป็น conll2002\ndef text2conll2002(text,pos=True):\n \"\"\"\n ใช้แปลงข้อความให้กลายเป็น conll2002\n \"\"\"\n text=toolner_to_tag(text)\n text=text.replace(\"''\",'\"')\n text=text.replace(\"’\",'\"').replace(\"‘\",'\"')#.replace('\"',\"\")\n tag=tokenizer.tokenize(text)\n j=0\n conll2002=\"\"\n for tagopen,text,tagclose in tag:\n word_cut=word_tokenize(text,engine=thaicut) # ใช้ตัวตัดคำ newmm\n i=0\n txt5=\"\"\n while i<len(word_cut):\n if word_cut[i]==\"''\" or word_cut[i]=='\"':pass\n elif i==0 and tagopen!='word':\n txt5+=word_cut[i]\n txt5+='\\t'+'B-'+tagopen\n elif tagopen!='word':\n txt5+=word_cut[i]\n txt5+='\\t'+'I-'+tagopen\n else:\n txt5+=word_cut[i]\n txt5+='\\t'+'O'\n txt5+='\\n'\n #j+=1\n i+=1\n conll2002+=txt5\n if pos==False:\n return conll2002\n return postag(conll2002)\n# ใช้สำหรับกำกับ pos tag เพื่อใช้กับ NER\n# print(text2conll2002(t,pos=False))\ndef postag(text):\n listtxt=[i for i in text.split('\\n') if i!='']\n list_word=[]\n for data in listtxt:\n list_word.append(data.split('\\t')[0])\n #print(text)\n list_word=pos_tag(list_word,engine=\"perceptron\", corpus=\"orchid_ud\")\n text=\"\"\n i=0\n for data in listtxt:\n text+=data.split('\\t')[0]+'\\t'+list_word[i][1]+'\\t'+data.split('\\t')[1]+'\\n'\n i+=1\n return text\n# เขียนไฟล์ข้อมูล conll2002\ndef write_conll2002(file_name,data):\n \"\"\"\n ใช้สำหรับเขียนไฟล์\n \"\"\"\n with codecs.open(file_name, \"w\", \"utf-8-sig\") as temp:\n temp.write(data)\n return True\n# อ่านข้อมูลจากไฟล์\ndef get_data(fileopen):\n\t\"\"\"\n สำหรับใช้อ่านทั้งหมดทั้งในไฟล์ทีละรรทัดออกมาเป็น list\n \"\"\"\n\twith codecs.open(fileopen, 'r',encoding='utf-8-sig') as f:\n\t\tlines = f.read().splitlines()\n\treturn [a for a in tqdm(lines) if Unique(a)] # เอาไม่ซ้ำกัน\n\ndef alldata(lists):\n text=\"\"\n for data in lists:\n text+=text2conll2002(data)\n text+='\\n'\n return text\n\ndef alldata_list(lists):\n data_all=[]\n for data in lists:\n data_num=[]\n try:\n txt=text2conll2002(data,pos=True).split('\\n')\n for d in txt:\n tt=d.split('\\t')\n if d!=\"\":\n if len(tt)==3:\n data_num.append((tt[0],tt[1],tt[2]))\n else:\n data_num.append((tt[0],tt[1]))\n #print(data_num)\n data_all.append(data_num)\n except:\n print(data)\n #print(data_all)\n return data_all\n\ndef alldata_list_str(lists):\n\tstring=\"\"\n\tfor data in lists:\n\t\tstring1=\"\"\n\t\tfor j in data:\n\t\t\tstring1+=j[0]+\"\t\"+j[1]+\"\t\"+j[2]+\"\\n\"\n\t\tstring1+=\"\\n\"\n\t\tstring+=string1\n\treturn string\n\ndef get_data_tag(listd):\n\tlist_all=[]\n\tc=[]\n\tfor i in listd:\n\t\tif i !='':\n\t\t\tc.append((i.split(\"\\t\")[0],i.split(\"\\t\")[1],i.split(\"\\t\")[2]))\n\t\telse:\n\t\t\tlist_all.append(c)\n\t\t\tc=[]\n\treturn list_all\ndef getall(lista):\n ll=[]\n for i in tqdm(lista):\n o=True\n for j in ll:\n if re.sub(\"\\[(.*?)\\]\",\"\",i)==re.sub(\"\\[(.*?)\\]\",\"\",j):\n o=False\n break\n if o==True:\n ll.append(i)\n return ll", "_____no_output_____" ], [ "data1=getall(get_data(file_name+\".txt\"))\nprint(len(data1))\n'''\n'''\n#del datatofile[0]\ndatatofile=alldata_list(data1)\ntt=[]\n#datatofile.reverse()\nimport random\n#random.shuffle(datatofile)\nprint(len(datatofile))\n#training_samples = datatofile[:int(len(datatofile) * 0.8)]\n#test_samples = datatofile[int(len(datatofile) * 0.8):]\n'''training_samples = datatofile[:2822]\ntest_samples = datatofile[2822:]'''\n#print(test_samples[0])\n#tag=TrainChunker(training_samples,test_samples) # Train\n\n#run(training_samples,test_samples)", "100%|██████████| 6457/6457 [00:00<00:00, 12023.70it/s]\n100%|██████████| 6349/6349 [03:18<00:00, 20.21it/s] \n" ], [ "#import dill\n#with open('train.data', 'rb') as file:\n# datatofile = dill.load(file)", "_____no_output_____" ], [ "with open(file_name+\"-pos.conll\",\"w\") as f:\n i=0\n while i<len(datatofile):\n for j in datatofile[i]:\n f.write(j[0]+\"\\t\"+j[1]+\"\\t\"+j[2]+\"\\n\")\n if i+1<len(datatofile):\n f.write(\"\\n\")\n i+=1\n\nwith open(file_name+\".conll\",\"w\") as f:\n i=0\n while i<len(datatofile):\n for j in datatofile[i]:\n f.write(j[0]+\"\\t\"+j[2]+\"\\n\")\n if i+1<len(datatofile):\n f.write(\"\\n\")\n i+=1", "_____no_output_____" ], [ "def isThai(chr):\n cVal = ord(chr)\n if(cVal >= 3584 and cVal <= 3711):\n return True\n return False\ndef isThaiWord(word):\n t=True\n for i in word:\n l=isThai(i)\n if l!=True and i!='.':\n t=False\n break\n return t\n\ndef is_stopword(word):\n return word in stopwords\ndef is_s(word):\n if word == \" \" or word ==\"\\t\" or word==\"\":\n return True\n else:\n return False\n\ndef lennum(word,num):\n if len(word)==num:\n return True\n return False\ndef doc2features(doc, i):\n word = doc[i][0]\n postag = doc[i][1]\n # Features from current word\n features={\n 'word.word': word,\n 'word.stopword': is_stopword(word),\n 'word.isthai':isThaiWord(word),\n 'word.isspace':word.isspace(),\n 'postag':postag,\n 'word.isdigit()': word.isdigit()\n }\n if word.isdigit() and len(word)==5:\n features['word.islen5']=True\n if i > 0:\n prevword = doc[i-1][0]\n postag1 = doc[i-1][1]\n features['word.prevword'] = prevword\n features['word.previsspace']=prevword.isspace()\n features['word.previsthai']=isThaiWord(prevword)\n features['word.prevstopword']=is_stopword(prevword)\n features['word.prepostag'] = postag1\n features['word.prevwordisdigit'] = prevword.isdigit()\n else:\n features['BOS'] = True # Special \"Beginning of Sequence\" tag\n # Features from next word\n if i < len(doc)-1:\n nextword = doc[i+1][0]\n postag1 = doc[i+1][1]\n features['word.nextword'] = nextword\n features['word.nextisspace']=nextword.isspace()\n features['word.nextpostag'] = postag1\n features['word.nextisthai']=isThaiWord(nextword)\n features['word.nextstopword']=is_stopword(nextword)\n features['word.nextwordisdigit'] = nextword.isdigit()\n else:\n features['EOS'] = True # Special \"End of Sequence\" tag\n return features\n\ndef extract_features(doc):\n return [doc2features(doc, i) for i in range(len(doc))]\n\ndef get_labels(doc):\n return [tag for (token,postag,tag) in doc]", "_____no_output_____" ], [ "X_data = [extract_features(doc) for doc in tqdm(datatofile)]\ny_data = [get_labels(doc) for doc in tqdm(datatofile)]", "100%|██████████| 6349/6349 [00:11<00:00, 549.11it/s]\n100%|██████████| 6349/6349 [00:00<00:00, 214278.19it/s]\n" ], [ "X, X_test, y, y_test = train_test_split(X_data, y_data, test_size=0.2)", "/home/wannaphong/anaconda3/lib/python3.7/site-packages/sklearn/metrics/classification.py:1437: UndefinedMetricWarning: F-score is ill-defined and being set to 0.0 in labels with no predicted samples.\n 'precision', 'predicted', average, warn_for)\n/home/wannaphong/anaconda3/lib/python3.7/site-packages/sklearn/metrics/classification.py:1439: UndefinedMetricWarning: F-score is ill-defined and being set to 0.0 in labels with no true samples.\n 'recall', 'true', average, warn_for)\n" ], [ "crf = sklearn_crfsuite.CRF(\n algorithm='lbfgs',\n c1=0.1,\n c2=0.1,\n max_iterations=500,\n all_possible_transitions=True,\n model_filename=file_name+\"-pos.model0\"\n)\ncrf.fit(X, y);\n\nlabels = list(crf.classes_)\nlabels.remove('O')\ny_pred = crf.predict(X_test)\ne=metrics.flat_f1_score(y_test, y_pred,\n average='weighted', labels=labels)\nprint(e)\nsorted_labels = sorted(\n labels,\n key=lambda name: (name[1:], name[0])\n)\nprint(metrics.flat_classification_report(\n y_test, y_pred, labels=sorted_labels, digits=3\n))", "_____no_output_____" ], [ "#del X_data[0]\n#del y_data[0]", "_____no_output_____" ], [ "!export PYTHONIOENCODING=utf-8", "_____no_output_____" ], [ "import sklearn_crfsuite\ncrf2 = sklearn_crfsuite.CRF(\n algorithm='lbfgs',\n c1=0.1,\n c2=0.1,\n max_iterations=500,\n all_possible_transitions=True,\n model_filename=file_name+\".model\"\n)\ncrf2.fit(X_data, y_data);", "_____no_output_____" ], [ "import dill\nwith open(\"train.data\", \"wb\") as dill_file:\n dill.dump(datatofile, dill_file)", "_____no_output_____" ], [ "# cross_validate\n\"\"\"\nimport dill\nwith open(\"datatrain.data\", \"wb\") as dill_file:\n dill.dump(datatofile, dill_file)\nf1_scorer = make_scorer(metrics.flat_f1_score, average='macro') \n\nscores = cross_validate(crf, X, y, scoring=f1_scorer, cv=5)\n# save data\nprint(scores)\n\"\"\"", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "[1, 2]*3", "_____no_output_____" ], [ "import random\na = random.randint(1, 7)\nb = random.random(1, 7)", "_____no_output_____" ], [ "print(a)", "1\n" ], [ "print(b)", "0.39243641919448247\n" ], [ "v = 1 + 1\nt = \"apa\" * 3\nz = False + True", "_____no_output_____" ], [ "v", "_____no_output_____" ], [ "t", "_____no_output_____" ], [ "z", "_____no_output_____" ], [ "x = 10\nif (x/2 >5):\n print(\"yes\")", "_____no_output_____" ], [ "sum = 0\nnumber = 1\nwhile sum < 10000:\n sum = sum + number\n print(sum)", "1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n12\n13\n14\n15\n16\n17\n18\n19\n20\n21\n22\n23\n24\n25\n26\n27\n28\n29\n30\n31\n32\n33\n34\n35\n36\n37\n38\n39\n40\n41\n42\n43\n44\n45\n46\n47\n48\n49\n50\n51\n52\n53\n54\n55\n56\n57\n58\n59\n60\n61\n62\n63\n64\n65\n66\n67\n68\n69\n70\n71\n72\n73\n74\n75\n76\n77\n78\n79\n80\n81\n82\n83\n84\n85\n86\n87\n88\n89\n90\n91\n92\n93\n94\n95\n96\n97\n98\n99\n100\n101\n102\n103\n104\n105\n106\n107\n108\n109\n110\n111\n112\n113\n114\n115\n116\n117\n118\n119\n120\n121\n122\n123\n124\n125\n126\n127\n128\n129\n130\n131\n132\n133\n134\n135\n136\n137\n138\n139\n140\n141\n142\n143\n144\n145\n146\n147\n148\n149\n150\n151\n152\n153\n154\n155\n156\n157\n158\n159\n160\n161\n162\n163\n164\n165\n166\n167\n168\n169\n170\n171\n172\n173\n174\n175\n176\n177\n178\n179\n180\n181\n182\n183\n184\n185\n186\n187\n188\n189\n190\n191\n192\n193\n194\n195\n196\n197\n198\n199\n200\n201\n202\n203\n204\n205\n206\n207\n208\n209\n210\n211\n212\n213\n214\n215\n216\n217\n218\n219\n220\n221\n222\n223\n224\n225\n226\n227\n228\n229\n230\n231\n232\n233\n234\n235\n236\n237\n238\n239\n240\n241\n242\n243\n244\n245\n246\n247\n248\n249\n250\n251\n252\n253\n254\n255\n256\n257\n258\n259\n260\n261\n262\n263\n264\n265\n266\n267\n268\n269\n270\n271\n272\n273\n274\n275\n276\n277\n278\n279\n280\n281\n282\n283\n284\n285\n286\n287\n288\n289\n290\n291\n292\n293\n294\n295\n296\n297\n298\n299\n300\n301\n302\n303\n304\n305\n306\n307\n308\n309\n310\n311\n312\n313\n314\n315\n316\n317\n318\n319\n320\n321\n322\n323\n324\n325\n326\n327\n328\n329\n330\n331\n332\n333\n334\n335\n336\n337\n338\n339\n340\n341\n342\n343\n344\n345\n346\n347\n348\n349\n350\n351\n352\n353\n354\n355\n356\n357\n358\n359\n360\n361\n362\n363\n364\n365\n366\n367\n368\n369\n370\n371\n372\n373\n374\n375\n376\n377\n378\n379\n380\n381\n382\n383\n384\n385\n386\n387\n388\n389\n390\n391\n392\n393\n394\n395\n396\n397\n398\n399\n400\n401\n402\n403\n404\n405\n406\n407\n408\n409\n410\n411\n412\n413\n414\n415\n416\n417\n418\n419\n420\n421\n422\n423\n424\n425\n426\n427\n428\n429\n430\n431\n432\n433\n434\n435\n436\n437\n438\n439\n440\n441\n442\n443\n444\n445\n446\n447\n448\n449\n450\n451\n452\n453\n454\n455\n456\n457\n458\n459\n460\n461\n462\n463\n464\n465\n466\n467\n468\n469\n470\n471\n472\n473\n474\n475\n476\n477\n478\n479\n480\n481\n482\n483\n484\n485\n486\n487\n488\n489\n490\n491\n492\n493\n494\n495\n496\n497\n498\n499\n500\n501\n502\n503\n504\n505\n506\n507\n508\n509\n510\n511\n512\n513\n514\n515\n516\n517\n518\n519\n520\n521\n522\n523\n524\n525\n526\n527\n528\n529\n530\n531\n532\n533\n534\n535\n536\n537\n538\n539\n540\n541\n542\n543\n544\n545\n546\n547\n548\n549\n550\n551\n552\n553\n554\n555\n556\n557\n558\n559\n560\n561\n562\n563\n564\n565\n566\n567\n568\n569\n570\n571\n572\n573\n574\n575\n576\n577\n578\n579\n580\n581\n582\n583\n584\n585\n586\n587\n588\n589\n590\n591\n592\n593\n594\n595\n596\n597\n598\n599\n600\n601\n602\n603\n604\n605\n606\n607\n608\n609\n610\n611\n612\n613\n614\n615\n616\n617\n618\n619\n620\n621\n622\n623\n624\n625\n626\n627\n628\n629\n630\n631\n632\n633\n634\n635\n636\n637\n638\n639\n640\n641\n642\n643\n644\n645\n646\n647\n648\n649\n650\n651\n652\n653\n654\n655\n656\n657\n658\n659\n660\n661\n662\n663\n664\n665\n666\n667\n668\n669\n670\n671\n672\n673\n674\n675\n676\n677\n678\n679\n680\n681\n682\n683\n684\n685\n686\n687\n688\n689\n690\n691\n692\n693\n694\n695\n696\n697\n698\n699\n700\n701\n702\n703\n704\n705\n706\n707\n708\n709\n710\n711\n712\n713\n714\n715\n716\n717\n718\n719\n720\n721\n722\n723\n724\n725\n726\n727\n728\n729\n730\n731\n732\n733\n734\n735\n736\n737\n738\n739\n740\n741\n742\n743\n744\n745\n746\n747\n748\n749\n750\n751\n752\n753\n754\n755\n756\n757\n758\n759\n760\n761\n762\n763\n764\n765\n766\n767\n768\n769\n770\n771\n772\n773\n774\n775\n776\n777\n778\n779\n780\n781\n782\n783\n784\n785\n786\n787\n788\n789\n790\n791\n792\n793\n794\n795\n796\n797\n798\n799\n800\n801\n802\n803\n804\n805\n806\n807\n808\n809\n810\n811\n812\n813\n814\n815\n816\n817\n818\n819\n820\n821\n822\n823\n824\n825\n826\n827\n828\n829\n830\n831\n832\n833\n834\n835\n836\n837\n838\n839\n840\n841\n842\n843\n844\n845\n846\n847\n848\n849\n850\n851\n852\n853\n854\n855\n856\n857\n858\n859\n860\n861\n862\n863\n864\n865\n866\n867\n868\n869\n870\n871\n872\n873\n874\n875\n876\n877\n878\n879\n880\n881\n882\n883\n884\n885\n886\n887\n888\n889\n890\n891\n892\n893\n894\n895\n896\n897\n898\n899\n900\n901\n902\n903\n904\n905\n906\n907\n908\n909\n910\n911\n912\n913\n914\n915\n916\n917\n918\n919\n920\n921\n922\n923\n924\n925\n926\n927\n928\n929\n930\n931\n932\n933\n934\n935\n936\n937\n938\n939\n940\n941\n942\n943\n944\n945\n946\n947\n948\n949\n950\n951\n952\n953\n954\n955\n956\n957\n958\n959\n960\n961\n962\n963\n964\n965\n966\n967\n968\n969\n970\n971\n972\n973\n974\n975\n976\n977\n978\n979\n980\n981\n982\n983\n984\n985\n986\n987\n988\n989\n990\n991\n992\n993\n994\n995\n996\n997\n998\n999\n1000\n1001\n1002\n1003\n1004\n1005\n1006\n1007\n1008\n1009\n1010\n1011\n1012\n1013\n1014\n1015\n1016\n1017\n1018\n1019\n1020\n1021\n1022\n1023\n1024\n1025\n1026\n1027\n1028\n1029\n1030\n1031\n1032\n1033\n1034\n1035\n1036\n1037\n1038\n1039\n1040\n1041\n1042\n1043\n1044\n1045\n1046\n1047\n1048\n1049\n1050\n1051\n1052\n1053\n1054\n1055\n1056\n1057\n1058\n1059\n1060\n1061\n1062\n1063\n1064\n1065\n1066\n1067\n1068\n1069\n1070\n1071\n1072\n1073\n1074\n1075\n1076\n1077\n1078\n1079\n1080\n1081\n1082\n1083\n1084\n1085\n1086\n1087\n1088\n1089\n1090\n1091\n1092\n1093\n1094\n1095\n1096\n1097\n1098\n1099\n1100\n1101\n1102\n1103\n1104\n1105\n1106\n1107\n1108\n1109\n1110\n1111\n1112\n1113\n1114\n1115\n1116\n1117\n1118\n1119\n1120\n1121\n1122\n1123\n1124\n1125\n1126\n1127\n1128\n1129\n1130\n1131\n1132\n1133\n1134\n1135\n1136\n1137\n1138\n1139\n1140\n1141\n1142\n1143\n1144\n1145\n1146\n1147\n1148\n1149\n1150\n1151\n1152\n1153\n1154\n1155\n1156\n1157\n1158\n1159\n1160\n1161\n1162\n1163\n1164\n1165\n1166\n1167\n1168\n1169\n1170\n1171\n1172\n1173\n1174\n1175\n1176\n1177\n1178\n1179\n1180\n1181\n1182\n1183\n1184\n1185\n1186\n1187\n1188\n1189\n1190\n1191\n1192\n1193\n1194\n1195\n1196\n1197\n1198\n1199\n1200\n1201\n1202\n1203\n1204\n1205\n1206\n1207\n1208\n1209\n1210\n1211\n1212\n1213\n1214\n1215\n1216\n1217\n1218\n1219\n1220\n1221\n1222\n1223\n1224\n1225\n1226\n1227\n1228\n1229\n1230\n1231\n1232\n1233\n1234\n1235\n1236\n1237\n1238\n1239\n1240\n1241\n1242\n1243\n1244\n1245\n1246\n1247\n1248\n1249\n1250\n1251\n1252\n1253\n1254\n1255\n1256\n1257\n1258\n1259\n1260\n1261\n1262\n1263\n1264\n1265\n1266\n1267\n1268\n1269\n1270\n1271\n1272\n1273\n1274\n1275\n1276\n1277\n1278\n1279\n1280\n1281\n1282\n1283\n1284\n1285\n1286\n1287\n1288\n1289\n1290\n1291\n1292\n1293\n1294\n1295\n1296\n1297\n1298\n1299\n1300\n1301\n1302\n1303\n1304\n1305\n1306\n1307\n1308\n1309\n1310\n1311\n1312\n1313\n1314\n1315\n1316\n1317\n1318\n1319\n1320\n1321\n1322\n1323\n1324\n1325\n1326\n1327\n1328\n1329\n1330\n1331\n1332\n1333\n1334\n1335\n1336\n1337\n1338\n1339\n1340\n1341\n1342\n1343\n1344\n1345\n1346\n1347\n1348\n1349\n1350\n1351\n1352\n1353\n1354\n1355\n1356\n1357\n1358\n1359\n1360\n1361\n1362\n1363\n1364\n1365\n1366\n1367\n1368\n1369\n1370\n1371\n1372\n1373\n1374\n1375\n1376\n1377\n1378\n1379\n1380\n1381\n1382\n1383\n1384\n1385\n1386\n1387\n1388\n1389\n1390\n1391\n1392\n1393\n1394\n1395\n1396\n1397\n1398\n1399\n1400\n1401\n1402\n1403\n1404\n1405\n1406\n1407\n1408\n1409\n1410\n1411\n1412\n1413\n1414\n1415\n1416\n1417\n1418\n1419\n1420\n1421\n1422\n1423\n1424\n1425\n1426\n1427\n1428\n1429\n1430\n1431\n1432\n1433\n1434\n1435\n1436\n1437\n1438\n1439\n1440\n1441\n1442\n1443\n1444\n1445\n1446\n1447\n1448\n1449\n1450\n1451\n1452\n1453\n1454\n1455\n1456\n1457\n1458\n1459\n1460\n1461\n1462\n1463\n1464\n1465\n1466\n1467\n1468\n1469\n1470\n1471\n1472\n1473\n1474\n1475\n1476\n1477\n1478\n1479\n1480\n1481\n1482\n1483\n1484\n1485\n1486\n1487\n1488\n1489\n1490\n1491\n1492\n1493\n1494\n1495\n1496\n1497\n1498\n1499\n1500\n1501\n1502\n1503\n1504\n1505\n1506\n1507\n1508\n1509\n1510\n1511\n1512\n1513\n1514\n1515\n1516\n1517\n1518\n1519\n1520\n1521\n1522\n1523\n1524\n1525\n1526\n1527\n1528\n1529\n1530\n1531\n1532\n1533\n1534\n1535\n1536\n1537\n1538\n1539\n1540\n1541\n1542\n1543\n1544\n1545\n1546\n1547\n1548\n1549\n1550\n1551\n1552\n1553\n1554\n1555\n1556\n1557\n1558\n1559\n1560\n1561\n1562\n1563\n1564\n1565\n1566\n1567\n1568\n1569\n1570\n1571\n1572\n1573\n1574\n1575\n1576\n1577\n1578\n1579\n1580\n1581\n1582\n1583\n1584\n1585\n1586\n1587\n1588\n1589\n1590\n1591\n1592\n1593\n1594\n1595\n1596\n1597\n1598\n1599\n1600\n1601\n1602\n1603\n1604\n1605\n1606\n1607\n1608\n1609\n1610\n1611\n1612\n1613\n1614\n1615\n1616\n1617\n1618\n1619\n1620\n1621\n1622\n1623\n1624\n1625\n1626\n1627\n1628\n1629\n1630\n1631\n1632\n1633\n1634\n1635\n1636\n1637\n1638\n1639\n1640\n1641\n1642\n1643\n1644\n1645\n1646\n1647\n1648\n1649\n1650\n1651\n1652\n1653\n1654\n1655\n1656\n1657\n1658\n1659\n1660\n1661\n1662\n1663\n1664\n1665\n1666\n1667\n1668\n1669\n1670\n1671\n1672\n1673\n1674\n1675\n1676\n1677\n1678\n1679\n1680\n1681\n1682\n1683\n1684\n1685\n1686\n1687\n1688\n1689\n1690\n1691\n1692\n1693\n1694\n1695\n1696\n1697\n1698\n1699\n1700\n1701\n1702\n1703\n1704\n1705\n1706\n1707\n1708\n1709\n1710\n1711\n1712\n1713\n1714\n1715\n1716\n1717\n1718\n1719\n1720\n1721\n1722\n1723\n1724\n1725\n1726\n1727\n1728\n1729\n1730\n1731\n1732\n1733\n1734\n1735\n1736\n1737\n1738\n1739\n1740\n1741\n1742\n1743\n1744\n1745\n1746\n1747\n1748\n1749\n1750\n1751\n1752\n1753\n1754\n1755\n1756\n1757\n1758\n1759\n1760\n1761\n1762\n1763\n1764\n1765\n1766\n1767\n1768\n1769\n1770\n1771\n1772\n1773\n1774\n1775\n1776\n1777\n1778\n1779\n1780\n1781\n1782\n1783\n1784\n1785\n1786\n1787\n1788\n1789\n1790\n1791\n1792\n1793\n1794\n1795\n1796\n1797\n1798\n1799\n1800\n1801\n1802\n1803\n1804\n1805\n1806\n1807\n1808\n1809\n1810\n1811\n1812\n1813\n1814\n1815\n1816\n1817\n1818\n1819\n1820\n1821\n1822\n1823\n1824\n1825\n1826\n1827\n1828\n1829\n1830\n1831\n1832\n1833\n1834\n1835\n1836\n1837\n1838\n1839\n1840\n1841\n1842\n1843\n1844\n1845\n1846\n1847\n1848\n1849\n1850\n1851\n1852\n1853\n1854\n1855\n1856\n1857\n1858\n1859\n1860\n1861\n1862\n1863\n1864\n1865\n1866\n1867\n1868\n1869\n1870\n1871\n1872\n1873\n1874\n1875\n1876\n1877\n1878\n1879\n1880\n1881\n1882\n1883\n1884\n1885\n1886\n1887\n1888\n1889\n1890\n1891\n" ], [ "for i in range(10):\n if i%3:\n print(i)\n ", "1\n2\n4\n5\n7\n8\n" ], [ "def AddNumbers(a, b):\n a = a+b\n return a", "_____no_output_____" ], [ "a = 3\nb = 7\n\nprint(AddNumbers(a,b), a)", "10 3\n" ], [ "1/20 * 3/5", "_____no_output_____" ], [ "(7/10 - 1/20)*3/5", "_____no_output_____" ], [ "8/400", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Introduction\nThis notebook can be used to both train a discriminator on the AG news dataset and steering text generation in the direction of each of the four classes of this dataset, namely world, sports, business and sci/tech. \n\nMy code uses and builds on a text generation plug-and-play model developed by the Uber AI research team, which can be found here: https://github.com/uber-research/PPLM. \nI had a lot of problems setting up the code provided by the original paper, since the package versions is old and has a lot of incompatability issues. I spent loads of times trying to set up pip and/or anaconda environments to run their code, but there was always an issue. \nTherefore, I developed this in Google Colab, which seems to be the only place where I can run their code without problems. **I strongly recommend you running this in Google Colab as well**. Thus, my code is kind of hard to use exactly because the original PPLM code is hard to use. I forked the PPLM repo and removed lots of unecessary stuff, only keeping the parts I'm using in this notebook. Also, I added my newly trained discriminator model. \n \nBy running this entire notebook cell for cell, you both train the discriminator and performs the generation experiment. However, since I've already trained this very discriminator, you can skip those cells. You can also skip the cells corresponding to saving models and results to disk. I've marked the \"mandatory\" cells with the comment \"# MUST BE RUN\" for this purpose.\n\n## Main functionality\nThis notebook essentially just runs my experiment setup using the newly trained discriminator to steer text generation in the direction of the discriminator classes text. \n \nThe main function is named *text_generation*, which can be used to generate a user-chosen amount of samples using either an unperturbed model or perturbed model. In the latter case, the user might choose which class he/she wished to steer text generation towards. I should also say that it's not quite new functionality, it's based on some of the PPLM code modified to suit my experiment.\n\n## Termology used throughout:\n- Model setting: using a general language model (GPT-2) together with the discriminator fixed on optimizing for one specific class.\n- Perturbed and unperturbed: This is essentially whether a discriminator has been used in the text generation. For instance, unperturbated text is \"clean\", meaning unsteered, while perturbated text is steered in a class direction.", "_____no_output_____" ], [ "# Setup: import the code base from github, and install requirements", "_____no_output_____" ] ], [ [ "# MUST BE RUN\n!git clone https://github.com/eskilhamre/PPLM.git", "Cloning into 'PPLM'...\nremote: Enumerating objects: 297, done.\u001b[K\nremote: Counting objects: 100% (47/47), done.\u001b[K\nremote: Compressing objects: 100% (47/47), done.\u001b[K\nremote: Total 297 (delta 20), reused 14 (delta 0), pack-reused 250\u001b[K\nReceiving objects: 100% (297/297), 2.46 MiB | 19.66 MiB/s, done.\nResolving deltas: 100% (125/125), done.\n" ], [ "# MUST BE RUN\nimport os\nos.chdir('PPLM')", "_____no_output_____" ], [ "# MUST BE RUN\n!pip install -r requirements.txt", "Collecting torch==1.7.0\n Downloading torch-1.7.0-cp37-cp37m-manylinux1_x86_64.whl (776.7 MB)\n\u001b[K |████████████████████████████████| 776.7 MB 4.3 kB/s \n\u001b[?25hCollecting nltk==3.4.5\n Downloading nltk-3.4.5.zip (1.5 MB)\n\u001b[K |████████████████████████████████| 1.5 MB 32.4 MB/s \n\u001b[?25hCollecting colorama==0.4.4\n Downloading colorama-0.4.4-py2.py3-none-any.whl (16 kB)\nCollecting transformers==3.4.0\n Downloading transformers-3.4.0-py3-none-any.whl (1.3 MB)\n\u001b[K |████████████████████████████████| 1.3 MB 64.5 MB/s \n\u001b[?25hCollecting torchtext==0.3.1\n Downloading torchtext-0.3.1-py3-none-any.whl (62 kB)\n\u001b[K |████████████████████████████████| 62 kB 1.3 MB/s \n\u001b[?25hRequirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from -r requirements.txt (line 6)) (1.1.5)\nRequirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from torch==1.7.0->-r requirements.txt (line 1)) (1.19.5)\nRequirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch==1.7.0->-r requirements.txt (line 1)) (3.10.0.2)\nCollecting dataclasses\n Downloading dataclasses-0.6-py3-none-any.whl (14 kB)\nRequirement already satisfied: future in /usr/local/lib/python3.7/dist-packages (from torch==1.7.0->-r requirements.txt (line 1)) (0.16.0)\nRequirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from nltk==3.4.5->-r requirements.txt (line 2)) (1.15.0)\nCollecting sacremoses\n Downloading sacremoses-0.0.46-py3-none-any.whl (895 kB)\n\u001b[K |████████████████████████████████| 895 kB 77.6 MB/s \n\u001b[?25hRequirement already satisfied: packaging in /usr/local/lib/python3.7/dist-packages (from transformers==3.4.0->-r requirements.txt (line 4)) (21.2)\nRequirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers==3.4.0->-r requirements.txt (line 4)) (3.3.2)\nCollecting tokenizers==0.9.2\n Downloading tokenizers-0.9.2-cp37-cp37m-manylinux1_x86_64.whl (2.9 MB)\n\u001b[K |████████████████████████████████| 2.9 MB 51.0 MB/s \n\u001b[?25hRequirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers==3.4.0->-r requirements.txt (line 4)) (2.23.0)\nCollecting sentencepiece!=0.1.92\n Downloading sentencepiece-0.1.96-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB)\n\u001b[K |████████████████████████████████| 1.2 MB 61.0 MB/s \n\u001b[?25hRequirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.7/dist-packages (from transformers==3.4.0->-r requirements.txt (line 4)) (4.62.3)\nRequirement already satisfied: protobuf in /usr/local/lib/python3.7/dist-packages (from transformers==3.4.0->-r requirements.txt (line 4)) (3.17.3)\nRequirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.7/dist-packages (from transformers==3.4.0->-r requirements.txt (line 4)) (2019.12.20)\nRequirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas->-r requirements.txt (line 6)) (2.8.2)\nRequirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/dist-packages (from pandas->-r requirements.txt (line 6)) (2018.9)\nRequirement already satisfied: pyparsing<3,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging->transformers==3.4.0->-r requirements.txt (line 4)) (2.4.7)\nRequirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers==3.4.0->-r requirements.txt (line 4)) (2.10)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers==3.4.0->-r requirements.txt (line 4)) (1.24.3)\nRequirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->transformers==3.4.0->-r requirements.txt (line 4)) (2021.10.8)\nRequirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers==3.4.0->-r requirements.txt (line 4)) (3.0.4)\nRequirement already satisfied: joblib in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers==3.4.0->-r requirements.txt (line 4)) (1.1.0)\nRequirement already satisfied: click in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers==3.4.0->-r requirements.txt (line 4)) (7.1.2)\nBuilding wheels for collected packages: nltk\n Building wheel for nltk (setup.py) ... \u001b[?25l\u001b[?25hdone\n Created wheel for nltk: filename=nltk-3.4.5-py3-none-any.whl size=1449922 sha256=06f23b40b68c3b0c3f1a527c02b60b16d5edf0d9f578661b8d10e7464f40289e\n Stored in directory: /root/.cache/pip/wheels/48/8b/7f/473521e0c731c6566d631b281f323842bbda9bd819eb9a3ead\nSuccessfully built nltk\nInstalling collected packages: dataclasses, torch, tokenizers, sentencepiece, sacremoses, transformers, torchtext, nltk, colorama\n Attempting uninstall: torch\n Found existing installation: torch 1.10.0+cu111\n Uninstalling torch-1.10.0+cu111:\n Successfully uninstalled torch-1.10.0+cu111\n Attempting uninstall: torchtext\n Found existing installation: torchtext 0.11.0\n Uninstalling torchtext-0.11.0:\n Successfully uninstalled torchtext-0.11.0\n Attempting uninstall: nltk\n Found existing installation: nltk 3.2.5\n Uninstalling nltk-3.2.5:\n Successfully uninstalled nltk-3.2.5\n\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\ntorchvision 0.11.1+cu111 requires torch==1.10.0, but you have torch 1.7.0 which is incompatible.\u001b[0m\nSuccessfully installed colorama-0.4.4 dataclasses-0.6 nltk-3.4.5 sacremoses-0.0.46 sentencepiece-0.1.96 tokenizers-0.9.2 torch-1.7.0 torchtext-0.3.1 transformers-3.4.0\n" ] ], [ [ "# Train a discriminator on AG news dataset\n## First, download the general lanaguage model used to train the discriminator", "_____no_output_____" ] ], [ [ "from transformers.modeling_gpt2 import GPT2LMHeadModel\n# This downloads GPT-2 Medium, it takes a little while\n_ = GPT2LMHeadModel.from_pretrained(\"gpt2-medium\")", "_____no_output_____" ] ], [ [ "## Import the dataset\nThe data can be found at: https://www.kaggle.com/amananandrai/ag-news-classification-dataset/version/2?select=train.csv\n\nThe PPLM interface requires the data to be a tsv file containing the entire dataset, where the first column is the labels and seconds column the text. Thus, we have to prepare the dataset for this. \n\nFirst, download the dataset following the link above, and upload both files in the set below.", "_____no_output_____" ] ], [ [ "import pandas as pd\nfrom google.colab import files\nimport torch\ntorch.cuda.is_available()", "_____no_output_____" ], [ "uploaded = files.upload()", "_____no_output_____" ], [ "data_fp = \"./ag-news-data.tsv\" # where we want to store our prepared dataset", "_____no_output_____" ], [ "def prepare_dataset(text_index,\n label_index, \n label_names=None\n ):\n train = pd.read_csv(\"train.csv\")\n test = pd.read_csv(\"test.csv\")\n all_data = pd.concat([train, test])\n all_data = all_data.iloc[:, [label_index, text_index]]\n\n if label_names:\n labels_map = {i+1: label_name for i, label_name in enumerate(label_names)} # here assuming labels are numerated 1,...,n, which is the case for AG news\n all_data.iloc[:, 0] = all_data.iloc[:, 0].map(labels_map) # exchange label numbers by their name\n\n return all_data", "_____no_output_____" ], [ "idx2class = [\"world\", \"sports\", \"business\", \"sci/tech\"]\ndata = prepare_dataset(2, 0, idx2class)\ndata.to_csv(data_fp, sep='\\t', index=False, header=False)", "_____no_output_____" ], [ "from run_pplm_discrim_train import train_discriminator\n\n# ensure reproducible discriminator\ntorch.manual_seed(444)\nnp.random.seed(444)\n\ndiscriminator, disc_info = train_discriminator(\n dataset=\"generic\",\n dataset_fp=data_fp,\n pretrained_model=\"gpt2-medium\",\n epochs=8,\n learning_rate=0.0001,\n batch_size=128,\n log_interval=10,\n save_model=True,\n cached=False,\n no_cuda=False,\n output_fp='models/',\n idx2class=idx2class\n)", "Preprocessing generic dataset...\n" ] ], [ [ "We achieve about 90% accuracy on unseen data, which is pretty good in my opinion. I haven't studied the training accuracy (so I can't say the following for sure), but I don't think we're neither underfitting or overfitting here. This is good stuff! \nAlso, the validation/test accuracy seems to stagnate on ~90%, so more epochs would probably be in no use.", "_____no_output_____" ], [ "## Training the discriminator is done, let's download it", "_____no_output_____" ] ], [ [ "classifier_name = \"models/news_classifierhead.pt\"\ntorch.save(discriminator.get_classifier().state_dict(), \"models/news_classifierhead.pt\")\nfiles.download(classifier_name)", "_____no_output_____" ] ], [ [ "At this point, I put the newly generated model in the discrim_models/ folder, and updated my Github code to include this model. I went back to the beginning of the notebook and recloned the repo.", "_____no_output_____" ], [ "# Scoring the generated samples\nWhen doing manual comparison between text generated from different model settings, I'm only interested in comparing only the best sample for each model setting. The idea is to generate lots of samples using the same setting, and picking the best one based on some type of scoring. \nWhat do I mean by best samples? I'm automating this evaluation as means of scoring and ranking the sentences in a similar way as described in the PPLM paper; \n- fluency is measured by the general language model likelihood p(sentence). In scoring, I utilize the fact that the lower the language model loss(sentence), the higher the p(sentence). I use GPT-1 for this, as in the PPLM paper (in the paper however, they use GPT-1 to calculate perplexity, and as I understand it this should correspond to loss.)\n- diversity of words is measured by the mean of the (length normalized) Dist-1, Dist-2 and Dist-3 score, (the PPLM paper was inspired by the way they use this metric in this paper: https://arxiv.org/pdf/1510.03055.pdf)\n\n", "_____no_output_____" ] ], [ [ "# MUST BE RUN\nfrom transformers.modeling_gpt2 import GPT2LMHeadModel\nfrom transformers import GPT2Tokenizer, OpenAIGPTLMHeadModel, OpenAIGPTTokenizer\nfrom nltk import ngrams\nimport numpy as np", "_____no_output_____" ] ], [ [ "### Instantiate models used for scoring samples", "_____no_output_____" ] ], [ [ "# MUST BE RUN\ndevice = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\n# tokenizer and language model used for calculating fluency / \"perplexity\"\ngpt1_tokenizer = OpenAIGPTTokenizer.from_pretrained(\"openai-gpt\")\ngpt1_model = OpenAIGPTLMHeadModel.from_pretrained(\"openai-gpt\")\ngpt1_model.eval()\ngpt1_model.to(device)\ndevice", "_____no_output_____" ], [ "# MUST BE RUN\n\n##############\n# This is to be used on all generated sentences (to be aggregated wrt. model setting), not used for selection\n##############\n\ndef lm_score(sentence):\n \"\"\"\n Calculates the language model total loss of the sentence.\n Code heavily inspired from: https://github.com/huggingface/transformers/issues/1009\n The total loss is equivalent to\n - [ log P(x1 | <|endoftext|>) + log P(x2 | x1, <|endoftext|>) + ... ]\n which means it corresponds to perplexity, and can be used as such in comparisons.\n \"\"\"\n tokens = gpt1_tokenizer.encode(sentence)\n input_ids = torch.tensor(tokens).unsqueeze(0)\n input_ids = input_ids.to(device)\n\n with torch.no_grad():\n outputs = gpt1_model(input_ids, labels=input_ids)\n loss, logits = outputs[:2]\n return loss.item() # * len(tokens) don't multiply with length, this would prefer shorter sentences\n\n###############\n# This is used for selecting the best sample when model setting and prefix is fixed\n###############\n\ndef dist_n_score(sentence, n):\n \"\"\"Calculates the number of distinct n-grams in the sentence, normalized by the sentence length\"\"\"\n if len(sentence.split()) < n:\n raise ValueError(\"Cannot find ngram of sentence with less than n words\")\n \n sentence = sentence.lower().strip()\n dist_n_grams = set()\n for n_gram in ngrams(sentence.split(), n):\n dist_n_grams.add(n_gram)\n \n return len(dist_n_grams) / len(sentence.split())\n\n\ndef dist_score(sentence):\n \"\"\"\n Calculcates the dist-1, dist-2 and dist-3 score of the sentence, as well as their mean\n \"\"\"\n sentence = sentence.lower().strip()\n sentence = sentence.replace(\".\", \"\").replace(\",\", \"\").replace(\"\\n\", \"\")\n \n dist_scores = [dist_n_score(sentence, n) for n in range(1, 4)]\n dist_1, dist_2, dist_3 = dist_scores\n return np.mean(dist_scores), dist_1, dist_2, dist_3\n\n\nsentences =['there is a book on the desk', 'there is a plane on the desk', 'there is a book in the desk', \"desk desk desk desk cat cat\"]\nprint([lm_score(s) for s in sentences])\nprint([dist_score(s)[0] for s in sentences])", "[3.0594890117645264, 4.118381977081299, 3.2676448822021484, 9.095804214477539]\n[0.8571428571428572, 0.8571428571428572, 0.8571428571428572, 0.4444444444444444]\n" ] ], [ [ "We can see that the most sensible of the four sentences receives lowest language model score (thus higher probability). Also, we can see that the non-sensible sentence receives both bad language model and dist-score.", "_____no_output_____" ], [ "# Text generation", "_____no_output_____" ] ], [ [ "# MUST BE RUN\nfrom run_pplm import run_pplm_example, generate_text_pplm, get_classifier, PPLM_DISCRIM", "_____no_output_____" ], [ "# MUST BE RUN\n\ndef text_generation(\n model,\n tokenizer,\n discrim=\"news\",\n class_label=\"sports\",\n prefix_text=\"Last summer\",\n perturb=False,\n num_samples=3,\n device=\"cuda\",\n length=150,\n stepsize=0.04,\n num_iterations=10,\n window_length=0, # 0 corresponds to entire sequence\n gamma=1.0,\n gm_scale=0.95,\n kl_scale=0.01,\n verbosity_level=1 # REGULAR\n):\n \"\"\"\n Used to generate a user-specified number of samples, with optional use of the discriminator\n to perturbate the generated samples in the direction of it's gradient.\n\n This is a modified version of the PPML text generation function to suit my experiment.\n\n Only supports generating text using discriminator models (BoW models not supported)\n The default hyper parameters chosen here are the same as in the PPLM Colab demo, since\n this seems to work great for discriminators.\n\n Returns a list of generated text samples and their corresponding attribute model losses\n \"\"\"\n\n # we pass the discriminator even if we want unpertubated text, since it's used for attribute scoring\n discrim_model, class_id = get_classifier(\n discrim,\n class_label,\n device\n )\n\n # encode prefix text\n tokenized_cond_text = tokenizer.encode(\n tokenizer.bos_token + prefix_text,\n add_special_tokens=False\n )\n\n gen_text_samples = []\n discrim_losses = []\n\n if device == 'cuda':\n torch.cuda.empty_cache()\n\n for i in range(num_samples):\n gen_tok_text, discrim_loss, _ = generate_text_pplm(\n model=model,\n tokenizer=tokenizer,\n context=tokenized_cond_text,\n device=device,\n perturb=perturb,\n classifier=discrim_model,\n class_label=class_id,\n loss_type=PPLM_DISCRIM, # BoW not supported as of now\n length=length,\n stepsize=stepsize,\n sample=True,\n num_iterations=num_iterations,\n horizon_length=1,\n window_length=window_length,\n gamma=gamma,\n gm_scale=gm_scale,\n kl_scale=kl_scale,\n verbosity_level=verbosity_level\n )\n\n \n gen_text = tokenizer.decode(gen_tok_text[0][1:]) # decode generated text\n gen_text_samples.append(gen_text)\n discrim_losses.append(discrim_loss.item()) #.data.cpu().numpy())\n \n if device == \"cuda\":\n torch.cuda.empty_cache()\n\n return gen_text_samples, discrim_losses\n\n\ndef select_best(gen_text_samples, discrim_losses):\n \"\"\"\n Given the outout from the text_generation function, filters away 3/4 of the \n generated samples based on mean dist-score, and rank the remaining 1/4 based\n on discriminator losses.\n \n Returns the best sample based smallest discriminator loss (the one maximizing \n the attribute, according to the discriminator)\n \"\"\"\n if len(gen_text_samples) < 4:\n raise ValueError(\"Cannot filter away 3/4 of less than 4 samples\")\n \n n_keep = 1 * len(gen_text_samples) // 4 # number of samples to keep\n\n # filter out the 3/4 samples with lowest mean dist-score\n mean_dists = [dist_score(sample)[0] for sample in gen_text_samples]\n idx_to_keep = np.argpartition(mean_dists, -n_keep)[-n_keep:] # indices of samples with highest mean dist score\n samples = np.array([gen_text_samples, discrim_losses, mean_dists]).T\n filtered_samples = samples[idx_to_keep]\n \n # fetch best sample among the remaining ones\n best_idx = np.argmin(filtered_samples[:, 1]) # index of sample with minimal discrim loss\n best_sample, smallest_loss, mean_dist = filtered_samples[best_idx]\n return best_sample, smallest_loss, mean_dist", "_____no_output_____" ] ], [ [ "## Import the base model used to sample from", "_____no_output_____" ] ], [ [ "# MUST BE RUN\npretrained_model = \"gpt2-medium\"\n\nmodel = GPT2LMHeadModel.from_pretrained(\n pretrained_model,\n output_hidden_states=True\n)\nmodel.to(device)\nmodel.eval()\n\n# Freeze GPT-2 weigths\nfor param in model.parameters():\n param.requires_grad = False\n\ntokenizer = GPT2Tokenizer.from_pretrained(pretrained_model)", "_____no_output_____" ] ], [ [ "## Sample from all combinations of model setting and prefix sentences\nFirst, let's create some data structures to gather relevant information from the sampling process.", "_____no_output_____" ] ], [ [ "# MUST BE RUN\n\n# most relevant hyper params wrt. speed\ngenerated_len = 120\nnum_samples = 12\n\nprefixes = [\n \"Last week\",\n \"The potato\",\n \"Breaking news:\",\n \"In the last year\",\n \"The president of the country\",\n]\n\nmodel_settings = [\"world\", \"sports\", \"business\", \"sci/tech\"] # the classes of the discriminator", "_____no_output_____" ], [ "# MUST BE RUN\n\n# data structures of generated results\ngen_samples = {model_setting: [] for model_setting in [\"unpert\"] + model_settings.copy()} # contains all generated samples\ncomparisons = {model_setting: {prefix: dict() for prefix in prefixes} for model_setting in model_settings} # contains the best samples for each model setting and prefix combo", "_____no_output_____" ] ], [ [ "The cell below runs the *entire* sampling process, it took ~5 hours to run on Googles Compute Engine backend using their GPUs. \n \nHere I decided that generate unperturbed text for each model setting. This might seem silly and redundant, since the unperturbed text is not affected by this choice. \nAnd while that is partly true, I did this to be able to calculate the discriminator losses of the generated text, so that I can select the \"best\" sample wrt. to the classes (even though it's best by chance). I thought this is only fair: the perturbed model gets many chances to generate a \"good\" sample (in the eyes of the discriminator), so the unperturbed model should also have this. \nAlso, I didn't find a easy way of using the discriminator to just score the text sample wrt. a class right of the bat. This is partly due to the fact that discriminator is actually trained on the transformer output.", "_____no_output_____" ] ], [ [ "# MUST BE RUN\n\n# since we're sampling, set seed for reproducibility\ntorch.manual_seed(444)\nnp.random.seed(444)\n\nn_combinations = len(prefixes) * len(model_settings)\ni = 1\nfor prefix_sentence in prefixes:\n for j, model_setting in enumerate(model_settings):\n print(f\"\\n\\nRun {i:3d}/{n_combinations:3d} : optimizing for class: {model_setting}, with prefix: {prefix_sentence}\\n\")\n \n unpert_text_samples, unpert_discrim_losses = text_generation(\n model,\n tokenizer,\n device=device,\n length=generated_len,\n num_samples=num_samples,\n prefix_text=prefix_sentence,\n discrim=\"news\",\n class_label=model_setting,\n perturb=False\n )\n\n pert_text_samples, pert_discrim_losses = text_generation(\n model,\n tokenizer,\n device=device,\n length=generated_len,\n num_samples=num_samples,\n prefix_text=prefix_sentence,\n discrim=\"news\",\n class_label=model_setting,\n perturb=True\n )\n # store generated samples\n if j == 0:\n gen_samples[\"unpert\"].extend(unpert_text_samples) # only store unpertubated generation once per prefix\n gen_samples[model_setting].extend(pert_text_samples)\n\n # save the best sample, it's discriminator loss and mean dist-score for both the perturbated and unperturbated samples\n comparisons[model_setting][prefix_sentence][\"unpert\"] = list(select_best(unpert_text_samples, unpert_discrim_losses))\n comparisons[model_setting][prefix_sentence][\"pert\"] = list(select_best(pert_text_samples, pert_discrim_losses))\n\n i += 1\n", "\n\nRun 1/ 15 : optimizing for class: sports, with prefix: Last week\n\n<|endoftext|>Last week's\n<|endoftext|>Last week's announcement\n<|endoftext|>Last week's announcement that\n" ] ], [ [ "## Generation analysis\nFirst, let's download the generated samples.", "_____no_output_____" ] ], [ [ "import json\n\nwith open(\"all-samples.json\", \"w\") as fp:\n json.dump(gen_samples, fp)\n files.download(\"all-samples.json\")\n\nwith open(\"comparisons.json\", \"w\") as fp:\n json.dump(comparisons, fp)\n files.download(\"comparisons.json\"\")", "_____no_output_____" ] ], [ [ "## Let's extract the metrics from the generated samples\nIn the code below, for each model setting, I calculate the perplexity score and dist-1, dist-2, and dist-3 scores for all samples. I then accumulate the mean and standard deviations of the scores wrt. to each model setting, to study how well each model setting actually performed in the experiment above.", "_____no_output_____" ] ], [ [ "# MUST BE RUN\n\nmetrics_means_dict = {}\nmetrics_stds_dict = {}\n\nfor model_setting, samples in gen_samples.items():\n perplexities = [lm_score(sample) for sample in samples]\n dist_scores = [dist_score(sample)[1:] for sample in samples] # stored as (mean_dist_score, dist-1, dist-2, dist-3), ignore mean\n all_metrics = np.c_[np.array(perplexities), np.array(dist_scores)]\n metrics_means_dict[model_setting] = np.mean(all_metrics, axis=0)\n metrics_stds_dict[model_setting] = np.std(all_metrics, axis=0)\n\n# structure the statistics neatly dataframes\nmetrics_means_df = pd.DataFrame(data=metrics_means_dict, index=[\"perplexity\", \"dist-1\", \"dist-2\", \"dist-3\"])\nmetrics_means_df = pd.DataFrame(data=metrics_means_dict, index=[\"perplexity\", \"dist-1\", \"dist-2\", \"dist-3\"])", "_____no_output_____" ], [ "# save the extracted statistics as csv files\nmetrics_means_df.to_csv(\"metrics-means.csv\")\nmetrics_std_df.to_csv(\"metrics-std.csv\")\nfiles.download(\"metrics-means.csv\")\nfiles.download(\"metrics-std.csv)", "_____no_output_____" ] ], [ [ "## Let's see the best examples for each model setting and prefix", "_____no_output_____" ] ], [ [ "# MUST BE RUN\n\nfor model_setting, prefix_dict in comparisons.items():\n print(f\"Model setting: {model_setting}\\n\")\n for prefix_sentence in prefix_dict.keys():\n unpert_sample, unpert_loss, unpert_mean_dist = prefix_dict[prefix_sentence][\"unpert\"]\n pert_sample, pert_loss, pert_mean_dist = prefix_dict[prefix_sentence][\"pert\"]\n\n print(f\"Prefix is: {prefix_sentence}\\n\")\n print(f\"Unperturbated:\\nSample: {unpert_sample}\\nDiscrim loss: {unpert_loss:2.2f} | Mean dist-n score: {unpert_mean_dist:2.1f}\\n\")\n print(f\" Perturbated:\\nSample: {pert_sample}\\nDiscrim loss: {pert_loss:2.2f} | Mean dist-n score: {pert_mean_dist:2.1f}\")\n print(\"\\n\\n\")\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# 1002_Saure und alkalische Lösungen in unserer Umwelt\n\n**8h**", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%run ../Python_files/util_data_storage_and_load.py", "_____no_output_____" ], [ "%run ../Python_files/load_dicts.py", "_____no_output_____" ], [ "%run ../Python_files/util.py", "_____no_output_____" ], [ "import numpy as np\nfrom numpy.linalg import inv", "_____no_output_____" ], [ "# load link flow data\n\nimport json\n\nwith open('../temp_files/link_day_minute_Apr_dict_JSON_adjusted.json', 'r') as json_file:\n link_day_minute_Apr_dict_JSON = json.load(json_file)", "_____no_output_____" ], [ "# week_day_Apr_list = [2, 3, 4, 5, 6, 9, 10, 11, 12, 13, 16, 17, 18, 19, 20, 23, 24, 25, 26, 27, 30]\n\n# testing set 1\nweek_day_Apr_list_1 = [20, 23, 24, 25, 26, 27, 30]\n\n# testing set 2\nweek_day_Apr_list_2 = [11, 12, 13, 16, 17, 18, 19]\n\n# testing set 3\nweek_day_Apr_list_3 = [2, 3, 4, 5, 6, 9, 10]", "_____no_output_____" ], [ "link_flow_testing_set_Apr_MD_1 = []\nfor link_idx in range(24):\n for day in week_day_Apr_list_1: \n key = 'link_' + str(link_idx) + '_' + str(day)\n link_flow_testing_set_Apr_MD_1.append(link_day_minute_Apr_dict_JSON[key] ['MD_flow'])\n \nlink_flow_testing_set_Apr_MD_2 = []\nfor link_idx in range(24):\n for day in week_day_Apr_list_2: \n key = 'link_' + str(link_idx) + '_' + str(day)\n link_flow_testing_set_Apr_MD_2.append(link_day_minute_Apr_dict_JSON[key] ['MD_flow'])\n \nlink_flow_testing_set_Apr_MD_3 = []\nfor link_idx in range(24):\n for day in week_day_Apr_list_3: \n key = 'link_' + str(link_idx) + '_' + str(day)\n link_flow_testing_set_Apr_MD_3.append(link_day_minute_Apr_dict_JSON[key] ['MD_flow'])", "_____no_output_____" ], [ "testing_set_1 = np.matrix(link_flow_testing_set_Apr_MD_1)\ntesting_set_1 = np.matrix.reshape(testing_set_1, 24, 7)\ntesting_set_1 = np.nan_to_num(testing_set_1)\ny = np.array(np.transpose(testing_set_1))\ny = y[np.all(y != 0, axis=1)]\ntesting_set_1 = np.transpose(y)\ntesting_set_1 = np.matrix(testing_set_1)\n\ntesting_set_2 = np.matrix(link_flow_testing_set_Apr_MD_2)\ntesting_set_2 = np.matrix.reshape(testing_set_2, 24, 7)\ntesting_set_2 = np.nan_to_num(testing_set_2)\ny = np.array(np.transpose(testing_set_2))\ny = y[np.all(y != 0, axis=1)]\ntesting_set_2 = np.transpose(y)\ntesting_set_2 = np.matrix(testing_set_2)\n\ntesting_set_3 = np.matrix(link_flow_testing_set_Apr_MD_3)\ntesting_set_3 = np.matrix.reshape(testing_set_3, 24, 7)\ntesting_set_3 = np.nan_to_num(testing_set_3)\ny = np.array(np.transpose(testing_set_3))\ny = y[np.all(y != 0, axis=1)]\ntesting_set_3 = np.transpose(y)\ntesting_set_3 = np.matrix(testing_set_3)", "_____no_output_____" ], [ "np.size(testing_set_2, 0), np.size(testing_set_3, 1)", "_____no_output_____" ], [ "testing_set_3[:,:1]", "_____no_output_____" ], [ "# write testing sets to file\n\nzdump([testing_set_1, testing_set_2, testing_set_3], '../temp_files/testing_sets_Apr_MD.pkz')", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "Florida updates their data daily at 11am and 6pm EDT", "_____no_output_____" ] ], [ [ "from selenium import webdriver\nimport time\nimport pandas as pd\nimport pendulum\nimport re\nimport yaml\nfrom selenium.webdriver.chrome.options import Options\nchrome_options = Options()\n#chrome_options.add_argument(\"--disable-extensions\")\n#chrome_options.add_argument(\"--disable-gpu\")\n#chrome_options.add_argument(\"--no-sandbox) # linux only\nchrome_options.add_argument(\"--start-maximized\")\n# chrome_options.add_argument(\"--headless\")\nchrome_options.add_argument(\"user-agent=[Mozilla/5.0 (Macintosh; Intel Mac OS X 10.14; rv:73.0) Gecko/20100101 Firefox/73.0]\")", "_____no_output_____" ], [ "with open('config.yaml', 'r') as f:\n config = yaml.safe_load(f.read())", "_____no_output_____" ], [ "state = 'FL'", "_____no_output_____" ], [ "scrape_timestamp = pendulum.now().strftime('%Y%m%d%H%M%S')", "_____no_output_____" ], [ "# FL positive cases by county\nurl = 'https://arcg.is/0nHO11'", "_____no_output_____" ], [ "def fetch():\n driver = webdriver.Chrome('../20190611 - Parts recommendation/chromedriver', options=chrome_options)\n\n driver.get(url)\n time.sleep(5)\n\n # class topBoxH1Text is used in the summary box.\n datatbl = driver.find_element_by_id('ember219').find_elements_by_class_name('external-html')\n\n data = [re.search('^(.*) \\((\\d*)\\)*', row.text).groups() for row in datatbl]\n\n page_source = driver.page_source\n driver.close()\n\n return pd.DataFrame(data, columns=['county','positive_cases']), page_source", "_____no_output_____" ], [ "def save(df, source):\n df.to_csv(f\"{config['data_folder']}/{state}_county_{scrape_timestamp}.txt\", sep='|', index=False)\n\n with open(f\"{config['data_source_backup_folder']}/{state}_county_{scrape_timestamp}.html\", 'w') as f:\n f.write(source)", "_____no_output_____" ], [ "def run():\n df, source = fetch()\n save(df, source)", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/sapinspys/DS-Unit-2-Regression-Classification/blob/master/DS7_Sprint_Challenge_5_Regression_Classification.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "_Lambda School Data Science, Unit 2_\n \n# Regression & Classification Sprint Challenge\n\nTo demonstrate mastery on your Sprint Challenge, do all the required, numbered instructions in this notebook.\n\nTo earn a score of \"3\", also do all the stretch goals.\n\nYou are permitted and encouraged to do as much data exploration as you want.", "_____no_output_____" ], [ "### Part 1, Classification\n- 1.1. Begin with baselines for classification\n- 1.2. Do train/test split. Arrange data into X features matrix and y target vector\n- 1.3. Use scikit-learn to fit a logistic regression model\n- 1.4. Report classification metric: accuracy\n\n### Part 2, Regression\n- 2.1. Begin with baselines for regression\n- 2.2. Do train/validate/test split. \n- 2.3. Arrange data into X features matrix and y target vector\n- 2.4. Do one-hot encoding\n- 2.5. Use scikit-learn to fit a linear regression (or ridge regression) model\n- 2.6. Report validation MAE and $R^2$\n\n### Stretch Goals, Regression\n- Make visualizations to explore relationships between features and target\n- Try at least 3 feature combinations. You may select features manually, or automatically\n- Report validation MAE and $R^2$ for each feature combination you try\n- Report test MAE and $R^2$ for your final model\n- Print or plot the coefficients for the features in your model", "_____no_output_____" ] ], [ [ "# If you're in Colab...\nimport sys\nin_colab = 'google.colab' in sys.modules\n\nif in_colab:\n !pip install category_encoders==2.0.0\n !pip install pandas-profiling==2.3.0\n !pip install plotly==4.1.1", "Collecting category_encoders==2.0.0\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/6e/a1/f7a22f144f33be78afeb06bfa78478e8284a64263a3c09b1ef54e673841e/category_encoders-2.0.0-py2.py3-none-any.whl (87kB)\n\u001b[K |████████████████████████████████| 92kB 5.7MB/s \n\u001b[?25hRequirement already satisfied: statsmodels>=0.6.1 in /usr/local/lib/python3.6/dist-packages (from category_encoders==2.0.0) (0.10.1)\nRequirement already satisfied: scipy>=0.19.0 in /usr/local/lib/python3.6/dist-packages (from category_encoders==2.0.0) (1.3.1)\nRequirement already satisfied: patsy>=0.4.1 in /usr/local/lib/python3.6/dist-packages (from category_encoders==2.0.0) (0.5.1)\nRequirement already satisfied: numpy>=1.11.3 in /usr/local/lib/python3.6/dist-packages (from category_encoders==2.0.0) (1.16.5)\nRequirement already satisfied: scikit-learn>=0.20.0 in /usr/local/lib/python3.6/dist-packages (from category_encoders==2.0.0) (0.21.3)\nRequirement already satisfied: pandas>=0.21.1 in /usr/local/lib/python3.6/dist-packages (from category_encoders==2.0.0) (0.24.2)\nRequirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from patsy>=0.4.1->category_encoders==2.0.0) (1.12.0)\nRequirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from scikit-learn>=0.20.0->category_encoders==2.0.0) (0.13.2)\nRequirement already satisfied: python-dateutil>=2.5.0 in /usr/local/lib/python3.6/dist-packages (from pandas>=0.21.1->category_encoders==2.0.0) (2.5.3)\nRequirement already satisfied: pytz>=2011k in /usr/local/lib/python3.6/dist-packages (from pandas>=0.21.1->category_encoders==2.0.0) (2018.9)\nInstalling collected packages: category-encoders\nSuccessfully installed category-encoders-2.0.0\nCollecting pandas-profiling==2.3.0\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/2c/2f/aae19e2173c10a9bb7fee5f5cad35dbe53a393960fc91abc477dcc4661e8/pandas-profiling-2.3.0.tar.gz (127kB)\n\u001b[K |████████████████████████████████| 133kB 4.8MB/s \n\u001b[?25hRequirement already satisfied: pandas>=0.19 in /usr/local/lib/python3.6/dist-packages (from pandas-profiling==2.3.0) (0.24.2)\nRequirement already satisfied: matplotlib>=1.4 in /usr/local/lib/python3.6/dist-packages (from pandas-profiling==2.3.0) (3.0.3)\nRequirement already satisfied: jinja2>=2.8 in /usr/local/lib/python3.6/dist-packages (from pandas-profiling==2.3.0) (2.10.1)\nRequirement already satisfied: missingno>=0.4.2 in /usr/local/lib/python3.6/dist-packages (from pandas-profiling==2.3.0) (0.4.2)\nCollecting htmlmin>=0.1.12 (from pandas-profiling==2.3.0)\n Downloading https://files.pythonhosted.org/packages/b3/e7/fcd59e12169de19f0131ff2812077f964c6b960e7c09804d30a7bf2ab461/htmlmin-0.1.12.tar.gz\nCollecting phik>=0.9.8 (from pandas-profiling==2.3.0)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/45/ad/24a16fa4ba612fb96a3c4bb115a5b9741483f53b66d3d3afd987f20fa227/phik-0.9.8-py3-none-any.whl (606kB)\n\u001b[K |████████████████████████████████| 614kB 36.9MB/s \n\u001b[?25hCollecting confuse>=1.0.0 (from pandas-profiling==2.3.0)\n Downloading https://files.pythonhosted.org/packages/4c/6f/90e860cba937c174d8b3775729ccc6377eb91f52ad4eeb008e7252a3646d/confuse-1.0.0.tar.gz\nRequirement already satisfied: astropy in /usr/local/lib/python3.6/dist-packages (from pandas-profiling==2.3.0) (3.0.5)\nRequirement already satisfied: pytz>=2011k in /usr/local/lib/python3.6/dist-packages (from pandas>=0.19->pandas-profiling==2.3.0) (2018.9)\nRequirement already satisfied: numpy>=1.12.0 in /usr/local/lib/python3.6/dist-packages (from pandas>=0.19->pandas-profiling==2.3.0) (1.16.5)\nRequirement already satisfied: python-dateutil>=2.5.0 in /usr/local/lib/python3.6/dist-packages (from pandas>=0.19->pandas-profiling==2.3.0) (2.5.3)\nRequirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib>=1.4->pandas-profiling==2.3.0) (2.4.2)\nRequirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.6/dist-packages (from matplotlib>=1.4->pandas-profiling==2.3.0) (0.10.0)\nRequirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib>=1.4->pandas-profiling==2.3.0) (1.1.0)\nRequirement already satisfied: MarkupSafe>=0.23 in /usr/local/lib/python3.6/dist-packages (from jinja2>=2.8->pandas-profiling==2.3.0) (1.1.1)\nRequirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (from missingno>=0.4.2->pandas-profiling==2.3.0) (1.3.1)\nRequirement already satisfied: seaborn in /usr/local/lib/python3.6/dist-packages (from missingno>=0.4.2->pandas-profiling==2.3.0) (0.9.0)\nRequirement already satisfied: numba>=0.38.1 in /usr/local/lib/python3.6/dist-packages (from phik>=0.9.8->pandas-profiling==2.3.0) (0.40.1)\nRequirement already satisfied: jupyter-client>=5.2.3 in /usr/local/lib/python3.6/dist-packages (from phik>=0.9.8->pandas-profiling==2.3.0) (5.3.1)\nRequirement already satisfied: nbconvert>=5.3.1 in /usr/local/lib/python3.6/dist-packages (from phik>=0.9.8->pandas-profiling==2.3.0) (5.6.0)\nCollecting pytest>=4.0.2 (from phik>=0.9.8->pandas-profiling==2.3.0)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/2f/19/d5f71752f71451ccc5ed5f6739e9da4a235f38783fdaf3629cae41b2ca7b/pytest-5.1.2-py3-none-any.whl (224kB)\n\u001b[K |████████████████████████████████| 225kB 43.5MB/s \n\u001b[?25hCollecting pytest-pylint>=0.13.0 (from phik>=0.9.8->pandas-profiling==2.3.0)\n Downloading https://files.pythonhosted.org/packages/64/dc/6f35f114844fb12e38d60c4f3d2441a55baff7043ad4e013777dff55746c/pytest_pylint-0.14.1-py3-none-any.whl\nRequirement already satisfied: pyyaml in /usr/local/lib/python3.6/dist-packages (from confuse>=1.0.0->pandas-profiling==2.3.0) (3.13)\nRequirement already satisfied: six>=1.5 in /usr/local/lib/python3.6/dist-packages (from python-dateutil>=2.5.0->pandas>=0.19->pandas-profiling==2.3.0) (1.12.0)\nRequirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from kiwisolver>=1.0.1->matplotlib>=1.4->pandas-profiling==2.3.0) (41.2.0)\nRequirement already satisfied: llvmlite>=0.25.0dev0 in /usr/local/lib/python3.6/dist-packages (from numba>=0.38.1->phik>=0.9.8->pandas-profiling==2.3.0) (0.29.0)\nRequirement already satisfied: jupyter-core in /usr/local/lib/python3.6/dist-packages (from jupyter-client>=5.2.3->phik>=0.9.8->pandas-profiling==2.3.0) (4.5.0)\nRequirement already satisfied: traitlets in /usr/local/lib/python3.6/dist-packages (from jupyter-client>=5.2.3->phik>=0.9.8->pandas-profiling==2.3.0) (4.3.2)\nRequirement already satisfied: tornado>=4.1 in /usr/local/lib/python3.6/dist-packages (from jupyter-client>=5.2.3->phik>=0.9.8->pandas-profiling==2.3.0) (4.5.3)\nRequirement already satisfied: pyzmq>=13 in /usr/local/lib/python3.6/dist-packages (from jupyter-client>=5.2.3->phik>=0.9.8->pandas-profiling==2.3.0) (17.0.0)\nRequirement already satisfied: bleach in /usr/local/lib/python3.6/dist-packages (from nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (3.1.0)\nRequirement already satisfied: pandocfilters>=1.4.1 in /usr/local/lib/python3.6/dist-packages (from nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (1.4.2)\nRequirement already satisfied: mistune<2,>=0.8.1 in /usr/local/lib/python3.6/dist-packages (from nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (0.8.4)\nRequirement already satisfied: defusedxml in /usr/local/lib/python3.6/dist-packages (from nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (0.6.0)\nRequirement already satisfied: pygments in /usr/local/lib/python3.6/dist-packages (from nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (2.1.3)\nRequirement already satisfied: nbformat>=4.4 in /usr/local/lib/python3.6/dist-packages (from nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (4.4.0)\nRequirement already satisfied: entrypoints>=0.2.2 in /usr/local/lib/python3.6/dist-packages (from nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (0.3)\nRequirement already satisfied: testpath in /usr/local/lib/python3.6/dist-packages (from nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (0.4.2)\nRequirement already satisfied: py>=1.5.0 in /usr/local/lib/python3.6/dist-packages (from pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0) (1.8.0)\nRequirement already satisfied: more-itertools>=4.0.0 in /usr/local/lib/python3.6/dist-packages (from pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0) (7.2.0)\nCollecting pluggy<1.0,>=0.12 (from pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0)\n Downloading https://files.pythonhosted.org/packages/92/c7/48439f7d5fd6bddb4c04b850bb862b42e3e2b98570040dfaf68aedd8114b/pluggy-0.13.0-py2.py3-none-any.whl\nRequirement already satisfied: wcwidth in /usr/local/lib/python3.6/dist-packages (from pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0) (0.1.7)\nRequirement already satisfied: attrs>=17.4.0 in /usr/local/lib/python3.6/dist-packages (from pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0) (19.1.0)\nRequirement already satisfied: atomicwrites>=1.0 in /usr/local/lib/python3.6/dist-packages (from pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0) (1.3.0)\nRequirement already satisfied: importlib-metadata>=0.12; python_version < \"3.8\" in /usr/local/lib/python3.6/dist-packages (from pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0) (0.20)\nRequirement already satisfied: packaging in /usr/local/lib/python3.6/dist-packages (from pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0) (19.1)\nCollecting pylint>=1.4.5 (from pytest-pylint>=0.13.0->phik>=0.9.8->pandas-profiling==2.3.0)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/60/c2/b3f73f4ac008bef6e75bca4992f3963b3f85942e0277237721ef1c151f0d/pylint-2.3.1-py3-none-any.whl (765kB)\n\u001b[K |████████████████████████████████| 768kB 45.4MB/s \n\u001b[?25hRequirement already satisfied: decorator in /usr/local/lib/python3.6/dist-packages (from traitlets->jupyter-client>=5.2.3->phik>=0.9.8->pandas-profiling==2.3.0) (4.4.0)\nRequirement already satisfied: ipython-genutils in /usr/local/lib/python3.6/dist-packages (from traitlets->jupyter-client>=5.2.3->phik>=0.9.8->pandas-profiling==2.3.0) (0.2.0)\nRequirement already satisfied: webencodings in /usr/local/lib/python3.6/dist-packages (from bleach->nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (0.5.1)\nRequirement already satisfied: jsonschema!=2.5.0,>=2.4 in /usr/local/lib/python3.6/dist-packages (from nbformat>=4.4->nbconvert>=5.3.1->phik>=0.9.8->pandas-profiling==2.3.0) (2.6.0)\nRequirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.6/dist-packages (from importlib-metadata>=0.12; python_version < \"3.8\"->pytest>=4.0.2->phik>=0.9.8->pandas-profiling==2.3.0) (0.6.0)\nCollecting mccabe<0.7,>=0.6 (from pylint>=1.4.5->pytest-pylint>=0.13.0->phik>=0.9.8->pandas-profiling==2.3.0)\n Downloading https://files.pythonhosted.org/packages/87/89/479dc97e18549e21354893e4ee4ef36db1d237534982482c3681ee6e7b57/mccabe-0.6.1-py2.py3-none-any.whl\nCollecting isort<5,>=4.2.5 (from pylint>=1.4.5->pytest-pylint>=0.13.0->phik>=0.9.8->pandas-profiling==2.3.0)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/e5/b0/c121fd1fa3419ea9bfd55c7f9c4fedfec5143208d8c7ad3ce3db6c623c21/isort-4.3.21-py2.py3-none-any.whl (42kB)\n\u001b[K |████████████████████████████████| 51kB 15.4MB/s \n\u001b[?25hCollecting astroid<3,>=2.2.0 (from pylint>=1.4.5->pytest-pylint>=0.13.0->phik>=0.9.8->pandas-profiling==2.3.0)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/d5/ad/7221a62a2dbce5c3b8c57fd18e1052c7331adc19b3f27f1561aa6e620db2/astroid-2.2.5-py3-none-any.whl (193kB)\n\u001b[K |████████████████████████████████| 194kB 51.9MB/s \n\u001b[?25hCollecting typed-ast>=1.3.0; implementation_name == \"cpython\" (from astroid<3,>=2.2.0->pylint>=1.4.5->pytest-pylint>=0.13.0->phik>=0.9.8->pandas-profiling==2.3.0)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/31/d3/9d1802c161626d0278bafb1ffb32f76b9d01e123881bbf9d91e8ccf28e18/typed_ast-1.4.0-cp36-cp36m-manylinux1_x86_64.whl (736kB)\n\u001b[K |████████████████████████████████| 737kB 36.1MB/s \n\u001b[?25hRequirement already satisfied: wrapt in /usr/local/lib/python3.6/dist-packages (from astroid<3,>=2.2.0->pylint>=1.4.5->pytest-pylint>=0.13.0->phik>=0.9.8->pandas-profiling==2.3.0) (1.11.2)\nCollecting lazy-object-proxy (from astroid<3,>=2.2.0->pylint>=1.4.5->pytest-pylint>=0.13.0->phik>=0.9.8->pandas-profiling==2.3.0)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/0e/26/534a6d32572a9dbca11619321535c0a7ab34688545d9d67c2c204b9e3a3d/lazy_object_proxy-1.4.2-cp36-cp36m-manylinux1_x86_64.whl (49kB)\n\u001b[K |████████████████████████████████| 51kB 19.5MB/s \n\u001b[?25hBuilding wheels for collected packages: pandas-profiling, htmlmin, confuse\n Building wheel for pandas-profiling (setup.py) ... \u001b[?25l\u001b[?25hdone\n Created wheel for pandas-profiling: filename=pandas_profiling-2.3.0-py2.py3-none-any.whl size=145035 sha256=f4c4cc93c1f36b718263d47260574d6248e35272a2d88e5aac952bbb727ba835\n Stored in directory: /root/.cache/pip/wheels/ce/c7/f1/dbfef4848ebb048cb1d4a22d1ed0c62d8ff2523747235e19fe\n Building wheel for htmlmin (setup.py) ... \u001b[?25l\u001b[?25hdone\n Created wheel for htmlmin: filename=htmlmin-0.1.12-cp36-none-any.whl size=27084 sha256=dbd22fdb9a39619cc8bcabb02d1707c95b42799116385e67cb2da7dcb7030c8b\n Stored in directory: /root/.cache/pip/wheels/43/07/ac/7c5a9d708d65247ac1f94066cf1db075540b85716c30255459\n Building wheel for confuse (setup.py) ... \u001b[?25l\u001b[?25hdone\n Created wheel for confuse: filename=confuse-1.0.0-cp36-none-any.whl size=17486 sha256=c99643983359d92dc05809b720add2f90d43f5a2c8089c63fac41b37cf71b962\n Stored in directory: /root/.cache/pip/wheels/b0/b2/96/2074eee7dbf7b7df69d004c9b6ac4e32dad04fb7666cf943bd\nSuccessfully built pandas-profiling htmlmin confuse\n\u001b[31mERROR: datascience 0.10.6 has requirement folium==0.2.1, but you'll have folium 0.8.3 which is incompatible.\u001b[0m\nInstalling collected packages: htmlmin, pluggy, pytest, mccabe, isort, typed-ast, lazy-object-proxy, astroid, pylint, pytest-pylint, phik, confuse, pandas-profiling\n Found existing installation: pluggy 0.7.1\n Uninstalling pluggy-0.7.1:\n Successfully uninstalled pluggy-0.7.1\n Found existing installation: pytest 3.6.4\n Uninstalling pytest-3.6.4:\n Successfully uninstalled pytest-3.6.4\n Found existing installation: pandas-profiling 1.4.1\n Uninstalling pandas-profiling-1.4.1:\n Successfully uninstalled pandas-profiling-1.4.1\nSuccessfully installed astroid-2.2.5 confuse-1.0.0 htmlmin-0.1.12 isort-4.3.21 lazy-object-proxy-1.4.2 mccabe-0.6.1 pandas-profiling-2.3.0 phik-0.9.8 pluggy-0.13.0 pylint-2.3.1 pytest-5.1.2 pytest-pylint-0.14.1 typed-ast-1.4.0\nRequirement already satisfied: plotly==4.1.1 in /usr/local/lib/python3.6/dist-packages (4.1.1)\nRequirement already satisfied: retrying>=1.3.3 in /usr/local/lib/python3.6/dist-packages (from plotly==4.1.1) (1.3.3)\nRequirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from plotly==4.1.1) (1.12.0)\n" ] ], [ [ "# Part 1, Classification: Predict Blood Donations 🚑\nOur dataset is from a mobile blood donation vehicle in Taiwan. The Blood Transfusion Service Center drives to different universities and collects blood as part of a blood drive.\n\nThe goal is to predict whether the donor made a donation in March 2007, using information about each donor's history.\n\nGood data-driven systems for tracking and predicting donations and supply needs can improve the entire supply chain, making sure that more patients get the blood transfusions they need.", "_____no_output_____" ] ], [ [ "import pandas as pd\n\ndonors = pd.read_csv('https://archive.ics.uci.edu/ml/machine-learning-databases/blood-transfusion/transfusion.data')\nassert donors.shape == (748,5)\n\ndonors = donors.rename(columns={\n 'Recency (months)': 'months_since_last_donation', \n 'Frequency (times)': 'number_of_donations', \n 'Monetary (c.c. blood)': 'total_volume_donated', \n 'Time (months)': 'months_since_first_donation', \n 'whether he/she donated blood in March 2007': 'made_donation_in_march_2007'\n})\n\ndonors.head()", "_____no_output_____" ] ], [ [ "## 1.1. Begin with baselines\n\nWhat accuracy score would you get here with a \"majority class baseline\"?\n \n(You don't need to split the data into train and test sets yet. You can answer this question either with a scikit-learn function or with a pandas function.)", "_____no_output_____" ] ], [ [ "y_train = donors.made_donation_in_march_2007\ny_train.value_counts(normalize=True)", "_____no_output_____" ], [ "# We can cross-check this using scikit-learn\nmajority_class = y_train.mode()[0]\nmajority_class", "_____no_output_____" ], [ "y_pred = [majority_class] * len(y_train)\nlen(y_pred)", "_____no_output_____" ], [ "from sklearn.metrics import accuracy_score\naccuracy_score(y_train, y_pred)", "_____no_output_____" ] ], [ [ "## 1.2. Do train/test split. Arrange data into X features matrix and y target vector\n\nDo these steps in either order.\n\nSplit randomly. Use scikit-learn's train/test split function. Include 75% of the data in the train set, and hold out 25% for the test set.", "_____no_output_____" ] ], [ [ "train_features = donors[donors.columns.difference(['made_donation_in_march_2007'])]\nprint(train_features.shape)\ntrain_features.head()", "(748, 4)\n" ], [ "train_labels = donors['made_donation_in_march_2007']\nprint(train_labels.shape)\ntrain_labels.head()", "(748,)\n" ], [ "from sklearn.model_selection import train_test_split\n\nX_train = train_features\ny_train = train_labels\n\nX_train, X_val, y_train, y_val = train_test_split(\n X_train, y_train, train_size=0.75, test_size=0.25, \n stratify=y_train\n) \n\nX_train.shape, X_val.shape, y_train.shape, y_val.shape", "_____no_output_____" ], [ "X_train.isnull().sum()", "_____no_output_____" ], [ "train_labels.isnull().sum()", "_____no_output_____" ] ], [ [ "## 1.3. Use scikit-learn to fit a logistic regression model\n\nYou may use any number of features", "_____no_output_____" ] ], [ [ "# No categorical features, only numerical, no need for OHE\nfrom sklearn.linear_model import LogisticRegression\n\n# Right now we include all numerical features\n\n# If we want to split into less features:\n# X_train_subset = X_train[features]\n# X_val_subset = X_val[features]\n\nmodel = LogisticRegression(\n solver='lbfgs', multi_class='auto', max_iter=1000, n_jobs=-1) # Optimized from class\n\nmodel.fit(X_train, y_train)", "_____no_output_____" ] ], [ [ "## 1.4. Report classification metric: accuracy\n\nWhat is your model's accuracy on the test set?\n\nDon't worry if your model doesn't beat the mean baseline. That's okay!\n\n_\"The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data.\"_ —[John Tukey](https://en.wikiquote.org/wiki/John_Tukey)\n\n", "_____no_output_____" ] ], [ [ "print('Validation Accuracy', model.score(X_val, y_val))", "Validation Accuracy 0.7807486631016043\n" ] ], [ [ "# Part 2, Regression: Predict home prices in Ames, Iowa 🏠\n\nYou'll use historical housing data. There's a data dictionary at the bottom of the notebook. \n\nRun this code cell to load the dataset:\n\n\n\n", "_____no_output_____" ] ], [ [ "import pandas as pd\nURL = 'https://drive.google.com/uc?export=download&id=1522WlEW6HFss36roD_Cd9nybqSuiVcCK'\nhomes = pd.read_csv(URL)\nassert homes.shape == (2904, 47)\n\nhomes.head()", "_____no_output_____" ] ], [ [ "## 2.1. Begin with baselines\n\nWhat is the Mean Absolute Error and R^2 score for a mean baseline?", "_____no_output_____" ] ], [ [ "from sklearn.metrics import mean_absolute_error, mean_squared_error, r2_score\nimport numpy as np\n\ny_train = homes['SalePrice']\ny_pred_train = [y_train.mean()] * len(y_train)\n\nprint('Mean Baseline:')\nprint('Train Mean Absolute Error:', mean_absolute_error(y_train, y_pred_train))\nprint('Train R^2 Score:', r2_score(y_train, y_pred_train))", "Mean Baseline:\nTrain Mean Absolute Error: 58149.92774120811\nTrain R^2 Score: 0.0\n" ], [ "# 0% indicates that the model explains none of the variability of the response data around its mean.", "_____no_output_____" ] ], [ [ "## 2.2. Do train/test split\n\nTrain on houses sold in the years 2006 - 2008. (1,920 rows)\n\nValidate on house sold in 2009. (644 rows)\n\nTest on houses sold in 2010. (340 rows)", "_____no_output_____" ] ], [ [ "train = homes[(homes.Yr_Sold >= 2006) & (homes.Yr_Sold <= 2008)]\nprint(train.shape)", "(1920, 47)\n" ], [ "val = homes[homes.Yr_Sold == 2009]\nprint(val.shape)", "(644, 47)\n" ], [ "test = homes[homes.Yr_Sold == 2010]\nprint(test.shape)", "(340, 47)\n" ] ], [ [ "## 2.3. Arrange data into X features matrix and y target vector\n\nSelect at least one numeric feature and at least one categorical feature.\n\nOtherwise, you many choose whichever features and however many you want.", "_____no_output_____" ] ], [ [ "homes.isnull().sum()", "_____no_output_____" ], [ "target = 'SalePrice'\n\nnumeric_features = train[train.columns.difference(['SalePrice'])].select_dtypes(include='number').columns.tolist()\n\ncardinality = train.select_dtypes(exclude='number').nunique()\nlow_cardinality_features = cardinality[cardinality <= 10].index.tolist()\n\nfeatures = numeric_features + low_cardinality_features\nfeatures", "_____no_output_____" ] ], [ [ "## 2.4. Do one-hot encoding\n\nEncode your categorical feature(s).", "_____no_output_____" ] ], [ [ "X_train = train[features]\ny_train = train[target]\n\nX_val = val[features]\ny_val = val[target]\n\nX_test = test[features]\ny_test = test[target]", "_____no_output_____" ], [ "import category_encoders as ce\nfrom sklearn.preprocessing import StandardScaler\n\nX_train_subset = X_train[features]\nX_val_subset = X_val[features]\nX_test_subset = X_test[features]\n\nencoder = ce.OneHotEncoder(use_cat_names=True)\nX_train_encoded = encoder.fit_transform(X_train_subset)\nX_val_encoded = encoder.transform(X_val_subset)\nX_test_encoded = encoder.transform(X_test_subset)\n\nscaler = StandardScaler()\nX_train_scaled = scaler.fit_transform(X_train_encoded)\nX_val_scaled = scaler.transform(X_val_encoded)\nX_test_scaled = scaler.fit_transform(X_test_encoded)\n\nprint(X_train_scaled.shape, X_val_scaled.shape, X_test_scaled.shape)", "(1920, 157) (644, 157) (340, 157)\n" ] ], [ [ "## 2.5. Use scikit-learn to fit a linear regression (or ridge regression) model\nFit your model.", "_____no_output_____" ] ], [ [ "# STRETCH GOAL: Try at least 3 feature combinations. You may select features manually, or automatically\n# STRETCH GOAL: Report validation MAE and R2 for each feature combination you try\nfrom sklearn.linear_model import RidgeCV\nfrom sklearn.feature_selection import f_regression, SelectKBest\n\nfor k in range(1, len(X_train_encoded.columns)+1):\n print(f'{k} features')\n \n selector = SelectKBest(score_func=f_regression, k=k)\n X_train_selected = selector.fit_transform(X_train_scaled, y_train)\n X_test_selected = selector.transform(X_test_scaled)\n\n model = RidgeCV()\n model.fit(X_train_selected, y_train)\n \n y_pred = model.predict(X_test_selected)\n mae = mean_absolute_error(y_test, y_pred)\n print(f'Test MAE: ${mean_absolute_error(y_test, y_pred):,.0f}')\n print(f'Test R2: {r2_score(y_test, y_pred):,.3f} \\n')", "1 features\nTest MAE: $35,831\nTest R2: 0.570 \n\n2 features\nTest MAE: $30,063\nTest R2: 0.700 \n\n3 features\nTest MAE: $27,450\nTest R2: 0.741 \n\n4 features\nTest MAE: $27,428\nTest R2: 0.744 \n\n5 features\nTest MAE: $27,420\nTest R2: 0.745 \n\n6 features\nTest MAE: $25,676\nTest R2: 0.770 \n\n7 features\nTest MAE: $25,671\nTest R2: 0.771 \n\n8 features\nTest MAE: $25,330\nTest R2: 0.775 \n\n9 features\nTest MAE: $25,349\nTest R2: 0.777 \n\n10 features\nTest MAE: $23,751\nTest R2: 0.810 \n\n11 features\nTest MAE: $23,823\nTest R2: 0.810 \n\n12 features\nTest MAE: $23,898\nTest R2: 0.812 \n\n13 features\nTest MAE: $23,929\nTest R2: 0.813 \n\n14 features\nTest MAE: $23,601\nTest R2: 0.821 \n\n15 features\nTest MAE: $23,521\nTest R2: 0.822 \n\n16 features\nTest MAE: $23,486\nTest R2: 0.823 \n\n17 features\nTest MAE: $23,503\nTest R2: 0.823 \n\n18 features\nTest MAE: $23,457\nTest R2: 0.822 \n\n19 features\nTest MAE: $23,438\nTest R2: 0.823 \n\n20 features\nTest MAE: $23,409\nTest R2: 0.823 \n\n21 features\nTest MAE: $23,388\nTest R2: 0.825 \n\n22 features\nTest MAE: $23,382\nTest R2: 0.825 \n\n23 features\nTest MAE: $23,473\nTest R2: 0.823 \n\n24 features\nTest MAE: $23,560\nTest R2: 0.821 \n\n25 features\nTest MAE: $23,506\nTest R2: 0.821 \n\n26 features\nTest MAE: $23,526\nTest R2: 0.821 \n\n27 features\nTest MAE: $23,562\nTest R2: 0.820 \n\n28 features\nTest MAE: $23,575\nTest R2: 0.820 \n\n29 features\nTest MAE: $23,225\nTest R2: 0.825 \n\n30 features\nTest MAE: $23,055\nTest R2: 0.827 \n\n31 features\nTest MAE: $23,055\nTest R2: 0.827 \n\n32 features\nTest MAE: $22,967\nTest R2: 0.827 \n\n33 features\nTest MAE: $22,217\nTest R2: 0.835 \n\n34 features\nTest MAE: $22,245\nTest R2: 0.834 \n\n35 features\nTest MAE: $22,217\nTest R2: 0.834 \n\n36 features\nTest MAE: $22,207\nTest R2: 0.835 \n\n37 features\nTest MAE: $22,155\nTest R2: 0.835 \n\n38 features\nTest MAE: $22,133\nTest R2: 0.835 \n\n39 features\nTest MAE: $22,133\nTest R2: 0.835 \n\n40 features\nTest MAE: $22,123\nTest R2: 0.835 \n\n41 features\nTest MAE: $22,166\nTest R2: 0.835 \n\n42 features\nTest MAE: $22,108\nTest R2: 0.836 \n\n43 features\nTest MAE: $21,839\nTest R2: 0.840 \n\n44 features\nTest MAE: $21,782\nTest R2: 0.841 \n\n45 features\nTest MAE: $21,841\nTest R2: 0.841 \n\n46 features\nTest MAE: $21,995\nTest R2: 0.840 \n\n47 features\nTest MAE: $22,005\nTest R2: 0.840 \n\n48 features\nTest MAE: $21,981\nTest R2: 0.840 \n\n49 features\nTest MAE: $21,968\nTest R2: 0.840 \n\n50 features\nTest MAE: $21,968\nTest R2: 0.840 \n\n51 features\nTest MAE: $21,924\nTest R2: 0.840 \n\n52 features\nTest MAE: $21,901\nTest R2: 0.842 \n\n53 features\nTest MAE: $21,863\nTest R2: 0.842 \n\n54 features\nTest MAE: $22,138\nTest R2: 0.841 \n\n55 features\nTest MAE: $21,932\nTest R2: 0.846 \n\n56 features\nTest MAE: $21,789\nTest R2: 0.848 \n\n57 features\nTest MAE: $21,681\nTest R2: 0.851 \n\n58 features\nTest MAE: $21,691\nTest R2: 0.851 \n\n59 features\nTest MAE: $21,315\nTest R2: 0.854 \n\n60 features\nTest MAE: $21,381\nTest R2: 0.853 \n\n61 features\nTest MAE: $21,411\nTest R2: 0.853 \n\n62 features\nTest MAE: $21,413\nTest R2: 0.853 \n\n63 features\nTest MAE: $21,329\nTest R2: 0.854 \n\n64 features\nTest MAE: $21,322\nTest R2: 0.854 \n\n65 features\nTest MAE: $21,338\nTest R2: 0.854 \n\n66 features\nTest MAE: $21,249\nTest R2: 0.854 \n\n67 features\nTest MAE: $21,240\nTest R2: 0.855 \n\n68 features\nTest MAE: $21,135\nTest R2: 0.855 \n\n69 features\nTest MAE: $21,122\nTest R2: 0.855 \n\n70 features\nTest MAE: $21,071\nTest R2: 0.855 \n\n71 features\nTest MAE: $20,721\nTest R2: 0.860 \n\n72 features\nTest MAE: $20,711\nTest R2: 0.860 \n\n73 features\nTest MAE: $20,711\nTest R2: 0.860 \n\n74 features\nTest MAE: $20,801\nTest R2: 0.859 \n\n75 features\nTest MAE: $20,662\nTest R2: 0.861 \n\n76 features\nTest MAE: $20,688\nTest R2: 0.861 \n\n77 features\nTest MAE: $20,696\nTest R2: 0.861 \n\n78 features\nTest MAE: $20,698\nTest R2: 0.861 \n\n79 features\nTest MAE: $20,602\nTest R2: 0.862 \n\n80 features\nTest MAE: $20,479\nTest R2: 0.862 \n\n81 features\nTest MAE: $20,473\nTest R2: 0.862 \n\n82 features\nTest MAE: $20,457\nTest R2: 0.862 \n\n83 features\nTest MAE: $20,461\nTest R2: 0.862 \n\n84 features\nTest MAE: $20,462\nTest R2: 0.862 \n\n85 features\nTest MAE: $20,461\nTest R2: 0.862 \n\n86 features\nTest MAE: $20,467\nTest R2: 0.862 \n\n87 features\nTest MAE: $20,457\nTest R2: 0.862 \n\n88 features\nTest MAE: $20,584\nTest R2: 0.859 \n\n89 features\nTest MAE: $20,590\nTest R2: 0.859 \n\n90 features\nTest MAE: $20,570\nTest R2: 0.859 \n\n91 features\nTest MAE: $20,565\nTest R2: 0.859 \n\n92 features\nTest MAE: $20,603\nTest R2: 0.859 \n\n93 features\nTest MAE: $20,560\nTest R2: 0.860 \n\n94 features\nTest MAE: $20,585\nTest R2: 0.860 \n\n95 features\nTest MAE: $20,574\nTest R2: 0.860 \n\n96 features\nTest MAE: $20,573\nTest R2: 0.860 \n\n97 features\nTest MAE: $20,576\nTest R2: 0.860 \n\n98 features\nTest MAE: $20,654\nTest R2: 0.860 \n\n99 features\nTest MAE: $20,751\nTest R2: 0.859 \n\n100 features\nTest MAE: $20,826\nTest R2: 0.858 \n\n101 features\nTest MAE: $20,838\nTest R2: 0.858 \n\n102 features\nTest MAE: $20,839\nTest R2: 0.858 \n\n103 features\nTest MAE: $20,904\nTest R2: 0.858 \n\n104 features\nTest MAE: $20,929\nTest R2: 0.858 \n\n105 features\nTest MAE: $20,948\nTest R2: 0.858 \n\n106 features\nTest MAE: $20,961\nTest R2: 0.858 \n\n107 features\nTest MAE: $20,962\nTest R2: 0.858 \n\n108 features\nTest MAE: $20,962\nTest R2: 0.858 \n\n109 features\nTest MAE: $20,965\nTest R2: 0.858 \n\n110 features\nTest MAE: $20,987\nTest R2: 0.857 \n\n111 features\nTest MAE: $20,972\nTest R2: 0.857 \n\n112 features\nTest MAE: $20,939\nTest R2: 0.857 \n\n113 features\nTest MAE: $20,932\nTest R2: 0.858 \n\n114 features\nTest MAE: $21,055\nTest R2: 0.854 \n\n115 features\nTest MAE: $21,086\nTest R2: 0.853 \n\n116 features\nTest MAE: $21,086\nTest R2: 0.853 \n\n117 features\nTest MAE: $21,134\nTest R2: 0.853 \n\n118 features\nTest MAE: $21,134\nTest R2: 0.853 \n\n119 features\nTest MAE: $21,143\nTest R2: 0.853 \n\n120 features\nTest MAE: $21,135\nTest R2: 0.853 \n\n121 features\nTest MAE: $21,135\nTest R2: 0.853 \n\n122 features\nTest MAE: $21,132\nTest R2: 0.853 \n\n123 features\nTest MAE: $21,176\nTest R2: 0.853 \n\n124 features\nTest MAE: $21,176\nTest R2: 0.853 \n\n125 features\nTest MAE: $21,154\nTest R2: 0.853 \n\n126 features\nTest MAE: $21,155\nTest R2: 0.853 \n\n127 features\nTest MAE: $21,158\nTest R2: 0.853 \n\n128 features\nTest MAE: $21,163\nTest R2: 0.853 \n\n129 features\nTest MAE: $21,189\nTest R2: 0.856 \n\n130 features\nTest MAE: $21,106\nTest R2: 0.856 \n\n131 features\nTest MAE: $21,121\nTest R2: 0.856 \n\n132 features\nTest MAE: $21,126\nTest R2: 0.854 \n\n133 features\nTest MAE: $21,160\nTest R2: 0.854 \n\n134 features\nTest MAE: $21,157\nTest R2: 0.854 \n\n135 features\nTest MAE: $21,157\nTest R2: 0.854 \n\n136 features\nTest MAE: $21,156\nTest R2: 0.854 \n\n137 features\nTest MAE: $21,178\nTest R2: 0.853 \n\n138 features\nTest MAE: $21,182\nTest R2: 0.853 \n\n139 features\nTest MAE: $21,011\nTest R2: 0.853 \n\n140 features\nTest MAE: $20,984\nTest R2: 0.854 \n\n141 features\nTest MAE: $20,985\nTest R2: 0.854 \n\n142 features\nTest MAE: $20,916\nTest R2: 0.855 \n\n143 features\nTest MAE: $20,917\nTest R2: 0.855 \n\n144 features\nTest MAE: $20,912\nTest R2: 0.855 \n\n145 features\nTest MAE: $21,049\nTest R2: 0.852 \n\n146 features\nTest MAE: $21,055\nTest R2: 0.851 \n\n147 features\nTest MAE: $20,995\nTest R2: 0.853 \n\n148 features\nTest MAE: $20,598\nTest R2: 0.859 \n\n149 features\nTest MAE: $20,593\nTest R2: 0.859 \n\n150 features\nTest MAE: $20,609\nTest R2: 0.859 \n\n151 features\nTest MAE: $20,570\nTest R2: 0.860 \n\n152 features\nTest MAE: $20,566\nTest R2: 0.860 \n\n153 features\nTest MAE: $20,669\nTest R2: 0.858 \n\n154 features\nTest MAE: $20,647\nTest R2: 0.858 \n\n155 features\nTest MAE: $20,585\nTest R2: 0.858 \n\n156 features\nTest MAE: $20,621\nTest R2: 0.858 \n\n157 features\nTest MAE: $20,645\nTest R2: 0.858 \n\n" ] ], [ [ "## 2.6. Report validation MAE and $R^2$\n\nWhat is your model's Mean Absolute Error and $R^2$ score on the validation set?", "_____no_output_____" ] ], [ [ "# STRETCH GOAL: Report test MAE and $R^2$ for your final model\nselector = SelectKBest(score_func=f_regression, k=157)\n\nX_train_selected = selector.fit_transform(X_train_scaled, y_train)\nX_val_selected = selector.transform(X_val_scaled)\n\nmodel = RidgeCV()\nmodel.fit(X_train_selected, y_train)\n\ny_pred_val = model.predict(X_val_selected)\nprint('Mean Baseline:')\nprint('Val Mean Absolute Error:', mean_absolute_error(y_val, y_pred_val))\nprint('Val R^2 Score:', r2_score(y_val, y_pred_val))", "Mean Baseline:\nVal Mean Absolute Error: 19865.0311356575\nVal R^2 Score: 0.8731097413449\n" ], [ "# STRETCH GOAL: Print or plot the coefficients for the features in your model\ncoefficients = pd.Series(model.coef_, X_train_encoded.columns)\nplt.figure(figsize=(10,30))\ncoefficients.sort_values().plot.barh(color='blue');", "_____no_output_____" ] ], [ [ "# Stretch Goals, Regression\n- Make visualizations to explore relationships between features and target\n- Try at least 3 feature combinations. You may select features manually, or automatically\n- Report validation MAE and $R^2$ for each feature combination you try\n- Report test MAE and $R^2$ for your final model\n- Print or plot the coefficients for the features in your model", "_____no_output_____" ] ], [ [ "#STRETCH GOAL: Make visualizations to explore relationships between features and target\n%matplotlib inline\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\nfor col in sorted(low_cardinality_features):\n sns.catplot(x=col, y='SalePrice', data=train, kind='bar', color='grey')\n plt.xticks(rotation=45)\n plt.show()", "_____no_output_____" ] ], [ [ "# Data Dictionary \n\nHere's a description of the data fields:\n\n```\n1st_Flr_SF: First Floor square feet\n\nBedroom_AbvGr: Bedrooms above grade (does NOT include basement bedrooms)\n\nBldg_Type: Type of dwelling\n\t\t\n 1Fam\tSingle-family Detached\t\n 2FmCon\tTwo-family Conversion; originally built as one-family dwelling\n Duplx\tDuplex\n TwnhsE\tTownhouse End Unit\n TwnhsI\tTownhouse Inside Unit\n \nBsmt_Half_Bath: Basement half bathrooms\n\nBsmt_Full_Bath: Basement full bathrooms\n\nCentral_Air: Central air conditioning\n\n N\tNo\n Y\tYes\n\t\t\nCondition_1: Proximity to various conditions\n\t\n Artery\tAdjacent to arterial street\n Feedr\tAdjacent to feeder street\t\n Norm\tNormal\t\n RRNn\tWithin 200' of North-South Railroad\n RRAn\tAdjacent to North-South Railroad\n PosN\tNear positive off-site feature--park, greenbelt, etc.\n PosA\tAdjacent to postive off-site feature\n RRNe\tWithin 200' of East-West Railroad\n RRAe\tAdjacent to East-West Railroad\n\t\nCondition_2: Proximity to various conditions (if more than one is present)\n\t\t\n Artery\tAdjacent to arterial street\n Feedr\tAdjacent to feeder street\t\n Norm\tNormal\t\n RRNn\tWithin 200' of North-South Railroad\n RRAn\tAdjacent to North-South Railroad\n PosN\tNear positive off-site feature--park, greenbelt, etc.\n PosA\tAdjacent to postive off-site feature\n RRNe\tWithin 200' of East-West Railroad\n RRAe\tAdjacent to East-West Railroad\n \n Electrical: Electrical system\n\n SBrkr\tStandard Circuit Breakers & Romex\n FuseA\tFuse Box over 60 AMP and all Romex wiring (Average)\t\n FuseF\t60 AMP Fuse Box and mostly Romex wiring (Fair)\n FuseP\t60 AMP Fuse Box and mostly knob & tube wiring (poor)\n Mix\tMixed\n \n Exter_Cond: Evaluates the present condition of the material on the exterior\n\t\t\n Ex\tExcellent\n Gd\tGood\n TA\tAverage/Typical\n Fa\tFair\n Po\tPoor\n \n Exter_Qual: Evaluates the quality of the material on the exterior \n\t\t\n Ex\tExcellent\n Gd\tGood\n TA\tAverage/Typical\n Fa\tFair\n Po\tPoor\n\t\t\nExterior_1st: Exterior covering on house\n\n AsbShng\tAsbestos Shingles\n AsphShn\tAsphalt Shingles\n BrkComm\tBrick Common\n BrkFace\tBrick Face\n CBlock\tCinder Block\n CemntBd\tCement Board\n HdBoard\tHard Board\n ImStucc\tImitation Stucco\n MetalSd\tMetal Siding\n Other\tOther\n Plywood\tPlywood\n PreCast\tPreCast\t\n Stone\tStone\n Stucco\tStucco\n VinylSd\tVinyl Siding\n Wd Sdng\tWood Siding\n WdShing\tWood Shingles\n\t\nExterior_2nd: Exterior covering on house (if more than one material)\n\n AsbShng\tAsbestos Shingles\n AsphShn\tAsphalt Shingles\n BrkComm\tBrick Common\n BrkFace\tBrick Face\n CBlock\tCinder Block\n CemntBd\tCement Board\n HdBoard\tHard Board\n ImStucc\tImitation Stucco\n MetalSd\tMetal Siding\n Other\tOther\n Plywood\tPlywood\n PreCast\tPreCast\n Stone\tStone\n Stucco\tStucco\n VinylSd\tVinyl Siding\n Wd Sdng\tWood Siding\n WdShing\tWood Shingles\n \nFoundation: Type of foundation\n\t\t\n BrkTil\tBrick & Tile\n CBlock\tCinder Block\n PConc\tPoured Contrete\t\n Slab\tSlab\n Stone\tStone\n Wood\tWood\n\t\t\nFull_Bath: Full bathrooms above grade\n\nFunctional: Home functionality (Assume typical unless deductions are warranted)\n\n Typ\tTypical Functionality\n Min1\tMinor Deductions 1\n Min2\tMinor Deductions 2\n Mod\tModerate Deductions\n Maj1\tMajor Deductions 1\n Maj2\tMajor Deductions 2\n Sev\tSeverely Damaged\n Sal\tSalvage only\n\t\t\nGr_Liv_Area: Above grade (ground) living area square feet\n \nHalf_Bath: Half baths above grade\n\nHeating: Type of heating\n\t\t\n Floor\tFloor Furnace\n GasA\tGas forced warm air furnace\n GasW\tGas hot water or steam heat\n Grav\tGravity furnace\t\n OthW\tHot water or steam heat other than gas\n Wall\tWall furnace\n\t\t\nHeating_QC: Heating quality and condition\n\n Ex\tExcellent\n Gd\tGood\n TA\tAverage/Typical\n Fa\tFair\n Po\tPoor\n\nHouse_Style: Style of dwelling\n\t\n 1Story\tOne story\n 1.5Fin\tOne and one-half story: 2nd level finished\n 1.5Unf\tOne and one-half story: 2nd level unfinished\n 2Story\tTwo story\n 2.5Fin\tTwo and one-half story: 2nd level finished\n 2.5Unf\tTwo and one-half story: 2nd level unfinished\n SFoyer\tSplit Foyer\n SLvl\tSplit Level\n\nKitchen_AbvGr: Kitchens above grade\n\nKitchen_Qual: Kitchen quality\n\n Ex\tExcellent\n Gd\tGood\n TA\tTypical/Average\n Fa\tFair\n Po\tPoor\n\nLandContour: Flatness of the property\n\n Lvl\tNear Flat/Level\t\n Bnk\tBanked - Quick and significant rise from street grade to building\n HLS\tHillside - Significant slope from side to side\n Low\tDepression\n\t\t\nLand_Slope: Slope of property\n\t\t\n Gtl\tGentle slope\n Mod\tModerate Slope\t\n Sev\tSevere Slope\n\nLot_Area: Lot size in square feet\n\nLot_Config: Lot configuration\n\n Inside\tInside lot\n Corner\tCorner lot\n CulDSac\tCul-de-sac\n FR2\tFrontage on 2 sides of property\n FR3\tFrontage on 3 sides of property\n\nLot_Shape: General shape of property\n\n Reg\tRegular\t\n IR1\tSlightly irregular\n IR2\tModerately Irregular\n IR3\tIrregular\n\nMS_SubClass: Identifies the type of dwelling involved in the sale.\t\n\n 20\t1-STORY 1946 & NEWER ALL STYLES\n 30\t1-STORY 1945 & OLDER\n 40\t1-STORY W/FINISHED ATTIC ALL AGES\n 45\t1-1/2 STORY - UNFINISHED ALL AGES\n 50\t1-1/2 STORY FINISHED ALL AGES\n 60\t2-STORY 1946 & NEWER\n 70\t2-STORY 1945 & OLDER\n 75\t2-1/2 STORY ALL AGES\n 80\tSPLIT OR MULTI-LEVEL\n 85\tSPLIT FOYER\n 90\tDUPLEX - ALL STYLES AND AGES\n 120\t1-STORY PUD (Planned Unit Development) - 1946 & NEWER\n 150\t1-1/2 STORY PUD - ALL AGES\n 160\t2-STORY PUD - 1946 & NEWER\n 180\tPUD - MULTILEVEL - INCL SPLIT LEV/FOYER\n 190\t2 FAMILY CONVERSION - ALL STYLES AND AGES\n\nMS_Zoning: Identifies the general zoning classification of the sale.\n\t\t\n A\tAgriculture\n C\tCommercial\n FV\tFloating Village Residential\n I\tIndustrial\n RH\tResidential High Density\n RL\tResidential Low Density\n RP\tResidential Low Density Park \n RM\tResidential Medium Density\n\nMas_Vnr_Type: Masonry veneer type\n\n BrkCmn\tBrick Common\n BrkFace\tBrick Face\n CBlock\tCinder Block\n None\tNone\n Stone\tStone\n\nMo_Sold: Month Sold (MM)\n\nNeighborhood: Physical locations within Ames city limits\n\n Blmngtn\tBloomington Heights\n Blueste\tBluestem\n BrDale\tBriardale\n BrkSide\tBrookside\n ClearCr\tClear Creek\n CollgCr\tCollege Creek\n Crawfor\tCrawford\n Edwards\tEdwards\n Gilbert\tGilbert\n IDOTRR\tIowa DOT and Rail Road\n MeadowV\tMeadow Village\n Mitchel\tMitchell\n Names\tNorth Ames\n NoRidge\tNorthridge\n NPkVill\tNorthpark Villa\n NridgHt\tNorthridge Heights\n NWAmes\tNorthwest Ames\n OldTown\tOld Town\n SWISU\tSouth & West of Iowa State University\n Sawyer\tSawyer\n SawyerW\tSawyer West\n Somerst\tSomerset\n StoneBr\tStone Brook\n Timber\tTimberland\n Veenker\tVeenker\n\t\t\t\nOverall_Cond: Rates the overall condition of the house\n\n 10\tVery Excellent\n 9\tExcellent\n 8\tVery Good\n 7\tGood\n 6\tAbove Average\t\n 5\tAverage\n 4\tBelow Average\t\n 3\tFair\n 2\tPoor\n 1\tVery Poor\n\nOverall_Qual: Rates the overall material and finish of the house\n\n 10\tVery Excellent\n 9\tExcellent\n 8\tVery Good\n 7\tGood\n 6\tAbove Average\n 5\tAverage\n 4\tBelow Average\n 3\tFair\n 2\tPoor\n 1\tVery Poor\n\nPaved_Drive: Paved driveway\n\n Y\tPaved \n P\tPartial Pavement\n N\tDirt/Gravel\n\nRoof_Matl: Roof material\n\n ClyTile\tClay or Tile\n CompShg\tStandard (Composite) Shingle\n Membran\tMembrane\n Metal\tMetal\n Roll\tRoll\n Tar&Grv\tGravel & Tar\n WdShake\tWood Shakes\n WdShngl\tWood Shingles\n\nRoof_Style: Type of roof\n\n Flat\tFlat\n Gable\tGable\n Gambrel\tGabrel (Barn)\n Hip\tHip\n Mansard\tMansard\n Shed\tShed\n\nSalePrice: the sales price for each house\n\nSale_Condition: Condition of sale\n\n Normal\tNormal Sale\n Abnorml\tAbnormal Sale - trade, foreclosure, short sale\n AdjLand\tAdjoining Land Purchase\n Alloca\tAllocation - two linked properties with separate deeds, typically condo with a garage unit\t\n Family\tSale between family members\n Partial\tHome was not completed when last assessed (associated with New Homes)\n\nSale_Type: Type of sale\n\t\t\n WD \tWarranty Deed - Conventional\n CWD\tWarranty Deed - Cash\n VWD\tWarranty Deed - VA Loan\n New\tHome just constructed and sold\n COD\tCourt Officer Deed/Estate\n Con\tContract 15% Down payment regular terms\n ConLw\tContract Low Down payment and low interest\n ConLI\tContract Low Interest\n ConLD\tContract Low Down\n Oth\tOther\n\t\nStreet: Type of road access to property\n\n Grvl\tGravel\t\n Pave\tPaved\n \t\nTotRms_AbvGrd: Total rooms above grade (does not include bathrooms)\n\nUtilities: Type of utilities available\n\t\t\n AllPub\tAll public Utilities (E,G,W,& S)\t\n NoSewr\tElectricity, Gas, and Water (Septic Tank)\n NoSeWa\tElectricity and Gas Only\n ELO\tElectricity only\t\n\t\nYear_Built: Original construction date\n\nYear_Remod/Add: Remodel date (same as construction date if no remodeling or additions)\n\t\t\t\t\t\t\nYr_Sold: Year Sold (YYYY)\t\n\n```", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import cv2\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport glob\nimport pandas as pd\nimport os", "_____no_output_____" ], [ "def imshow(img):\n img = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)\n plt.imshow(img)", "_____no_output_____" ], [ "def get_lane_mask(sample,lane_idx):\n points_lane = []\n h_max = np.max(data['h_samples'][sample])\n h_min = np.min(data['h_samples'][sample])\n x_idx = data['lanes'][sample][lane_idx]\n y_idx = data['h_samples'][sample]\n for x,y in zip(x_idx,y_idx):\n offset = (y-h_min)/20\n # print(offset)\n if x>-100:\n points_lane.append([x-offset/2,y])\n x_idx_=x_idx.copy()\n y_idx_=y_idx.copy()\n x_idx_.reverse()\n y_idx_.reverse()\n for x,y in zip(x_idx_,y_idx_):\n offset = (y-h_min)/20\n # print(offset)\n if x>-100:\n points_lane.append([x+offset/2,y])\n return points_lane", "_____no_output_____" ], [ "def create_lane_mask(img_raw,sample):\n colors = [[255,0,0],[0,255,0],[0,0,255],[0,255,255]]\n laneMask = np.zeros(img_raw.shape, dtype=np.uint8)\n for lane_idx in range(len(data.lanes[sample])):\n points_lane = get_lane_mask(sample,lane_idx)\n if len(points_lane)>0: \n pts = np.array(points_lane, np.int32)\n pts = pts.reshape((-1,1,2))\n laneMask = cv2.fillPoly(laneMask,[pts],colors[lane_idx])\n colors = [[255,0,0],[0,255,0],[0,0,255],[0,255,255]]\n # create grey-scale label image\n label = np.zeros((720,1280),dtype = np.uint8)\n for i in range(len(colors)):\n label[np.where((laneMask == colors[i]).all(axis = 2))] = i+1\n else: continue\n return(img_raw, label)", "_____no_output_____" ], [ "data = pd.read_json(os.path.join(data_dir, 'label_data.json'), lines=True)\ndata.info()\nprint(len(data.raw_file))\ndata", "<class 'pandas.core.frame.DataFrame'>\nRangeIndex: 2858 entries, 0 to 2857\nData columns (total 3 columns):\n # Column Non-Null Count Dtype \n--- ------ -------------- ----- \n 0 lanes 2858 non-null object\n 1 h_samples 2858 non-null object\n 2 raw_file 2858 non-null object\ndtypes: object(3)\nmemory usage: 67.1+ KB\n2858\n" ], [ "print(len(data.raw_file))", "2858\n" ], [ "for i in range(len(data.raw_file)):\n img_path = data.raw_file[i]\n img_path = os.path.join(data_dir,img_path)\n print('Reading from: ', img_path)\n path_list = img_path.split('/')[:-1]\n mask_path_dir = os.path.join(*path_list)\n\n img_raw = cv2.imread(img_path)\n img_, mask = create_lane_mask(img_raw,i)\n \"\"\"\n fig = plt.figure(figsize=(15,20))\n plt.subplot(211)\n imshow(img_raw)\n plt.subplot(212)\n print(mask.shape)\n plt.imshow(mask)\n \"\"\"\n mask_path_dir = mask_path_dir.replace('clips', 'masks')\n print('Saving to: ', mask_path_dir)\n try:\n os.makedirs(mask_path_dir)\n except:\n pass\n\n for i in range(1, 21):\n cv2.imwrite(os.path.join( mask_path_dir, f'{i}.tiff'), mask)\n # i = i+1\n", "Reading from: dataset/clips/0313-1/6040/20.jpg\nSaving to: dataset/masks/0313-1/6040\nReading from: dataset/clips/0313-1/5320/20.jpg\nSaving to: dataset/masks/0313-1/5320\nReading from: dataset/clips/0313-1/23700/20.jpg\nSaving to: dataset/masks/0313-1/23700\nReading from: dataset/clips/0313-1/51660/20.jpg\nSaving to: dataset/masks/0313-1/51660\nReading from: dataset/clips/0313-1/25680/20.jpg\nSaving to: dataset/masks/0313-1/25680\nReading from: dataset/clips/0313-1/36000/20.jpg\nSaving to: dataset/masks/0313-1/36000\nReading from: dataset/clips/0313-1/9460/20.jpg\nSaving to: dataset/masks/0313-1/9460\nReading from: dataset/clips/0313-1/6180/20.jpg\nSaving to: dataset/masks/0313-1/6180\nReading from: dataset/clips/0313-1/10100/20.jpg\nSaving to: dataset/masks/0313-1/10100\nReading from: dataset/clips/0313-1/39720/20.jpg\nSaving to: dataset/masks/0313-1/39720\nReading from: dataset/clips/0313-1/7780/20.jpg\nSaving to: dataset/masks/0313-1/7780\nReading from: dataset/clips/0313-1/31680/20.jpg\nSaving to: dataset/masks/0313-1/31680\nReading from: dataset/clips/0313-1/11980/20.jpg\nSaving to: dataset/masks/0313-1/11980\nReading from: dataset/clips/0313-1/3300/20.jpg\nSaving to: dataset/masks/0313-1/3300\nReading from: dataset/clips/0313-1/32520/20.jpg\nSaving to: dataset/masks/0313-1/32520\nReading from: dataset/clips/0313-1/47160/20.jpg\nSaving to: dataset/masks/0313-1/47160\nReading from: dataset/clips/0313-1/15220/20.jpg\nSaving to: dataset/masks/0313-1/15220\nReading from: dataset/clips/0313-1/19320/20.jpg\nSaving to: dataset/masks/0313-1/19320\nReading from: dataset/clips/0313-1/9600/20.jpg\nSaving to: dataset/masks/0313-1/9600\nReading from: dataset/clips/0313-1/3960/20.jpg\nSaving to: dataset/masks/0313-1/3960\nReading from: dataset/clips/0313-1/42420/20.jpg\nSaving to: dataset/masks/0313-1/42420\nReading from: dataset/clips/0313-1/300/20.jpg\nSaving to: dataset/masks/0313-1/300\nReading from: dataset/clips/0313-1/9860/20.jpg\nSaving to: dataset/masks/0313-1/9860\nReading from: dataset/clips/0313-1/25440/20.jpg\nSaving to: dataset/masks/0313-1/25440\nReading from: dataset/clips/0313-1/35400/20.jpg\nSaving to: dataset/masks/0313-1/35400\nReading from: dataset/clips/0313-1/29100/20.jpg\nSaving to: dataset/masks/0313-1/29100\nReading from: dataset/clips/0313-1/33180/20.jpg\nSaving to: dataset/masks/0313-1/33180\nReading from: dataset/clips/0313-1/16100/20.jpg\nSaving to: dataset/masks/0313-1/16100\nReading from: dataset/clips/0313-1/42000/20.jpg\nSaving to: dataset/masks/0313-1/42000\nReading from: dataset/clips/0313-1/600/20.jpg\nSaving to: dataset/masks/0313-1/600\nReading from: dataset/clips/0313-1/46680/20.jpg\nSaving to: dataset/masks/0313-1/46680\nReading from: dataset/clips/0313-1/7640/20.jpg\nSaving to: dataset/masks/0313-1/7640\nReading from: dataset/clips/0313-1/43020/20.jpg\nSaving to: dataset/masks/0313-1/43020\nReading from: dataset/clips/0313-1/29880/20.jpg\nSaving to: dataset/masks/0313-1/29880\nReading from: dataset/clips/0313-1/24840/20.jpg\nSaving to: dataset/masks/0313-1/24840\nReading from: dataset/clips/0313-1/9160/20.jpg\nSaving to: dataset/masks/0313-1/9160\nReading from: dataset/clips/0313-1/6420/20.jpg\nSaving to: dataset/masks/0313-1/6420\nReading from: dataset/clips/0313-1/48300/20.jpg\nSaving to: dataset/masks/0313-1/48300\nReading from: dataset/clips/0313-1/4240/20.jpg\nSaving to: dataset/masks/0313-1/4240\nReading from: dataset/clips/0313-1/12600/20.jpg\nSaving to: dataset/masks/0313-1/12600\nReading from: dataset/clips/0313-1/15680/20.jpg\nSaving to: dataset/masks/0313-1/15680\nReading from: dataset/clips/0313-1/9580/20.jpg\nSaving to: dataset/masks/0313-1/9580\nReading from: dataset/clips/0313-1/14620/20.jpg\nSaving to: dataset/masks/0313-1/14620\nReading from: dataset/clips/0313-1/49500/20.jpg\nSaving to: dataset/masks/0313-1/49500\nReading from: dataset/clips/0313-1/10960/20.jpg\nSaving to: dataset/masks/0313-1/10960\nReading from: dataset/clips/0313-1/12200/20.jpg\nSaving to: dataset/masks/0313-1/12200\nReading from: dataset/clips/0313-1/15080/20.jpg\nSaving to: dataset/masks/0313-1/15080\nReading from: dataset/clips/0313-1/46560/20.jpg\nSaving to: dataset/masks/0313-1/46560\nReading from: dataset/clips/0313-1/13240/20.jpg\nSaving to: dataset/masks/0313-1/13240\nReading from: dataset/clips/0313-1/16880/20.jpg\nSaving to: dataset/masks/0313-1/16880\nReading from: dataset/clips/0313-1/40860/20.jpg\nSaving to: dataset/masks/0313-1/40860\nReading from: dataset/clips/0313-1/7000/20.jpg\nSaving to: dataset/masks/0313-1/7000\nReading from: dataset/clips/0313-1/29400/20.jpg\nSaving to: dataset/masks/0313-1/29400\nReading from: dataset/clips/0313-1/7020/20.jpg\nSaving to: dataset/masks/0313-1/7020\nReading from: dataset/clips/0313-1/10680/20.jpg\nSaving to: dataset/masks/0313-1/10680\nReading from: dataset/clips/0313-1/9020/20.jpg\nSaving to: dataset/masks/0313-1/9020\nReading from: dataset/clips/0313-1/27660/20.jpg\nSaving to: dataset/masks/0313-1/27660\nReading from: dataset/clips/0313-1/43200/20.jpg\nSaving to: dataset/masks/0313-1/43200\nReading from: dataset/clips/0313-1/9080/20.jpg\nSaving to: dataset/masks/0313-1/9080\nReading from: dataset/clips/0313-1/42060/20.jpg\nSaving to: dataset/masks/0313-1/42060\nReading from: dataset/clips/0313-1/17060/20.jpg\nSaving to: dataset/masks/0313-1/17060\nReading from: dataset/clips/0313-1/4740/20.jpg\nSaving to: dataset/masks/0313-1/4740\nReading from: dataset/clips/0313-1/25380/20.jpg\nSaving to: dataset/masks/0313-1/25380\nReading from: dataset/clips/0313-1/23460/20.jpg\nSaving to: dataset/masks/0313-1/23460\nReading from: dataset/clips/0313-1/5780/20.jpg\nSaving to: dataset/masks/0313-1/5780\nReading from: dataset/clips/0313-1/16900/20.jpg\nSaving to: dataset/masks/0313-1/16900\nReading from: dataset/clips/0313-1/31800/20.jpg\nSaving to: dataset/masks/0313-1/31800\nReading from: dataset/clips/0313-1/16520/20.jpg\nSaving to: dataset/masks/0313-1/16520\nReading from: dataset/clips/0313-1/38820/20.jpg\nSaving to: dataset/masks/0313-1/38820\nReading from: dataset/clips/0313-1/4300/20.jpg\nSaving to: dataset/masks/0313-1/4300\nReading from: dataset/clips/0313-1/23520/20.jpg\nSaving to: dataset/masks/0313-1/23520\nReading from: dataset/clips/0313-1/52260/20.jpg\nSaving to: dataset/masks/0313-1/52260\nReading from: dataset/clips/0313-1/9900/20.jpg\nSaving to: dataset/masks/0313-1/9900\nReading from: dataset/clips/0313-1/38040/20.jpg\nSaving to: dataset/masks/0313-1/38040\nReading from: dataset/clips/0313-1/28440/20.jpg\nSaving to: dataset/masks/0313-1/28440\nReading from: dataset/clips/0313-1/47700/20.jpg\nSaving to: dataset/masks/0313-1/47700\nReading from: dataset/clips/0313-1/8000/20.jpg\nSaving to: dataset/masks/0313-1/8000\nReading from: dataset/clips/0313-1/16300/20.jpg\nSaving to: dataset/masks/0313-1/16300\nReading from: dataset/clips/0313-1/31560/20.jpg\nSaving to: dataset/masks/0313-1/31560\nReading from: dataset/clips/0313-1/52560/20.jpg\nSaving to: dataset/masks/0313-1/52560\nReading from: dataset/clips/0313-1/5700/20.jpg\nSaving to: dataset/masks/0313-1/5700\nReading from: dataset/clips/0313-1/17480/20.jpg\nSaving to: dataset/masks/0313-1/17480\nReading from: dataset/clips/0313-1/17280/20.jpg\nSaving to: dataset/masks/0313-1/17280\nReading from: dataset/clips/0313-1/15300/20.jpg\nSaving to: dataset/masks/0313-1/15300\nReading from: dataset/clips/0313-1/33480/20.jpg\nSaving to: dataset/masks/0313-1/33480\nReading from: dataset/clips/0313-1/17340/20.jpg\nSaving to: dataset/masks/0313-1/17340\nReading from: dataset/clips/0313-1/180/20.jpg\nSaving to: dataset/masks/0313-1/180\nReading from: dataset/clips/0313-1/27900/20.jpg\nSaving to: dataset/masks/0313-1/27900\nReading from: dataset/clips/0313-1/45720/20.jpg\nSaving to: dataset/masks/0313-1/45720\nReading from: dataset/clips/0313-1/2460/20.jpg\nSaving to: dataset/masks/0313-1/2460\nReading from: dataset/clips/0313-1/17460/20.jpg\nSaving to: dataset/masks/0313-1/17460\nReading from: dataset/clips/0313-1/44460/20.jpg\nSaving to: dataset/masks/0313-1/44460\nReading from: dataset/clips/0313-1/22260/20.jpg\nSaving to: dataset/masks/0313-1/22260\nReading from: dataset/clips/0313-1/31380/20.jpg\nSaving to: dataset/masks/0313-1/31380\nReading from: dataset/clips/0313-1/48540/20.jpg\nSaving to: dataset/masks/0313-1/48540\nReading from: dataset/clips/0313-1/23580/20.jpg\nSaving to: dataset/masks/0313-1/23580\nReading from: dataset/clips/0313-1/18840/20.jpg\nSaving to: dataset/masks/0313-1/18840\nReading from: dataset/clips/0313-1/8840/20.jpg\nSaving to: dataset/masks/0313-1/8840\nReading from: dataset/clips/0313-1/42840/20.jpg\nSaving to: dataset/masks/0313-1/42840\nReading from: dataset/clips/0313-1/49860/20.jpg\nSaving to: dataset/masks/0313-1/49860\nReading from: dataset/clips/0313-1/51180/20.jpg\nSaving to: dataset/masks/0313-1/51180\nReading from: dataset/clips/0313-1/38940/20.jpg\nSaving to: dataset/masks/0313-1/38940\nReading from: dataset/clips/0313-1/16360/20.jpg\nSaving to: dataset/masks/0313-1/16360\nReading from: dataset/clips/0313-1/47880/20.jpg\nSaving to: dataset/masks/0313-1/47880\nReading from: dataset/clips/0313-1/28080/20.jpg\nSaving to: dataset/masks/0313-1/28080\nReading from: dataset/clips/0313-1/18900/20.jpg\nSaving to: dataset/masks/0313-1/18900\nReading from: dataset/clips/0313-1/1200/20.jpg\nSaving to: dataset/masks/0313-1/1200\nReading from: dataset/clips/0313-1/10840/20.jpg\nSaving to: dataset/masks/0313-1/10840\nReading from: dataset/clips/0313-1/17120/20.jpg\nSaving to: dataset/masks/0313-1/17120\nReading from: dataset/clips/0313-1/2340/20.jpg\nSaving to: dataset/masks/0313-1/2340\nReading from: dataset/clips/0313-1/50160/20.jpg\nSaving to: dataset/masks/0313-1/50160\nReading from: dataset/clips/0313-1/43980/20.jpg\nSaving to: dataset/masks/0313-1/43980\nReading from: dataset/clips/0313-1/48360/20.jpg\nSaving to: dataset/masks/0313-1/48360\nReading from: dataset/clips/0313-1/11440/20.jpg\nSaving to: dataset/masks/0313-1/11440\nReading from: dataset/clips/0313-1/2760/20.jpg\nSaving to: dataset/masks/0313-1/2760\nReading from: dataset/clips/0313-1/23820/20.jpg\nSaving to: dataset/masks/0313-1/23820\nReading from: dataset/clips/0313-1/14860/20.jpg\nSaving to: dataset/masks/0313-1/14860\nReading from: dataset/clips/0313-1/39780/20.jpg\nSaving to: dataset/masks/0313-1/39780\nReading from: dataset/clips/0313-1/46860/20.jpg\nSaving to: dataset/masks/0313-1/46860\nReading from: dataset/clips/0313-1/23880/20.jpg\nSaving to: dataset/masks/0313-1/23880\nReading from: dataset/clips/0313-1/50040/20.jpg\nSaving to: dataset/masks/0313-1/50040\nReading from: dataset/clips/0313-1/11840/20.jpg\nSaving to: dataset/masks/0313-1/11840\nReading from: dataset/clips/0313-1/20280/20.jpg\nSaving to: dataset/masks/0313-1/20280\nReading from: dataset/clips/0313-1/1680/20.jpg\nSaving to: dataset/masks/0313-1/1680\nReading from: dataset/clips/0313-1/42240/20.jpg\nSaving to: dataset/masks/0313-1/42240\nReading from: dataset/clips/0313-1/47340/20.jpg\nSaving to: dataset/masks/0313-1/47340\nReading from: dataset/clips/0313-1/50820/20.jpg\nSaving to: dataset/masks/0313-1/50820\nReading from: dataset/clips/0313-1/17240/20.jpg\nSaving to: dataset/masks/0313-1/17240\nReading from: dataset/clips/0313-1/4920/20.jpg\nSaving to: dataset/masks/0313-1/4920\nReading from: dataset/clips/0313-1/14400/20.jpg\nSaving to: dataset/masks/0313-1/14400\nReading from: dataset/clips/0313-1/14140/20.jpg\nSaving to: dataset/masks/0313-1/14140\nReading from: dataset/clips/0313-1/21300/20.jpg\nSaving to: dataset/masks/0313-1/21300\nReading from: dataset/clips/0313-1/12820/20.jpg\nSaving to: dataset/masks/0313-1/12820\nReading from: dataset/clips/0313-1/6700/20.jpg\nSaving to: dataset/masks/0313-1/6700\nReading from: dataset/clips/0313-1/5440/20.jpg\nSaving to: dataset/masks/0313-1/5440\nReading from: dataset/clips/0313-1/41220/20.jpg\nSaving to: dataset/masks/0313-1/41220\nReading from: dataset/clips/0313-1/13180/20.jpg\nSaving to: dataset/masks/0313-1/13180\nReading from: dataset/clips/0313-1/24420/20.jpg\nSaving to: dataset/masks/0313-1/24420\nReading from: dataset/clips/0313-1/20160/20.jpg\nSaving to: dataset/masks/0313-1/20160\nReading from: dataset/clips/0313-1/13120/20.jpg\nSaving to: dataset/masks/0313-1/13120\nReading from: dataset/clips/0313-1/8220/20.jpg\nSaving to: dataset/masks/0313-1/8220\nReading from: dataset/clips/0313-1/42300/20.jpg\nSaving to: dataset/masks/0313-1/42300\nReading from: dataset/clips/0313-1/52740/20.jpg\nSaving to: dataset/masks/0313-1/52740\nReading from: dataset/clips/0313-1/3480/20.jpg\nSaving to: dataset/masks/0313-1/3480\nReading from: dataset/clips/0313-1/17220/20.jpg\nSaving to: dataset/masks/0313-1/17220\nReading from: dataset/clips/0313-1/15500/20.jpg\nSaving to: dataset/masks/0313-1/15500\nReading from: dataset/clips/0313-1/26640/20.jpg\nSaving to: dataset/masks/0313-1/26640\nReading from: dataset/clips/0313-1/4340/20.jpg\nSaving to: dataset/masks/0313-1/4340\nReading from: dataset/clips/0313-1/8860/20.jpg\nSaving to: dataset/masks/0313-1/8860\nReading from: dataset/clips/0313-1/26280/20.jpg\nSaving to: dataset/masks/0313-1/26280\nReading from: dataset/clips/0313-1/30660/20.jpg\nSaving to: dataset/masks/0313-1/30660\nReading from: dataset/clips/0313-1/9360/20.jpg\nSaving to: dataset/masks/0313-1/9360\nReading from: dataset/clips/0313-1/6600/20.jpg\nSaving to: dataset/masks/0313-1/6600\nReading from: dataset/clips/0313-1/7440/20.jpg\nSaving to: dataset/masks/0313-1/7440\nReading from: dataset/clips/0313-1/29160/20.jpg\nSaving to: dataset/masks/0313-1/29160\nReading from: dataset/clips/0313-1/25800/20.jpg\nSaving to: dataset/masks/0313-1/25800\nReading from: dataset/clips/0313-1/11600/20.jpg\nSaving to: dataset/masks/0313-1/11600\nReading from: dataset/clips/0313-1/11120/20.jpg\nSaving to: dataset/masks/0313-1/11120\nReading from: dataset/clips/0313-1/7060/20.jpg\nSaving to: dataset/masks/0313-1/7060\nReading from: dataset/clips/0313-1/16040/20.jpg\nSaving to: dataset/masks/0313-1/16040\nReading from: dataset/clips/0313-1/12980/20.jpg\nSaving to: dataset/masks/0313-1/12980\nReading from: dataset/clips/0313-1/27420/20.jpg\nSaving to: dataset/masks/0313-1/27420\nReading from: dataset/clips/0313-1/15100/20.jpg\nSaving to: dataset/masks/0313-1/15100\nReading from: dataset/clips/0313-1/42600/20.jpg\nSaving to: dataset/masks/0313-1/42600\nReading from: dataset/clips/0313-1/27480/20.jpg\nSaving to: dataset/masks/0313-1/27480\nReading from: dataset/clips/0313-1/33300/20.jpg\nSaving to: dataset/masks/0313-1/33300\nReading from: dataset/clips/0313-1/6800/20.jpg\nSaving to: dataset/masks/0313-1/6800\nReading from: dataset/clips/0313-1/32040/20.jpg\nSaving to: dataset/masks/0313-1/32040\nReading from: dataset/clips/0313-1/8100/20.jpg\nSaving to: dataset/masks/0313-1/8100\nReading from: dataset/clips/0313-1/37320/20.jpg\nSaving to: dataset/masks/0313-1/37320\nReading from: dataset/clips/0313-1/5580/20.jpg\nSaving to: dataset/masks/0313-1/5580\nReading from: dataset/clips/0313-1/36180/20.jpg\nSaving to: dataset/masks/0313-1/36180\nReading from: dataset/clips/0313-1/12880/20.jpg\nSaving to: dataset/masks/0313-1/12880\nReading from: dataset/clips/0313-1/46080/20.jpg\nSaving to: dataset/masks/0313-1/46080\nReading from: dataset/clips/0313-1/19800/20.jpg\nSaving to: dataset/masks/0313-1/19800\nReading from: dataset/clips/0313-1/49620/20.jpg\nSaving to: dataset/masks/0313-1/49620\nReading from: dataset/clips/0313-1/17620/20.jpg\nSaving to: dataset/masks/0313-1/17620\nReading from: dataset/clips/0313-1/7980/20.jpg\nSaving to: dataset/masks/0313-1/7980\nReading from: dataset/clips/0313-1/14660/20.jpg\nSaving to: dataset/masks/0313-1/14660\nReading from: dataset/clips/0313-1/22920/20.jpg\nSaving to: dataset/masks/0313-1/22920\nReading from: dataset/clips/0313-1/36780/20.jpg\nSaving to: dataset/masks/0313-1/36780\nReading from: dataset/clips/0313-1/15860/20.jpg\nSaving to: dataset/masks/0313-1/15860\nReading from: dataset/clips/0313-1/5920/20.jpg\nSaving to: dataset/masks/0313-1/5920\nReading from: dataset/clips/0313-1/1020/20.jpg\nSaving to: dataset/masks/0313-1/1020\nReading from: dataset/clips/0313-1/41700/20.jpg\nSaving to: dataset/masks/0313-1/41700\nReading from: dataset/clips/0313-1/41760/20.jpg\nSaving to: dataset/masks/0313-1/41760\nReading from: dataset/clips/0313-1/18600/20.jpg\nSaving to: dataset/masks/0313-1/18600\nReading from: dataset/clips/0313-1/51540/20.jpg\nSaving to: dataset/masks/0313-1/51540\nReading from: dataset/clips/0313-1/16680/20.jpg\nSaving to: dataset/masks/0313-1/16680\nReading from: dataset/clips/0313-1/4380/20.jpg\nSaving to: dataset/masks/0313-1/4380\nReading from: dataset/clips/0313-1/5720/20.jpg\nSaving to: dataset/masks/0313-1/5720\nReading from: dataset/clips/0313-1/42120/20.jpg\nSaving to: dataset/masks/0313-1/42120\nReading from: dataset/clips/0313-1/7660/20.jpg\nSaving to: dataset/masks/0313-1/7660\nReading from: dataset/clips/0313-1/26100/20.jpg\nSaving to: dataset/masks/0313-1/26100\nReading from: dataset/clips/0313-1/17600/20.jpg\nSaving to: dataset/masks/0313-1/17600\nReading from: dataset/clips/0313-1/39480/20.jpg\nSaving to: dataset/masks/0313-1/39480\nReading from: dataset/clips/0313-1/17260/20.jpg\nSaving to: dataset/masks/0313-1/17260\nReading from: dataset/clips/0313-1/13660/20.jpg\nSaving to: dataset/masks/0313-1/13660\nReading from: dataset/clips/0313-1/9960/20.jpg\nSaving to: dataset/masks/0313-1/9960\nReading from: dataset/clips/0313-1/2160/20.jpg\nSaving to: dataset/masks/0313-1/2160\nReading from: dataset/clips/0313-1/39840/20.jpg\nSaving to: dataset/masks/0313-1/39840\nReading from: dataset/clips/0313-1/21060/20.jpg\nSaving to: dataset/masks/0313-1/21060\nReading from: dataset/clips/0313-1/27120/20.jpg\nSaving to: dataset/masks/0313-1/27120\nReading from: dataset/clips/0313-1/48840/20.jpg\nSaving to: dataset/masks/0313-1/48840\nReading from: dataset/clips/0313-1/8540/20.jpg\nSaving to: dataset/masks/0313-1/8540\nReading from: dataset/clips/0313-1/23280/20.jpg\nSaving to: dataset/masks/0313-1/23280\nReading from: dataset/clips/0313-1/1140/20.jpg\nSaving to: dataset/masks/0313-1/1140\nReading from: dataset/clips/0313-1/34980/20.jpg\nSaving to: dataset/masks/0313-1/34980\nReading from: dataset/clips/0313-1/38640/20.jpg\nSaving to: dataset/masks/0313-1/38640\nReading from: dataset/clips/0313-1/23160/20.jpg\nSaving to: dataset/masks/0313-1/23160\nReading from: dataset/clips/0313-1/12140/20.jpg\nSaving to: dataset/masks/0313-1/12140\nReading from: dataset/clips/0313-1/52140/20.jpg\nSaving to: dataset/masks/0313-1/52140\nReading from: dataset/clips/0313-1/15340/20.jpg\nSaving to: dataset/masks/0313-1/15340\nReading from: dataset/clips/0313-1/52860/20.jpg\nSaving to: dataset/masks/0313-1/52860\nReading from: dataset/clips/0313-1/12220/20.jpg\nSaving to: dataset/masks/0313-1/12220\nReading from: dataset/clips/0313-1/17140/20.jpg\nSaving to: dataset/masks/0313-1/17140\nReading from: dataset/clips/0313-1/41640/20.jpg\nSaving to: dataset/masks/0313-1/41640\nReading from: dataset/clips/0313-1/31020/20.jpg\nSaving to: dataset/masks/0313-1/31020\nReading from: dataset/clips/0313-1/8740/20.jpg\nSaving to: dataset/masks/0313-1/8740\nReading from: dataset/clips/0313-1/5880/20.jpg\nSaving to: dataset/masks/0313-1/5880\nReading from: dataset/clips/0313-1/16120/20.jpg\nSaving to: dataset/masks/0313-1/16120\nReading from: dataset/clips/0313-1/13580/20.jpg\nSaving to: dataset/masks/0313-1/13580\nReading from: dataset/clips/0313-1/8420/20.jpg\nSaving to: dataset/masks/0313-1/8420\nReading from: dataset/clips/0313-1/10340/20.jpg\nSaving to: dataset/masks/0313-1/10340\nReading from: dataset/clips/0313-1/3060/20.jpg\nSaving to: dataset/masks/0313-1/3060\nReading from: dataset/clips/0313-1/30900/20.jpg\nSaving to: dataset/masks/0313-1/30900\nReading from: dataset/clips/0313-1/5140/20.jpg\nSaving to: dataset/masks/0313-1/5140\nReading from: dataset/clips/0313-1/14820/20.jpg\nSaving to: dataset/masks/0313-1/14820\nReading from: dataset/clips/0313-1/45000/20.jpg\nSaving to: dataset/masks/0313-1/45000\nReading from: dataset/clips/0313-1/720/20.jpg\nSaving to: dataset/masks/0313-1/720\nReading from: dataset/clips/0313-1/9200/20.jpg\nSaving to: dataset/masks/0313-1/9200\nReading from: dataset/clips/0313-1/13200/20.jpg\nSaving to: dataset/masks/0313-1/13200\nReading from: dataset/clips/0313-1/51420/20.jpg\nSaving to: dataset/masks/0313-1/51420\nReading from: dataset/clips/0313-1/13360/20.jpg\nSaving to: dataset/masks/0313-1/13360\nReading from: dataset/clips/0313-1/24780/20.jpg\nSaving to: dataset/masks/0313-1/24780\nReading from: dataset/clips/0313-1/37260/20.jpg\nSaving to: dataset/masks/0313-1/37260\nReading from: dataset/clips/0313-1/4560/20.jpg\nSaving to: dataset/masks/0313-1/4560\nReading from: dataset/clips/0313-1/24720/20.jpg\nSaving to: dataset/masks/0313-1/24720\nReading from: dataset/clips/0313-1/11620/20.jpg\nSaving to: dataset/masks/0313-1/11620\nReading from: dataset/clips/0313-1/14380/20.jpg\nSaving to: dataset/masks/0313-1/14380\nReading from: dataset/clips/0313-1/12580/20.jpg\nSaving to: dataset/masks/0313-1/12580\nReading from: dataset/clips/0313-1/9700/20.jpg\nSaving to: dataset/masks/0313-1/9700\nReading from: dataset/clips/0313-1/15440/20.jpg\nSaving to: dataset/masks/0313-1/15440\nReading from: dataset/clips/0313-1/26700/20.jpg\nSaving to: dataset/masks/0313-1/26700\nReading from: dataset/clips/0313-1/47280/20.jpg\nSaving to: dataset/masks/0313-1/47280\nReading from: dataset/clips/0313-1/33900/20.jpg\nSaving to: dataset/masks/0313-1/33900\nReading from: dataset/clips/0313-1/30300/20.jpg\nSaving to: dataset/masks/0313-1/30300\nReading from: dataset/clips/0313-1/7500/20.jpg\nSaving to: dataset/masks/0313-1/7500\nReading from: dataset/clips/0313-1/40620/20.jpg\nSaving to: dataset/masks/0313-1/40620\nReading from: dataset/clips/0313-1/21240/20.jpg\nSaving to: dataset/masks/0313-1/21240\nReading from: dataset/clips/0313-1/6280/20.jpg\nSaving to: dataset/masks/0313-1/6280\nReading from: dataset/clips/0313-1/4440/20.jpg\nSaving to: dataset/masks/0313-1/4440\nReading from: dataset/clips/0313-1/16700/20.jpg\nSaving to: dataset/masks/0313-1/16700\nReading from: dataset/clips/0313-1/29220/20.jpg\nSaving to: dataset/masks/0313-1/29220\nReading from: dataset/clips/0313-1/18960/20.jpg\nSaving to: dataset/masks/0313-1/18960\nReading from: dataset/clips/0313-1/17180/20.jpg\nSaving to: dataset/masks/0313-1/17180\nReading from: dataset/clips/0313-1/34680/20.jpg\nSaving to: dataset/masks/0313-1/34680\nReading from: dataset/clips/0313-1/49740/20.jpg\nSaving to: dataset/masks/0313-1/49740\nReading from: dataset/clips/0313-1/32880/20.jpg\nSaving to: dataset/masks/0313-1/32880\nReading from: dataset/clips/0313-1/9320/20.jpg\nSaving to: dataset/masks/0313-1/9320\nReading from: dataset/clips/0313-1/14160/20.jpg\nSaving to: dataset/masks/0313-1/14160\nReading from: dataset/clips/0313-1/5240/20.jpg\nSaving to: dataset/masks/0313-1/5240\nReading from: dataset/clips/0313-1/20820/20.jpg\nSaving to: dataset/masks/0313-1/20820\nReading from: dataset/clips/0313-1/19440/20.jpg\nSaving to: dataset/masks/0313-1/19440\nReading from: dataset/clips/0313-1/9300/20.jpg\nSaving to: dataset/masks/0313-1/9300\nReading from: dataset/clips/0313-1/10420/20.jpg\nSaving to: dataset/masks/0313-1/10420\nReading from: dataset/clips/0313-1/41940/20.jpg\nSaving to: dataset/masks/0313-1/41940\nReading from: dataset/clips/0313-1/25980/20.jpg\nSaving to: dataset/masks/0313-1/25980\nReading from: dataset/clips/0313-1/16440/20.jpg\nSaving to: dataset/masks/0313-1/16440\nReading from: dataset/clips/0313-1/6660/20.jpg\nSaving to: dataset/masks/0313-1/6660\nReading from: dataset/clips/0313-1/51060/20.jpg\nSaving to: dataset/masks/0313-1/51060\nReading from: dataset/clips/0313-1/16780/20.jpg\nSaving to: dataset/masks/0313-1/16780\nReading from: dataset/clips/0313-1/21480/20.jpg\nSaving to: dataset/masks/0313-1/21480\nReading from: dataset/clips/0313-1/36600/20.jpg\nSaving to: dataset/masks/0313-1/36600\nReading from: dataset/clips/0313-1/34920/20.jpg\nSaving to: dataset/masks/0313-1/34920\nReading from: dataset/clips/0313-1/15660/20.jpg\nSaving to: dataset/masks/0313-1/15660\nReading from: dataset/clips/0313-1/47580/20.jpg\nSaving to: dataset/masks/0313-1/47580\nReading from: dataset/clips/0313-1/12320/20.jpg\nSaving to: dataset/masks/0313-1/12320\nReading from: dataset/clips/0313-1/3940/20.jpg\nSaving to: dataset/masks/0313-1/3940\nReading from: dataset/clips/0313-1/6100/20.jpg\nSaving to: dataset/masks/0313-1/6100\nReading from: dataset/clips/0313-1/40380/20.jpg\nSaving to: dataset/masks/0313-1/40380\nReading from: dataset/clips/0313-1/15380/20.jpg\nSaving to: dataset/masks/0313-1/15380\nReading from: dataset/clips/0313-1/7940/20.jpg\nSaving to: dataset/masks/0313-1/7940\nReading from: dataset/clips/0313-1/4180/20.jpg\nSaving to: dataset/masks/0313-1/4180\nReading from: dataset/clips/0313-1/36900/20.jpg\nSaving to: dataset/masks/0313-1/36900\nReading from: dataset/clips/0313-1/25140/20.jpg\nSaving to: dataset/masks/0313-1/25140\nReading from: dataset/clips/0313-1/28680/20.jpg\nSaving to: dataset/masks/0313-1/28680\nReading from: dataset/clips/0313-1/12680/20.jpg\nSaving to: dataset/masks/0313-1/12680\nReading from: dataset/clips/0313-1/48060/20.jpg\nSaving to: dataset/masks/0313-1/48060\nReading from: dataset/clips/0313-1/14120/20.jpg\nSaving to: dataset/masks/0313-1/14120\nReading from: dataset/clips/0313-1/15140/20.jpg\nSaving to: dataset/masks/0313-1/15140\nReading from: dataset/clips/0313-1/24060/20.jpg\nSaving to: dataset/masks/0313-1/24060\nReading from: dataset/clips/0313-1/47040/20.jpg\nSaving to: dataset/masks/0313-1/47040\nReading from: dataset/clips/0313-1/9720/20.jpg\nSaving to: dataset/masks/0313-1/9720\nReading from: dataset/clips/0313-1/11800/20.jpg\nSaving to: dataset/masks/0313-1/11800\nReading from: dataset/clips/0313-1/960/20.jpg\nSaving to: dataset/masks/0313-1/960\nReading from: dataset/clips/0313-1/51360/20.jpg\nSaving to: dataset/masks/0313-1/51360\nReading from: dataset/clips/0313-1/4860/20.jpg\nSaving to: dataset/masks/0313-1/4860\nReading from: dataset/clips/0313-1/9880/20.jpg\nSaving to: dataset/masks/0313-1/9880\nReading from: dataset/clips/0313-1/6680/20.jpg\nSaving to: dataset/masks/0313-1/6680\nReading from: dataset/clips/0313-1/48240/20.jpg\nSaving to: dataset/masks/0313-1/48240\nReading from: dataset/clips/0313-1/6820/20.jpg\nSaving to: dataset/masks/0313-1/6820\nReading from: dataset/clips/0313-1/16160/20.jpg\nSaving to: dataset/masks/0313-1/16160\nReading from: dataset/clips/0313-1/8260/20.jpg\nSaving to: dataset/masks/0313-1/8260\nReading from: dataset/clips/0313-1/29940/20.jpg\nSaving to: dataset/masks/0313-1/29940\nReading from: dataset/clips/0313-1/33240/20.jpg\nSaving to: dataset/masks/0313-1/33240\nReading from: dataset/clips/0313-1/40080/20.jpg\nSaving to: dataset/masks/0313-1/40080\nReading from: dataset/clips/0313-1/17660/20.jpg\nSaving to: dataset/masks/0313-1/17660\nReading from: dataset/clips/0313-1/3540/20.jpg\nSaving to: dataset/masks/0313-1/3540\nReading from: dataset/clips/0313-1/15840/20.jpg\nSaving to: dataset/masks/0313-1/15840\nReading from: dataset/clips/0313-1/11560/20.jpg\nSaving to: dataset/masks/0313-1/11560\nReading from: dataset/clips/0313-1/12000/20.jpg\nSaving to: dataset/masks/0313-1/12000\nReading from: dataset/clips/0313-1/39120/20.jpg\nSaving to: dataset/masks/0313-1/39120\nReading from: dataset/clips/0313-1/49320/20.jpg\nSaving to: dataset/masks/0313-1/49320\nReading from: dataset/clips/0313-1/49080/20.jpg\nSaving to: dataset/masks/0313-1/49080\nReading from: dataset/clips/0313-1/25020/20.jpg\nSaving to: dataset/masks/0313-1/25020\nReading from: dataset/clips/0313-1/14020/20.jpg\nSaving to: dataset/masks/0313-1/14020\nReading from: dataset/clips/0313-1/45540/20.jpg\nSaving to: dataset/masks/0313-1/45540\nReading from: dataset/clips/0313-1/51840/20.jpg\nSaving to: dataset/masks/0313-1/51840\nReading from: dataset/clips/0313-1/16860/20.jpg\nSaving to: dataset/masks/0313-1/16860\nReading from: dataset/clips/0313-1/4780/20.jpg\nSaving to: dataset/masks/0313-1/4780\nReading from: dataset/clips/0313-1/22320/20.jpg\nSaving to: dataset/masks/0313-1/22320\nReading from: dataset/clips/0313-1/52980/20.jpg\nSaving to: dataset/masks/0313-1/52980\nReading from: dataset/clips/0313-1/11140/20.jpg\nSaving to: dataset/masks/0313-1/11140\nReading from: dataset/clips/0313-1/51960/20.jpg\nSaving to: dataset/masks/0313-1/51960\nReading from: dataset/clips/0313-1/11820/20.jpg\nSaving to: dataset/masks/0313-1/11820\nReading from: dataset/clips/0313-1/52620/20.jpg\nSaving to: dataset/masks/0313-1/52620\nReading from: dataset/clips/0313-1/50880/20.jpg\nSaving to: dataset/masks/0313-1/50880\nReading from: dataset/clips/0313-1/30120/20.jpg\nSaving to: dataset/masks/0313-1/30120\nReading from: dataset/clips/0313-1/14640/20.jpg\nSaving to: dataset/masks/0313-1/14640\nReading from: dataset/clips/0313-1/50700/20.jpg\nSaving to: dataset/masks/0313-1/50700\nReading from: dataset/clips/0313-1/7520/20.jpg\nSaving to: dataset/masks/0313-1/7520\nReading from: dataset/clips/0313-1/33720/20.jpg\nSaving to: dataset/masks/0313-1/33720\nReading from: dataset/clips/0313-1/12260/20.jpg\nSaving to: dataset/masks/0313-1/12260\nReading from: dataset/clips/0313-1/15800/20.jpg\nSaving to: dataset/masks/0313-1/15800\nReading from: dataset/clips/0313-1/27780/20.jpg\nSaving to: dataset/masks/0313-1/27780\nReading from: dataset/clips/0313-1/9120/20.jpg\nSaving to: dataset/masks/0313-1/9120\nReading from: dataset/clips/0313-1/7100/20.jpg\nSaving to: dataset/masks/0313-1/7100\nReading from: dataset/clips/0313-1/32100/20.jpg\nSaving to: dataset/masks/0313-1/32100\nReading from: dataset/clips/0313-1/35820/20.jpg\nSaving to: dataset/masks/0313-1/35820\nReading from: dataset/clips/0313-1/5040/20.jpg\nSaving to: dataset/masks/0313-1/5040\nReading from: dataset/clips/0313-1/22860/20.jpg\nSaving to: dataset/masks/0313-1/22860\nReading from: dataset/clips/0313-1/11900/20.jpg\nSaving to: dataset/masks/0313-1/11900\nReading from: dataset/clips/0313-1/22380/20.jpg\nSaving to: dataset/masks/0313-1/22380\nReading from: dataset/clips/0313-1/8820/20.jpg\nSaving to: dataset/masks/0313-1/8820\nReading from: dataset/clips/0313-1/16980/20.jpg\nSaving to: dataset/masks/0313-1/16980\nReading from: dataset/clips/0313-1/40560/20.jpg\nSaving to: dataset/masks/0313-1/40560\nReading from: dataset/clips/0313-1/6220/20.jpg\nSaving to: dataset/masks/0313-1/6220\nReading from: dataset/clips/0313-1/12020/20.jpg\nSaving to: dataset/masks/0313-1/12020\nReading from: dataset/clips/0313-1/32400/20.jpg\nSaving to: dataset/masks/0313-1/32400\nReading from: dataset/clips/0313-1/21840/20.jpg\nSaving to: dataset/masks/0313-1/21840\nReading from: dataset/clips/0313-1/15940/20.jpg\nSaving to: dataset/masks/0313-1/15940\nReading from: dataset/clips/0313-1/14320/20.jpg\nSaving to: dataset/masks/0313-1/14320\nReading from: dataset/clips/0313-1/37680/20.jpg\nSaving to: dataset/masks/0313-1/37680\nReading from: dataset/clips/0313-1/8580/20.jpg\nSaving to: dataset/masks/0313-1/8580\nReading from: dataset/clips/0313-1/7040/20.jpg\nSaving to: dataset/masks/0313-1/7040\nReading from: dataset/clips/0313-1/8660/20.jpg\nSaving to: dataset/masks/0313-1/8660\nReading from: dataset/clips/0313-1/5280/20.jpg\nSaving to: dataset/masks/0313-1/5280\nReading from: dataset/clips/0313-1/9180/20.jpg\nSaving to: dataset/masks/0313-1/9180\nReading from: dataset/clips/0313-1/27720/20.jpg\nSaving to: dataset/masks/0313-1/27720\nReading from: dataset/clips/0313-1/14800/20.jpg\nSaving to: dataset/masks/0313-1/14800\nReading from: dataset/clips/0313-1/39240/20.jpg\nSaving to: dataset/masks/0313-1/39240\nReading from: dataset/clips/0313-1/21600/20.jpg\nSaving to: dataset/masks/0313-1/21600\nReading from: dataset/clips/0313-1/34740/20.jpg\nSaving to: dataset/masks/0313-1/34740\nReading from: dataset/clips/0313-1/240/20.jpg\nSaving to: dataset/masks/0313-1/240\nReading from: dataset/clips/0313-1/11520/20.jpg\nSaving to: dataset/masks/0313-1/11520\nReading from: dataset/clips/0313-1/41520/20.jpg\nSaving to: dataset/masks/0313-1/41520\nReading from: dataset/clips/0313-1/13380/20.jpg\nSaving to: dataset/masks/0313-1/13380\nReading from: dataset/clips/0313-1/12280/20.jpg\nSaving to: dataset/masks/0313-1/12280\nReading from: dataset/clips/0313-1/12160/20.jpg\nSaving to: dataset/masks/0313-1/12160\nReading from: dataset/clips/0313-1/15360/20.jpg\nSaving to: dataset/masks/0313-1/15360\nReading from: dataset/clips/0313-1/4520/20.jpg\nSaving to: dataset/masks/0313-1/4520\nReading from: dataset/clips/0313-1/14180/20.jpg\nSaving to: dataset/masks/0313-1/14180\nReading from: dataset/clips/0313-1/26340/20.jpg\nSaving to: dataset/masks/0313-1/26340\nReading from: dataset/clips/0313-1/37740/20.jpg\nSaving to: dataset/masks/0313-1/37740\nReading from: dataset/clips/0313-1/6460/20.jpg\nSaving to: dataset/masks/0313-1/6460\nReading from: dataset/clips/0313-1/6380/20.jpg\nSaving to: dataset/masks/0313-1/6380\nReading from: dataset/clips/0313-1/9340/20.jpg\nSaving to: dataset/masks/0313-1/9340\nReading from: dataset/clips/0313-1/43080/20.jpg\nSaving to: dataset/masks/0313-1/43080\nReading from: dataset/clips/0313-1/11740/20.jpg\nSaving to: dataset/masks/0313-1/11740\nReading from: dataset/clips/0313-1/11220/20.jpg\nSaving to: dataset/masks/0313-1/11220\nReading from: dataset/clips/0313-1/16620/20.jpg\nSaving to: dataset/masks/0313-1/16620\nReading from: dataset/clips/0313-1/27240/20.jpg\nSaving to: dataset/masks/0313-1/27240\nReading from: dataset/clips/0313-1/6240/20.jpg\nSaving to: dataset/masks/0313-1/6240\nReading from: dataset/clips/0313-1/26460/20.jpg\nSaving to: dataset/masks/0313-1/26460\nReading from: dataset/clips/0313-1/43560/20.jpg\nSaving to: dataset/masks/0313-1/43560\nReading from: dataset/clips/0313-1/8520/20.jpg\nSaving to: dataset/masks/0313-1/8520\nReading from: dataset/clips/0313-1/660/20.jpg\nSaving to: dataset/masks/0313-1/660\nReading from: dataset/clips/0313-1/10640/20.jpg\nSaving to: dataset/masks/0313-1/10640\nReading from: dataset/clips/0313-1/46620/20.jpg\nSaving to: dataset/masks/0313-1/46620\nReading from: dataset/clips/0313-1/24900/20.jpg\nSaving to: dataset/masks/0313-1/24900\nReading from: dataset/clips/0313-1/5200/20.jpg\nSaving to: dataset/masks/0313-1/5200\nReading from: dataset/clips/0313-1/14880/20.jpg\nSaving to: dataset/masks/0313-1/14880\nReading from: dataset/clips/0313-1/9640/20.jpg\nSaving to: dataset/masks/0313-1/9640\nReading from: dataset/clips/0313-1/38400/20.jpg\nSaving to: dataset/masks/0313-1/38400\nReading from: dataset/clips/0313-1/9280/20.jpg\nSaving to: dataset/masks/0313-1/9280\nReading from: dataset/clips/0313-1/12960/20.jpg\nSaving to: dataset/masks/0313-1/12960\nReading from: dataset/clips/0313-1/12900/20.jpg\nSaving to: dataset/masks/0313-1/12900\nReading from: dataset/clips/0313-1/18540/20.jpg\nSaving to: dataset/masks/0313-1/18540\nReading from: dataset/clips/0313-1/5800/20.jpg\nSaving to: dataset/masks/0313-1/5800\nReading from: dataset/clips/0313-1/2640/20.jpg\nSaving to: dataset/masks/0313-1/2640\nReading from: dataset/clips/0313-1/4420/20.jpg\nSaving to: dataset/masks/0313-1/4420\nReading from: dataset/clips/0313-1/16420/20.jpg\nSaving to: dataset/masks/0313-1/16420\nReading from: dataset/clips/0313-1/3120/20.jpg\nSaving to: dataset/masks/0313-1/3120\nReading from: dataset/clips/0313-1/13340/20.jpg\nSaving to: dataset/masks/0313-1/13340\nReading from: dataset/clips/0313-1/15160/20.jpg\nSaving to: dataset/masks/0313-1/15160\nReading from: dataset/clips/0313-1/17640/20.jpg\nSaving to: dataset/masks/0313-1/17640\nReading from: dataset/clips/0313-1/21360/20.jpg\nSaving to: dataset/masks/0313-1/21360\nReading from: dataset/clips/0313-1/27360/20.jpg\nSaving to: dataset/masks/0313-1/27360\nReading from: dataset/clips/0313-1/17440/20.jpg\nSaving to: dataset/masks/0313-1/17440\nReading from: dataset/clips/0313-1/6980/20.jpg\nSaving to: dataset/masks/0313-1/6980\nReading from: dataset/clips/0313-1/15400/20.jpg\nSaving to: dataset/masks/0313-1/15400\nReading from: dataset/clips/0313-1/8900/20.jpg\nSaving to: dataset/masks/0313-1/8900\nReading from: dataset/clips/0313-1/25200/20.jpg\nSaving to: dataset/masks/0313-1/25200\nReading from: dataset/clips/0313-1/36480/20.jpg\nSaving to: dataset/masks/0313-1/36480\nReading from: dataset/clips/0313-1/11660/20.jpg\nSaving to: dataset/masks/0313-1/11660\nReading from: dataset/clips/0313-1/47220/20.jpg\nSaving to: dataset/masks/0313-1/47220\nReading from: dataset/clips/0313-1/51600/20.jpg\nSaving to: dataset/masks/0313-1/51600\nReading from: dataset/clips/0313-1/22740/20.jpg\nSaving to: dataset/masks/0313-1/22740\nReading from: dataset/clips/0313-1/35100/20.jpg\nSaving to: dataset/masks/0313-1/35100\nReading from: dataset/clips/0313-1/2700/20.jpg\nSaving to: dataset/masks/0313-1/2700\nReading from: dataset/clips/0313-1/32700/20.jpg\nSaving to: dataset/masks/0313-1/32700\nReading from: dataset/clips/0313-1/39180/20.jpg\nSaving to: dataset/masks/0313-1/39180\nReading from: dataset/clips/0313-1/11200/20.jpg\nSaving to: dataset/masks/0313-1/11200\nReading from: dataset/clips/0313-1/9920/20.jpg\nSaving to: dataset/masks/0313-1/9920\nReading from: dataset/clips/0313-1/49800/20.jpg\nSaving to: dataset/masks/0313-1/49800\nReading from: dataset/clips/0313-1/14060/20.jpg\nSaving to: dataset/masks/0313-1/14060\nReading from: dataset/clips/0313-1/11960/20.jpg\nSaving to: dataset/masks/0313-1/11960\nReading from: dataset/clips/0313-1/22140/20.jpg\nSaving to: dataset/masks/0313-1/22140\nReading from: dataset/clips/0313-1/13520/20.jpg\nSaving to: dataset/masks/0313-1/13520\nReading from: dataset/clips/0313-1/4640/20.jpg\nSaving to: dataset/masks/0313-1/4640\nReading from: dataset/clips/0313-1/11160/20.jpg\nSaving to: dataset/masks/0313-1/11160\nReading from: dataset/clips/0313-1/5100/20.jpg\nSaving to: dataset/masks/0313-1/5100\nReading from: dataset/clips/0313-1/19020/20.jpg\nSaving to: dataset/masks/0313-1/19020\nReading from: dataset/clips/0313-1/17520/20.jpg\nSaving to: dataset/masks/0313-1/17520\nReading from: dataset/clips/0313-1/22560/20.jpg\nSaving to: dataset/masks/0313-1/22560\nReading from: dataset/clips/0313-1/3660/20.jpg\nSaving to: dataset/masks/0313-1/3660\nReading from: dataset/clips/0313-1/43740/20.jpg\nSaving to: dataset/masks/0313-1/43740\nReading from: dataset/clips/0313-1/28500/20.jpg\nSaving to: dataset/masks/0313-1/28500\nReading from: dataset/clips/0313-1/10480/20.jpg\nSaving to: dataset/masks/0313-1/10480\nReading from: dataset/clips/0313-1/21660/20.jpg\nSaving to: dataset/masks/0313-1/21660\nReading from: dataset/clips/0313-1/15980/20.jpg\nSaving to: dataset/masks/0313-1/15980\nReading from: dataset/clips/0313-1/10120/20.jpg\nSaving to: dataset/masks/0313-1/10120\nReading from: dataset/clips/0313-1/7380/20.jpg\nSaving to: dataset/masks/0313-1/7380\nReading from: dataset/clips/0313-1/13680/20.jpg\nSaving to: dataset/masks/0313-1/13680\nReading from: dataset/clips/0313-1/19740/20.jpg\nSaving to: dataset/masks/0313-1/19740\nReading from: dataset/clips/0313-1/8120/20.jpg\nSaving to: dataset/masks/0313-1/8120\nReading from: dataset/clips/0313-1/17800/20.jpg\nSaving to: dataset/masks/0313-1/17800\nReading from: dataset/clips/0313-1/15120/20.jpg\nSaving to: dataset/masks/0313-1/15120\nReading from: dataset/clips/0313-1/4040/20.jpg\nSaving to: dataset/masks/0313-1/4040\nReading from: dataset/clips/0313-1/13060/20.jpg\nSaving to: dataset/masks/0313-1/13060\nReading from: dataset/clips/0313-1/13720/20.jpg\nSaving to: dataset/masks/0313-1/13720\nReading from: dataset/clips/0313-1/12700/20.jpg\nSaving to: dataset/masks/0313-1/12700\nReading from: dataset/clips/0313-1/39960/20.jpg\nSaving to: dataset/masks/0313-1/39960\nReading from: dataset/clips/0313-1/14600/20.jpg\nSaving to: dataset/masks/0313-1/14600\nReading from: dataset/clips/0313-1/8640/20.jpg\nSaving to: dataset/masks/0313-1/8640\nReading from: dataset/clips/0313-1/12100/20.jpg\nSaving to: dataset/masks/0313-1/12100\nReading from: dataset/clips/0313-1/17020/20.jpg\nSaving to: dataset/masks/0313-1/17020\nReading from: dataset/clips/0313-1/14260/20.jpg\nSaving to: dataset/masks/0313-1/14260\nReading from: dataset/clips/0313-1/5000/20.jpg\nSaving to: dataset/masks/0313-1/5000\nReading from: dataset/clips/0313-1/17380/20.jpg\nSaving to: dataset/masks/0313-1/17380\nReading from: dataset/clips/0313-1/16280/20.jpg\nSaving to: dataset/masks/0313-1/16280\nReading from: dataset/clips/0313-1/18420/20.jpg\nSaving to: dataset/masks/0313-1/18420\nReading from: dataset/clips/0313-1/45900/20.jpg\nSaving to: dataset/masks/0313-1/45900\nReading from: dataset/clips/0313-1/41280/20.jpg\nSaving to: dataset/masks/0313-1/41280\nReading from: dataset/clips/0313-1/15900/20.jpg\nSaving to: dataset/masks/0313-1/15900\nReading from: dataset/clips/0313-1/14520/20.jpg\nSaving to: dataset/masks/0313-1/14520\nReading from: dataset/clips/0313-1/12420/20.jpg\nSaving to: dataset/masks/0313-1/12420\nReading from: dataset/clips/0313-1/9680/20.jpg\nSaving to: dataset/masks/0313-1/9680\nReading from: dataset/clips/0313-1/4680/20.jpg\nSaving to: dataset/masks/0313-1/4680\nReading from: dataset/clips/0313-1/8700/20.jpg\nSaving to: dataset/masks/0313-1/8700\nReading from: dataset/clips/0313-1/9560/20.jpg\nSaving to: dataset/masks/0313-1/9560\nReading from: dataset/clips/0313-1/42540/20.jpg\nSaving to: dataset/masks/0313-1/42540\nReading from: dataset/clips/0313-1/14840/20.jpg\nSaving to: dataset/masks/0313-1/14840\nReading from: dataset/clips/0313-1/49020/20.jpg\nSaving to: dataset/masks/0313-1/49020\nReading from: dataset/clips/0313-1/25080/20.jpg\nSaving to: dataset/masks/0313-1/25080\nReading from: dataset/clips/0313-1/26160/20.jpg\nSaving to: dataset/masks/0313-1/26160\nReading from: dataset/clips/0313-1/42900/20.jpg\nSaving to: dataset/masks/0313-1/42900\nReading from: dataset/clips/0313-1/42720/20.jpg\nSaving to: dataset/masks/0313-1/42720\nReading from: dataset/clips/0313-1/11080/20.jpg\nSaving to: dataset/masks/0313-1/11080\nReading from: dataset/clips/0313-1/13320/20.jpg\nSaving to: dataset/masks/0313-1/13320\nReading from: dataset/clips/0313-1/4580/20.jpg\nSaving to: dataset/masks/0313-1/4580\nReading from: dataset/clips/0313-1/21420/20.jpg\nSaving to: dataset/masks/0313-1/21420\nReading from: dataset/clips/0313-1/35040/20.jpg\nSaving to: dataset/masks/0313-1/35040\nReading from: dataset/clips/0313-1/6620/20.jpg\nSaving to: dataset/masks/0313-1/6620\nReading from: dataset/clips/0313-1/13480/20.jpg\nSaving to: dataset/masks/0313-1/13480\nReading from: dataset/clips/0313-1/36840/20.jpg\nSaving to: dataset/masks/0313-1/36840\nReading from: dataset/clips/0313-1/4540/20.jpg\nSaving to: dataset/masks/0313-1/4540\nReading from: dataset/clips/0313-1/14300/20.jpg\nSaving to: dataset/masks/0313-1/14300\nReading from: dataset/clips/0313-1/17360/20.jpg\nSaving to: dataset/masks/0313-1/17360\nReading from: dataset/clips/0313-1/40320/20.jpg\nSaving to: dataset/masks/0313-1/40320\nReading from: dataset/clips/0313-1/50940/20.jpg\nSaving to: dataset/masks/0313-1/50940\nReading from: dataset/clips/0313-1/33780/20.jpg\nSaving to: dataset/masks/0313-1/33780\nReading from: dataset/clips/0313-1/13560/20.jpg\nSaving to: dataset/masks/0313-1/13560\nReading from: dataset/clips/0313-1/29520/20.jpg\nSaving to: dataset/masks/0313-1/29520\nReading from: dataset/clips/0313-1/15740/20.jpg\nSaving to: dataset/masks/0313-1/15740\nReading from: dataset/clips/0313-1/3900/20.jpg\nSaving to: dataset/masks/0313-1/3900\nReading from: dataset/clips/0313-1/10660/20.jpg\nSaving to: dataset/masks/0313-1/10660\nReading from: dataset/clips/0313-1/6400/20.jpg\nSaving to: dataset/masks/0313-1/6400\nReading from: dataset/clips/0313-1/23340/20.jpg\nSaving to: dataset/masks/0313-1/23340\nReading from: dataset/clips/0313-1/11720/20.jpg\nSaving to: dataset/masks/0313-1/11720\nReading from: dataset/clips/0313-1/44040/20.jpg\nSaving to: dataset/masks/0313-1/44040\nReading from: dataset/clips/0313-1/45240/20.jpg\nSaving to: dataset/masks/0313-1/45240\nReading from: dataset/clips/0313-1/39360/20.jpg\nSaving to: dataset/masks/0313-1/39360\nReading from: dataset/clips/0313-1/11240/20.jpg\nSaving to: dataset/masks/0313-1/11240\nReading from: dataset/clips/0313-1/37800/20.jpg\nSaving to: dataset/masks/0313-1/37800\nReading from: dataset/clips/0313-1/10260/20.jpg\nSaving to: dataset/masks/0313-1/10260\nReading from: dataset/clips/0313-1/10240/20.jpg\nSaving to: dataset/masks/0313-1/10240\nReading from: dataset/clips/0313-1/46920/20.jpg\nSaving to: dataset/masks/0313-1/46920\nReading from: dataset/clips/0313-1/38220/20.jpg\nSaving to: dataset/masks/0313-1/38220\nReading from: dataset/clips/0313-1/24540/20.jpg\nSaving to: dataset/masks/0313-1/24540\nReading from: dataset/clips/0313-1/60/20.jpg\nSaving to: dataset/masks/0313-1/60\nReading from: dataset/clips/0313-1/46980/20.jpg\nSaving to: dataset/masks/0313-1/46980\nReading from: dataset/clips/0313-1/25320/20.jpg\nSaving to: dataset/masks/0313-1/25320\nReading from: dataset/clips/0313-1/10920/20.jpg\nSaving to: dataset/masks/0313-1/10920\nReading from: dataset/clips/0313-1/1860/20.jpg\nSaving to: dataset/masks/0313-1/1860\nReading from: dataset/clips/0313-1/40740/20.jpg\nSaving to: dataset/masks/0313-1/40740\nReading from: dataset/clips/0313-1/11400/20.jpg\nSaving to: dataset/masks/0313-1/11400\nReading from: dataset/clips/0313-1/13960/20.jpg\nSaving to: dataset/masks/0313-1/13960\nReading from: dataset/clips/0313-1/10440/20.jpg\nSaving to: dataset/masks/0313-1/10440\nReading from: dataset/clips/0313-1/28980/20.jpg\nSaving to: dataset/masks/0313-1/28980\nReading from: dataset/clips/0313-1/4320/20.jpg\nSaving to: dataset/masks/0313-1/4320\nReading from: dataset/clips/0313-1/7260/20.jpg\nSaving to: dataset/masks/0313-1/7260\nReading from: dataset/clips/0313-1/17100/20.jpg\nSaving to: dataset/masks/0313-1/17100\nReading from: dataset/clips/0313-1/14340/20.jpg\nSaving to: dataset/masks/0313-1/14340\nReading from: dataset/clips/0313-1/9540/20.jpg\nSaving to: dataset/masks/0313-1/9540\nReading from: dataset/clips/0313-1/7200/20.jpg\nSaving to: dataset/masks/0313-1/7200\nReading from: dataset/clips/0313-1/16180/20.jpg\nSaving to: dataset/masks/0313-1/16180\nReading from: dataset/clips/0313-1/37920/20.jpg\nSaving to: dataset/masks/0313-1/37920\nReading from: dataset/clips/0313-1/36660/20.jpg\nSaving to: dataset/masks/0313-1/36660\nReading from: dataset/clips/0313-1/49980/20.jpg\nSaving to: dataset/masks/0313-1/49980\nReading from: dataset/clips/0313-1/33420/20.jpg\nSaving to: dataset/masks/0313-1/33420\nReading from: dataset/clips/0313-1/22080/20.jpg\nSaving to: dataset/masks/0313-1/22080\nReading from: dataset/clips/0313-1/13420/20.jpg\nSaving to: dataset/masks/0313-1/13420\nReading from: dataset/clips/0313-1/46260/20.jpg\nSaving to: dataset/masks/0313-1/46260\nReading from: dataset/clips/0313-1/49380/20.jpg\nSaving to: dataset/masks/0313-1/49380\nReading from: dataset/clips/0313-1/53100/20.jpg\nSaving to: dataset/masks/0313-1/53100\nReading from: dataset/clips/0313-1/4800/20.jpg\nSaving to: dataset/masks/0313-1/4800\nReading from: dataset/clips/0313-1/17820/20.jpg\nSaving to: dataset/masks/0313-1/17820\nReading from: dataset/clips/0313-1/11500/20.jpg\nSaving to: dataset/masks/0313-1/11500\nReading from: dataset/clips/0313-1/33840/20.jpg\nSaving to: dataset/masks/0313-1/33840\nReading from: dataset/clips/0313-1/17780/20.jpg\nSaving to: dataset/masks/0313-1/17780\nReading from: dataset/clips/0313-1/8940/20.jpg\nSaving to: dataset/masks/0313-1/8940\nReading from: dataset/clips/0313-1/7860/20.jpg\nSaving to: dataset/masks/0313-1/7860\nReading from: dataset/clips/0313-1/49140/20.jpg\nSaving to: dataset/masks/0313-1/49140\nReading from: dataset/clips/0313-1/18300/20.jpg\nSaving to: dataset/masks/0313-1/18300\nReading from: dataset/clips/0313-1/11920/20.jpg\nSaving to: dataset/masks/0313-1/11920\nReading from: dataset/clips/0313-1/9940/20.jpg\nSaving to: dataset/masks/0313-1/9940\nReading from: dataset/clips/0313-1/16460/20.jpg\nSaving to: dataset/masks/0313-1/16460\nReading from: dataset/clips/0313-1/14280/20.jpg\nSaving to: dataset/masks/0313-1/14280\nReading from: dataset/clips/0313-1/9400/20.jpg\nSaving to: dataset/masks/0313-1/9400\nReading from: dataset/clips/0313-1/24660/20.jpg\nSaving to: dataset/masks/0313-1/24660\nReading from: dataset/clips/0313-1/9820/20.jpg\nSaving to: dataset/masks/0313-1/9820\nReading from: dataset/clips/0313-1/6120/20.jpg\nSaving to: dataset/masks/0313-1/6120\nReading from: dataset/clips/0313-1/24120/20.jpg\nSaving to: dataset/masks/0313-1/24120\nReading from: dataset/clips/0313-1/23040/20.jpg\nSaving to: dataset/masks/0313-1/23040\nReading from: dataset/clips/0313-1/30960/20.jpg\nSaving to: dataset/masks/0313-1/30960\nReading from: dataset/clips/0313-1/7600/20.jpg\nSaving to: dataset/masks/0313-1/7600\nReading from: dataset/clips/0313-1/43800/20.jpg\nSaving to: dataset/masks/0313-1/43800\nReading from: dataset/clips/0313-1/360/20.jpg\nSaving to: dataset/masks/0313-1/360\nReading from: dataset/clips/0313-1/3720/20.jpg\nSaving to: dataset/masks/0313-1/3720\nReading from: dataset/clips/0313-1/14040/20.jpg\nSaving to: dataset/masks/0313-1/14040\nReading from: dataset/clips/0313-1/5220/20.jpg\nSaving to: dataset/masks/0313-1/5220\nReading from: dataset/clips/0313-1/5820/20.jpg\nSaving to: dataset/masks/0313-1/5820\nReading from: dataset/clips/0313-1/23220/20.jpg\nSaving to: dataset/masks/0313-1/23220\nReading from: dataset/clips/0313-1/4760/20.jpg\nSaving to: dataset/masks/0313-1/4760\nReading from: dataset/clips/0313-1/35880/20.jpg\nSaving to: dataset/masks/0313-1/35880\nReading from: dataset/clips/0313-1/7360/20.jpg\nSaving to: dataset/masks/0313-1/7360\nReading from: dataset/clips/0313-1/5480/20.jpg\nSaving to: dataset/masks/0313-1/5480\nReading from: dataset/clips/0313-1/44820/20.jpg\nSaving to: dataset/masks/0313-1/44820\nReading from: dataset/clips/0313-1/12080/20.jpg\nSaving to: dataset/masks/0313-1/12080\nReading from: dataset/clips/0313-1/5620/20.jpg\nSaving to: dataset/masks/0313-1/5620\nReading from: dataset/clips/0313-1/27840/20.jpg\nSaving to: dataset/masks/0313-1/27840\nReading from: dataset/clips/0313-1/40680/20.jpg\nSaving to: dataset/masks/0313-1/40680\nReading from: dataset/clips/0313-1/6520/20.jpg\nSaving to: dataset/masks/0313-1/6520\nReading from: dataset/clips/0313-1/35340/20.jpg\nSaving to: dataset/masks/0313-1/35340\nReading from: dataset/clips/0313-1/37500/20.jpg\nSaving to: dataset/masks/0313-1/37500\nReading from: dataset/clips/0313-1/45060/20.jpg\nSaving to: dataset/masks/0313-1/45060\nReading from: dataset/clips/0313-1/43680/20.jpg\nSaving to: dataset/masks/0313-1/43680\nReading from: dataset/clips/0313-1/2520/20.jpg\nSaving to: dataset/masks/0313-1/2520\nReading from: dataset/clips/0313-1/14580/20.jpg\nSaving to: dataset/masks/0313-1/14580\nReading from: dataset/clips/0313-1/7480/20.jpg\nSaving to: dataset/masks/0313-1/7480\nReading from: dataset/clips/0313-1/8720/20.jpg\nSaving to: dataset/masks/0313-1/8720\nReading from: dataset/clips/0313-1/26940/20.jpg\nSaving to: dataset/masks/0313-1/26940\nReading from: dataset/clips/0313-1/10080/20.jpg\nSaving to: dataset/masks/0313-1/10080\nReading from: dataset/clips/0313-1/51240/20.jpg\nSaving to: dataset/masks/0313-1/51240\nReading from: dataset/clips/0313-1/29040/20.jpg\nSaving to: dataset/masks/0313-1/29040\nReading from: dataset/clips/0313-1/18480/20.jpg\nSaving to: dataset/masks/0313-1/18480\nReading from: dataset/clips/0313-1/5020/20.jpg\nSaving to: dataset/masks/0313-1/5020\nReading from: dataset/clips/0313-1/27540/20.jpg\nSaving to: dataset/masks/0313-1/27540\nReading from: dataset/clips/0313-1/8600/20.jpg\nSaving to: dataset/masks/0313-1/8600\nReading from: dataset/clips/0313-1/16260/20.jpg\nSaving to: dataset/masks/0313-1/16260\nReading from: dataset/clips/0313-1/27600/20.jpg\nSaving to: dataset/masks/0313-1/27600\nReading from: dataset/clips/0313-1/43500/20.jpg\nSaving to: dataset/masks/0313-1/43500\nReading from: dataset/clips/0313-1/4140/20.jpg\nSaving to: dataset/masks/0313-1/4140\nReading from: dataset/clips/0313-1/22200/20.jpg\nSaving to: dataset/masks/0313-1/22200\nReading from: dataset/clips/0313-1/46380/20.jpg\nSaving to: dataset/masks/0313-1/46380\nReading from: dataset/clips/0313-1/40920/20.jpg\nSaving to: dataset/masks/0313-1/40920\nReading from: dataset/clips/0313-1/14780/20.jpg\nSaving to: dataset/masks/0313-1/14780\nReading from: dataset/clips/0313-1/4600/20.jpg\nSaving to: dataset/masks/0313-1/4600\nReading from: dataset/clips/0313-1/40020/20.jpg\nSaving to: dataset/masks/0313-1/40020\nReading from: dataset/clips/0313-1/15640/20.jpg\nSaving to: dataset/masks/0313-1/15640\nReading from: dataset/clips/0313-1/15920/20.jpg\nSaving to: dataset/masks/0313-1/15920\nReading from: dataset/clips/0313-1/24360/20.jpg\nSaving to: dataset/masks/0313-1/24360\nReading from: dataset/clips/0313-1/16740/20.jpg\nSaving to: dataset/masks/0313-1/16740\nReading from: dataset/clips/0313-1/7680/20.jpg\nSaving to: dataset/masks/0313-1/7680\nReading from: dataset/clips/0313-1/51300/20.jpg\nSaving to: dataset/masks/0313-1/51300\nReading from: dataset/clips/0313-1/10200/20.jpg\nSaving to: dataset/masks/0313-1/10200\nReading from: dataset/clips/0313-1/11040/20.jpg\nSaving to: dataset/masks/0313-1/11040\nReading from: dataset/clips/0313-1/10160/20.jpg\nSaving to: dataset/masks/0313-1/10160\nReading from: dataset/clips/0313-1/11640/20.jpg\nSaving to: dataset/masks/0313-1/11640\nReading from: dataset/clips/0313-1/19560/20.jpg\nSaving to: dataset/masks/0313-1/19560\nReading from: dataset/clips/0313-1/3360/20.jpg\nSaving to: dataset/masks/0313-1/3360\nReading from: dataset/clips/0313-1/31500/20.jpg\nSaving to: dataset/masks/0313-1/31500\nReading from: dataset/clips/0313-1/6340/20.jpg\nSaving to: dataset/masks/0313-1/6340\nReading from: dataset/clips/0313-1/8780/20.jpg\nSaving to: dataset/masks/0313-1/8780\nReading from: dataset/clips/0313-1/10020/20.jpg\nSaving to: dataset/masks/0313-1/10020\nReading from: dataset/clips/0313-1/34860/20.jpg\nSaving to: dataset/masks/0313-1/34860\nReading from: dataset/clips/0313-1/29640/20.jpg\nSaving to: dataset/masks/0313-1/29640\nReading from: dataset/clips/0313-1/17320/20.jpg\nSaving to: dataset/masks/0313-1/17320\nReading from: dataset/clips/0313-1/50220/20.jpg\nSaving to: dataset/masks/0313-1/50220\nReading from: dataset/clips/0313-1/30720/20.jpg\nSaving to: dataset/masks/0313-1/30720\nReading from: dataset/clips/0313-1/2880/20.jpg\nSaving to: dataset/masks/0313-1/2880\nReading from: dataset/clips/0313-1/35700/20.jpg\nSaving to: dataset/masks/0313-1/35700\nReading from: dataset/clips/0313-1/30480/20.jpg\nSaving to: dataset/masks/0313-1/30480\nReading from: dataset/clips/0313-1/30360/20.jpg\nSaving to: dataset/masks/0313-1/30360\nReading from: dataset/clips/0313-1/14240/20.jpg\nSaving to: dataset/masks/0313-1/14240\nReading from: dataset/clips/0313-1/44700/20.jpg\nSaving to: dataset/masks/0313-1/44700\nReading from: dataset/clips/0313-1/25620/20.jpg\nSaving to: dataset/masks/0313-1/25620\nReading from: dataset/clips/0313-1/10980/20.jpg\nSaving to: dataset/masks/0313-1/10980\nReading from: dataset/clips/0313-1/7140/20.jpg\nSaving to: dataset/masks/0313-1/7140\nReading from: dataset/clips/0313-1/45660/20.jpg\nSaving to: dataset/masks/0313-1/45660\nReading from: dataset/clips/0313-1/7120/20.jpg\nSaving to: dataset/masks/0313-1/7120\nReading from: dataset/clips/0313-1/18720/20.jpg\nSaving to: dataset/masks/0313-1/18720\nReading from: dataset/clips/0313-1/39300/20.jpg\nSaving to: dataset/masks/0313-1/39300\nReading from: dataset/clips/0313-1/37860/20.jpg\nSaving to: dataset/masks/0313-1/37860\nReading from: dataset/clips/0313-1/5560/20.jpg\nSaving to: dataset/masks/0313-1/5560\nReading from: dataset/clips/0313-1/28920/20.jpg\nSaving to: dataset/masks/0313-1/28920\nReading from: dataset/clips/0313-1/33600/20.jpg\nSaving to: dataset/masks/0313-1/33600\nReading from: dataset/clips/0313-1/41340/20.jpg\nSaving to: dataset/masks/0313-1/41340\nReading from: dataset/clips/0313-1/43260/20.jpg\nSaving to: dataset/masks/0313-1/43260\nReading from: dataset/clips/0313-1/6720/20.jpg\nSaving to: dataset/masks/0313-1/6720\nReading from: dataset/clips/0313-1/27180/20.jpg\nSaving to: dataset/masks/0313-1/27180\nReading from: dataset/clips/0313-1/7900/20.jpg\nSaving to: dataset/masks/0313-1/7900\nReading from: dataset/clips/0313-1/36300/20.jpg\nSaving to: dataset/masks/0313-1/36300\nReading from: dataset/clips/0313-1/10940/20.jpg\nSaving to: dataset/masks/0313-1/10940\nReading from: dataset/clips/0313-1/45780/20.jpg\nSaving to: dataset/masks/0313-1/45780\nReading from: dataset/clips/0313-1/23940/20.jpg\nSaving to: dataset/masks/0313-1/23940\nReading from: dataset/clips/0313-1/38460/20.jpg\nSaving to: dataset/masks/0313-1/38460\nReading from: dataset/clips/0313-1/11480/20.jpg\nSaving to: dataset/masks/0313-1/11480\nReading from: dataset/clips/0313-1/5900/20.jpg\nSaving to: dataset/masks/0313-1/5900\nReading from: dataset/clips/0313-1/12400/20.jpg\nSaving to: dataset/masks/0313-1/12400\nReading from: dataset/clips/0313-1/20100/20.jpg\nSaving to: dataset/masks/0313-1/20100\nReading from: dataset/clips/0313-1/15600/20.jpg\nSaving to: dataset/masks/0313-1/15600\nReading from: dataset/clips/0313-1/33120/20.jpg\nSaving to: dataset/masks/0313-1/33120\nReading from: dataset/clips/0313-1/53040/20.jpg\nSaving to: dataset/masks/0313-1/53040\nReading from: dataset/clips/0313-1/6860/20.jpg\nSaving to: dataset/masks/0313-1/6860\nReading from: dataset/clips/0313-1/36540/20.jpg\nSaving to: dataset/masks/0313-1/36540\nReading from: dataset/clips/0313-1/420/20.jpg\nSaving to: dataset/masks/0313-1/420\nReading from: dataset/clips/0313-1/16540/20.jpg\nSaving to: dataset/masks/0313-1/16540\nReading from: dataset/clips/0313-1/40200/20.jpg\nSaving to: dataset/masks/0313-1/40200\nReading from: dataset/clips/0313-1/38880/20.jpg\nSaving to: dataset/masks/0313-1/38880\nReading from: dataset/clips/0313-1/13780/20.jpg\nSaving to: dataset/masks/0313-1/13780\nReading from: dataset/clips/0313-1/12920/20.jpg\nSaving to: dataset/masks/0313-1/12920\nReading from: dataset/clips/0313-1/17420/20.jpg\nSaving to: dataset/masks/0313-1/17420\nReading from: dataset/clips/0313-1/11700/20.jpg\nSaving to: dataset/masks/0313-1/11700\nReading from: dataset/clips/0313-1/23640/20.jpg\nSaving to: dataset/masks/0313-1/23640\nReading from: dataset/clips/0313-1/19680/20.jpg\nSaving to: dataset/masks/0313-1/19680\nReading from: dataset/clips/0313-1/38280/20.jpg\nSaving to: dataset/masks/0313-1/38280\nReading from: dataset/clips/0313-1/21900/20.jpg\nSaving to: dataset/masks/0313-1/21900\nReading from: dataset/clips/0313-1/11300/20.jpg\nSaving to: dataset/masks/0313-1/11300\nReading from: dataset/clips/0313-1/15760/20.jpg\nSaving to: dataset/masks/0313-1/15760\nReading from: dataset/clips/0313-1/13020/20.jpg\nSaving to: dataset/masks/0313-1/13020\nReading from: dataset/clips/0313-1/8200/20.jpg\nSaving to: dataset/masks/0313-1/8200\nReading from: dataset/clips/0313-1/17980/20.jpg\nSaving to: dataset/masks/0313-1/17980\nReading from: dataset/clips/0313-1/25500/20.jpg\nSaving to: dataset/masks/0313-1/25500\nReading from: dataset/clips/0313-1/4220/20.jpg\nSaving to: dataset/masks/0313-1/4220\nReading from: dataset/clips/0313-1/17860/20.jpg\nSaving to: dataset/masks/0313-1/17860\nReading from: dataset/clips/0313-1/14000/20.jpg\nSaving to: dataset/masks/0313-1/14000\nReading from: dataset/clips/0313-1/12620/20.jpg\nSaving to: dataset/masks/0313-1/12620\nReading from: dataset/clips/0313-1/13140/20.jpg\nSaving to: dataset/masks/0313-1/13140\nReading from: dataset/clips/0313-1/52680/20.jpg\nSaving to: dataset/masks/0313-1/52680\nReading from: dataset/clips/0313-1/17080/20.jpg\nSaving to: dataset/masks/0313-1/17080\nReading from: dataset/clips/0313-1/4940/20.jpg\nSaving to: dataset/masks/0313-1/4940\nReading from: dataset/clips/0313-1/29820/20.jpg\nSaving to: dataset/masks/0313-1/29820\nReading from: dataset/clips/0313-1/1500/20.jpg\nSaving to: dataset/masks/0313-1/1500\nReading from: dataset/clips/0313-1/12360/20.jpg\nSaving to: dataset/masks/0313-1/12360\nReading from: dataset/clips/0313-1/7920/20.jpg\nSaving to: dataset/masks/0313-1/7920\nReading from: dataset/clips/0313-1/5660/20.jpg\nSaving to: dataset/masks/0313-1/5660\nReading from: dataset/clips/0313-1/17680/20.jpg\nSaving to: dataset/masks/0313-1/17680\nReading from: dataset/clips/0313-1/8340/20.jpg\nSaving to: dataset/masks/0313-1/8340\nReading from: dataset/clips/0313-1/30600/20.jpg\nSaving to: dataset/masks/0313-1/30600\nReading from: dataset/clips/0313-1/32760/20.jpg\nSaving to: dataset/masks/0313-1/32760\nReading from: dataset/clips/0313-1/4700/20.jpg\nSaving to: dataset/masks/0313-1/4700\nReading from: dataset/clips/0313-1/25860/20.jpg\nSaving to: dataset/masks/0313-1/25860\nReading from: dataset/clips/0313-1/4080/20.jpg\nSaving to: dataset/masks/0313-1/4080\nReading from: dataset/clips/0313-1/16020/20.jpg\nSaving to: dataset/masks/0313-1/16020\nReading from: dataset/clips/0313-1/16080/20.jpg\nSaving to: dataset/masks/0313-1/16080\nReading from: dataset/clips/0313-1/10780/20.jpg\nSaving to: dataset/masks/0313-1/10780\nReading from: dataset/clips/0313-1/8560/20.jpg\nSaving to: dataset/masks/0313-1/8560\nReading from: dataset/clips/0313-1/9420/20.jpg\nSaving to: dataset/masks/0313-1/9420\nReading from: dataset/clips/0313-1/4820/20.jpg\nSaving to: dataset/masks/0313-1/4820\nReading from: dataset/clips/0313-1/17920/20.jpg\nSaving to: dataset/masks/0313-1/17920\nReading from: dataset/clips/0313-1/8160/20.jpg\nSaving to: dataset/masks/0313-1/8160\nReading from: dataset/clips/0313-1/3980/20.jpg\nSaving to: dataset/masks/0313-1/3980\nReading from: dataset/clips/0313-1/5060/20.jpg\nSaving to: dataset/masks/0313-1/5060\nReading from: dataset/clips/0313-1/8960/20.jpg\nSaving to: dataset/masks/0313-1/8960\nReading from: dataset/clips/0313-1/14760/20.jpg\nSaving to: dataset/masks/0313-1/14760\nReading from: dataset/clips/0313-1/22800/20.jpg\nSaving to: dataset/masks/0313-1/22800\nReading from: dataset/clips/0313-1/37200/20.jpg\nSaving to: dataset/masks/0313-1/37200\nReading from: dataset/clips/0313-1/21540/20.jpg\nSaving to: dataset/masks/0313-1/21540\nReading from: dataset/clips/0313-1/13460/20.jpg\nSaving to: dataset/masks/0313-1/13460\nReading from: dataset/clips/0313-1/6740/20.jpg\nSaving to: dataset/masks/0313-1/6740\nReading from: dataset/clips/0313-1/6440/20.jpg\nSaving to: dataset/masks/0313-1/6440\nReading from: dataset/clips/0313-1/15960/20.jpg\nSaving to: dataset/masks/0313-1/15960\nReading from: dataset/clips/0313-1/32640/20.jpg\nSaving to: dataset/masks/0313-1/32640\nReading from: dataset/clips/0313-1/13980/20.jpg\nSaving to: dataset/masks/0313-1/13980\nReading from: dataset/clips/0313-1/34320/20.jpg\nSaving to: dataset/masks/0313-1/34320\nReading from: dataset/clips/0313-1/17500/20.jpg\nSaving to: dataset/masks/0313-1/17500\nReading from: dataset/clips/0313-1/5680/20.jpg\nSaving to: dataset/masks/0313-1/5680\nReading from: dataset/clips/0313-1/8360/20.jpg\nSaving to: dataset/masks/0313-1/8360\nReading from: dataset/clips/0313-1/4960/20.jpg\nSaving to: dataset/masks/0313-1/4960\nReading from: dataset/clips/0313-1/28140/20.jpg\nSaving to: dataset/masks/0313-1/28140\nReading from: dataset/clips/0313-1/10800/20.jpg\nSaving to: dataset/masks/0313-1/10800\nReading from: dataset/clips/0313-1/43440/20.jpg\nSaving to: dataset/masks/0313-1/43440\nReading from: dataset/clips/0313-1/7700/20.jpg\nSaving to: dataset/masks/0313-1/7700\nReading from: dataset/clips/0313-1/16720/20.jpg\nSaving to: dataset/masks/0313-1/16720\nReading from: dataset/clips/0313-1/10000/20.jpg\nSaving to: dataset/masks/0313-1/10000\nReading from: dataset/clips/0313-1/37620/20.jpg\nSaving to: dataset/masks/0313-1/37620\nReading from: dataset/clips/0313-1/6840/20.jpg\nSaving to: dataset/masks/0313-1/6840\nReading from: dataset/clips/0313-1/11020/20.jpg\nSaving to: dataset/masks/0313-1/11020\nReading from: dataset/clips/0313-1/39060/20.jpg\nSaving to: dataset/masks/0313-1/39060\nReading from: dataset/clips/0313-1/37980/20.jpg\nSaving to: dataset/masks/0313-1/37980\nReading from: dataset/clips/0313-1/1620/20.jpg\nSaving to: dataset/masks/0313-1/1620\nReading from: dataset/clips/0313-1/38100/20.jpg\nSaving to: dataset/masks/0313-1/38100\nReading from: dataset/clips/0313-1/32940/20.jpg\nSaving to: dataset/masks/0313-1/32940\nReading from: dataset/clips/0313-1/29280/20.jpg\nSaving to: dataset/masks/0313-1/29280\nReading from: dataset/clips/0313-1/13440/20.jpg\nSaving to: dataset/masks/0313-1/13440\nReading from: dataset/clips/0313-1/6920/20.jpg\nSaving to: dataset/masks/0313-1/6920\nReading from: dataset/clips/0313-1/6140/20.jpg\nSaving to: dataset/masks/0313-1/6140\nReading from: dataset/clips/0313-1/27000/20.jpg\nSaving to: dataset/masks/0313-1/27000\nReading from: dataset/clips/0313-1/16220/20.jpg\nSaving to: dataset/masks/0313-1/16220\nReading from: dataset/clips/0313-1/15200/20.jpg\nSaving to: dataset/masks/0313-1/15200\nReading from: dataset/clips/0313-1/26400/20.jpg\nSaving to: dataset/masks/0313-1/26400\nReading from: dataset/clips/0313-1/7400/20.jpg\nSaving to: dataset/masks/0313-1/7400\nReading from: dataset/clips/0313-1/12240/20.jpg\nSaving to: dataset/masks/0313-1/12240\nReading from: dataset/clips/0313-1/14200/20.jpg\nSaving to: dataset/masks/0313-1/14200\nReading from: dataset/clips/0313-1/38160/20.jpg\nSaving to: dataset/masks/0313-1/38160\nReading from: dataset/clips/0313-1/35220/20.jpg\nSaving to: dataset/masks/0313-1/35220\nReading from: dataset/clips/0313-1/16820/20.jpg\nSaving to: dataset/masks/0313-1/16820\nReading from: dataset/clips/0313-1/39900/20.jpg\nSaving to: dataset/masks/0313-1/39900\nReading from: dataset/clips/0313-1/4160/20.jpg\nSaving to: dataset/masks/0313-1/4160\nReading from: dataset/clips/0313-1/24180/20.jpg\nSaving to: dataset/masks/0313-1/24180\nReading from: dataset/clips/0313-1/11420/20.jpg\nSaving to: dataset/masks/0313-1/11420\nReading from: dataset/clips/0313-1/13640/20.jpg\nSaving to: dataset/masks/0313-1/13640\nReading from: dataset/clips/0313-1/7080/20.jpg\nSaving to: dataset/masks/0313-1/7080\nReading from: dataset/clips/0313-1/40440/20.jpg\nSaving to: dataset/masks/0313-1/40440\nReading from: dataset/clips/0313-1/10620/20.jpg\nSaving to: dataset/masks/0313-1/10620\nReading from: dataset/clips/0313-1/15280/20.jpg\nSaving to: dataset/masks/0313-1/15280\nReading from: dataset/clips/0313-1/3180/20.jpg\nSaving to: dataset/masks/0313-1/3180\nReading from: dataset/clips/0313-1/15480/20.jpg\nSaving to: dataset/masks/0313-1/15480\nReading from: dataset/clips/0313-1/44940/20.jpg\nSaving to: dataset/masks/0313-1/44940\nReading from: dataset/clips/0313-1/15540/20.jpg\nSaving to: dataset/masks/0313-1/15540\nReading from: dataset/clips/0313-1/49440/20.jpg\nSaving to: dataset/masks/0313-1/49440\nReading from: dataset/clips/0313-1/480/20.jpg\nSaving to: dataset/masks/0313-1/480\nReading from: dataset/clips/0313-1/51120/20.jpg\nSaving to: dataset/masks/0313-1/51120\nReading from: dataset/clips/0313-1/29340/20.jpg\nSaving to: dataset/masks/0313-1/29340\nReading from: dataset/clips/0313-1/8380/20.jpg\nSaving to: dataset/masks/0313-1/8380\nReading from: dataset/clips/0313-1/48000/20.jpg\nSaving to: dataset/masks/0313-1/48000\nReading from: dataset/clips/0313-1/9740/20.jpg\nSaving to: dataset/masks/0313-1/9740\nReading from: dataset/clips/0313-1/37020/20.jpg\nSaving to: dataset/masks/0313-1/37020\nReading from: dataset/clips/0313-1/9260/20.jpg\nSaving to: dataset/masks/0313-1/9260\nReading from: dataset/clips/0313-1/48480/20.jpg\nSaving to: dataset/masks/0313-1/48480\nReading from: dataset/clips/0313-1/20460/20.jpg\nSaving to: dataset/masks/0313-1/20460\nReading from: dataset/clips/0313-1/31980/20.jpg\nSaving to: dataset/masks/0313-1/31980\nReading from: dataset/clips/0313-1/19500/20.jpg\nSaving to: dataset/masks/0313-1/19500\nReading from: dataset/clips/0313-1/40500/20.jpg\nSaving to: dataset/masks/0313-1/40500\nReading from: dataset/clips/0313-1/17560/20.jpg\nSaving to: dataset/masks/0313-1/17560\nReading from: dataset/clips/0313-1/16960/20.jpg\nSaving to: dataset/masks/0313-1/16960\nReading from: dataset/clips/0313-1/30420/20.jpg\nSaving to: dataset/masks/0313-1/30420\nReading from: dataset/clips/0313-1/14500/20.jpg\nSaving to: dataset/masks/0313-1/14500\nReading from: dataset/clips/0313-1/43380/20.jpg\nSaving to: dataset/masks/0313-1/43380\nReading from: dataset/clips/0313-1/5080/20.jpg\nSaving to: dataset/masks/0313-1/5080\nReading from: dataset/clips/0313-1/17300/20.jpg\nSaving to: dataset/masks/0313-1/17300\nReading from: dataset/clips/0313-1/20880/20.jpg\nSaving to: dataset/masks/0313-1/20880\nReading from: dataset/clips/0313-1/36060/20.jpg\nSaving to: dataset/masks/0313-1/36060\nReading from: dataset/clips/0313-1/4000/20.jpg\nSaving to: dataset/masks/0313-1/4000\nReading from: dataset/clips/0313-1/8440/20.jpg\nSaving to: dataset/masks/0313-1/8440\nReading from: dataset/clips/0313-1/15720/20.jpg\nSaving to: dataset/masks/0313-1/15720\nReading from: dataset/clips/0313-1/41460/20.jpg\nSaving to: dataset/masks/0313-1/41460\nReading from: dataset/clips/0313-1/5840/20.jpg\nSaving to: dataset/masks/0313-1/5840\nReading from: dataset/clips/0313-1/42180/20.jpg\nSaving to: dataset/masks/0313-1/42180\nReading from: dataset/clips/0313-1/4980/20.jpg\nSaving to: dataset/masks/0313-1/4980\nReading from: dataset/clips/0313-1/10320/20.jpg\nSaving to: dataset/masks/0313-1/10320\nReading from: dataset/clips/0313-1/41040/20.jpg\nSaving to: dataset/masks/0313-1/41040\nReading from: dataset/clips/0313-1/4460/20.jpg\nSaving to: dataset/masks/0313-1/4460\nReading from: dataset/clips/0313-1/13280/20.jpg\nSaving to: dataset/masks/0313-1/13280\nReading from: dataset/clips/0313-1/48960/20.jpg\nSaving to: dataset/masks/0313-1/48960\nReading from: dataset/clips/0313-1/5500/20.jpg\nSaving to: dataset/masks/0313-1/5500\nReading from: dataset/clips/0313-1/31440/20.jpg\nSaving to: dataset/masks/0313-1/31440\nReading from: dataset/clips/0313-1/17700/20.jpg\nSaving to: dataset/masks/0313-1/17700\nReading from: dataset/clips/0313-1/2580/20.jpg\nSaving to: dataset/masks/0313-1/2580\nReading from: dataset/clips/0313-1/4620/20.jpg\nSaving to: dataset/masks/0313-1/4620\nReading from: dataset/clips/0313-1/37080/20.jpg\nSaving to: dataset/masks/0313-1/37080\nReading from: dataset/clips/0313-1/18240/20.jpg\nSaving to: dataset/masks/0313-1/18240\nReading from: dataset/clips/0313-1/29460/20.jpg\nSaving to: dataset/masks/0313-1/29460\nReading from: dataset/clips/0313-1/8300/20.jpg\nSaving to: dataset/masks/0313-1/8300\nReading from: dataset/clips/0313-1/9040/20.jpg\nSaving to: dataset/masks/0313-1/9040\nReading from: dataset/clips/0313-1/8480/20.jpg\nSaving to: dataset/masks/0313-1/8480\nReading from: dataset/clips/0313-1/8620/20.jpg\nSaving to: dataset/masks/0313-1/8620\nReading from: dataset/clips/0313-1/33060/20.jpg\nSaving to: dataset/masks/0313-1/33060\nReading from: dataset/clips/0313-1/26880/20.jpg\nSaving to: dataset/masks/0313-1/26880\nReading from: dataset/clips/0313-1/5400/20.jpg\nSaving to: dataset/masks/0313-1/5400\nReading from: dataset/clips/0313-1/47820/20.jpg\nSaving to: dataset/masks/0313-1/47820\nReading from: dataset/clips/0313-1/14900/20.jpg\nSaving to: dataset/masks/0313-1/14900\nReading from: dataset/clips/0313-1/5380/20.jpg\nSaving to: dataset/masks/0313-1/5380\nReading from: dataset/clips/0313-1/23100/20.jpg\nSaving to: dataset/masks/0313-1/23100\nReading from: dataset/clips/0313-1/33660/20.jpg\nSaving to: dataset/masks/0313-1/33660\nReading from: dataset/clips/0313-1/29580/20.jpg\nSaving to: dataset/masks/0313-1/29580\nReading from: dataset/clips/0313-1/12740/20.jpg\nSaving to: dataset/masks/0313-1/12740\nReading from: dataset/clips/0313-1/7620/20.jpg\nSaving to: dataset/masks/0313-1/7620\nReading from: dataset/clips/0313-1/8140/20.jpg\nSaving to: dataset/masks/0313-1/8140\nReading from: dataset/clips/0313-1/10220/20.jpg\nSaving to: dataset/masks/0313-1/10220\nReading from: dataset/clips/0313-1/4500/20.jpg\nSaving to: dataset/masks/0313-1/4500\nReading from: dataset/clips/0313-1/5640/20.jpg\nSaving to: dataset/masks/0313-1/5640\nReading from: dataset/clips/0313-1/15060/20.jpg\nSaving to: dataset/masks/0313-1/15060\nReading from: dataset/clips/0313-1/8280/20.jpg\nSaving to: dataset/masks/0313-1/8280\nReading from: dataset/clips/0313-1/9440/20.jpg\nSaving to: dataset/masks/0313-1/9440\nReading from: dataset/clips/0313-1/50520/20.jpg\nSaving to: dataset/masks/0313-1/50520\nReading from: dataset/clips/0313-1/10900/20.jpg\nSaving to: dataset/masks/0313-1/10900\nReading from: dataset/clips/0313-1/6200/20.jpg\nSaving to: dataset/masks/0313-1/6200\nReading from: dataset/clips/0313-1/6960/20.jpg\nSaving to: dataset/masks/0313-1/6960\nReading from: dataset/clips/0313-1/19860/20.jpg\nSaving to: dataset/masks/0313-1/19860\nReading from: dataset/clips/0313-1/37380/20.jpg\nSaving to: dataset/masks/0313-1/37380\nReading from: dataset/clips/0313-1/11280/20.jpg\nSaving to: dataset/masks/0313-1/11280\nReading from: dataset/clips/0313-1/35520/20.jpg\nSaving to: dataset/masks/0313-1/35520\nReading from: dataset/clips/0313-1/1440/20.jpg\nSaving to: dataset/masks/0313-1/1440\nReading from: dataset/clips/0313-1/9620/20.jpg\nSaving to: dataset/masks/0313-1/9620\nReading from: dataset/clips/0313-1/16940/20.jpg\nSaving to: dataset/masks/0313-1/16940\nReading from: dataset/clips/0313-1/35640/20.jpg\nSaving to: dataset/masks/0313-1/35640\nReading from: dataset/clips/0313-1/51720/20.jpg\nSaving to: dataset/masks/0313-1/51720\nReading from: dataset/clips/0313-1/21720/20.jpg\nSaving to: dataset/masks/0313-1/21720\nReading from: dataset/clips/0313-1/10380/20.jpg\nSaving to: dataset/masks/0313-1/10380\nReading from: dataset/clips/0313-1/17540/20.jpg\nSaving to: dataset/masks/0313-1/17540\nReading from: dataset/clips/0313-1/11780/20.jpg\nSaving to: dataset/masks/0313-1/11780\nReading from: dataset/clips/0313-1/24240/20.jpg\nSaving to: dataset/masks/0313-1/24240\nReading from: dataset/clips/0313-1/48120/20.jpg\nSaving to: dataset/masks/0313-1/48120\nReading from: dataset/clips/0313-1/31920/20.jpg\nSaving to: dataset/masks/0313-1/31920\nReading from: dataset/clips/0313-1/40140/20.jpg\nSaving to: dataset/masks/0313-1/40140\nReading from: dataset/clips/0313-1/30780/20.jpg\nSaving to: dataset/masks/0313-1/30780\nReading from: dataset/clips/0313-1/4720/20.jpg\nSaving to: dataset/masks/0313-1/4720\nReading from: dataset/clips/0313-1/10280/20.jpg\nSaving to: dataset/masks/0313-1/10280\nReading from: dataset/clips/0313-1/7740/20.jpg\nSaving to: dataset/masks/0313-1/7740\nReading from: dataset/clips/0313-1/1800/20.jpg\nSaving to: dataset/masks/0313-1/1800\nReading from: dataset/clips/0313-1/10500/20.jpg\nSaving to: dataset/masks/0313-1/10500\nReading from: dataset/clips/0313-1/8460/20.jpg\nSaving to: dataset/masks/0313-1/8460\nReading from: dataset/clips/0313-1/12060/20.jpg\nSaving to: dataset/masks/0313-1/12060\nReading from: dataset/clips/0313-1/10140/20.jpg\nSaving to: dataset/masks/0313-1/10140\nReading from: dataset/clips/0313-1/9760/20.jpg\nSaving to: dataset/masks/0313-1/9760\nReading from: dataset/clips/0313-1/7760/20.jpg\nSaving to: dataset/masks/0313-1/7760\nReading from: dataset/clips/0313-1/11260/20.jpg\nSaving to: dataset/masks/0313-1/11260\nReading from: dataset/clips/0313-1/30180/20.jpg\nSaving to: dataset/masks/0313-1/30180\nReading from: dataset/clips/0313-1/43140/20.jpg\nSaving to: dataset/masks/0313-1/43140\nReading from: dataset/clips/0313-1/14420/20.jpg\nSaving to: dataset/masks/0313-1/14420\nReading from: dataset/clips/0313-1/41100/20.jpg\nSaving to: dataset/masks/0313-1/41100\nReading from: dataset/clips/0313-1/44160/20.jpg\nSaving to: dataset/masks/0313-1/44160\nReading from: dataset/clips/0313-1/7320/20.jpg\nSaving to: dataset/masks/0313-1/7320\nReading from: dataset/clips/0313-1/50280/20.jpg\nSaving to: dataset/masks/0313-1/50280\nReading from: dataset/clips/0313-1/10180/20.jpg\nSaving to: dataset/masks/0313-1/10180\nReading from: dataset/clips/0313-1/32340/20.jpg\nSaving to: dataset/masks/0313-1/32340\nReading from: dataset/clips/0313-1/7340/20.jpg\nSaving to: dataset/masks/0313-1/7340\nReading from: dataset/clips/0313-1/18180/20.jpg\nSaving to: dataset/masks/0313-1/18180\nReading from: dataset/clips/0313-1/16800/20.jpg\nSaving to: dataset/masks/0313-1/16800\nReading from: dataset/clips/0313-1/39000/20.jpg\nSaving to: dataset/masks/0313-1/39000\nReading from: dataset/clips/0313-1/13000/20.jpg\nSaving to: dataset/masks/0313-1/13000\nReading from: dataset/clips/0313-1/11180/20.jpg\nSaving to: dataset/masks/0313-1/11180\nReading from: dataset/clips/0313-1/6580/20.jpg\nSaving to: dataset/masks/0313-1/6580\nReading from: dataset/clips/0313-1/16000/20.jpg\nSaving to: dataset/masks/0313-1/16000\nReading from: dataset/clips/0313-1/9980/20.jpg\nSaving to: dataset/masks/0313-1/9980\nReading from: dataset/clips/0313-1/12300/20.jpg\nSaving to: dataset/masks/0313-1/12300\nReading from: dataset/clips/0313-1/10520/20.jpg\nSaving to: dataset/masks/0313-1/10520\nReading from: dataset/clips/0313-1/12940/20.jpg\nSaving to: dataset/masks/0313-1/12940\nReading from: dataset/clips/0313-1/44760/20.jpg\nSaving to: dataset/masks/0313-1/44760\nReading from: dataset/clips/0313-1/13500/20.jpg\nSaving to: dataset/masks/0313-1/13500\nReading from: dataset/clips/0313-1/9000/20.jpg\nSaving to: dataset/masks/0313-1/9000\nReading from: dataset/clips/0313-1/50460/20.jpg\nSaving to: dataset/masks/0313-1/50460\nReading from: dataset/clips/0313-1/6260/20.jpg\nSaving to: dataset/masks/0313-1/6260\nReading from: dataset/clips/0313-1/8080/20.jpg\nSaving to: dataset/masks/0313-1/8080\nReading from: dataset/clips/0313-1/34620/20.jpg\nSaving to: dataset/masks/0313-1/34620\nReading from: dataset/clips/0313-1/16380/20.jpg\nSaving to: dataset/masks/0313-1/16380\nReading from: dataset/clips/0313-1/19080/20.jpg\nSaving to: dataset/masks/0313-1/19080\nReading from: dataset/clips/0313-1/48600/20.jpg\nSaving to: dataset/masks/0313-1/48600\nReading from: dataset/clips/0313-1/35580/20.jpg\nSaving to: dataset/masks/0313-1/35580\nReading from: dataset/clips/0313-1/13820/20.jpg\nSaving to: dataset/masks/0313-1/13820\nReading from: dataset/clips/0313-1/6760/20.jpg\nSaving to: dataset/masks/0313-1/6760\nReading from: dataset/clips/0313-1/27960/20.jpg\nSaving to: dataset/masks/0313-1/27960\nReading from: dataset/clips/0313-1/11460/20.jpg\nSaving to: dataset/masks/0313-1/11460\nReading from: dataset/clips/0313-1/7960/20.jpg\nSaving to: dataset/masks/0313-1/7960\nReading from: dataset/clips/0313-1/17840/20.jpg\nSaving to: dataset/masks/0313-1/17840\nReading from: dataset/clips/0313-1/35160/20.jpg\nSaving to: dataset/masks/0313-1/35160\nReading from: dataset/clips/0313-1/34140/20.jpg\nSaving to: dataset/masks/0313-1/34140\nReading from: dataset/clips/0313-1/3840/20.jpg\nSaving to: dataset/masks/0313-1/3840\nReading from: dataset/clips/0313-1/13800/20.jpg\nSaving to: dataset/masks/0313-1/13800\nReading from: dataset/clips/0313-1/36240/20.jpg\nSaving to: dataset/masks/0313-1/36240\nReading from: dataset/clips/0313-1/15040/20.jpg\nSaving to: dataset/masks/0313-1/15040\nReading from: dataset/clips/0313-1/50580/20.jpg\nSaving to: dataset/masks/0313-1/50580\nReading from: dataset/clips/0313-1/46200/20.jpg\nSaving to: dataset/masks/0313-1/46200\nReading from: dataset/clips/0313-1/48900/20.jpg\nSaving to: dataset/masks/0313-1/48900\nReading from: dataset/clips/0313-1/10820/20.jpg\nSaving to: dataset/masks/0313-1/10820\nReading from: dataset/clips/0313-1/7240/20.jpg\nSaving to: dataset/masks/0313-1/7240\nReading from: dataset/clips/0313-1/26520/20.jpg\nSaving to: dataset/masks/0313-1/26520\nReading from: dataset/clips/0313-1/9660/20.jpg\nSaving to: dataset/masks/0313-1/9660\nReading from: dataset/clips/0313-1/28200/20.jpg\nSaving to: dataset/masks/0313-1/28200\nReading from: dataset/clips/0313-1/7280/20.jpg\nSaving to: dataset/masks/0313-1/7280\nReading from: dataset/clips/0313-1/24480/20.jpg\nSaving to: dataset/masks/0313-1/24480\nReading from: dataset/clips/0313-1/11940/20.jpg\nSaving to: dataset/masks/0313-1/11940\nReading from: dataset/clips/0313-1/48720/20.jpg\nSaving to: dataset/masks/0313-1/48720\nReading from: dataset/clips/0313-1/4480/20.jpg\nSaving to: dataset/masks/0313-1/4480\nReading from: dataset/clips/0313-1/16320/20.jpg\nSaving to: dataset/masks/0313-1/16320\nReading from: dataset/clips/0313-1/13260/20.jpg\nSaving to: dataset/masks/0313-1/13260\nReading from: dataset/clips/0313-1/52920/20.jpg\nSaving to: dataset/masks/0313-1/52920\nReading from: dataset/clips/0313-1/8060/20.jpg\nSaving to: dataset/masks/0313-1/8060\nReading from: dataset/clips/0313-1/47940/20.jpg\nSaving to: dataset/masks/0313-1/47940\nReading from: dataset/clips/0313-1/9100/20.jpg\nSaving to: dataset/masks/0313-1/9100\nReading from: dataset/clips/0313-1/26820/20.jpg\nSaving to: dataset/masks/0313-1/26820\nReading from: dataset/clips/0313-1/17900/20.jpg\nSaving to: dataset/masks/0313-1/17900\nReading from: dataset/clips/0313-1/8020/20.jpg\nSaving to: dataset/masks/0313-1/8020\nReading from: dataset/clips/0313-1/8500/20.jpg\nSaving to: dataset/masks/0313-1/8500\nReading from: dataset/clips/0313-1/47760/20.jpg\nSaving to: dataset/masks/0313-1/47760\nReading from: dataset/clips/0313-1/34800/20.jpg\nSaving to: dataset/masks/0313-1/34800\nReading from: dataset/clips/0313-1/15880/20.jpg\nSaving to: dataset/masks/0313-1/15880\nReading from: dataset/clips/0313-1/2820/20.jpg\nSaving to: dataset/masks/0313-1/2820\nReading from: dataset/clips/0313-1/17740/20.jpg\nSaving to: dataset/masks/0313-1/17740\nReading from: dataset/clips/0313-1/21780/20.jpg\nSaving to: dataset/masks/0313-1/21780\nReading from: dataset/clips/0313-1/5420/20.jpg\nSaving to: dataset/masks/0313-1/5420\nReading from: dataset/clips/0313-1/4060/20.jpg\nSaving to: dataset/masks/0313-1/4060\nReading from: dataset/clips/0313-1/4400/20.jpg\nSaving to: dataset/masks/0313-1/4400\nReading from: dataset/clips/0313-1/5360/20.jpg\nSaving to: dataset/masks/0313-1/5360\nReading from: dataset/clips/0313-1/16640/20.jpg\nSaving to: dataset/masks/0313-1/16640\nReading from: dataset/clips/0313-1/43920/20.jpg\nSaving to: dataset/masks/0313-1/43920\nReading from: dataset/clips/0313-1/43620/20.jpg\nSaving to: dataset/masks/0313-1/43620\nReading from: dataset/clips/0313-1/6360/20.jpg\nSaving to: dataset/masks/0313-1/6360\nReading from: dataset/clips/0313-1/10460/20.jpg\nSaving to: dataset/masks/0313-1/10460\nReading from: dataset/clips/0313-1/49200/20.jpg\nSaving to: dataset/masks/0313-1/49200\nReading from: dataset/clips/0313-1/9240/20.jpg\nSaving to: dataset/masks/0313-1/9240\nReading from: dataset/clips/0313-1/28380/20.jpg\nSaving to: dataset/masks/0313-1/28380\nReading from: dataset/clips/0313-1/3920/20.jpg\nSaving to: dataset/masks/0313-1/3920\nReading from: dataset/clips/0313-1/44880/20.jpg\nSaving to: dataset/masks/0313-1/44880\nReading from: dataset/clips/0313-1/17580/20.jpg\nSaving to: dataset/masks/0313-1/17580\nReading from: dataset/clips/0313-1/5600/20.jpg\nSaving to: dataset/masks/0313-1/5600\nReading from: dataset/clips/0313-1/5340/20.jpg\nSaving to: dataset/masks/0313-1/5340\nReading from: dataset/clips/0313-1/8320/20.jpg\nSaving to: dataset/masks/0313-1/8320\nReading from: dataset/clips/0313-1/780/20.jpg\nSaving to: dataset/masks/0313-1/780\nReading from: dataset/clips/0313-1/7420/20.jpg\nSaving to: dataset/masks/0313-1/7420\nReading from: dataset/clips/0313-1/14080/20.jpg\nSaving to: dataset/masks/0313-1/14080\nReading from: dataset/clips/0313-1/34560/20.jpg\nSaving to: dataset/masks/0313-1/34560\nReading from: dataset/clips/0313-1/45960/20.jpg\nSaving to: dataset/masks/0313-1/45960\nReading from: dataset/clips/0313-1/2400/20.jpg\nSaving to: dataset/masks/0313-1/2400\nReading from: dataset/clips/0313-1/13600/20.jpg\nSaving to: dataset/masks/0313-1/13600\nReading from: dataset/clips/0313-1/16060/20.jpg\nSaving to: dataset/masks/0313-1/16060\nReading from: dataset/clips/0313-1/5740/20.jpg\nSaving to: dataset/masks/0313-1/5740\nReading from: dataset/clips/0313-1/13040/20.jpg\nSaving to: dataset/masks/0313-1/13040\nReading from: dataset/clips/0313-1/16760/20.jpg\nSaving to: dataset/masks/0313-1/16760\nReading from: dataset/clips/0313-1/15780/20.jpg\nSaving to: dataset/masks/0313-1/15780\nReading from: dataset/clips/0313-1/40980/20.jpg\nSaving to: dataset/masks/0313-1/40980\nReading from: dataset/clips/0313-1/46140/20.jpg\nSaving to: dataset/masks/0313-1/46140\nReading from: dataset/clips/0313-1/14940/20.jpg\nSaving to: dataset/masks/0313-1/14940\nReading from: dataset/clips/0313-1/12760/20.jpg\nSaving to: dataset/masks/0313-1/12760\nReading from: dataset/clips/0313-1/7840/20.jpg\nSaving to: dataset/masks/0313-1/7840\nReading from: dataset/clips/0313-1/42780/20.jpg\nSaving to: dataset/masks/0313-1/42780\nReading from: dataset/clips/0313-1/26220/20.jpg\nSaving to: dataset/masks/0313-1/26220\nReading from: dataset/clips/0313-1/25260/20.jpg\nSaving to: dataset/masks/0313-1/25260\nReading from: dataset/clips/0313-1/49680/20.jpg\nSaving to: dataset/masks/0313-1/49680\nReading from: dataset/clips/0313-1/540/20.jpg\nSaving to: dataset/masks/0313-1/540\nReading from: dataset/clips/0313-1/23400/20.jpg\nSaving to: dataset/masks/0313-1/23400\nReading from: dataset/clips/0313-1/48180/20.jpg\nSaving to: dataset/masks/0313-1/48180\nReading from: dataset/clips/0313-1/39660/20.jpg\nSaving to: dataset/masks/0313-1/39660\nReading from: dataset/clips/0313-1/16500/20.jpg\nSaving to: dataset/masks/0313-1/16500\nReading from: dataset/clips/0313-1/18660/20.jpg\nSaving to: dataset/masks/0313-1/18660\nReading from: dataset/clips/0313-1/7720/20.jpg\nSaving to: dataset/masks/0313-1/7720\nReading from: dataset/clips/0313-1/46320/20.jpg\nSaving to: dataset/masks/0313-1/46320\nReading from: dataset/clips/0313-1/37440/20.jpg\nSaving to: dataset/masks/0313-1/37440\nReading from: dataset/clips/0313-1/31740/20.jpg\nSaving to: dataset/masks/0313-1/31740\nReading from: dataset/clips/0313-1/11060/20.jpg\nSaving to: dataset/masks/0313-1/11060\nReading from: dataset/clips/0313-1/24600/20.jpg\nSaving to: dataset/masks/0313-1/24600\nReading from: dataset/clips/0313-1/30000/20.jpg\nSaving to: dataset/masks/0313-1/30000\nReading from: dataset/clips/0313-1/43860/20.jpg\nSaving to: dataset/masks/0313-1/43860\nReading from: dataset/clips/0313-1/13620/20.jpg\nSaving to: dataset/masks/0313-1/13620\nReading from: dataset/clips/0313-1/49920/20.jpg\nSaving to: dataset/masks/0313-1/49920\nReading from: dataset/clips/0313-1/49560/20.jpg\nSaving to: dataset/masks/0313-1/49560\nReading from: dataset/clips/0313-1/11100/20.jpg\nSaving to: dataset/masks/0313-1/11100\nReading from: dataset/clips/0313-1/36720/20.jpg\nSaving to: dataset/masks/0313-1/36720\nReading from: dataset/clips/0313-1/6780/20.jpg\nSaving to: dataset/masks/0313-1/6780\nReading from: dataset/clips/0313-1/28860/20.jpg\nSaving to: dataset/masks/0313-1/28860\nReading from: dataset/clips/0313-1/22680/20.jpg\nSaving to: dataset/masks/0313-1/22680\nReading from: dataset/clips/0313-1/9800/20.jpg\nSaving to: dataset/masks/0313-1/9800\nReading from: dataset/clips/0313-1/24300/20.jpg\nSaving to: dataset/masks/0313-1/24300\nReading from: dataset/clips/0313-1/38760/20.jpg\nSaving to: dataset/masks/0313-1/38760\nReading from: dataset/clips/0313-1/14920/20.jpg\nSaving to: dataset/masks/0313-1/14920\nReading from: dataset/clips/0313-1/14460/20.jpg\nSaving to: dataset/masks/0313-1/14460\nReading from: dataset/clips/0313-1/12720/20.jpg\nSaving to: dataset/masks/0313-1/12720\nReading from: dataset/clips/0313-1/21180/20.jpg\nSaving to: dataset/masks/0313-1/21180\nReading from: dataset/clips/0313-1/42480/20.jpg\nSaving to: dataset/masks/0313-1/42480\nReading from: dataset/clips/0313-1/15420/20.jpg\nSaving to: dataset/masks/0313-1/15420\nReading from: dataset/clips/0313-1/22440/20.jpg\nSaving to: dataset/masks/0313-1/22440\nReading from: dataset/clips/0313-1/13700/20.jpg\nSaving to: dataset/masks/0313-1/13700\nReading from: dataset/clips/0313-1/44580/20.jpg\nSaving to: dataset/masks/0313-1/44580\nReading from: dataset/clips/0313-1/20640/20.jpg\nSaving to: dataset/masks/0313-1/20640\nReading from: dataset/clips/0313-1/18780/20.jpg\nSaving to: dataset/masks/0313-1/18780\nReading from: dataset/clips/0313-1/10040/20.jpg\nSaving to: dataset/masks/0313-1/10040\nReading from: dataset/clips/0313-1/10540/20.jpg\nSaving to: dataset/masks/0313-1/10540\nReading from: dataset/clips/0313-1/52080/20.jpg\nSaving to: dataset/masks/0313-1/52080\nReading from: dataset/clips/0313-1/29760/20.jpg\nSaving to: dataset/masks/0313-1/29760\nReading from: dataset/clips/0313-1/15320/20.jpg\nSaving to: dataset/masks/0313-1/15320\nReading from: dataset/clips/0313-1/6540/20.jpg\nSaving to: dataset/masks/0313-1/6540\nReading from: dataset/clips/0313-1/46500/20.jpg\nSaving to: dataset/masks/0313-1/46500\nReading from: dataset/clips/0313-1/36360/20.jpg\nSaving to: dataset/masks/0313-1/36360\nReading from: dataset/clips/0313-1/35760/20.jpg\nSaving to: dataset/masks/0313-1/35760\nReading from: dataset/clips/0313-1/14540/20.jpg\nSaving to: dataset/masks/0313-1/14540\nReading from: dataset/clips/0313-1/9520/20.jpg\nSaving to: dataset/masks/0313-1/9520\nReading from: dataset/clips/0313-1/1260/20.jpg\nSaving to: dataset/masks/0313-1/1260\nReading from: dataset/clips/0313-1/45840/20.jpg\nSaving to: dataset/masks/0313-1/45840\nReading from: dataset/clips/0313-1/20520/20.jpg\nSaving to: dataset/masks/0313-1/20520\nReading from: dataset/clips/0313-1/16480/20.jpg\nSaving to: dataset/masks/0313-1/16480\nReading from: dataset/clips/0313-1/3240/20.jpg\nSaving to: dataset/masks/0313-1/3240\nReading from: dataset/clips/0313-1/48660/20.jpg\nSaving to: dataset/masks/0313-1/48660\nReading from: dataset/clips/0313-1/20400/20.jpg\nSaving to: dataset/masks/0313-1/20400\nReading from: dataset/clips/0313-1/13540/20.jpg\nSaving to: dataset/masks/0313-1/13540\nReading from: dataset/clips/0313-1/21120/20.jpg\nSaving to: dataset/masks/0313-1/21120\nReading from: dataset/clips/0313-1/6560/20.jpg\nSaving to: dataset/masks/0313-1/6560\nReading from: dataset/clips/0313-1/22980/20.jpg\nSaving to: dataset/masks/0313-1/22980\nReading from: dataset/clips/0313-1/8400/20.jpg\nSaving to: dataset/masks/0313-1/8400\nReading from: dataset/clips/0313-1/17940/20.jpg\nSaving to: dataset/masks/0313-1/17940\nReading from: dataset/clips/0313-1/17960/20.jpg\nSaving to: dataset/masks/0313-1/17960\nReading from: dataset/clips/0313-1/11580/20.jpg\nSaving to: dataset/masks/0313-1/11580\nReading from: dataset/clips/0313-1/20700/20.jpg\nSaving to: dataset/masks/0313-1/20700\nReading from: dataset/clips/0313-1/52380/20.jpg\nSaving to: dataset/masks/0313-1/52380\nReading from: dataset/clips/0313-1/6500/20.jpg\nSaving to: dataset/masks/0313-1/6500\nReading from: dataset/clips/0313-1/30240/20.jpg\nSaving to: dataset/masks/0313-1/30240\nReading from: dataset/clips/0313-1/25560/20.jpg\nSaving to: dataset/masks/0313-1/25560\nReading from: dataset/clips/0313-1/14480/20.jpg\nSaving to: dataset/masks/0313-1/14480\nReading from: dataset/clips/0313-1/12040/20.jpg\nSaving to: dataset/masks/0313-1/12040\nReading from: dataset/clips/0313-1/1320/20.jpg\nSaving to: dataset/masks/0313-1/1320\nReading from: dataset/clips/0313-1/20940/20.jpg\nSaving to: dataset/masks/0313-1/20940\nReading from: dataset/clips/0313-1/4260/20.jpg\nSaving to: dataset/masks/0313-1/4260\nReading from: dataset/clips/0313-1/42960/20.jpg\nSaving to: dataset/masks/0313-1/42960\nReading from: dataset/clips/0313-1/41400/20.jpg\nSaving to: dataset/masks/0313-1/41400\nReading from: dataset/clips/0313-1/10860/20.jpg\nSaving to: dataset/masks/0313-1/10860\nReading from: dataset/clips/0313-1/47640/20.jpg\nSaving to: dataset/masks/0313-1/47640\nReading from: dataset/clips/0313-1/5940/20.jpg\nSaving to: dataset/masks/0313-1/5940\nReading from: dataset/clips/0313-1/34020/20.jpg\nSaving to: dataset/masks/0313-1/34020\nReading from: dataset/clips/0313-1/6900/20.jpg\nSaving to: dataset/masks/0313-1/6900\nReading from: dataset/clips/0313-1/13940/20.jpg\nSaving to: dataset/masks/0313-1/13940\nReading from: dataset/clips/0313-1/13300/20.jpg\nSaving to: dataset/masks/0313-1/13300\nReading from: dataset/clips/0313-1/31620/20.jpg\nSaving to: dataset/masks/0313-1/31620\nReading from: dataset/clips/0313-1/40260/20.jpg\nSaving to: dataset/masks/0313-1/40260\nReading from: dataset/clips/0313-1/4880/20.jpg\nSaving to: dataset/masks/0313-1/4880\nReading from: dataset/clips/0313-1/41880/20.jpg\nSaving to: dataset/masks/0313-1/41880\nReading from: dataset/clips/0313-1/34200/20.jpg\nSaving to: dataset/masks/0313-1/34200\nReading from: dataset/clips/0313-1/45300/20.jpg\nSaving to: dataset/masks/0313-1/45300\nReading from: dataset/clips/0313-1/20580/20.jpg\nSaving to: dataset/masks/0313-1/20580\nReading from: dataset/clips/0313-1/47520/20.jpg\nSaving to: dataset/masks/0313-1/47520\nReading from: dataset/clips/0313-1/30540/20.jpg\nSaving to: dataset/masks/0313-1/30540\nReading from: dataset/clips/0313-1/42360/20.jpg\nSaving to: dataset/masks/0313-1/42360\nReading from: dataset/clips/0313-1/5520/20.jpg\nSaving to: dataset/masks/0313-1/5520\nReading from: dataset/clips/0313-1/4280/20.jpg\nSaving to: dataset/masks/0313-1/4280\nReading from: dataset/clips/0313-1/12520/20.jpg\nSaving to: dataset/masks/0313-1/12520\nReading from: dataset/clips/0313-1/6480/20.jpg\nSaving to: dataset/masks/0313-1/6480\nReading from: dataset/clips/0313-1/47400/20.jpg\nSaving to: dataset/masks/0313-1/47400\nReading from: dataset/clips/0313-1/24000/20.jpg\nSaving to: dataset/masks/0313-1/24000\nReading from: dataset/clips/0313-1/10600/20.jpg\nSaving to: dataset/masks/0313-1/10600\nReading from: dataset/clips/0313-1/7880/20.jpg\nSaving to: dataset/masks/0313-1/7880\nReading from: dataset/clips/0313-1/17760/20.jpg\nSaving to: dataset/masks/0313-1/17760\nReading from: dataset/clips/0313-1/27300/20.jpg\nSaving to: dataset/masks/0313-1/27300\nReading from: dataset/clips/0313-1/14960/20.jpg\nSaving to: dataset/masks/0313-1/14960\nReading from: dataset/clips/0313-1/3780/20.jpg\nSaving to: dataset/masks/0313-1/3780\nReading from: dataset/clips/0313-1/26580/20.jpg\nSaving to: dataset/masks/0313-1/26580\nReading from: dataset/clips/0313-1/14720/20.jpg\nSaving to: dataset/masks/0313-1/14720\nReading from: dataset/clips/0313-1/34080/20.jpg\nSaving to: dataset/masks/0313-1/34080\nReading from: dataset/clips/0313-1/46740/20.jpg\nSaving to: dataset/masks/0313-1/46740\nReading from: dataset/clips/0313-1/9220/20.jpg\nSaving to: dataset/masks/0313-1/9220\nReading from: dataset/clips/0313-1/52500/20.jpg\nSaving to: dataset/masks/0313-1/52500\nReading from: dataset/clips/0313-1/20040/20.jpg\nSaving to: dataset/masks/0313-1/20040\nReading from: dataset/clips/0313-1/7540/20.jpg\nSaving to: dataset/masks/0313-1/7540\nReading from: dataset/clips/0313-1/13860/20.jpg\nSaving to: dataset/masks/0313-1/13860\nReading from: dataset/clips/0313-1/10880/20.jpg\nSaving to: dataset/masks/0313-1/10880\nReading from: dataset/clips/0313-1/5760/20.jpg\nSaving to: dataset/masks/0313-1/5760\nReading from: dataset/clips/0313-1/6320/20.jpg\nSaving to: dataset/masks/0313-1/6320\nReading from: dataset/clips/0313-1/6020/20.jpg\nSaving to: dataset/masks/0313-1/6020\nReading from: dataset/clips/0313-1/12780/20.jpg\nSaving to: dataset/masks/0313-1/12780\nReading from: dataset/clips/0313-1/9840/20.jpg\nSaving to: dataset/masks/0313-1/9840\nReading from: dataset/clips/0313-1/44280/20.jpg\nSaving to: dataset/masks/0313-1/44280\nReading from: dataset/clips/0313-1/9480/20.jpg\nSaving to: dataset/masks/0313-1/9480\nReading from: dataset/clips/0313-1/4360/20.jpg\nSaving to: dataset/masks/0313-1/4360\nReading from: dataset/clips/0313-1/14440/20.jpg\nSaving to: dataset/masks/0313-1/14440\nReading from: dataset/clips/0313-1/8980/20.jpg\nSaving to: dataset/masks/0313-1/8980\nReading from: dataset/clips/0313-1/13760/20.jpg\nSaving to: dataset/masks/0313-1/13760\nReading from: dataset/clips/0313-1/39420/20.jpg\nSaving to: dataset/masks/0313-1/39420\nReading from: dataset/clips/0313-1/5160/20.jpg\nSaving to: dataset/masks/0313-1/5160\nReading from: dataset/clips/0313-1/44640/20.jpg\nSaving to: dataset/masks/0313-1/44640\nReading from: dataset/clips/0313-1/4840/20.jpg\nSaving to: dataset/masks/0313-1/4840\nReading from: dataset/clips/0313-1/16400/20.jpg\nSaving to: dataset/masks/0313-1/16400\nReading from: dataset/clips/0313-1/45480/20.jpg\nSaving to: dataset/masks/0313-1/45480\nReading from: dataset/clips/0313-1/45360/20.jpg\nSaving to: dataset/masks/0313-1/45360\nReading from: dataset/clips/0313-1/18060/20.jpg\nSaving to: dataset/masks/0313-1/18060\nReading from: dataset/clips/0313-1/47460/20.jpg\nSaving to: dataset/masks/0313-1/47460\nReading from: dataset/clips/0313-1/7180/20.jpg\nSaving to: dataset/masks/0313-1/7180\nReading from: dataset/clips/0313-1/10760/20.jpg\nSaving to: dataset/masks/0313-1/10760\nReading from: dataset/clips/0313-1/8800/20.jpg\nSaving to: dataset/masks/0313-1/8800\nReading from: dataset/clips/0313-1/1920/20.jpg\nSaving to: dataset/masks/0313-1/1920\nReading from: dataset/clips/0313-1/9500/20.jpg\nSaving to: dataset/masks/0313-1/9500\nReading from: dataset/clips/0313-1/11000/20.jpg\nSaving to: dataset/masks/0313-1/11000\nReading from: dataset/clips/0313-1/43320/20.jpg\nSaving to: dataset/masks/0313-1/43320\nReading from: dataset/clips/0313-1/5980/20.jpg\nSaving to: dataset/masks/0313-1/5980\nReading from: dataset/clips/0313-1/10360/20.jpg\nSaving to: dataset/masks/0313-1/10360\nReading from: dataset/clips/0313-1/39600/20.jpg\nSaving to: dataset/masks/0313-1/39600\nReading from: dataset/clips/0313-1/1980/20.jpg\nSaving to: dataset/masks/0313-1/1980\nReading from: dataset/clips/0313-1/10300/20.jpg\nSaving to: dataset/masks/0313-1/10300\nReading from: dataset/clips/0313-1/31080/20.jpg\nSaving to: dataset/masks/0313-1/31080\nReading from: dataset/clips/0313-1/19200/20.jpg\nSaving to: dataset/masks/0313-1/19200\nReading from: dataset/clips/0313-1/11340/20.jpg\nSaving to: dataset/masks/0313-1/11340\nReading from: dataset/clips/0313-1/9140/20.jpg\nSaving to: dataset/masks/0313-1/9140\nReading from: dataset/clips/0313-1/16660/20.jpg\nSaving to: dataset/masks/0313-1/16660\nReading from: dataset/clips/0313-1/10560/20.jpg\nSaving to: dataset/masks/0313-1/10560\nReading from: dataset/clips/0313-1/31140/20.jpg\nSaving to: dataset/masks/0313-1/31140\nReading from: dataset/clips/0313-1/15240/20.jpg\nSaving to: dataset/masks/0313-1/15240\nReading from: dataset/clips/0313-1/50340/20.jpg\nSaving to: dataset/masks/0313-1/50340\nReading from: dataset/clips/0313-1/20220/20.jpg\nSaving to: dataset/masks/0313-1/20220\nReading from: dataset/clips/0313-1/5540/20.jpg\nSaving to: dataset/masks/0313-1/5540\nReading from: dataset/clips/0313-1/14100/20.jpg\nSaving to: dataset/masks/0313-1/14100\nReading from: dataset/clips/0313-1/10740/20.jpg\nSaving to: dataset/masks/0313-1/10740\nReading from: dataset/clips/0313-1/10400/20.jpg\nSaving to: dataset/masks/0313-1/10400\nReading from: dataset/clips/0313-1/7160/20.jpg\nSaving to: dataset/masks/0313-1/7160\nReading from: dataset/clips/0313-1/44100/20.jpg\nSaving to: dataset/masks/0313-1/44100\nReading from: dataset/clips/0313-1/38520/20.jpg\nSaving to: dataset/masks/0313-1/38520\nReading from: dataset/clips/0313-1/19980/20.jpg\nSaving to: dataset/masks/0313-1/19980\nReading from: dataset/clips/0313-1/14700/20.jpg\nSaving to: dataset/masks/0313-1/14700\nReading from: dataset/clips/0313-1/34500/20.jpg\nSaving to: dataset/masks/0313-1/34500\nReading from: dataset/clips/0313-1/28800/20.jpg\nSaving to: dataset/masks/0313-1/28800\nReading from: dataset/clips/0313-1/19620/20.jpg\nSaving to: dataset/masks/0313-1/19620\nReading from: dataset/clips/0313-1/7560/20.jpg\nSaving to: dataset/masks/0313-1/7560\nReading from: dataset/clips/0313-1/18360/20.jpg\nSaving to: dataset/masks/0313-1/18360\nReading from: dataset/clips/0313-1/48780/20.jpg\nSaving to: dataset/masks/0313-1/48780\nReading from: dataset/clips/0313-1/4100/20.jpg\nSaving to: dataset/masks/0313-1/4100\nReading from: dataset/clips/0313-1/15020/20.jpg\nSaving to: dataset/masks/0313-1/15020\nReading from: dataset/clips/0313-1/12660/20.jpg\nSaving to: dataset/masks/0313-1/12660\nReading from: dataset/clips/0313-1/32820/20.jpg\nSaving to: dataset/masks/0313-1/32820\nReading from: dataset/clips/0313-1/12120/20.jpg\nSaving to: dataset/masks/0313-1/12120\nReading from: dataset/clips/0313-1/51900/20.jpg\nSaving to: dataset/masks/0313-1/51900\nReading from: dataset/clips/0313-1/42660/20.jpg\nSaving to: dataset/masks/0313-1/42660\nReading from: dataset/clips/0313-1/20760/20.jpg\nSaving to: dataset/masks/0313-1/20760\nReading from: dataset/clips/0313-1/36960/20.jpg\nSaving to: dataset/masks/0313-1/36960\nReading from: dataset/clips/0313-1/17880/20.jpg\nSaving to: dataset/masks/0313-1/17880\nReading from: dataset/clips/0313-1/35280/20.jpg\nSaving to: dataset/masks/0313-1/35280\nReading from: dataset/clips/0313-1/6060/20.jpg\nSaving to: dataset/masks/0313-1/6060\nReading from: dataset/clips/0313-1/36420/20.jpg\nSaving to: dataset/masks/0313-1/36420\nReading from: dataset/clips/0313-1/51480/20.jpg\nSaving to: dataset/masks/0313-1/51480\nReading from: dataset/clips/0313-1/50760/20.jpg\nSaving to: dataset/masks/0313-1/50760\nReading from: dataset/clips/0313-1/50400/20.jpg\nSaving to: dataset/masks/0313-1/50400\nReading from: dataset/clips/0313-1/39540/20.jpg\nSaving to: dataset/masks/0313-1/39540\nReading from: dataset/clips/0313-1/27060/20.jpg\nSaving to: dataset/masks/0313-1/27060\nReading from: dataset/clips/0313-1/34440/20.jpg\nSaving to: dataset/masks/0313-1/34440\nReading from: dataset/clips/0313-1/11880/20.jpg\nSaving to: dataset/masks/0313-1/11880\nReading from: dataset/clips/0313-1/12340/20.jpg\nSaving to: dataset/masks/0313-1/12340\nReading from: dataset/clips/0313-1/16240/20.jpg\nSaving to: dataset/masks/0313-1/16240\nReading from: dataset/clips/0313-1/46020/20.jpg\nSaving to: dataset/masks/0313-1/46020\nReading from: dataset/clips/0313-1/51000/20.jpg\nSaving to: dataset/masks/0313-1/51000\nReading from: dataset/clips/0313-1/6640/20.jpg\nSaving to: dataset/masks/0313-1/6640\nReading from: dataset/clips/0313-1/16140/20.jpg\nSaving to: dataset/masks/0313-1/16140\nReading from: dataset/clips/0313-1/4020/20.jpg\nSaving to: dataset/masks/0313-1/4020\nReading from: dataset/clips/0313-1/16200/20.jpg\nSaving to: dataset/masks/0313-1/16200\nReading from: dataset/clips/0313-1/34260/20.jpg\nSaving to: dataset/masks/0313-1/34260\nReading from: dataset/clips/0313-1/16840/20.jpg\nSaving to: dataset/masks/0313-1/16840\nReading from: dataset/clips/0313-1/15820/20.jpg\nSaving to: dataset/masks/0313-1/15820\nReading from: dataset/clips/0313-1/28020/20.jpg\nSaving to: dataset/masks/0313-1/28020\nReading from: dataset/clips/0313-1/33540/20.jpg\nSaving to: dataset/masks/0313-1/33540\nReading from: dataset/clips/0313-1/15620/20.jpg\nSaving to: dataset/masks/0313-1/15620\nReading from: dataset/clips/0313-1/12460/20.jpg\nSaving to: dataset/masks/0313-1/12460\nReading from: dataset/clips/0313-1/44220/20.jpg\nSaving to: dataset/masks/0313-1/44220\nReading from: dataset/clips/0313-1/15580/20.jpg\nSaving to: dataset/masks/0313-1/15580\nReading from: dataset/clips/0313-1/15260/20.jpg\nSaving to: dataset/masks/0313-1/15260\nReading from: dataset/clips/0313-1/41820/20.jpg\nSaving to: dataset/masks/0313-1/41820\nReading from: dataset/clips/0313-1/41580/20.jpg\nSaving to: dataset/masks/0313-1/41580\nReading from: dataset/clips/0313-1/4200/20.jpg\nSaving to: dataset/masks/0313-1/4200\nReading from: dataset/clips/0313-1/13920/20.jpg\nSaving to: dataset/masks/0313-1/13920\nReading from: dataset/clips/0313-1/7800/20.jpg\nSaving to: dataset/masks/0313-1/7800\nReading from: dataset/clips/0313-1/12860/20.jpg\nSaving to: dataset/masks/0313-1/12860\nReading from: dataset/clips/0313-1/12840/20.jpg\nSaving to: dataset/masks/0313-1/12840\nReading from: dataset/clips/0313-1/11380/20.jpg\nSaving to: dataset/masks/0313-1/11380\nReading from: dataset/clips/0313-1/11320/20.jpg\nSaving to: dataset/masks/0313-1/11320\nReading from: dataset/clips/0313-1/8180/20.jpg\nSaving to: dataset/masks/0313-1/8180\nReading from: dataset/clips/0313-1/17200/20.jpg\nSaving to: dataset/masks/0313-1/17200\nReading from: dataset/clips/0313-1/40800/20.jpg\nSaving to: dataset/masks/0313-1/40800\nReading from: dataset/clips/0313-1/6080/20.jpg\nSaving to: dataset/masks/0313-1/6080\nReading from: dataset/clips/0313-1/17720/20.jpg\nSaving to: dataset/masks/0313-1/17720\nReading from: dataset/clips/0313-1/29700/20.jpg\nSaving to: dataset/masks/0313-1/29700\nReading from: dataset/clips/0313-1/5120/20.jpg\nSaving to: dataset/masks/0313-1/5120\nReading from: dataset/clips/0313-1/45180/20.jpg\nSaving to: dataset/masks/0313-1/45180\nReading from: dataset/clips/0313-1/49260/20.jpg\nSaving to: dataset/masks/0313-1/49260\nReading from: dataset/clips/0313-1/30840/20.jpg\nSaving to: dataset/masks/0313-1/30840\nReading from: dataset/clips/0313-1/5860/20.jpg\nSaving to: dataset/masks/0313-1/5860\nReading from: dataset/clips/0313-1/14220/20.jpg\nSaving to: dataset/masks/0313-1/14220\nReading from: dataset/clips/0313-1/46800/20.jpg\nSaving to: dataset/masks/0313-1/46800\nReading from: dataset/clips/0313-1/19140/20.jpg\nSaving to: dataset/masks/0313-1/19140\nReading from: dataset/clips/0313-1/6000/20.jpg\nSaving to: dataset/masks/0313-1/6000\nReading from: dataset/clips/0313-1/16600/20.jpg\nSaving to: dataset/masks/0313-1/16600\nReading from: dataset/clips/0313-1/28620/20.jpg\nSaving to: dataset/masks/0313-1/28620\nReading from: dataset/clips/0313-1/51780/20.jpg\nSaving to: dataset/masks/0313-1/51780\nReading from: dataset/clips/0313-1/16580/20.jpg\nSaving to: dataset/masks/0313-1/16580\nReading from: dataset/clips/0313-1/44520/20.jpg\nSaving to: dataset/masks/0313-1/44520\nReading from: dataset/clips/0313-1/24960/20.jpg\nSaving to: dataset/masks/0313-1/24960\nReading from: dataset/clips/0313-1/5460/20.jpg\nSaving to: dataset/masks/0313-1/5460\nReading from: dataset/clips/0313-1/12500/20.jpg\nSaving to: dataset/masks/0313-1/12500\nReading from: dataset/clips/0313-1/15560/20.jpg\nSaving to: dataset/masks/0313-1/15560\nReading from: dataset/clips/0313-1/5260/20.jpg\nSaving to: dataset/masks/0313-1/5260\nReading from: dataset/clips/0313-1/32280/20.jpg\nSaving to: dataset/masks/0313-1/32280\nReading from: dataset/clips/0313-1/45420/20.jpg\nSaving to: dataset/masks/0313-1/45420\nReading from: dataset/clips/0313-1/25920/20.jpg\nSaving to: dataset/masks/0313-1/25920\nReading from: dataset/clips/0313-1/18000/20.jpg\nSaving to: dataset/masks/0313-1/18000\nReading from: dataset/clips/0313-1/900/20.jpg\nSaving to: dataset/masks/0313-1/900\nReading from: dataset/clips/0313-1/31200/20.jpg\nSaving to: dataset/masks/0313-1/31200\nReading from: dataset/clips/0313-1/28740/20.jpg\nSaving to: dataset/masks/0313-1/28740\nReading from: dataset/clips/0313-1/35940/20.jpg\nSaving to: dataset/masks/0313-1/35940\nReading from: dataset/clips/0313-1/26040/20.jpg\nSaving to: dataset/masks/0313-1/26040\nReading from: dataset/clips/0313-1/13080/20.jpg\nSaving to: dataset/masks/0313-1/13080\nReading from: dataset/clips/0313-1/44400/20.jpg\nSaving to: dataset/masks/0313-1/44400\nReading from: dataset/clips/0313-1/22620/20.jpg\nSaving to: dataset/masks/0313-1/22620\nReading from: dataset/clips/0313-1/11760/20.jpg\nSaving to: dataset/masks/0313-1/11760\nReading from: dataset/clips/0313-1/52020/20.jpg\nSaving to: dataset/masks/0313-1/52020\nReading from: dataset/clips/0313-1/38340/20.jpg\nSaving to: dataset/masks/0313-1/38340\nReading from: dataset/clips/0313-1/28260/20.jpg\nSaving to: dataset/masks/0313-1/28260\nReading from: dataset/clips/0313-1/840/20.jpg\nSaving to: dataset/masks/0313-1/840\nReading from: dataset/clips/0313-1/1380/20.jpg\nSaving to: dataset/masks/0313-1/1380\nReading from: dataset/clips/0313-1/23760/20.jpg\nSaving to: dataset/masks/0313-1/23760\nReading from: dataset/clips/0313-1/15700/20.jpg\nSaving to: dataset/masks/0313-1/15700\nReading from: dataset/clips/0313-1/12540/20.jpg\nSaving to: dataset/masks/0313-1/12540\nReading from: dataset/clips/0313-1/5960/20.jpg\nSaving to: dataset/masks/0313-1/5960\nReading from: dataset/clips/0313-1/46440/20.jpg\nSaving to: dataset/masks/0313-1/46440\nReading from: dataset/clips/0313-1/16560/20.jpg\nSaving to: dataset/masks/0313-1/16560\nReading from: dataset/clips/0313-1/12440/20.jpg\nSaving to: dataset/masks/0313-1/12440\nReading from: dataset/clips/0313-1/31320/20.jpg\nSaving to: dataset/masks/0313-1/31320\nReading from: dataset/clips/0313-1/2100/20.jpg\nSaving to: dataset/masks/0313-1/2100\nReading from: dataset/clips/0313-1/45600/20.jpg\nSaving to: dataset/masks/0313-1/45600\nReading from: dataset/clips/0313-1/14680/20.jpg\nSaving to: dataset/masks/0313-1/14680\nReading from: dataset/clips/0313-1/30060/20.jpg\nSaving to: dataset/masks/0313-1/30060\nReading from: dataset/clips/0313-1/38580/20.jpg\nSaving to: dataset/masks/0313-1/38580\nReading from: dataset/clips/0313-1/38700/20.jpg\nSaving to: dataset/masks/0313-1/38700\nReading from: dataset/clips/0313-1/22500/20.jpg\nSaving to: dataset/masks/0313-1/22500\nReading from: dataset/clips/0313-1/50100/20.jpg\nSaving to: dataset/masks/0313-1/50100\nReading from: dataset/clips/0313-1/13160/20.jpg\nSaving to: dataset/masks/0313-1/13160\nReading from: dataset/clips/0313-1/28560/20.jpg\nSaving to: dataset/masks/0313-1/28560\nReading from: dataset/clips/0313-1/52440/20.jpg\nSaving to: dataset/masks/0313-1/52440\nReading from: dataset/clips/0313-1/6300/20.jpg\nSaving to: dataset/masks/0313-1/6300\nReading from: dataset/clips/0313-1/18120/20.jpg\nSaving to: dataset/masks/0313-1/18120\nReading from: dataset/clips/0313-1/10720/20.jpg\nSaving to: dataset/masks/0313-1/10720\nReading from: dataset/clips/0313-1/15180/20.jpg\nSaving to: dataset/masks/0313-1/15180\nReading from: dataset/clips/0313-1/12480/20.jpg\nSaving to: dataset/masks/0313-1/12480\nReading from: dataset/clips/0313-1/35460/20.jpg\nSaving to: dataset/masks/0313-1/35460\nReading from: dataset/clips/0313-1/15000/20.jpg\nSaving to: dataset/masks/0313-1/15000\nReading from: dataset/clips/0313-1/9380/20.jpg\nSaving to: dataset/masks/0313-1/9380\nReading from: dataset/clips/0313-1/48420/20.jpg\nSaving to: dataset/masks/0313-1/48420\nReading from: dataset/clips/0313-1/20340/20.jpg\nSaving to: dataset/masks/0313-1/20340\nReading from: dataset/clips/0313-1/33360/20.jpg\nSaving to: dataset/masks/0313-1/33360\nReading from: dataset/clips/0313-1/13740/20.jpg\nSaving to: dataset/masks/0313-1/13740\nReading from: dataset/clips/0313-1/37140/20.jpg\nSaving to: dataset/masks/0313-1/37140\nReading from: dataset/clips/0313-1/21000/20.jpg\nSaving to: dataset/masks/0313-1/21000\nReading from: dataset/clips/0313-1/34380/20.jpg\nSaving to: dataset/masks/0313-1/34380\nReading from: dataset/clips/0313-1/17040/20.jpg\nSaving to: dataset/masks/0313-1/17040\nReading from: dataset/clips/0313-1/17400/20.jpg\nSaving to: dataset/masks/0313-1/17400\nReading from: dataset/clips/0313-1/12640/20.jpg\nSaving to: dataset/masks/0313-1/12640\nReading from: dataset/clips/0313-1/7820/20.jpg\nSaving to: dataset/masks/0313-1/7820\nReading from: dataset/clips/0313-1/9060/20.jpg\nSaving to: dataset/masks/0313-1/9060\nReading from: dataset/clips/0313-1/13220/20.jpg\nSaving to: dataset/masks/0313-1/13220\nReading from: dataset/clips/0313-1/120/20.jpg\nSaving to: dataset/masks/0313-1/120\nReading from: dataset/clips/0313-1/7220/20.jpg\nSaving to: dataset/masks/0313-1/7220\nReading from: dataset/clips/0313-1/15460/20.jpg\nSaving to: dataset/masks/0313-1/15460\nReading from: dataset/clips/0313-1/9780/20.jpg\nSaving to: dataset/masks/0313-1/9780\nReading from: dataset/clips/0313-1/6160/20.jpg\nSaving to: dataset/masks/0313-1/6160\nReading from: dataset/clips/0313-1/1560/20.jpg\nSaving to: dataset/masks/0313-1/1560\nReading from: dataset/clips/0313-1/8680/20.jpg\nSaving to: dataset/masks/0313-1/8680\nReading from: dataset/clips/0313-1/12800/20.jpg\nSaving to: dataset/masks/0313-1/12800\nReading from: dataset/clips/0313-1/13840/20.jpg\nSaving to: dataset/masks/0313-1/13840\nReading from: dataset/clips/0313-1/32160/20.jpg\nSaving to: dataset/masks/0313-1/32160\nReading from: dataset/clips/0313-1/8880/20.jpg\nSaving to: dataset/masks/0313-1/8880\nReading from: dataset/clips/0313-1/45120/20.jpg\nSaving to: dataset/masks/0313-1/45120\nReading from: dataset/clips/0313-1/8040/20.jpg\nSaving to: dataset/masks/0313-1/8040\nReading from: dataset/clips/0313-1/8760/20.jpg\nSaving to: dataset/masks/0313-1/8760\nReading from: dataset/clips/0313-1/6880/20.jpg\nSaving to: dataset/masks/0313-1/6880\nReading from: dataset/clips/0313-1/3420/20.jpg\nSaving to: dataset/masks/0313-1/3420\nReading from: dataset/clips/0313-1/4660/20.jpg\nSaving to: dataset/masks/0313-1/4660\nReading from: dataset/clips/0313-1/11540/20.jpg\nSaving to: dataset/masks/0313-1/11540\nReading from: dataset/clips/0313-1/19920/20.jpg\nSaving to: dataset/masks/0313-1/19920\nReading from: dataset/clips/0313-1/21960/20.jpg\nSaving to: dataset/masks/0313-1/21960\nReading from: dataset/clips/0313-1/41160/20.jpg\nSaving to: dataset/masks/0313-1/41160\nReading from: dataset/clips/0313-1/36120/20.jpg\nSaving to: dataset/masks/0313-1/36120\nReading from: dataset/clips/0313-1/7580/20.jpg\nSaving to: dataset/masks/0313-1/7580\nReading from: dataset/clips/0313-1/31860/20.jpg\nSaving to: dataset/masks/0313-1/31860\nReading from: dataset/clips/0313-1/12380/20.jpg\nSaving to: dataset/masks/0313-1/12380\nReading from: dataset/clips/0313-1/33960/20.jpg\nSaving to: dataset/masks/0313-1/33960\nReading from: dataset/clips/0313-1/11860/20.jpg\nSaving to: dataset/masks/0313-1/11860\nReading from: dataset/clips/0313-1/2280/20.jpg\nSaving to: dataset/masks/0313-1/2280\nReading from: dataset/clips/0313-1/52200/20.jpg\nSaving to: dataset/masks/0313-1/52200\nReading from: dataset/clips/0313-1/10060/20.jpg\nSaving to: dataset/masks/0313-1/10060\nReading from: dataset/clips/0313-1/4120/20.jpg\nSaving to: dataset/masks/0313-1/4120\nReading from: dataset/clips/0313-1/10580/20.jpg\nSaving to: dataset/masks/0313-1/10580\nReading from: dataset/clips/0313-1/5180/20.jpg\nSaving to: dataset/masks/0313-1/5180\nReading from: dataset/clips/0313-1/14980/20.jpg\nSaving to: dataset/masks/0313-1/14980\nReading from: dataset/clips/0313-1/5300/20.jpg\nSaving to: dataset/masks/0313-1/5300\nReading from: dataset/clips/0313-1/4900/20.jpg\nSaving to: dataset/masks/0313-1/4900\nReading from: dataset/clips/0313-1/6940/20.jpg\nSaving to: dataset/masks/0313-1/6940\nReading from: dataset/clips/0313-1/22020/20.jpg\nSaving to: dataset/masks/0313-1/22020\nReading from: dataset/clips/0313-1/26760/20.jpg\nSaving to: dataset/masks/0313-1/26760\nReading from: dataset/clips/0313-1/12180/20.jpg\nSaving to: dataset/masks/0313-1/12180\nReading from: dataset/clips/0313-1/11360/20.jpg\nSaving to: dataset/masks/0313-1/11360\nReading from: dataset/clips/0313-1/17000/20.jpg\nSaving to: dataset/masks/0313-1/17000\nReading from: dataset/clips/0313-1/2940/20.jpg\nSaving to: dataset/masks/0313-1/2940\nReading from: dataset/clips/0313-1/11680/20.jpg\nSaving to: dataset/masks/0313-1/11680\nReading from: dataset/clips/0313-1/52800/20.jpg\nSaving to: dataset/masks/0313-1/52800\nReading from: dataset/clips/0313-1/44340/20.jpg\nSaving to: dataset/masks/0313-1/44340\nReading from: dataset/clips/0313-1/1080/20.jpg\nSaving to: dataset/masks/0313-1/1080\nReading from: dataset/clips/0313-1/13400/20.jpg\nSaving to: dataset/masks/0313-1/13400\nReading from: dataset/clips/0313-1/28320/20.jpg\nSaving to: dataset/masks/0313-1/28320\nReading from: dataset/clips/0313-1/32220/20.jpg\nSaving to: dataset/masks/0313-1/32220\nReading from: dataset/clips/0313-1/2040/20.jpg\nSaving to: dataset/masks/0313-1/2040\nReading from: dataset/clips/0313-1/12560/20.jpg\nSaving to: dataset/masks/0313-1/12560\nReading from: dataset/clips/0313-1/13100/20.jpg\nSaving to: dataset/masks/0313-1/13100\nReading from: dataset/clips/0313-1/32580/20.jpg\nSaving to: dataset/masks/0313-1/32580\nReading from: dataset/clips/0313-1/13880/20.jpg\nSaving to: dataset/masks/0313-1/13880\nReading from: dataset/clips/0313-1/14740/20.jpg\nSaving to: dataset/masks/0313-1/14740\nReading from: dataset/clips/0313-1/13900/20.jpg\nSaving to: dataset/masks/0313-1/13900\nReading from: dataset/clips/0313-1/3000/20.jpg\nSaving to: dataset/masks/0313-1/3000\nReading from: dataset/clips/0313-1/33000/20.jpg\nSaving to: dataset/masks/0313-1/33000\nReading from: dataset/clips/0313-1/7460/20.jpg\nSaving to: dataset/masks/0313-1/7460\nReading from: dataset/clips/0313-1/37560/20.jpg\nSaving to: dataset/masks/0313-1/37560\nReading from: dataset/clips/0313-1/52320/20.jpg\nSaving to: dataset/masks/0313-1/52320\nReading from: dataset/clips/0313-1/15520/20.jpg\nSaving to: dataset/masks/0313-1/15520\nReading from: dataset/clips/0313-1/32460/20.jpg\nSaving to: dataset/masks/0313-1/32460\nReading from: dataset/clips/0313-1/17160/20.jpg\nSaving to: dataset/masks/0313-1/17160\nReading from: dataset/clips/0313-1/1740/20.jpg\nSaving to: dataset/masks/0313-1/1740\nReading from: dataset/clips/0313-1/50640/20.jpg\nSaving to: dataset/masks/0313-1/50640\nReading from: dataset/clips/0313-1/25740/20.jpg\nSaving to: dataset/masks/0313-1/25740\nReading from: dataset/clips/0313-1/19380/20.jpg\nSaving to: dataset/masks/0313-1/19380\nReading from: dataset/clips/0313-1/16920/20.jpg\nSaving to: dataset/masks/0313-1/16920\nReading from: dataset/clips/0313-1/31260/20.jpg\nSaving to: dataset/masks/0313-1/31260\nReading from: dataset/clips/0313-1/47100/20.jpg\nSaving to: dataset/masks/0313-1/47100\nReading from: dataset/clips/0313-1/2220/20.jpg\nSaving to: dataset/masks/0313-1/2220\nReading from: dataset/clips/0313-1/16340/20.jpg\nSaving to: dataset/masks/0313-1/16340\nReading from: dataset/clips/0313-1/19260/20.jpg\nSaving to: dataset/masks/0313-1/19260\nReading from: dataset/clips/0313-1/10700/20.jpg\nSaving to: dataset/masks/0313-1/10700\nReading from: dataset/clips/0313-1/14360/20.jpg\nSaving to: dataset/masks/0313-1/14360\nReading from: dataset/clips/0313-1/3600/20.jpg\nSaving to: dataset/masks/0313-1/3600\nReading from: dataset/clips/0313-1/7300/20.jpg\nSaving to: dataset/masks/0313-1/7300\nReading from: dataset/clips/0313-1/8920/20.jpg\nSaving to: dataset/masks/0313-1/8920\nReading from: dataset/clips/0313-1/8240/20.jpg\nSaving to: dataset/masks/0313-1/8240\nReading from: dataset/clips/0313-1/14560/20.jpg\nSaving to: dataset/masks/0313-1/14560\nReading from: dataset/clips/0313-2/12180/20.jpg\nSaving to: dataset/masks/0313-2/12180\nReading from: dataset/clips/0313-2/33280/20.jpg\nSaving to: dataset/masks/0313-2/33280\nReading from: dataset/clips/0313-2/850/20.jpg\nSaving to: dataset/masks/0313-2/850\nReading from: dataset/clips/0313-2/275/20.jpg\nSaving to: dataset/masks/0313-2/275\nReading from: dataset/clips/0313-2/38640/20.jpg\nSaving to: dataset/masks/0313-2/38640\nReading from: dataset/clips/0313-2/105/20.jpg\nSaving to: dataset/masks/0313-2/105\nReading from: dataset/clips/0313-2/39780/20.jpg\nSaving to: dataset/masks/0313-2/39780\nReading from: dataset/clips/0313-2/1275/20.jpg\nSaving to: dataset/masks/0313-2/1275\nReading from: dataset/clips/0313-2/39120/20.jpg\nSaving to: dataset/masks/0313-2/39120\nReading from: dataset/clips/0313-2/8460/20.jpg\nSaving to: dataset/masks/0313-2/8460\nReading from: dataset/clips/0313-2/35900/20.jpg\nSaving to: dataset/masks/0313-2/35900\nReading from: dataset/clips/0313-2/60600/20.jpg\nSaving to: dataset/masks/0313-2/60600\nReading from: dataset/clips/0313-2/22140/20.jpg\nSaving to: dataset/masks/0313-2/22140\nReading from: dataset/clips/0313-2/20100/20.jpg\nSaving to: dataset/masks/0313-2/20100\nReading from: dataset/clips/0313-2/1510/20.jpg\nSaving to: dataset/masks/0313-2/1510\nReading from: dataset/clips/0313-2/1530/20.jpg\nSaving to: dataset/masks/0313-2/1530\nReading from: dataset/clips/0313-2/32900/20.jpg\nSaving to: dataset/masks/0313-2/32900\nReading from: dataset/clips/0313-2/1705/20.jpg\nSaving to: dataset/masks/0313-2/1705\nReading from: dataset/clips/0313-2/24300/20.jpg\nSaving to: dataset/masks/0313-2/24300\nReading from: dataset/clips/0313-2/325/20.jpg\nSaving to: dataset/masks/0313-2/325\nReading from: dataset/clips/0313-2/38700/20.jpg\nSaving to: dataset/masks/0313-2/38700\nReading from: dataset/clips/0313-2/385/20.jpg\nSaving to: dataset/masks/0313-2/385\nReading from: dataset/clips/0313-2/28680/20.jpg\nSaving to: dataset/masks/0313-2/28680\nReading from: dataset/clips/0313-2/1380/20.jpg\nSaving to: dataset/masks/0313-2/1380\nReading from: dataset/clips/0313-2/5160/20.jpg\nSaving to: dataset/masks/0313-2/5160\nReading from: dataset/clips/0313-2/41780/20.jpg\nSaving to: dataset/masks/0313-2/41780\nReading from: dataset/clips/0313-2/32520/20.jpg\nSaving to: dataset/masks/0313-2/32520\nReading from: dataset/clips/0313-2/7080/20.jpg\nSaving to: dataset/masks/0313-2/7080\nReading from: dataset/clips/0313-2/35760/20.jpg\nSaving to: dataset/masks/0313-2/35760\nReading from: dataset/clips/0313-2/555/20.jpg\nSaving to: dataset/masks/0313-2/555\nReading from: dataset/clips/0313-2/1370/20.jpg\nSaving to: dataset/masks/0313-2/1370\nReading from: dataset/clips/0313-2/39660/20.jpg\nSaving to: dataset/masks/0313-2/39660\nReading from: dataset/clips/0313-2/11220/20.jpg\nSaving to: dataset/masks/0313-2/11220\nReading from: dataset/clips/0313-2/15540/20.jpg\nSaving to: dataset/masks/0313-2/15540\nReading from: dataset/clips/0313-2/2400/20.jpg\nSaving to: dataset/masks/0313-2/2400\nReading from: dataset/clips/0313-2/870/20.jpg\nSaving to: dataset/masks/0313-2/870\nReading from: dataset/clips/0313-2/95/20.jpg\nSaving to: dataset/masks/0313-2/95\nReading from: dataset/clips/0313-2/855/20.jpg\nSaving to: dataset/masks/0313-2/855\nReading from: dataset/clips/0313-2/30600/20.jpg\nSaving to: dataset/masks/0313-2/30600\nReading from: dataset/clips/0313-2/435/20.jpg\nSaving to: dataset/masks/0313-2/435\nReading from: dataset/clips/0313-2/19860/20.jpg\nSaving to: dataset/masks/0313-2/19860\nReading from: dataset/clips/0313-2/42040/20.jpg\nSaving to: dataset/masks/0313-2/42040\nReading from: dataset/clips/0313-2/995/20.jpg\nSaving to: dataset/masks/0313-2/995\nReading from: dataset/clips/0313-2/370/20.jpg\nSaving to: dataset/masks/0313-2/370\nReading from: dataset/clips/0313-2/39400/20.jpg\nSaving to: dataset/masks/0313-2/39400\nReading from: dataset/clips/0313-2/1410/20.jpg\nSaving to: dataset/masks/0313-2/1410\nReading from: dataset/clips/0313-2/1325/20.jpg\nSaving to: dataset/masks/0313-2/1325\nReading from: dataset/clips/0313-2/23520/20.jpg\nSaving to: dataset/masks/0313-2/23520\nReading from: dataset/clips/0313-2/13440/20.jpg\nSaving to: dataset/masks/0313-2/13440\nReading from: dataset/clips/0313-2/16500/20.jpg\nSaving to: dataset/masks/0313-2/16500\nReading from: dataset/clips/0313-2/34020/20.jpg\nSaving to: dataset/masks/0313-2/34020\nReading from: dataset/clips/0313-2/64020/20.jpg\nSaving to: dataset/masks/0313-2/64020\nReading from: dataset/clips/0313-2/1405/20.jpg\nSaving to: dataset/masks/0313-2/1405\nReading from: dataset/clips/0313-2/14340/20.jpg\nSaving to: dataset/masks/0313-2/14340\nReading from: dataset/clips/0313-2/40660/20.jpg\nSaving to: dataset/masks/0313-2/40660\nReading from: dataset/clips/0313-2/39380/20.jpg\nSaving to: dataset/masks/0313-2/39380\nReading from: dataset/clips/0313-2/32640/20.jpg\nSaving to: dataset/masks/0313-2/32640\nReading from: dataset/clips/0313-2/19260/20.jpg\nSaving to: dataset/masks/0313-2/19260\nReading from: dataset/clips/0313-2/1350/20.jpg\nSaving to: dataset/masks/0313-2/1350\nReading from: dataset/clips/0313-2/38060/20.jpg\nSaving to: dataset/masks/0313-2/38060\nReading from: dataset/clips/0313-2/11760/20.jpg\nSaving to: dataset/masks/0313-2/11760\nReading from: dataset/clips/0313-2/34180/20.jpg\nSaving to: dataset/masks/0313-2/34180\nReading from: dataset/clips/0313-2/3360/20.jpg\nSaving to: dataset/masks/0313-2/3360\nReading from: dataset/clips/0313-2/1135/20.jpg\nSaving to: dataset/masks/0313-2/1135\nReading from: dataset/clips/0313-2/21720/20.jpg\nSaving to: dataset/masks/0313-2/21720\nReading from: dataset/clips/0313-2/600/20.jpg\nSaving to: dataset/masks/0313-2/600\nReading from: dataset/clips/0313-2/11700/20.jpg\nSaving to: dataset/masks/0313-2/11700\nReading from: dataset/clips/0313-2/4920/20.jpg\nSaving to: dataset/masks/0313-2/4920\nReading from: dataset/clips/0313-2/31680/20.jpg\nSaving to: dataset/masks/0313-2/31680\nReading from: dataset/clips/0313-2/970/20.jpg\nSaving to: dataset/masks/0313-2/970\nReading from: dataset/clips/0313-2/3300/20.jpg\nSaving to: dataset/masks/0313-2/3300\nReading from: dataset/clips/0313-2/27120/20.jpg\nSaving to: dataset/masks/0313-2/27120\nReading from: dataset/clips/0313-2/1355/20.jpg\nSaving to: dataset/masks/0313-2/1355\nReading from: dataset/clips/0313-2/13200/20.jpg\nSaving to: dataset/masks/0313-2/13200\nReading from: dataset/clips/0313-2/37640/20.jpg\nSaving to: dataset/masks/0313-2/37640\nReading from: dataset/clips/0313-2/1465/20.jpg\nSaving to: dataset/masks/0313-2/1465\nReading from: dataset/clips/0313-2/40880/20.jpg\nSaving to: dataset/masks/0313-2/40880\nReading from: dataset/clips/0313-2/38960/20.jpg\nSaving to: dataset/masks/0313-2/38960\nReading from: dataset/clips/0313-2/550/20.jpg\nSaving to: dataset/masks/0313-2/550\nReading from: dataset/clips/0313-2/6480/20.jpg\nSaving to: dataset/masks/0313-2/6480\nReading from: dataset/clips/0313-2/27300/20.jpg\nSaving to: dataset/masks/0313-2/27300\nReading from: dataset/clips/0313-2/9180/20.jpg\nSaving to: dataset/masks/0313-2/9180\nReading from: dataset/clips/0313-2/33680/20.jpg\nSaving to: dataset/masks/0313-2/33680\nReading from: dataset/clips/0313-2/32700/20.jpg\nSaving to: dataset/masks/0313-2/32700\nReading from: dataset/clips/0313-2/20040/20.jpg\nSaving to: dataset/masks/0313-2/20040\nReading from: dataset/clips/0313-2/36080/20.jpg\nSaving to: dataset/masks/0313-2/36080\nReading from: dataset/clips/0313-2/20640/20.jpg\nSaving to: dataset/masks/0313-2/20640\nReading from: dataset/clips/0313-2/1670/20.jpg\nSaving to: dataset/masks/0313-2/1670\nReading from: dataset/clips/0313-2/39520/20.jpg\nSaving to: dataset/masks/0313-2/39520\nReading from: dataset/clips/0313-2/5580/20.jpg\nSaving to: dataset/masks/0313-2/5580\nReading from: dataset/clips/0313-2/55/20.jpg\nSaving to: dataset/masks/0313-2/55\nReading from: dataset/clips/0313-2/695/20.jpg\nSaving to: dataset/masks/0313-2/695\nReading from: dataset/clips/0313-2/41860/20.jpg\nSaving to: dataset/masks/0313-2/41860\nReading from: dataset/clips/0313-2/25140/20.jpg\nSaving to: dataset/masks/0313-2/25140\nReading from: dataset/clips/0313-2/37440/20.jpg\nSaving to: dataset/masks/0313-2/37440\nReading from: dataset/clips/0313-2/490/20.jpg\nSaving to: dataset/masks/0313-2/490\nReading from: dataset/clips/0313-2/70/20.jpg\nSaving to: dataset/masks/0313-2/70\nReading from: dataset/clips/0313-2/25980/20.jpg\nSaving to: dataset/masks/0313-2/25980\nReading from: dataset/clips/0313-2/13980/20.jpg\nSaving to: dataset/masks/0313-2/13980\nReading from: dataset/clips/0313-2/8220/20.jpg\nSaving to: dataset/masks/0313-2/8220\nReading from: dataset/clips/0313-2/40480/20.jpg\nSaving to: dataset/masks/0313-2/40480\nReading from: dataset/clips/0313-2/670/20.jpg\nSaving to: dataset/masks/0313-2/670\nReading from: dataset/clips/0313-2/41880/20.jpg\nSaving to: dataset/masks/0313-2/41880\nReading from: dataset/clips/0313-2/680/20.jpg\nSaving to: dataset/masks/0313-2/680\nReading from: dataset/clips/0313-2/42740/20.jpg\nSaving to: dataset/masks/0313-2/42740\nReading from: dataset/clips/0313-2/710/20.jpg\nSaving to: dataset/masks/0313-2/710\nReading from: dataset/clips/0313-2/38560/20.jpg\nSaving to: dataset/masks/0313-2/38560\nReading from: dataset/clips/0313-2/34080/20.jpg\nSaving to: dataset/masks/0313-2/34080\nReading from: dataset/clips/0313-2/27660/20.jpg\nSaving to: dataset/masks/0313-2/27660\nReading from: dataset/clips/0313-2/41100/20.jpg\nSaving to: dataset/masks/0313-2/41100\nReading from: dataset/clips/0313-2/27180/20.jpg\nSaving to: dataset/masks/0313-2/27180\nReading from: dataset/clips/0313-2/59640/20.jpg\nSaving to: dataset/masks/0313-2/59640\nReading from: dataset/clips/0313-2/150/20.jpg\nSaving to: dataset/masks/0313-2/150\nReading from: dataset/clips/0313-2/24960/20.jpg\nSaving to: dataset/masks/0313-2/24960\nReading from: dataset/clips/0313-2/65/20.jpg\nSaving to: dataset/masks/0313-2/65\nReading from: dataset/clips/0313-2/16740/20.jpg\nSaving to: dataset/masks/0313-2/16740\nReading from: dataset/clips/0313-2/1250/20.jpg\nSaving to: dataset/masks/0313-2/1250\nReading from: dataset/clips/0313-2/1295/20.jpg\nSaving to: dataset/masks/0313-2/1295\nReading from: dataset/clips/0313-2/17340/20.jpg\nSaving to: dataset/masks/0313-2/17340\nReading from: dataset/clips/0313-2/10080/20.jpg\nSaving to: dataset/masks/0313-2/10080\nReading from: dataset/clips/0313-2/36860/20.jpg\nSaving to: dataset/masks/0313-2/36860\nReading from: dataset/clips/0313-2/39900/20.jpg\nSaving to: dataset/masks/0313-2/39900\nReading from: dataset/clips/0313-2/6060/20.jpg\nSaving to: dataset/masks/0313-2/6060\nReading from: dataset/clips/0313-2/24600/20.jpg\nSaving to: dataset/masks/0313-2/24600\nReading from: dataset/clips/0313-2/61200/20.jpg\nSaving to: dataset/masks/0313-2/61200\nReading from: dataset/clips/0313-2/50/20.jpg\nSaving to: dataset/masks/0313-2/50\nReading from: dataset/clips/0313-2/12720/20.jpg\nSaving to: dataset/masks/0313-2/12720\nReading from: dataset/clips/0313-2/36840/20.jpg\nSaving to: dataset/masks/0313-2/36840\nReading from: dataset/clips/0313-2/485/20.jpg\nSaving to: dataset/masks/0313-2/485\nReading from: dataset/clips/0313-2/575/20.jpg\nSaving to: dataset/masks/0313-2/575\nReading from: dataset/clips/0313-2/34560/20.jpg\nSaving to: dataset/masks/0313-2/34560\nReading from: dataset/clips/0313-2/13920/20.jpg\nSaving to: dataset/masks/0313-2/13920\nReading from: dataset/clips/0313-2/31560/20.jpg\nSaving to: dataset/masks/0313-2/31560\nReading from: dataset/clips/0313-2/440/20.jpg\nSaving to: dataset/masks/0313-2/440\nReading from: dataset/clips/0313-2/22080/20.jpg\nSaving to: dataset/masks/0313-2/22080\nReading from: dataset/clips/0313-2/63780/20.jpg\nSaving to: dataset/masks/0313-2/63780\nReading from: dataset/clips/0313-2/1750/20.jpg\nSaving to: dataset/masks/0313-2/1750\nReading from: dataset/clips/0313-2/5940/20.jpg\nSaving to: dataset/masks/0313-2/5940\nReading from: dataset/clips/0313-2/60300/20.jpg\nSaving to: dataset/masks/0313-2/60300\nReading from: dataset/clips/0313-2/32800/20.jpg\nSaving to: dataset/masks/0313-2/32800\nReading from: dataset/clips/0313-2/41480/20.jpg\nSaving to: dataset/masks/0313-2/41480\nReading from: dataset/clips/0313-2/6120/20.jpg\nSaving to: dataset/masks/0313-2/6120\nReading from: dataset/clips/0313-2/15060/20.jpg\nSaving to: dataset/masks/0313-2/15060\nReading from: dataset/clips/0313-2/785/20.jpg\nSaving to: dataset/masks/0313-2/785\nReading from: dataset/clips/0313-2/15780/20.jpg\nSaving to: dataset/masks/0313-2/15780\nReading from: dataset/clips/0313-2/21180/20.jpg\nSaving to: dataset/masks/0313-2/21180\nReading from: dataset/clips/0313-2/17580/20.jpg\nSaving to: dataset/masks/0313-2/17580\nReading from: dataset/clips/0313-2/33580/20.jpg\nSaving to: dataset/masks/0313-2/33580\nReading from: dataset/clips/0313-2/1255/20.jpg\nSaving to: dataset/masks/0313-2/1255\nReading from: dataset/clips/0313-2/41080/20.jpg\nSaving to: dataset/masks/0313-2/41080\nReading from: dataset/clips/0313-2/33540/20.jpg\nSaving to: dataset/masks/0313-2/33540\nReading from: dataset/clips/0313-2/63120/20.jpg\nSaving to: dataset/masks/0313-2/63120\nReading from: dataset/clips/0313-2/715/20.jpg\nSaving to: dataset/masks/0313-2/715\nReading from: dataset/clips/0313-2/27240/20.jpg\nSaving to: dataset/masks/0313-2/27240\nReading from: dataset/clips/0313-2/2820/20.jpg\nSaving to: dataset/masks/0313-2/2820\nReading from: dataset/clips/0313-2/15420/20.jpg\nSaving to: dataset/masks/0313-2/15420\nReading from: dataset/clips/0313-2/425/20.jpg\nSaving to: dataset/masks/0313-2/425\nReading from: dataset/clips/0313-2/62460/20.jpg\nSaving to: dataset/masks/0313-2/62460\nReading from: dataset/clips/0313-2/59160/20.jpg\nSaving to: dataset/masks/0313-2/59160\nReading from: dataset/clips/0313-2/11040/20.jpg\nSaving to: dataset/masks/0313-2/11040\nReading from: dataset/clips/0313-2/290/20.jpg\nSaving to: dataset/masks/0313-2/290\nReading from: dataset/clips/0313-2/37340/20.jpg\nSaving to: dataset/masks/0313-2/37340\nReading from: dataset/clips/0313-2/39700/20.jpg\nSaving to: dataset/masks/0313-2/39700\nReading from: dataset/clips/0313-2/41160/20.jpg\nSaving to: dataset/masks/0313-2/41160\nReading from: dataset/clips/0313-2/41640/20.jpg\nSaving to: dataset/masks/0313-2/41640\nReading from: dataset/clips/0313-2/41760/20.jpg\nSaving to: dataset/masks/0313-2/41760\nReading from: dataset/clips/0313-2/39720/20.jpg\nSaving to: dataset/masks/0313-2/39720\nReading from: dataset/clips/0313-2/30420/20.jpg\nSaving to: dataset/masks/0313-2/30420\nReading from: dataset/clips/0313-2/32940/20.jpg\nSaving to: dataset/masks/0313-2/32940\nReading from: dataset/clips/0313-2/25680/20.jpg\nSaving to: dataset/masks/0313-2/25680\nReading from: dataset/clips/0313-2/18420/20.jpg\nSaving to: dataset/masks/0313-2/18420\nReading from: dataset/clips/0313-2/510/20.jpg\nSaving to: dataset/masks/0313-2/510\nReading from: dataset/clips/0313-2/20880/20.jpg\nSaving to: dataset/masks/0313-2/20880\nReading from: dataset/clips/0313-2/23160/20.jpg\nSaving to: dataset/masks/0313-2/23160\nReading from: dataset/clips/0313-2/1205/20.jpg\nSaving to: dataset/masks/0313-2/1205\nReading from: dataset/clips/0313-2/185/20.jpg\nSaving to: dataset/masks/0313-2/185\nReading from: dataset/clips/0313-2/16920/20.jpg\nSaving to: dataset/masks/0313-2/16920\nReading from: dataset/clips/0313-2/35580/20.jpg\nSaving to: dataset/masks/0313-2/35580\nReading from: dataset/clips/0313-2/1690/20.jpg\nSaving to: dataset/masks/0313-2/1690\nReading from: dataset/clips/0313-2/9300/20.jpg\nSaving to: dataset/masks/0313-2/9300\nReading from: dataset/clips/0313-2/36120/20.jpg\nSaving to: dataset/masks/0313-2/36120\nReading from: dataset/clips/0313-2/3120/20.jpg\nSaving to: dataset/masks/0313-2/3120\nReading from: dataset/clips/0313-2/1110/20.jpg\nSaving to: dataset/masks/0313-2/1110\nReading from: dataset/clips/0313-2/835/20.jpg\nSaving to: dataset/masks/0313-2/835\nReading from: dataset/clips/0313-2/64680/20.jpg\nSaving to: dataset/masks/0313-2/64680\nReading from: dataset/clips/0313-2/4680/20.jpg\nSaving to: dataset/masks/0313-2/4680\nReading from: dataset/clips/0313-2/18120/20.jpg\nSaving to: dataset/masks/0313-2/18120\nReading from: dataset/clips/0313-2/12000/20.jpg\nSaving to: dataset/masks/0313-2/12000\nReading from: dataset/clips/0313-2/40900/20.jpg\nSaving to: dataset/masks/0313-2/40900\nReading from: dataset/clips/0313-2/62280/20.jpg\nSaving to: dataset/masks/0313-2/62280\nReading from: dataset/clips/0313-2/1495/20.jpg\nSaving to: dataset/masks/0313-2/1495\nReading from: dataset/clips/0313-2/36420/20.jpg\nSaving to: dataset/masks/0313-2/36420\nReading from: dataset/clips/0313-2/43060/20.jpg\nSaving to: dataset/masks/0313-2/43060\nReading from: dataset/clips/0313-2/30060/20.jpg\nSaving to: dataset/masks/0313-2/30060\nReading from: dataset/clips/0313-2/35660/20.jpg\nSaving to: dataset/masks/0313-2/35660\nReading from: dataset/clips/0313-2/21840/20.jpg\nSaving to: dataset/masks/0313-2/21840\nReading from: dataset/clips/0313-2/7860/20.jpg\nSaving to: dataset/masks/0313-2/7860\nReading from: dataset/clips/0313-2/28860/20.jpg\nSaving to: dataset/masks/0313-2/28860\nReading from: dataset/clips/0313-2/35180/20.jpg\nSaving to: dataset/masks/0313-2/35180\nReading from: dataset/clips/0313-2/38780/20.jpg\nSaving to: dataset/masks/0313-2/38780\nReading from: dataset/clips/0313-2/39460/20.jpg\nSaving to: dataset/masks/0313-2/39460\nReading from: dataset/clips/0313-2/42760/20.jpg\nSaving to: dataset/masks/0313-2/42760\nReading from: dataset/clips/0313-2/495/20.jpg\nSaving to: dataset/masks/0313-2/495\nReading from: dataset/clips/0313-2/40980/20.jpg\nSaving to: dataset/masks/0313-2/40980\nReading from: dataset/clips/0313-2/810/20.jpg\nSaving to: dataset/masks/0313-2/810\nReading from: dataset/clips/0313-2/3480/20.jpg\nSaving to: dataset/masks/0313-2/3480\nReading from: dataset/clips/0313-2/17040/20.jpg\nSaving to: dataset/masks/0313-2/17040\nReading from: dataset/clips/0313-2/21300/20.jpg\nSaving to: dataset/masks/0313-2/21300\nReading from: dataset/clips/0313-2/34380/20.jpg\nSaving to: dataset/masks/0313-2/34380\nReading from: dataset/clips/0313-2/59340/20.jpg\nSaving to: dataset/masks/0313-2/59340\nReading from: dataset/clips/0313-2/40120/20.jpg\nSaving to: dataset/masks/0313-2/40120\nReading from: dataset/clips/0313-2/36300/20.jpg\nSaving to: dataset/masks/0313-2/36300\nReading from: dataset/clips/0313-2/20400/20.jpg\nSaving to: dataset/masks/0313-2/20400\nReading from: dataset/clips/0313-2/16260/20.jpg\nSaving to: dataset/masks/0313-2/16260\nReading from: dataset/clips/0313-2/42640/20.jpg\nSaving to: dataset/masks/0313-2/42640\nReading from: dataset/clips/0313-2/42560/20.jpg\nSaving to: dataset/masks/0313-2/42560\nReading from: dataset/clips/0313-2/655/20.jpg\nSaving to: dataset/masks/0313-2/655\nReading from: dataset/clips/0313-2/16980/20.jpg\nSaving to: dataset/masks/0313-2/16980\nReading from: dataset/clips/0313-2/61500/20.jpg\nSaving to: dataset/masks/0313-2/61500\nReading from: dataset/clips/0313-2/39360/20.jpg\nSaving to: dataset/masks/0313-2/39360\nReading from: dataset/clips/0313-2/3180/20.jpg\nSaving to: dataset/masks/0313-2/3180\nReading from: dataset/clips/0313-2/1550/20.jpg\nSaving to: dataset/masks/0313-2/1550\nReading from: dataset/clips/0313-2/3600/20.jpg\nSaving to: dataset/masks/0313-2/3600\nReading from: dataset/clips/0313-2/3240/20.jpg\nSaving to: dataset/masks/0313-2/3240\nReading from: dataset/clips/0313-2/175/20.jpg\nSaving to: dataset/masks/0313-2/175\nReading from: dataset/clips/0313-2/38520/20.jpg\nSaving to: dataset/masks/0313-2/38520\nReading from: dataset/clips/0313-2/640/20.jpg\nSaving to: dataset/masks/0313-2/640\nReading from: dataset/clips/0313-2/36620/20.jpg\nSaving to: dataset/masks/0313-2/36620\nReading from: dataset/clips/0313-2/24480/20.jpg\nSaving to: dataset/masks/0313-2/24480\nReading from: dataset/clips/0313-2/24360/20.jpg\nSaving to: dataset/masks/0313-2/24360\nReading from: dataset/clips/0313-2/32720/20.jpg\nSaving to: dataset/masks/0313-2/32720\nReading from: dataset/clips/0313-2/31980/20.jpg\nSaving to: dataset/masks/0313-2/31980\nReading from: dataset/clips/0313-2/10260/20.jpg\nSaving to: dataset/masks/0313-2/10260\nReading from: dataset/clips/0313-2/32600/20.jpg\nSaving to: dataset/masks/0313-2/32600\nReading from: dataset/clips/0313-2/36600/20.jpg\nSaving to: dataset/masks/0313-2/36600\nReading from: dataset/clips/0313-2/17160/20.jpg\nSaving to: dataset/masks/0313-2/17160\nReading from: dataset/clips/0313-2/24720/20.jpg\nSaving to: dataset/masks/0313-2/24720\nReading from: dataset/clips/0313-2/32500/20.jpg\nSaving to: dataset/masks/0313-2/32500\nReading from: dataset/clips/0313-2/350/20.jpg\nSaving to: dataset/masks/0313-2/350\nReading from: dataset/clips/0313-2/1430/20.jpg\nSaving to: dataset/masks/0313-2/1430\nReading from: dataset/clips/0313-2/28980/20.jpg\nSaving to: dataset/masks/0313-2/28980\nReading from: dataset/clips/0313-2/39640/20.jpg\nSaving to: dataset/masks/0313-2/39640\nReading from: dataset/clips/0313-2/39040/20.jpg\nSaving to: dataset/masks/0313-2/39040\nReading from: dataset/clips/0313-2/62340/20.jpg\nSaving to: dataset/masks/0313-2/62340\nReading from: dataset/clips/0313-2/12540/20.jpg\nSaving to: dataset/masks/0313-2/12540\nReading from: dataset/clips/0313-2/36520/20.jpg\nSaving to: dataset/masks/0313-2/36520\nReading from: dataset/clips/0313-2/39880/20.jpg\nSaving to: dataset/masks/0313-2/39880\nReading from: dataset/clips/0313-2/19320/20.jpg\nSaving to: dataset/masks/0313-2/19320\nReading from: dataset/clips/0313-2/34340/20.jpg\nSaving to: dataset/masks/0313-2/34340\nReading from: dataset/clips/0313-2/25440/20.jpg\nSaving to: dataset/masks/0313-2/25440\nReading from: dataset/clips/0313-2/23400/20.jpg\nSaving to: dataset/masks/0313-2/23400\nReading from: dataset/clips/0313-2/10140/20.jpg\nSaving to: dataset/masks/0313-2/10140\nReading from: dataset/clips/0313-2/1655/20.jpg\nSaving to: dataset/masks/0313-2/1655\nReading from: dataset/clips/0313-2/25800/20.jpg\nSaving to: dataset/masks/0313-2/25800\nReading from: dataset/clips/0313-2/25860/20.jpg\nSaving to: dataset/masks/0313-2/25860\nReading from: dataset/clips/0313-2/37360/20.jpg\nSaving to: dataset/masks/0313-2/37360\nReading from: dataset/clips/0313-2/33120/20.jpg\nSaving to: dataset/masks/0313-2/33120\nReading from: dataset/clips/0313-2/15300/20.jpg\nSaving to: dataset/masks/0313-2/15300\nReading from: dataset/clips/0313-2/10920/20.jpg\nSaving to: dataset/masks/0313-2/10920\nReading from: dataset/clips/0313-2/2940/20.jpg\nSaving to: dataset/masks/0313-2/2940\nReading from: dataset/clips/0313-2/1305/20.jpg\nSaving to: dataset/masks/0313-2/1305\nReading from: dataset/clips/0313-2/20/20.jpg\nSaving to: dataset/masks/0313-2/20\nReading from: dataset/clips/0313-2/7680/20.jpg\nSaving to: dataset/masks/0313-2/7680\nReading from: dataset/clips/0313-2/18300/20.jpg\nSaving to: dataset/masks/0313-2/18300\nReading from: dataset/clips/0313-2/235/20.jpg\nSaving to: dataset/masks/0313-2/235\nReading from: dataset/clips/0313-2/45/20.jpg\nSaving to: dataset/masks/0313-2/45\nReading from: dataset/clips/0313-2/37240/20.jpg\nSaving to: dataset/masks/0313-2/37240\nReading from: dataset/clips/0313-2/37160/20.jpg\nSaving to: dataset/masks/0313-2/37160\nReading from: dataset/clips/0313-2/59400/20.jpg\nSaving to: dataset/masks/0313-2/59400\nReading from: dataset/clips/0313-2/32780/20.jpg\nSaving to: dataset/masks/0313-2/32780\nReading from: dataset/clips/0313-2/40220/20.jpg\nSaving to: dataset/masks/0313-2/40220\nReading from: dataset/clips/0313-2/4500/20.jpg\nSaving to: dataset/masks/0313-2/4500\nReading from: dataset/clips/0313-2/35060/20.jpg\nSaving to: dataset/masks/0313-2/35060\nReading from: dataset/clips/0313-2/1715/20.jpg\nSaving to: dataset/masks/0313-2/1715\nReading from: dataset/clips/0313-2/4080/20.jpg\nSaving to: dataset/masks/0313-2/4080\nReading from: dataset/clips/0313-2/63600/20.jpg\nSaving to: dataset/masks/0313-2/63600\nReading from: dataset/clips/0313-2/32980/20.jpg\nSaving to: dataset/masks/0313-2/32980\nReading from: dataset/clips/0313-2/38880/20.jpg\nSaving to: dataset/masks/0313-2/38880\nReading from: dataset/clips/0313-2/18540/20.jpg\nSaving to: dataset/masks/0313-2/18540\nReading from: dataset/clips/0313-2/34920/20.jpg\nSaving to: dataset/masks/0313-2/34920\nReading from: dataset/clips/0313-2/6780/20.jpg\nSaving to: dataset/masks/0313-2/6780\nReading from: dataset/clips/0313-2/1270/20.jpg\nSaving to: dataset/masks/0313-2/1270\nReading from: dataset/clips/0313-2/12960/20.jpg\nSaving to: dataset/masks/0313-2/12960\nReading from: dataset/clips/0313-2/42880/20.jpg\nSaving to: dataset/masks/0313-2/42880\nReading from: dataset/clips/0313-2/415/20.jpg\nSaving to: dataset/masks/0313-2/415\nReading from: dataset/clips/0313-2/36880/20.jpg\nSaving to: dataset/masks/0313-2/36880\nReading from: dataset/clips/0313-2/840/20.jpg\nSaving to: dataset/masks/0313-2/840\nReading from: dataset/clips/0313-2/28500/20.jpg\nSaving to: dataset/masks/0313-2/28500\nReading from: dataset/clips/0313-2/28620/20.jpg\nSaving to: dataset/masks/0313-2/28620\nReading from: dataset/clips/0313-2/140/20.jpg\nSaving to: dataset/masks/0313-2/140\nReading from: dataset/clips/0313-2/18240/20.jpg\nSaving to: dataset/masks/0313-2/18240\nReading from: dataset/clips/0313-2/22320/20.jpg\nSaving to: dataset/masks/0313-2/22320\nReading from: dataset/clips/0313-2/41820/20.jpg\nSaving to: dataset/masks/0313-2/41820\nReading from: dataset/clips/0313-2/10740/20.jpg\nSaving to: dataset/masks/0313-2/10740\nReading from: dataset/clips/0313-2/955/20.jpg\nSaving to: dataset/masks/0313-2/955\nReading from: dataset/clips/0313-2/40780/20.jpg\nSaving to: dataset/masks/0313-2/40780\nReading from: dataset/clips/0313-2/59820/20.jpg\nSaving to: dataset/masks/0313-2/59820\nReading from: dataset/clips/0313-2/39140/20.jpg\nSaving to: dataset/masks/0313-2/39140\nReading from: dataset/clips/0313-2/1790/20.jpg\nSaving to: dataset/masks/0313-2/1790\nReading from: dataset/clips/0313-2/39920/20.jpg\nSaving to: dataset/masks/0313-2/39920\nReading from: dataset/clips/0313-2/64560/20.jpg\nSaving to: dataset/masks/0313-2/64560\nReading from: dataset/clips/0313-2/1175/20.jpg\nSaving to: dataset/masks/0313-2/1175\nReading from: dataset/clips/0313-2/36340/20.jpg\nSaving to: dataset/masks/0313-2/36340\nReading from: dataset/clips/0313-2/1120/20.jpg\nSaving to: dataset/masks/0313-2/1120\nReading from: dataset/clips/0313-2/35260/20.jpg\nSaving to: dataset/masks/0313-2/35260\nReading from: dataset/clips/0313-2/720/20.jpg\nSaving to: dataset/masks/0313-2/720\nReading from: dataset/clips/0313-2/38320/20.jpg\nSaving to: dataset/masks/0313-2/38320\nReading from: dataset/clips/0313-2/10500/20.jpg\nSaving to: dataset/masks/0313-2/10500\nReading from: dataset/clips/0313-2/505/20.jpg\nSaving to: dataset/masks/0313-2/505\nReading from: dataset/clips/0313-2/29640/20.jpg\nSaving to: dataset/masks/0313-2/29640\nReading from: dataset/clips/0313-2/40280/20.jpg\nSaving to: dataset/masks/0313-2/40280\nReading from: dataset/clips/0313-2/5280/20.jpg\nSaving to: dataset/masks/0313-2/5280\nReading from: dataset/clips/0313-2/39000/20.jpg\nSaving to: dataset/masks/0313-2/39000\nReading from: dataset/clips/0313-2/1145/20.jpg\nSaving to: dataset/masks/0313-2/1145\nReading from: dataset/clips/0313-2/11940/20.jpg\nSaving to: dataset/masks/0313-2/11940\nReading from: dataset/clips/0313-2/34480/20.jpg\nSaving to: dataset/masks/0313-2/34480\nReading from: dataset/clips/0313-2/39960/20.jpg\nSaving to: dataset/masks/0313-2/39960\nReading from: dataset/clips/0313-2/33600/20.jpg\nSaving to: dataset/masks/0313-2/33600\nReading from: dataset/clips/0313-2/1170/20.jpg\nSaving to: dataset/masks/0313-2/1170\nReading from: dataset/clips/0313-2/1545/20.jpg\nSaving to: dataset/masks/0313-2/1545\nReading from: dataset/clips/0313-2/15900/20.jpg\nSaving to: dataset/masks/0313-2/15900\nReading from: dataset/clips/0313-2/23340/20.jpg\nSaving to: dataset/masks/0313-2/23340\nReading from: dataset/clips/0313-2/4020/20.jpg\nSaving to: dataset/masks/0313-2/4020\nReading from: dataset/clips/0313-2/40820/20.jpg\nSaving to: dataset/masks/0313-2/40820\nReading from: dataset/clips/0313-2/36820/20.jpg\nSaving to: dataset/masks/0313-2/36820\nReading from: dataset/clips/0313-2/39500/20.jpg\nSaving to: dataset/masks/0313-2/39500\nReading from: dataset/clips/0313-2/5820/20.jpg\nSaving to: dataset/masks/0313-2/5820\nReading from: dataset/clips/0313-2/25320/20.jpg\nSaving to: dataset/masks/0313-2/25320\nReading from: dataset/clips/0313-2/41600/20.jpg\nSaving to: dataset/masks/0313-2/41600\nReading from: dataset/clips/0313-2/42540/20.jpg\nSaving to: dataset/masks/0313-2/42540\nReading from: dataset/clips/0313-2/1615/20.jpg\nSaving to: dataset/masks/0313-2/1615\nReading from: dataset/clips/0313-2/36400/20.jpg\nSaving to: dataset/masks/0313-2/36400\nReading from: dataset/clips/0313-2/35820/20.jpg\nSaving to: dataset/masks/0313-2/35820\nReading from: dataset/clips/0313-2/35740/20.jpg\nSaving to: dataset/masks/0313-2/35740\nReading from: dataset/clips/0313-2/1770/20.jpg\nSaving to: dataset/masks/0313-2/1770\nReading from: dataset/clips/0313-2/20760/20.jpg\nSaving to: dataset/masks/0313-2/20760\nReading from: dataset/clips/0313-2/36960/20.jpg\nSaving to: dataset/masks/0313-2/36960\nReading from: dataset/clips/0313-2/35380/20.jpg\nSaving to: dataset/masks/0313-2/35380\nReading from: dataset/clips/0313-2/1860/20.jpg\nSaving to: dataset/masks/0313-2/1860\nReading from: dataset/clips/0313-2/30720/20.jpg\nSaving to: dataset/masks/0313-2/30720\nReading from: dataset/clips/0313-2/37800/20.jpg\nSaving to: dataset/masks/0313-2/37800\nReading from: dataset/clips/0313-2/34220/20.jpg\nSaving to: dataset/masks/0313-2/34220\nReading from: dataset/clips/0313-2/39860/20.jpg\nSaving to: dataset/masks/0313-2/39860\nReading from: dataset/clips/0313-2/34740/20.jpg\nSaving to: dataset/masks/0313-2/34740\nReading from: dataset/clips/0313-2/38600/20.jpg\nSaving to: dataset/masks/0313-2/38600\nReading from: dataset/clips/0313-2/38660/20.jpg\nSaving to: dataset/masks/0313-2/38660\nReading from: dataset/clips/0313-2/13140/20.jpg\nSaving to: dataset/masks/0313-2/13140\nReading from: dataset/clips/0313-2/1535/20.jpg\nSaving to: dataset/masks/0313-2/1535\nReading from: dataset/clips/0313-2/1685/20.jpg\nSaving to: dataset/masks/0313-2/1685\nReading from: dataset/clips/0313-2/35620/20.jpg\nSaving to: dataset/masks/0313-2/35620\nReading from: dataset/clips/0313-2/16560/20.jpg\nSaving to: dataset/masks/0313-2/16560\nReading from: dataset/clips/0313-2/59460/20.jpg\nSaving to: dataset/masks/0313-2/59460\nReading from: dataset/clips/0313-2/20820/20.jpg\nSaving to: dataset/masks/0313-2/20820\nReading from: dataset/clips/0313-2/35680/20.jpg\nSaving to: dataset/masks/0313-2/35680\nReading from: dataset/clips/0313-2/32540/20.jpg\nSaving to: dataset/masks/0313-2/32540\nReading from: dataset/clips/0313-2/32560/20.jpg\nSaving to: dataset/masks/0313-2/32560\nReading from: dataset/clips/0313-2/26640/20.jpg\nSaving to: dataset/masks/0313-2/26640\nReading from: dataset/clips/0313-2/33360/20.jpg\nSaving to: dataset/masks/0313-2/33360\nReading from: dataset/clips/0313-2/33420/20.jpg\nSaving to: dataset/masks/0313-2/33420\nReading from: dataset/clips/0313-2/32420/20.jpg\nSaving to: dataset/masks/0313-2/32420\nReading from: dataset/clips/0313-2/32160/20.jpg\nSaving to: dataset/masks/0313-2/32160\nReading from: dataset/clips/0313-2/1435/20.jpg\nSaving to: dataset/masks/0313-2/1435\nReading from: dataset/clips/0313-2/36140/20.jpg\nSaving to: dataset/masks/0313-2/36140\nReading from: dataset/clips/0313-2/335/20.jpg\nSaving to: dataset/masks/0313-2/335\nReading from: dataset/clips/0313-2/39740/20.jpg\nSaving to: dataset/masks/0313-2/39740\nReading from: dataset/clips/0313-2/39980/20.jpg\nSaving to: dataset/masks/0313-2/39980\nReading from: dataset/clips/0313-2/32960/20.jpg\nSaving to: dataset/masks/0313-2/32960\nReading from: dataset/clips/0313-2/63300/20.jpg\nSaving to: dataset/masks/0313-2/63300\nReading from: dataset/clips/0313-2/35100/20.jpg\nSaving to: dataset/masks/0313-2/35100\nReading from: dataset/clips/0313-2/29460/20.jpg\nSaving to: dataset/masks/0313-2/29460\nReading from: dataset/clips/0313-2/31500/20.jpg\nSaving to: dataset/masks/0313-2/31500\nReading from: dataset/clips/0313-2/630/20.jpg\nSaving to: dataset/masks/0313-2/630\nReading from: dataset/clips/0313-2/160/20.jpg\nSaving to: dataset/masks/0313-2/160\nReading from: dataset/clips/0313-2/36160/20.jpg\nSaving to: dataset/masks/0313-2/36160\nReading from: dataset/clips/0313-2/9660/20.jpg\nSaving to: dataset/masks/0313-2/9660\nReading from: dataset/clips/0313-2/42520/20.jpg\nSaving to: dataset/masks/0313-2/42520\nReading from: dataset/clips/0313-2/63900/20.jpg\nSaving to: dataset/masks/0313-2/63900\nReading from: dataset/clips/0313-2/360/20.jpg\nSaving to: dataset/masks/0313-2/360\nReading from: dataset/clips/0313-2/39440/20.jpg\nSaving to: dataset/masks/0313-2/39440\nReading from: dataset/clips/0313-2/61440/20.jpg\nSaving to: dataset/masks/0313-2/61440\nReading from: dataset/clips/0313-2/330/20.jpg\nSaving to: dataset/masks/0313-2/330\nReading from: dataset/clips/0313-2/930/20.jpg\nSaving to: dataset/masks/0313-2/930\nReading from: dataset/clips/0313-2/34660/20.jpg\nSaving to: dataset/masks/0313-2/34660\nReading from: dataset/clips/0313-2/63420/20.jpg\nSaving to: dataset/masks/0313-2/63420\nReading from: dataset/clips/0313-2/820/20.jpg\nSaving to: dataset/masks/0313-2/820\nReading from: dataset/clips/0313-2/60/20.jpg\nSaving to: dataset/masks/0313-2/60\nReading from: dataset/clips/0313-2/23040/20.jpg\nSaving to: dataset/masks/0313-2/23040\nReading from: dataset/clips/0313-2/135/20.jpg\nSaving to: dataset/masks/0313-2/135\nReading from: dataset/clips/0313-2/1920/20.jpg\nSaving to: dataset/masks/0313-2/1920\nReading from: dataset/clips/0313-2/705/20.jpg\nSaving to: dataset/masks/0313-2/705\nReading from: dataset/clips/0313-2/190/20.jpg\nSaving to: dataset/masks/0313-2/190\nReading from: dataset/clips/0313-2/41200/20.jpg\nSaving to: dataset/masks/0313-2/41200\nReading from: dataset/clips/0313-2/19440/20.jpg\nSaving to: dataset/masks/0313-2/19440\nReading from: dataset/clips/0313-2/29160/20.jpg\nSaving to: dataset/masks/0313-2/29160\nReading from: dataset/clips/0313-2/625/20.jpg\nSaving to: dataset/masks/0313-2/625\nReading from: dataset/clips/0313-2/34300/20.jpg\nSaving to: dataset/masks/0313-2/34300\nReading from: dataset/clips/0313-2/775/20.jpg\nSaving to: dataset/masks/0313-2/775\nReading from: dataset/clips/0313-2/24420/20.jpg\nSaving to: dataset/masks/0313-2/24420\nReading from: dataset/clips/0313-2/13800/20.jpg\nSaving to: dataset/masks/0313-2/13800\nReading from: dataset/clips/0313-2/60420/20.jpg\nSaving to: dataset/masks/0313-2/60420\nReading from: dataset/clips/0313-2/35040/20.jpg\nSaving to: dataset/masks/0313-2/35040\nReading from: dataset/clips/0313-2/14220/20.jpg\nSaving to: dataset/masks/0313-2/14220\nReading from: dataset/clips/0313-2/22980/20.jpg\nSaving to: dataset/masks/0313-2/22980\nReading from: dataset/clips/0313-2/1755/20.jpg\nSaving to: dataset/masks/0313-2/1755\nReading from: dataset/clips/0313-2/4620/20.jpg\nSaving to: dataset/masks/0313-2/4620\nReading from: dataset/clips/0313-2/41980/20.jpg\nSaving to: dataset/masks/0313-2/41980\nReading from: dataset/clips/0313-2/40160/20.jpg\nSaving to: dataset/masks/0313-2/40160\nReading from: dataset/clips/0313-2/13020/20.jpg\nSaving to: dataset/masks/0313-2/13020\nReading from: dataset/clips/0313-2/41620/20.jpg\nSaving to: dataset/masks/0313-2/41620\nReading from: dataset/clips/0313-2/28560/20.jpg\nSaving to: dataset/masks/0313-2/28560\nReading from: dataset/clips/0313-2/64500/20.jpg\nSaving to: dataset/masks/0313-2/64500\nReading from: dataset/clips/0313-2/270/20.jpg\nSaving to: dataset/masks/0313-2/270\nReading from: dataset/clips/0313-2/64620/20.jpg\nSaving to: dataset/masks/0313-2/64620\nReading from: dataset/clips/0313-2/685/20.jpg\nSaving to: dataset/masks/0313-2/685\nReading from: dataset/clips/0313-2/8040/20.jpg\nSaving to: dataset/masks/0313-2/8040\nReading from: dataset/clips/0313-2/8700/20.jpg\nSaving to: dataset/masks/0313-2/8700\nReading from: dataset/clips/0313-2/35840/20.jpg\nSaving to: dataset/masks/0313-2/35840\nReading from: dataset/clips/0313-2/37120/20.jpg\nSaving to: dataset/masks/0313-2/37120\nReading from: dataset/clips/0313-2/16200/20.jpg\nSaving to: dataset/masks/0313-2/16200\nReading from: dataset/clips/0313-2/12360/20.jpg\nSaving to: dataset/masks/0313-2/12360\nReading from: dataset/clips/0313-2/41700/20.jpg\nSaving to: dataset/masks/0313-2/41700\nReading from: dataset/clips/0313-2/62640/20.jpg\nSaving to: dataset/masks/0313-2/62640\nReading from: dataset/clips/0313-2/1340/20.jpg\nSaving to: dataset/masks/0313-2/1340\nReading from: dataset/clips/0313-2/39180/20.jpg\nSaving to: dataset/masks/0313-2/39180\nReading from: dataset/clips/0313-2/1700/20.jpg\nSaving to: dataset/masks/0313-2/1700\nReading from: dataset/clips/0313-2/115/20.jpg\nSaving to: dataset/masks/0313-2/115\nReading from: dataset/clips/0313-2/29100/20.jpg\nSaving to: dataset/masks/0313-2/29100\nReading from: dataset/clips/0313-2/23100/20.jpg\nSaving to: dataset/masks/0313-2/23100\nReading from: dataset/clips/0313-2/21960/20.jpg\nSaving to: dataset/masks/0313-2/21960\nReading from: dataset/clips/0313-2/19920/20.jpg\nSaving to: dataset/masks/0313-2/19920\nReading from: dataset/clips/0313-2/60900/20.jpg\nSaving to: dataset/masks/0313-2/60900\nReading from: dataset/clips/0313-2/33740/20.jpg\nSaving to: dataset/masks/0313-2/33740\nReading from: dataset/clips/0313-2/33260/20.jpg\nSaving to: dataset/masks/0313-2/33260\nReading from: dataset/clips/0313-2/42920/20.jpg\nSaving to: dataset/masks/0313-2/42920\nReading from: dataset/clips/0313-2/5760/20.jpg\nSaving to: dataset/masks/0313-2/5760\nReading from: dataset/clips/0313-2/36480/20.jpg\nSaving to: dataset/masks/0313-2/36480\nReading from: dataset/clips/0313-2/20160/20.jpg\nSaving to: dataset/masks/0313-2/20160\nReading from: dataset/clips/0313-2/1570/20.jpg\nSaving to: dataset/masks/0313-2/1570\nReading from: dataset/clips/0313-2/62400/20.jpg\nSaving to: dataset/masks/0313-2/62400\nReading from: dataset/clips/0313-2/760/20.jpg\nSaving to: dataset/masks/0313-2/760\nReading from: dataset/clips/0313-2/250/20.jpg\nSaving to: dataset/masks/0313-2/250\nReading from: dataset/clips/0313-2/60120/20.jpg\nSaving to: dataset/masks/0313-2/60120\nReading from: dataset/clips/0313-2/2100/20.jpg\nSaving to: dataset/masks/0313-2/2100\nReading from: dataset/clips/0313-2/20940/20.jpg\nSaving to: dataset/masks/0313-2/20940\nReading from: dataset/clips/0313-2/17400/20.jpg\nSaving to: dataset/masks/0313-2/17400\nReading from: dataset/clips/0313-2/11280/20.jpg\nSaving to: dataset/masks/0313-2/11280\nReading from: dataset/clips/0313-2/29820/20.jpg\nSaving to: dataset/masks/0313-2/29820\nReading from: dataset/clips/0313-2/41960/20.jpg\nSaving to: dataset/masks/0313-2/41960\nReading from: dataset/clips/0313-2/63660/20.jpg\nSaving to: dataset/masks/0313-2/63660\nReading from: dataset/clips/0313-2/645/20.jpg\nSaving to: dataset/masks/0313-2/645\nReading from: dataset/clips/0313-2/42100/20.jpg\nSaving to: dataset/masks/0313-2/42100\nReading from: dataset/clips/0313-2/1025/20.jpg\nSaving to: dataset/masks/0313-2/1025\nReading from: dataset/clips/0313-2/35880/20.jpg\nSaving to: dataset/masks/0313-2/35880\nReading from: dataset/clips/0313-2/545/20.jpg\nSaving to: dataset/masks/0313-2/545\nReading from: dataset/clips/0313-2/20280/20.jpg\nSaving to: dataset/masks/0313-2/20280\nReading from: dataset/clips/0313-2/15120/20.jpg\nSaving to: dataset/masks/0313-2/15120\nReading from: dataset/clips/0313-2/36940/20.jpg\nSaving to: dataset/masks/0313-2/36940\nReading from: dataset/clips/0313-2/60000/20.jpg\nSaving to: dataset/masks/0313-2/60000\nReading from: dataset/clips/0313-2/635/20.jpg\nSaving to: dataset/masks/0313-2/635\nReading from: dataset/clips/0313-2/15960/20.jpg\nSaving to: dataset/masks/0313-2/15960\nReading from: dataset/clips/0313-2/1485/20.jpg\nSaving to: dataset/masks/0313-2/1485\nReading from: dataset/clips/0313-2/39820/20.jpg\nSaving to: dataset/masks/0313-2/39820\nReading from: dataset/clips/0313-2/31440/20.jpg\nSaving to: dataset/masks/0313-2/31440\nReading from: dataset/clips/0313-2/33040/20.jpg\nSaving to: dataset/masks/0313-2/33040\nReading from: dataset/clips/0313-2/32480/20.jpg\nSaving to: dataset/masks/0313-2/32480\nReading from: dataset/clips/0313-2/39160/20.jpg\nSaving to: dataset/masks/0313-2/39160\nReading from: dataset/clips/0313-2/26040/20.jpg\nSaving to: dataset/masks/0313-2/26040\nReading from: dataset/clips/0313-2/32620/20.jpg\nSaving to: dataset/masks/0313-2/32620\nReading from: dataset/clips/0313-2/1180/20.jpg\nSaving to: dataset/masks/0313-2/1180\nReading from: dataset/clips/0313-2/12840/20.jpg\nSaving to: dataset/masks/0313-2/12840\nReading from: dataset/clips/0313-2/830/20.jpg\nSaving to: dataset/masks/0313-2/830\nReading from: dataset/clips/0313-2/36460/20.jpg\nSaving to: dataset/masks/0313-2/36460\nReading from: dataset/clips/0313-2/6180/20.jpg\nSaving to: dataset/masks/0313-2/6180\nReading from: dataset/clips/0313-2/315/20.jpg\nSaving to: dataset/masks/0313-2/315\nReading from: dataset/clips/0313-2/1420/20.jpg\nSaving to: dataset/masks/0313-2/1420\nReading from: dataset/clips/0313-2/3060/20.jpg\nSaving to: dataset/masks/0313-2/3060\nReading from: dataset/clips/0313-2/36640/20.jpg\nSaving to: dataset/masks/0313-2/36640\nReading from: dataset/clips/0313-2/24180/20.jpg\nSaving to: dataset/masks/0313-2/24180\nReading from: dataset/clips/0313-2/41140/20.jpg\nSaving to: dataset/masks/0313-2/41140\nReading from: dataset/clips/0313-2/3000/20.jpg\nSaving to: dataset/masks/0313-2/3000\nReading from: dataset/clips/0313-2/890/20.jpg\nSaving to: dataset/masks/0313-2/890\nReading from: dataset/clips/0313-2/17640/20.jpg\nSaving to: dataset/masks/0313-2/17640\nReading from: dataset/clips/0313-2/1400/20.jpg\nSaving to: dataset/masks/0313-2/1400\nReading from: dataset/clips/0313-2/62100/20.jpg\nSaving to: dataset/masks/0313-2/62100\nReading from: dataset/clips/0313-2/1980/20.jpg\nSaving to: dataset/masks/0313-2/1980\nReading from: dataset/clips/0313-2/3660/20.jpg\nSaving to: dataset/masks/0313-2/3660\nReading from: dataset/clips/0313-2/2880/20.jpg\nSaving to: dataset/masks/0313-2/2880\nReading from: dataset/clips/0313-2/9900/20.jpg\nSaving to: dataset/masks/0313-2/9900\nReading from: dataset/clips/0313-2/24540/20.jpg\nSaving to: dataset/masks/0313-2/24540\nReading from: dataset/clips/0313-2/27000/20.jpg\nSaving to: dataset/masks/0313-2/27000\nReading from: dataset/clips/0313-2/33440/20.jpg\nSaving to: dataset/masks/0313-2/33440\nReading from: dataset/clips/0313-2/37740/20.jpg\nSaving to: dataset/masks/0313-2/37740\nReading from: dataset/clips/0313-2/8520/20.jpg\nSaving to: dataset/masks/0313-2/8520\nReading from: dataset/clips/0313-2/39020/20.jpg\nSaving to: dataset/masks/0313-2/39020\nReading from: dataset/clips/0313-2/6960/20.jpg\nSaving to: dataset/masks/0313-2/6960\nReading from: dataset/clips/0313-2/39340/20.jpg\nSaving to: dataset/masks/0313-2/39340\nReading from: dataset/clips/0313-2/1610/20.jpg\nSaving to: dataset/masks/0313-2/1610\nReading from: dataset/clips/0313-2/37080/20.jpg\nSaving to: dataset/masks/0313-2/37080\nReading from: dataset/clips/0313-2/37500/20.jpg\nSaving to: dataset/masks/0313-2/37500\nReading from: dataset/clips/0313-2/37760/20.jpg\nSaving to: dataset/masks/0313-2/37760\nReading from: dataset/clips/0313-2/8880/20.jpg\nSaving to: dataset/masks/0313-2/8880\nReading from: dataset/clips/0313-2/30180/20.jpg\nSaving to: dataset/masks/0313-2/30180\nReading from: dataset/clips/0313-2/30660/20.jpg\nSaving to: dataset/masks/0313-2/30660\nReading from: dataset/clips/0313-2/25620/20.jpg\nSaving to: dataset/masks/0313-2/25620\nReading from: dataset/clips/0313-2/34440/20.jpg\nSaving to: dataset/masks/0313-2/34440\nReading from: dataset/clips/0313-2/40000/20.jpg\nSaving to: dataset/masks/0313-2/40000\nReading from: dataset/clips/0313-2/34600/20.jpg\nSaving to: dataset/masks/0313-2/34600\nReading from: dataset/clips/0313-2/9120/20.jpg\nSaving to: dataset/masks/0313-2/9120\nReading from: dataset/clips/0313-2/33620/20.jpg\nSaving to: dataset/masks/0313-2/33620\nReading from: dataset/clips/0313-2/37100/20.jpg\nSaving to: dataset/masks/0313-2/37100\nReading from: dataset/clips/0313-2/4980/20.jpg\nSaving to: dataset/masks/0313-2/4980\nReading from: dataset/clips/0313-2/23280/20.jpg\nSaving to: dataset/masks/0313-2/23280\nReading from: dataset/clips/0313-2/1565/20.jpg\nSaving to: dataset/masks/0313-2/1565\nReading from: dataset/clips/0313-2/7800/20.jpg\nSaving to: dataset/masks/0313-2/7800\nReading from: dataset/clips/0313-2/965/20.jpg\nSaving to: dataset/masks/0313-2/965\nReading from: dataset/clips/0313-2/33840/20.jpg\nSaving to: dataset/masks/0313-2/33840\nReading from: dataset/clips/0313-2/28260/20.jpg\nSaving to: dataset/masks/0313-2/28260\nReading from: dataset/clips/0313-2/455/20.jpg\nSaving to: dataset/masks/0313-2/455\nReading from: dataset/clips/0313-2/42160/20.jpg\nSaving to: dataset/masks/0313-2/42160\nReading from: dataset/clips/0313-2/7020/20.jpg\nSaving to: dataset/masks/0313-2/7020\nReading from: dataset/clips/0313-2/4380/20.jpg\nSaving to: dataset/masks/0313-2/4380\nReading from: dataset/clips/0313-2/1280/20.jpg\nSaving to: dataset/masks/0313-2/1280\nReading from: dataset/clips/0313-2/40500/20.jpg\nSaving to: dataset/masks/0313-2/40500\nReading from: dataset/clips/0313-2/37880/20.jpg\nSaving to: dataset/masks/0313-2/37880\nReading from: dataset/clips/0313-2/515/20.jpg\nSaving to: dataset/masks/0313-2/515\nReading from: dataset/clips/0313-2/33760/20.jpg\nSaving to: dataset/masks/0313-2/33760\nReading from: dataset/clips/0313-2/9060/20.jpg\nSaving to: dataset/masks/0313-2/9060\nReading from: dataset/clips/0313-2/1425/20.jpg\nSaving to: dataset/masks/0313-2/1425\nReading from: dataset/clips/0313-2/20460/20.jpg\nSaving to: dataset/masks/0313-2/20460\nReading from: dataset/clips/0313-2/28740/20.jpg\nSaving to: dataset/masks/0313-2/28740\nReading from: dataset/clips/0313-2/42060/20.jpg\nSaving to: dataset/masks/0313-2/42060\nReading from: dataset/clips/0313-2/12900/20.jpg\nSaving to: dataset/masks/0313-2/12900\nReading from: dataset/clips/0313-2/725/20.jpg\nSaving to: dataset/masks/0313-2/725\nReading from: dataset/clips/0313-2/11580/20.jpg\nSaving to: dataset/masks/0313-2/11580\nReading from: dataset/clips/0313-2/12780/20.jpg\nSaving to: dataset/masks/0313-2/12780\nReading from: dataset/clips/0313-2/42660/20.jpg\nSaving to: dataset/masks/0313-2/42660\nReading from: dataset/clips/0313-2/1385/20.jpg\nSaving to: dataset/masks/0313-2/1385\nReading from: dataset/clips/0313-2/61320/20.jpg\nSaving to: dataset/masks/0313-2/61320\nReading from: dataset/clips/0313-2/38820/20.jpg\nSaving to: dataset/masks/0313-2/38820\nReading from: dataset/clips/0313-2/37980/20.jpg\nSaving to: dataset/masks/0313-2/37980\nReading from: dataset/clips/0313-2/1200/20.jpg\nSaving to: dataset/masks/0313-2/1200\nReading from: dataset/clips/0313-2/320/20.jpg\nSaving to: dataset/masks/0313-2/320\nReading from: dataset/clips/0313-2/33980/20.jpg\nSaving to: dataset/masks/0313-2/33980\nReading from: dataset/clips/0313-2/615/20.jpg\nSaving to: dataset/masks/0313-2/615\nReading from: dataset/clips/0313-2/795/20.jpg\nSaving to: dataset/masks/0313-2/795\nReading from: dataset/clips/0313-2/10200/20.jpg\nSaving to: dataset/masks/0313-2/10200\nReading from: dataset/clips/0313-2/42900/20.jpg\nSaving to: dataset/masks/0313-2/42900\nReading from: dataset/clips/0313-2/925/20.jpg\nSaving to: dataset/masks/0313-2/925\nReading from: dataset/clips/0313-2/60540/20.jpg\nSaving to: dataset/masks/0313-2/60540\nReading from: dataset/clips/0313-2/33000/20.jpg\nSaving to: dataset/masks/0313-2/33000\nReading from: dataset/clips/0313-2/4260/20.jpg\nSaving to: dataset/masks/0313-2/4260\nReading from: dataset/clips/0313-2/40100/20.jpg\nSaving to: dataset/masks/0313-2/40100\nReading from: dataset/clips/0313-2/35400/20.jpg\nSaving to: dataset/masks/0313-2/35400\nReading from: dataset/clips/0313-2/155/20.jpg\nSaving to: dataset/masks/0313-2/155\nReading from: dataset/clips/0313-2/1360/20.jpg\nSaving to: dataset/masks/0313-2/1360\nReading from: dataset/clips/0313-2/63720/20.jpg\nSaving to: dataset/masks/0313-2/63720\nReading from: dataset/clips/0313-2/63240/20.jpg\nSaving to: dataset/masks/0313-2/63240\nReading from: dataset/clips/0313-2/1760/20.jpg\nSaving to: dataset/masks/0313-2/1760\nReading from: dataset/clips/0313-2/38180/20.jpg\nSaving to: dataset/masks/0313-2/38180\nReading from: dataset/clips/0313-2/1740/20.jpg\nSaving to: dataset/masks/0313-2/1740\nReading from: dataset/clips/0313-2/39220/20.jpg\nSaving to: dataset/masks/0313-2/39220\nReading from: dataset/clips/0313-2/38480/20.jpg\nSaving to: dataset/masks/0313-2/38480\nReading from: dataset/clips/0313-2/39200/20.jpg\nSaving to: dataset/masks/0313-2/39200\nReading from: dataset/clips/0313-2/38420/20.jpg\nSaving to: dataset/masks/0313-2/38420\nReading from: dataset/clips/0313-2/1140/20.jpg\nSaving to: dataset/masks/0313-2/1140\nReading from: dataset/clips/0313-2/2580/20.jpg\nSaving to: dataset/masks/0313-2/2580\nReading from: dataset/clips/0313-2/22260/20.jpg\nSaving to: dataset/masks/0313-2/22260\nReading from: dataset/clips/0313-2/18180/20.jpg\nSaving to: dataset/masks/0313-2/18180\nReading from: dataset/clips/0313-2/24120/20.jpg\nSaving to: dataset/masks/0313-2/24120\nReading from: dataset/clips/0313-2/40940/20.jpg\nSaving to: dataset/masks/0313-2/40940\nReading from: dataset/clips/0313-2/37460/20.jpg\nSaving to: dataset/masks/0313-2/37460\nReading from: dataset/clips/0313-2/33660/20.jpg\nSaving to: dataset/masks/0313-2/33660\nReading from: dataset/clips/0313-2/36180/20.jpg\nSaving to: dataset/masks/0313-2/36180\nReading from: dataset/clips/0313-2/37260/20.jpg\nSaving to: dataset/masks/0313-2/37260\nReading from: dataset/clips/0313-2/610/20.jpg\nSaving to: dataset/masks/0313-2/610\nReading from: dataset/clips/0313-2/36100/20.jpg\nSaving to: dataset/masks/0313-2/36100\nReading from: dataset/clips/0313-2/38260/20.jpg\nSaving to: dataset/masks/0313-2/38260\nReading from: dataset/clips/0313-2/39760/20.jpg\nSaving to: dataset/masks/0313-2/39760\nReading from: dataset/clips/0313-2/36060/20.jpg\nSaving to: dataset/masks/0313-2/36060\nReading from: dataset/clips/0313-2/41740/20.jpg\nSaving to: dataset/masks/0313-2/41740\nReading from: dataset/clips/0313-2/9600/20.jpg\nSaving to: dataset/masks/0313-2/9600\nReading from: dataset/clips/0313-2/36380/20.jpg\nSaving to: dataset/masks/0313-2/36380\nReading from: dataset/clips/0313-2/33100/20.jpg\nSaving to: dataset/masks/0313-2/33100\nReading from: dataset/clips/0313-2/35520/20.jpg\nSaving to: dataset/masks/0313-2/35520\nReading from: dataset/clips/0313-2/33700/20.jpg\nSaving to: dataset/masks/0313-2/33700\nReading from: dataset/clips/0313-2/41920/20.jpg\nSaving to: dataset/masks/0313-2/41920\nReading from: dataset/clips/0313-2/36800/20.jpg\nSaving to: dataset/masks/0313-2/36800\nReading from: dataset/clips/0313-2/570/20.jpg\nSaving to: dataset/masks/0313-2/570\nReading from: dataset/clips/0313-2/35500/20.jpg\nSaving to: dataset/masks/0313-2/35500\nReading from: dataset/clips/0313-2/24000/20.jpg\nSaving to: dataset/masks/0313-2/24000\nReading from: dataset/clips/0313-2/225/20.jpg\nSaving to: dataset/masks/0313-2/225\nReading from: dataset/clips/0313-2/6300/20.jpg\nSaving to: dataset/masks/0313-2/6300\nReading from: dataset/clips/0313-2/31800/20.jpg\nSaving to: dataset/masks/0313-2/31800\nReading from: dataset/clips/0313-2/59700/20.jpg\nSaving to: dataset/masks/0313-2/59700\nReading from: dataset/clips/0313-2/13320/20.jpg\nSaving to: dataset/masks/0313-2/13320\nReading from: dataset/clips/0313-2/1795/20.jpg\nSaving to: dataset/masks/0313-2/1795\nReading from: dataset/clips/0313-2/7440/20.jpg\nSaving to: dataset/masks/0313-2/7440\nReading from: dataset/clips/0313-2/1045/20.jpg\nSaving to: dataset/masks/0313-2/1045\nReading from: dataset/clips/0313-2/40520/20.jpg\nSaving to: dataset/masks/0313-2/40520\nReading from: dataset/clips/0313-2/35860/20.jpg\nSaving to: dataset/masks/0313-2/35860\nReading from: dataset/clips/0313-2/9540/20.jpg\nSaving to: dataset/masks/0313-2/9540\nReading from: dataset/clips/0313-2/540/20.jpg\nSaving to: dataset/masks/0313-2/540\nReading from: dataset/clips/0313-2/33240/20.jpg\nSaving to: dataset/masks/0313-2/33240\nReading from: dataset/clips/0313-2/23700/20.jpg\nSaving to: dataset/masks/0313-2/23700\nReading from: dataset/clips/0313-2/34820/20.jpg\nSaving to: dataset/masks/0313-2/34820\nReading from: dataset/clips/0313-2/31260/20.jpg\nSaving to: dataset/masks/0313-2/31260\nReading from: dataset/clips/0313-2/39800/20.jpg\nSaving to: dataset/masks/0313-2/39800\nReading from: dataset/clips/0313-2/700/20.jpg\nSaving to: dataset/masks/0313-2/700\nReading from: dataset/clips/0313-2/11520/20.jpg\nSaving to: dataset/masks/0313-2/11520\nReading from: dataset/clips/0313-2/6000/20.jpg\nSaving to: dataset/masks/0313-2/6000\nReading from: dataset/clips/0313-2/40320/20.jpg\nSaving to: dataset/masks/0313-2/40320\nReading from: dataset/clips/0313-2/35120/20.jpg\nSaving to: dataset/masks/0313-2/35120\nReading from: dataset/clips/0313-2/300/20.jpg\nSaving to: dataset/masks/0313-2/300\nReading from: dataset/clips/0313-2/10980/20.jpg\nSaving to: dataset/masks/0313-2/10980\nReading from: dataset/clips/0313-2/28020/20.jpg\nSaving to: dataset/masks/0313-2/28020\nReading from: dataset/clips/0313-2/28080/20.jpg\nSaving to: dataset/masks/0313-2/28080\nReading from: dataset/clips/0313-2/6660/20.jpg\nSaving to: dataset/masks/0313-2/6660\nReading from: dataset/clips/0313-2/14820/20.jpg\nSaving to: dataset/masks/0313-2/14820\nReading from: dataset/clips/0313-2/6900/20.jpg\nSaving to: dataset/masks/0313-2/6900\nReading from: dataset/clips/0313-2/31920/20.jpg\nSaving to: dataset/masks/0313-2/31920\nReading from: dataset/clips/0313-2/19740/20.jpg\nSaving to: dataset/masks/0313-2/19740\nReading from: dataset/clips/0313-2/805/20.jpg\nSaving to: dataset/masks/0313-2/805\nReading from: dataset/clips/0313-2/28320/20.jpg\nSaving to: dataset/masks/0313-2/28320\nReading from: dataset/clips/0313-2/26700/20.jpg\nSaving to: dataset/masks/0313-2/26700\nReading from: dataset/clips/0313-2/34420/20.jpg\nSaving to: dataset/masks/0313-2/34420\nReading from: dataset/clips/0313-2/27360/20.jpg\nSaving to: dataset/masks/0313-2/27360\nReading from: dataset/clips/0313-2/33920/20.jpg\nSaving to: dataset/masks/0313-2/33920\nReading from: dataset/clips/0313-2/14580/20.jpg\nSaving to: dataset/masks/0313-2/14580\nReading from: dataset/clips/0313-2/605/20.jpg\nSaving to: dataset/masks/0313-2/605\nReading from: dataset/clips/0313-2/1580/20.jpg\nSaving to: dataset/masks/0313-2/1580\nReading from: dataset/clips/0313-2/38080/20.jpg\nSaving to: dataset/masks/0313-2/38080\nReading from: dataset/clips/0313-2/26340/20.jpg\nSaving to: dataset/masks/0313-2/26340\nReading from: dataset/clips/0313-2/2640/20.jpg\nSaving to: dataset/masks/0313-2/2640\nReading from: dataset/clips/0313-2/1445/20.jpg\nSaving to: dataset/masks/0313-2/1445\nReading from: dataset/clips/0313-2/12060/20.jpg\nSaving to: dataset/masks/0313-2/12060\nReading from: dataset/clips/0313-2/41260/20.jpg\nSaving to: dataset/masks/0313-2/41260\nReading from: dataset/clips/0313-2/15180/20.jpg\nSaving to: dataset/masks/0313-2/15180\nReading from: dataset/clips/0313-2/1095/20.jpg\nSaving to: dataset/masks/0313-2/1095\nReading from: dataset/clips/0313-2/23820/20.jpg\nSaving to: dataset/masks/0313-2/23820\nReading from: dataset/clips/0313-2/24240/20.jpg\nSaving to: dataset/masks/0313-2/24240\nReading from: dataset/clips/0313-2/34100/20.jpg\nSaving to: dataset/masks/0313-2/34100\nReading from: dataset/clips/0313-2/980/20.jpg\nSaving to: dataset/masks/0313-2/980\nReading from: dataset/clips/0313-2/33800/20.jpg\nSaving to: dataset/masks/0313-2/33800\nReading from: dataset/clips/0313-2/41840/20.jpg\nSaving to: dataset/masks/0313-2/41840\nReading from: dataset/clips/0313-2/63840/20.jpg\nSaving to: dataset/masks/0313-2/63840\nReading from: dataset/clips/0313-2/64740/20.jpg\nSaving to: dataset/masks/0313-2/64740\nReading from: dataset/clips/0313-2/5400/20.jpg\nSaving to: dataset/masks/0313-2/5400\nReading from: dataset/clips/0313-2/40140/20.jpg\nSaving to: dataset/masks/0313-2/40140\nReading from: dataset/clips/0313-2/35720/20.jpg\nSaving to: dataset/masks/0313-2/35720\nReading from: dataset/clips/0313-2/34800/20.jpg\nSaving to: dataset/masks/0313-2/34800\nReading from: dataset/clips/0313-2/36220/20.jpg\nSaving to: dataset/masks/0313-2/36220\nReading from: dataset/clips/0313-2/38380/20.jpg\nSaving to: dataset/masks/0313-2/38380\nReading from: dataset/clips/0313-2/410/20.jpg\nSaving to: dataset/masks/0313-2/410\nReading from: dataset/clips/0313-2/1460/20.jpg\nSaving to: dataset/masks/0313-2/1460\nReading from: dataset/clips/0313-2/5220/20.jpg\nSaving to: dataset/masks/0313-2/5220\nReading from: dataset/clips/0313-2/650/20.jpg\nSaving to: dataset/masks/0313-2/650\nReading from: dataset/clips/0313-2/30900/20.jpg\nSaving to: dataset/masks/0313-2/30900\nReading from: dataset/clips/0313-2/41320/20.jpg\nSaving to: dataset/masks/0313-2/41320\nReading from: dataset/clips/0313-2/36500/20.jpg\nSaving to: dataset/masks/0313-2/36500\nReading from: dataset/clips/0313-2/1520/20.jpg\nSaving to: dataset/masks/0313-2/1520\nReading from: dataset/clips/0313-2/905/20.jpg\nSaving to: dataset/masks/0313-2/905\nReading from: dataset/clips/0313-2/28140/20.jpg\nSaving to: dataset/masks/0313-2/28140\nReading from: dataset/clips/0313-2/8820/20.jpg\nSaving to: dataset/masks/0313-2/8820\nReading from: dataset/clips/0313-2/7320/20.jpg\nSaving to: dataset/masks/0313-2/7320\nReading from: dataset/clips/0313-2/19200/20.jpg\nSaving to: dataset/masks/0313-2/19200\nReading from: dataset/clips/0313-2/660/20.jpg\nSaving to: dataset/masks/0313-2/660\nReading from: dataset/clips/0313-2/40080/20.jpg\nSaving to: dataset/masks/0313-2/40080\nReading from: dataset/clips/0313-2/17700/20.jpg\nSaving to: dataset/masks/0313-2/17700\nReading from: dataset/clips/0313-2/170/20.jpg\nSaving to: dataset/masks/0313-2/170\nReading from: dataset/clips/0313-2/33520/20.jpg\nSaving to: dataset/masks/0313-2/33520\nReading from: dataset/clips/0313-2/59880/20.jpg\nSaving to: dataset/masks/0313-2/59880\nReading from: dataset/clips/0313-2/10440/20.jpg\nSaving to: dataset/masks/0313-2/10440\nReading from: dataset/clips/0313-2/22800/20.jpg\nSaving to: dataset/masks/0313-2/22800\nReading from: dataset/clips/0313-2/13740/20.jpg\nSaving to: dataset/masks/0313-2/13740\nReading from: dataset/clips/0313-2/910/20.jpg\nSaving to: dataset/masks/0313-2/910\nReading from: dataset/clips/0313-2/42780/20.jpg\nSaving to: dataset/masks/0313-2/42780\nReading from: dataset/clips/0313-2/9840/20.jpg\nSaving to: dataset/masks/0313-2/9840\nReading from: dataset/clips/0313-2/36660/20.jpg\nSaving to: dataset/masks/0313-2/36660\nReading from: dataset/clips/0313-2/42940/20.jpg\nSaving to: dataset/masks/0313-2/42940\nReading from: dataset/clips/0313-2/35780/20.jpg\nSaving to: dataset/masks/0313-2/35780\nReading from: dataset/clips/0313-2/2520/20.jpg\nSaving to: dataset/masks/0313-2/2520\nReading from: dataset/clips/0313-2/430/20.jpg\nSaving to: dataset/masks/0313-2/430\nReading from: dataset/clips/0313-2/23880/20.jpg\nSaving to: dataset/masks/0313-2/23880\nReading from: dataset/clips/0313-2/17820/20.jpg\nSaving to: dataset/masks/0313-2/17820\nReading from: dataset/clips/0313-2/4860/20.jpg\nSaving to: dataset/masks/0313-2/4860\nReading from: dataset/clips/0313-2/38400/20.jpg\nSaving to: dataset/masks/0313-2/38400\nReading from: dataset/clips/0313-2/63480/20.jpg\nSaving to: dataset/masks/0313-2/63480\nReading from: dataset/clips/0313-2/34140/20.jpg\nSaving to: dataset/masks/0313-2/34140\nReading from: dataset/clips/0313-2/23460/20.jpg\nSaving to: dataset/masks/0313-2/23460\nReading from: dataset/clips/0313-2/1245/20.jpg\nSaving to: dataset/masks/0313-2/1245\nReading from: dataset/clips/0313-2/1150/20.jpg\nSaving to: dataset/masks/0313-2/1150\nReading from: dataset/clips/0313-2/21060/20.jpg\nSaving to: dataset/masks/0313-2/21060\nReading from: dataset/clips/0313-2/30300/20.jpg\nSaving to: dataset/masks/0313-2/30300\nReading from: dataset/clips/0313-2/36680/20.jpg\nSaving to: dataset/masks/0313-2/36680\nReading from: dataset/clips/0313-2/35540/20.jpg\nSaving to: dataset/masks/0313-2/35540\nReading from: dataset/clips/0313-2/40920/20.jpg\nSaving to: dataset/masks/0313-2/40920\nReading from: dataset/clips/0313-2/34240/20.jpg\nSaving to: dataset/masks/0313-2/34240\nReading from: dataset/clips/0313-2/1665/20.jpg\nSaving to: dataset/masks/0313-2/1665\nReading from: dataset/clips/0313-2/1605/20.jpg\nSaving to: dataset/masks/0313-2/1605\nReading from: dataset/clips/0313-2/27780/20.jpg\nSaving to: dataset/masks/0313-2/27780\nReading from: dataset/clips/0313-2/36200/20.jpg\nSaving to: dataset/masks/0313-2/36200\nReading from: dataset/clips/0313-2/61380/20.jpg\nSaving to: dataset/masks/0313-2/61380\nReading from: dataset/clips/0313-2/585/20.jpg\nSaving to: dataset/masks/0313-2/585\nReading from: dataset/clips/0313-2/42840/20.jpg\nSaving to: dataset/masks/0313-2/42840\nReading from: dataset/clips/0313-2/2160/20.jpg\nSaving to: dataset/masks/0313-2/2160\nReading from: dataset/clips/0313-2/22020/20.jpg\nSaving to: dataset/masks/0313-2/22020\nReading from: dataset/clips/0313-2/59520/20.jpg\nSaving to: dataset/masks/0313-2/59520\nReading from: dataset/clips/0313-2/21420/20.jpg\nSaving to: dataset/masks/0313-2/21420\nReading from: dataset/clips/0313-2/6360/20.jpg\nSaving to: dataset/masks/0313-2/6360\nReading from: dataset/clips/0313-2/35480/20.jpg\nSaving to: dataset/masks/0313-2/35480\nReading from: dataset/clips/0313-2/6720/20.jpg\nSaving to: dataset/masks/0313-2/6720\nReading from: dataset/clips/0313-2/61020/20.jpg\nSaving to: dataset/masks/0313-2/61020\nReading from: dataset/clips/0313-2/41720/20.jpg\nSaving to: dataset/masks/0313-2/41720\nReading from: dataset/clips/0313-2/8280/20.jpg\nSaving to: dataset/masks/0313-2/8280\nReading from: dataset/clips/0313-2/14400/20.jpg\nSaving to: dataset/masks/0313-2/14400\nReading from: dataset/clips/0313-2/43020/20.jpg\nSaving to: dataset/masks/0313-2/43020\nReading from: dataset/clips/0313-2/37380/20.jpg\nSaving to: dataset/masks/0313-2/37380\nReading from: dataset/clips/0313-2/8160/20.jpg\nSaving to: dataset/masks/0313-2/8160\nReading from: dataset/clips/0313-2/36700/20.jpg\nSaving to: dataset/masks/0313-2/36700\nReading from: dataset/clips/0313-2/29700/20.jpg\nSaving to: dataset/masks/0313-2/29700\nReading from: dataset/clips/0313-2/40180/20.jpg\nSaving to: dataset/masks/0313-2/40180\nReading from: dataset/clips/0313-2/530/20.jpg\nSaving to: dataset/masks/0313-2/530\nReading from: dataset/clips/0313-2/39840/20.jpg\nSaving to: dataset/masks/0313-2/39840\nReading from: dataset/clips/0313-2/3900/20.jpg\nSaving to: dataset/masks/0313-2/3900\nReading from: dataset/clips/0313-2/61140/20.jpg\nSaving to: dataset/masks/0313-2/61140\nReading from: dataset/clips/0313-2/1710/20.jpg\nSaving to: dataset/masks/0313-2/1710\nReading from: dataset/clips/0313-2/1585/20.jpg\nSaving to: dataset/masks/0313-2/1585\nReading from: dataset/clips/0313-2/1345/20.jpg\nSaving to: dataset/masks/0313-2/1345\nReading from: dataset/clips/0313-2/20700/20.jpg\nSaving to: dataset/masks/0313-2/20700\nReading from: dataset/clips/0313-2/33220/20.jpg\nSaving to: dataset/masks/0313-2/33220\nReading from: dataset/clips/0313-2/9960/20.jpg\nSaving to: dataset/masks/0313-2/9960\nReading from: dataset/clips/0313-2/25/20.jpg\nSaving to: dataset/masks/0313-2/25\nReading from: dataset/clips/0313-2/14640/20.jpg\nSaving to: dataset/masks/0313-2/14640\nReading from: dataset/clips/0313-2/36900/20.jpg\nSaving to: dataset/masks/0313-2/36900\nReading from: dataset/clips/0313-2/16680/20.jpg\nSaving to: dataset/masks/0313-2/16680\nReading from: dataset/clips/0313-2/16080/20.jpg\nSaving to: dataset/masks/0313-2/16080\nReading from: dataset/clips/0313-2/37820/20.jpg\nSaving to: dataset/masks/0313-2/37820\nReading from: dataset/clips/0313-2/16140/20.jpg\nSaving to: dataset/masks/0313-2/16140\nReading from: dataset/clips/0313-2/8940/20.jpg\nSaving to: dataset/masks/0313-2/8940\nReading from: dataset/clips/0313-2/60360/20.jpg\nSaving to: dataset/masks/0313-2/60360\nReading from: dataset/clips/0313-2/255/20.jpg\nSaving to: dataset/masks/0313-2/255\nReading from: dataset/clips/0313-2/2340/20.jpg\nSaving to: dataset/masks/0313-2/2340\nReading from: dataset/clips/0313-2/30480/20.jpg\nSaving to: dataset/masks/0313-2/30480\nReading from: dataset/clips/0313-2/22500/20.jpg\nSaving to: dataset/masks/0313-2/22500\nReading from: dataset/clips/0313-2/34460/20.jpg\nSaving to: dataset/masks/0313-2/34460\nReading from: dataset/clips/0313-2/33480/20.jpg\nSaving to: dataset/masks/0313-2/33480\nReading from: dataset/clips/0313-2/34640/20.jpg\nSaving to: dataset/masks/0313-2/34640\nReading from: dataset/clips/0313-2/975/20.jpg\nSaving to: dataset/masks/0313-2/975\nReading from: dataset/clips/0313-2/26880/20.jpg\nSaving to: dataset/masks/0313-2/26880\nReading from: dataset/clips/0313-2/6540/20.jpg\nSaving to: dataset/masks/0313-2/6540\nReading from: dataset/clips/0313-2/3420/20.jpg\nSaving to: dataset/masks/0313-2/3420\nReading from: dataset/clips/0313-2/64080/20.jpg\nSaving to: dataset/masks/0313-2/64080\nReading from: dataset/clips/0313-2/34400/20.jpg\nSaving to: dataset/masks/0313-2/34400\nReading from: dataset/clips/0313-2/43000/20.jpg\nSaving to: dataset/masks/0313-2/43000\nReading from: dataset/clips/0313-2/35420/20.jpg\nSaving to: dataset/masks/0313-2/35420\nReading from: dataset/clips/0313-2/75/20.jpg\nSaving to: dataset/masks/0313-2/75\nReading from: dataset/clips/0313-2/3840/20.jpg\nSaving to: dataset/masks/0313-2/3840\nReading from: dataset/clips/0313-2/14700/20.jpg\nSaving to: dataset/masks/0313-2/14700\nReading from: dataset/clips/0313-2/40760/20.jpg\nSaving to: dataset/masks/0313-2/40760\nReading from: dataset/clips/0313-2/33500/20.jpg\nSaving to: dataset/masks/0313-2/33500\nReading from: dataset/clips/0313-2/41900/20.jpg\nSaving to: dataset/masks/0313-2/41900\nReading from: dataset/clips/0313-2/1130/20.jpg\nSaving to: dataset/masks/0313-2/1130\nReading from: dataset/clips/0313-2/205/20.jpg\nSaving to: dataset/masks/0313-2/205\nReading from: dataset/clips/0313-2/16320/20.jpg\nSaving to: dataset/masks/0313-2/16320\nReading from: dataset/clips/0313-2/33140/20.jpg\nSaving to: dataset/masks/0313-2/33140\nReading from: dataset/clips/0313-2/59280/20.jpg\nSaving to: dataset/masks/0313-2/59280\nReading from: dataset/clips/0313-2/1010/20.jpg\nSaving to: dataset/masks/0313-2/1010\nReading from: dataset/clips/0313-2/59220/20.jpg\nSaving to: dataset/masks/0313-2/59220\nReading from: dataset/clips/0313-2/37560/20.jpg\nSaving to: dataset/masks/0313-2/37560\nReading from: dataset/clips/0313-2/42980/20.jpg\nSaving to: dataset/masks/0313-2/42980\nReading from: dataset/clips/0313-2/960/20.jpg\nSaving to: dataset/masks/0313-2/960\nReading from: dataset/clips/0313-2/34000/20.jpg\nSaving to: dataset/masks/0313-2/34000\nReading from: dataset/clips/0313-2/875/20.jpg\nSaving to: dataset/masks/0313-2/875\nReading from: dataset/clips/0313-2/32880/20.jpg\nSaving to: dataset/masks/0313-2/32880\nReading from: dataset/clips/0313-2/18780/20.jpg\nSaving to: dataset/masks/0313-2/18780\nReading from: dataset/clips/0313-2/940/20.jpg\nSaving to: dataset/masks/0313-2/940\nReading from: dataset/clips/0313-2/1125/20.jpg\nSaving to: dataset/masks/0313-2/1125\nReading from: dataset/clips/0313-2/2040/20.jpg\nSaving to: dataset/masks/0313-2/2040\nReading from: dataset/clips/0313-2/18900/20.jpg\nSaving to: dataset/masks/0313-2/18900\nReading from: dataset/clips/0313-2/42340/20.jpg\nSaving to: dataset/masks/0313-2/42340\nReading from: dataset/clips/0313-2/36780/20.jpg\nSaving to: dataset/masks/0313-2/36780\nReading from: dataset/clips/0313-2/10860/20.jpg\nSaving to: dataset/masks/0313-2/10860\nReading from: dataset/clips/0313-2/42400/20.jpg\nSaving to: dataset/masks/0313-2/42400\nReading from: dataset/clips/0313-2/41460/20.jpg\nSaving to: dataset/masks/0313-2/41460\nReading from: dataset/clips/0313-2/1500/20.jpg\nSaving to: dataset/masks/0313-2/1500\nReading from: dataset/clips/0313-2/5460/20.jpg\nSaving to: dataset/masks/0313-2/5460\nReading from: dataset/clips/0313-2/28800/20.jpg\nSaving to: dataset/masks/0313-2/28800\nReading from: dataset/clips/0313-2/11400/20.jpg\nSaving to: dataset/masks/0313-2/11400\nReading from: dataset/clips/0313-2/31320/20.jpg\nSaving to: dataset/masks/0313-2/31320\nReading from: dataset/clips/0313-2/19500/20.jpg\nSaving to: dataset/masks/0313-2/19500\nReading from: dataset/clips/0313-2/750/20.jpg\nSaving to: dataset/masks/0313-2/750\nReading from: dataset/clips/0313-2/1595/20.jpg\nSaving to: dataset/masks/0313-2/1595\nReading from: dataset/clips/0313-2/17520/20.jpg\nSaving to: dataset/masks/0313-2/17520\nReading from: dataset/clips/0313-2/215/20.jpg\nSaving to: dataset/masks/0313-2/215\nReading from: dataset/clips/0313-2/33160/20.jpg\nSaving to: dataset/masks/0313-2/33160\nReading from: dataset/clips/0313-2/21780/20.jpg\nSaving to: dataset/masks/0313-2/21780\nReading from: dataset/clips/0313-2/35080/20.jpg\nSaving to: dataset/masks/0313-2/35080\nReading from: dataset/clips/0313-2/37860/20.jpg\nSaving to: dataset/masks/0313-2/37860\nReading from: dataset/clips/0313-2/42620/20.jpg\nSaving to: dataset/masks/0313-2/42620\nReading from: dataset/clips/0313-2/745/20.jpg\nSaving to: dataset/masks/0313-2/745\nReading from: dataset/clips/0313-2/35360/20.jpg\nSaving to: dataset/masks/0313-2/35360\nReading from: dataset/clips/0313-2/825/20.jpg\nSaving to: dataset/masks/0313-2/825\nReading from: dataset/clips/0313-2/4200/20.jpg\nSaving to: dataset/masks/0313-2/4200\nReading from: dataset/clips/0313-2/16020/20.jpg\nSaving to: dataset/masks/0313-2/16020\nReading from: dataset/clips/0313-2/4740/20.jpg\nSaving to: dataset/masks/0313-2/4740\nReading from: dataset/clips/0313-2/27480/20.jpg\nSaving to: dataset/masks/0313-2/27480\nReading from: dataset/clips/0313-2/355/20.jpg\nSaving to: dataset/masks/0313-2/355\nReading from: dataset/clips/0313-2/60240/20.jpg\nSaving to: dataset/masks/0313-2/60240\nReading from: dataset/clips/0313-2/26580/20.jpg\nSaving to: dataset/masks/0313-2/26580\nReading from: dataset/clips/0313-2/28380/20.jpg\nSaving to: dataset/masks/0313-2/28380\nReading from: dataset/clips/0313-2/22860/20.jpg\nSaving to: dataset/masks/0313-2/22860\nReading from: dataset/clips/0313-2/43040/20.jpg\nSaving to: dataset/masks/0313-2/43040\nReading from: dataset/clips/0313-2/11340/20.jpg\nSaving to: dataset/masks/0313-2/11340\nReading from: dataset/clips/0313-2/10560/20.jpg\nSaving to: dataset/masks/0313-2/10560\nReading from: dataset/clips/0313-2/4800/20.jpg\nSaving to: dataset/masks/0313-2/4800\nReading from: dataset/clips/0313-2/1320/20.jpg\nSaving to: dataset/masks/0313-2/1320\nReading from: dataset/clips/0313-2/15240/20.jpg\nSaving to: dataset/masks/0313-2/15240\nReading from: dataset/clips/0313-2/30840/20.jpg\nSaving to: dataset/masks/0313-2/30840\nReading from: dataset/clips/0313-2/20220/20.jpg\nSaving to: dataset/masks/0313-2/20220\nReading from: dataset/clips/0313-2/14280/20.jpg\nSaving to: dataset/masks/0313-2/14280\nReading from: dataset/clips/0313-2/41280/20.jpg\nSaving to: dataset/masks/0313-2/41280\nReading from: dataset/clips/0313-2/14940/20.jpg\nSaving to: dataset/masks/0313-2/14940\nReading from: dataset/clips/0313-2/790/20.jpg\nSaving to: dataset/masks/0313-2/790\nReading from: dataset/clips/0313-2/765/20.jpg\nSaving to: dataset/masks/0313-2/765\nReading from: dataset/clips/0313-2/445/20.jpg\nSaving to: dataset/masks/0313-2/445\nReading from: dataset/clips/0313-2/37000/20.jpg\nSaving to: dataset/masks/0313-2/37000\nReading from: dataset/clips/0313-2/42300/20.jpg\nSaving to: dataset/masks/0313-2/42300\nReading from: dataset/clips/0313-2/41420/20.jpg\nSaving to: dataset/masks/0313-2/41420\nReading from: dataset/clips/0313-2/40600/20.jpg\nSaving to: dataset/masks/0313-2/40600\nReading from: dataset/clips/0313-2/40260/20.jpg\nSaving to: dataset/masks/0313-2/40260\nReading from: dataset/clips/0313-2/180/20.jpg\nSaving to: dataset/masks/0313-2/180\nReading from: dataset/clips/0313-2/38240/20.jpg\nSaving to: dataset/masks/0313-2/38240\nReading from: dataset/clips/0313-2/1225/20.jpg\nSaving to: dataset/masks/0313-2/1225\nReading from: dataset/clips/0313-2/36360/20.jpg\nSaving to: dataset/masks/0313-2/36360\nReading from: dataset/clips/0313-2/25200/20.jpg\nSaving to: dataset/masks/0313-2/25200\nReading from: dataset/clips/0313-2/33940/20.jpg\nSaving to: dataset/masks/0313-2/33940\nReading from: dataset/clips/0313-2/10020/20.jpg\nSaving to: dataset/masks/0313-2/10020\nReading from: dataset/clips/0313-2/14460/20.jpg\nSaving to: dataset/masks/0313-2/14460\nReading from: dataset/clips/0313-2/61980/20.jpg\nSaving to: dataset/masks/0313-2/61980\nReading from: dataset/clips/0313-2/37220/20.jpg\nSaving to: dataset/masks/0313-2/37220\nReading from: dataset/clips/0313-2/23760/20.jpg\nSaving to: dataset/masks/0313-2/23760\nReading from: dataset/clips/0313-2/34680/20.jpg\nSaving to: dataset/masks/0313-2/34680\nReading from: dataset/clips/0313-2/35140/20.jpg\nSaving to: dataset/masks/0313-2/35140\nReading from: dataset/clips/0313-2/40380/20.jpg\nSaving to: dataset/masks/0313-2/40380\nReading from: dataset/clips/0313-2/38620/20.jpg\nSaving to: dataset/masks/0313-2/38620\nReading from: dataset/clips/0313-2/40580/20.jpg\nSaving to: dataset/masks/0313-2/40580\nReading from: dataset/clips/0313-2/1620/20.jpg\nSaving to: dataset/masks/0313-2/1620\nReading from: dataset/clips/0313-2/35340/20.jpg\nSaving to: dataset/masks/0313-2/35340\nReading from: dataset/clips/0313-2/64260/20.jpg\nSaving to: dataset/masks/0313-2/64260\nReading from: dataset/clips/0313-2/7260/20.jpg\nSaving to: dataset/masks/0313-2/7260\nReading from: dataset/clips/0313-2/19620/20.jpg\nSaving to: dataset/masks/0313-2/19620\nReading from: dataset/clips/0313-2/9240/20.jpg\nSaving to: dataset/masks/0313-2/9240\nReading from: dataset/clips/0313-2/22740/20.jpg\nSaving to: dataset/masks/0313-2/22740\nReading from: dataset/clips/0313-2/39100/20.jpg\nSaving to: dataset/masks/0313-2/39100\nReading from: dataset/clips/0313-2/39080/20.jpg\nSaving to: dataset/masks/0313-2/39080\nReading from: dataset/clips/0313-2/41240/20.jpg\nSaving to: dataset/masks/0313-2/41240\nReading from: dataset/clips/0313-2/29580/20.jpg\nSaving to: dataset/masks/0313-2/29580\nReading from: dataset/clips/0313-2/63180/20.jpg\nSaving to: dataset/masks/0313-2/63180\nReading from: dataset/clips/0313-2/34160/20.jpg\nSaving to: dataset/masks/0313-2/34160\nReading from: dataset/clips/0313-2/21120/20.jpg\nSaving to: dataset/masks/0313-2/21120\nReading from: dataset/clips/0313-2/1055/20.jpg\nSaving to: dataset/masks/0313-2/1055\nReading from: dataset/clips/0313-2/37700/20.jpg\nSaving to: dataset/masks/0313-2/37700\nReading from: dataset/clips/0313-2/41120/20.jpg\nSaving to: dataset/masks/0313-2/41120\nReading from: dataset/clips/0313-2/62220/20.jpg\nSaving to: dataset/masks/0313-2/62220\nReading from: dataset/clips/0313-2/35200/20.jpg\nSaving to: dataset/masks/0313-2/35200\nReading from: dataset/clips/0313-2/1720/20.jpg\nSaving to: dataset/masks/0313-2/1720\nReading from: dataset/clips/0313-2/41000/20.jpg\nSaving to: dataset/masks/0313-2/41000\nReading from: dataset/clips/0313-2/19020/20.jpg\nSaving to: dataset/masks/0313-2/19020\nReading from: dataset/clips/0313-2/15360/20.jpg\nSaving to: dataset/masks/0313-2/15360\nReading from: dataset/clips/0313-2/58680/20.jpg\nSaving to: dataset/masks/0313-2/58680\nReading from: dataset/clips/0313-2/1235/20.jpg\nSaving to: dataset/masks/0313-2/1235\nReading from: dataset/clips/0313-2/41300/20.jpg\nSaving to: dataset/masks/0313-2/41300\nReading from: dataset/clips/0313-2/18600/20.jpg\nSaving to: dataset/masks/0313-2/18600\nReading from: dataset/clips/0313-2/1185/20.jpg\nSaving to: dataset/masks/0313-2/1185\nReading from: dataset/clips/0313-2/1160/20.jpg\nSaving to: dataset/masks/0313-2/1160\nReading from: dataset/clips/0313-2/25740/20.jpg\nSaving to: dataset/masks/0313-2/25740\nReading from: dataset/clips/0313-2/35980/20.jpg\nSaving to: dataset/masks/0313-2/35980\nReading from: dataset/clips/0313-2/880/20.jpg\nSaving to: dataset/masks/0313-2/880\nReading from: dataset/clips/0313-2/7500/20.jpg\nSaving to: dataset/masks/0313-2/7500\nReading from: dataset/clips/0313-2/38120/20.jpg\nSaving to: dataset/masks/0313-2/38120\nReading from: dataset/clips/0313-2/1680/20.jpg\nSaving to: dataset/masks/0313-2/1680\nReading from: dataset/clips/0313-2/38100/20.jpg\nSaving to: dataset/masks/0313-2/38100\nReading from: dataset/clips/0313-2/40840/20.jpg\nSaving to: dataset/masks/0313-2/40840\nReading from: dataset/clips/0313-2/38460/20.jpg\nSaving to: dataset/masks/0313-2/38460\nReading from: dataset/clips/0313-2/42440/20.jpg\nSaving to: dataset/masks/0313-2/42440\nReading from: dataset/clips/0313-2/36580/20.jpg\nSaving to: dataset/masks/0313-2/36580\nReading from: dataset/clips/0313-2/38840/20.jpg\nSaving to: dataset/masks/0313-2/38840\nReading from: dataset/clips/0313-2/59100/20.jpg\nSaving to: dataset/masks/0313-2/59100\nReading from: dataset/clips/0313-2/38800/20.jpg\nSaving to: dataset/masks/0313-2/38800\nReading from: dataset/clips/0313-2/37900/20.jpg\nSaving to: dataset/masks/0313-2/37900\nReading from: dataset/clips/0313-2/38900/20.jpg\nSaving to: dataset/masks/0313-2/38900\nReading from: dataset/clips/0313-2/11160/20.jpg\nSaving to: dataset/masks/0313-2/11160\nReading from: dataset/clips/0313-2/41180/20.jpg\nSaving to: dataset/masks/0313-2/41180\nReading from: dataset/clips/0313-2/21900/20.jpg\nSaving to: dataset/masks/0313-2/21900\nReading from: dataset/clips/0313-2/1105/20.jpg\nSaving to: dataset/masks/0313-2/1105\nReading from: dataset/clips/0313-2/1780/20.jpg\nSaving to: dataset/masks/0313-2/1780\nReading from: dataset/clips/0313-2/14760/20.jpg\nSaving to: dataset/masks/0313-2/14760\nReading from: dataset/clips/0313-2/37920/20.jpg\nSaving to: dataset/masks/0313-2/37920\nReading from: dataset/clips/0313-2/14040/20.jpg\nSaving to: dataset/masks/0313-2/14040\nReading from: dataset/clips/0313-2/1000/20.jpg\nSaving to: dataset/masks/0313-2/1000\nReading from: dataset/clips/0313-2/60060/20.jpg\nSaving to: dataset/masks/0313-2/60060\nReading from: dataset/clips/0313-2/1440/20.jpg\nSaving to: dataset/masks/0313-2/1440\nReading from: dataset/clips/0313-2/1050/20.jpg\nSaving to: dataset/masks/0313-2/1050\nReading from: dataset/clips/0313-2/1480/20.jpg\nSaving to: dataset/masks/0313-2/1480\nReading from: dataset/clips/0313-2/310/20.jpg\nSaving to: dataset/masks/0313-2/310\nReading from: dataset/clips/0313-2/22560/20.jpg\nSaving to: dataset/masks/0313-2/22560\nReading from: dataset/clips/0313-2/9480/20.jpg\nSaving to: dataset/masks/0313-2/9480\nReading from: dataset/clips/0313-2/42600/20.jpg\nSaving to: dataset/masks/0313-2/42600\nReading from: dataset/clips/0313-2/33380/20.jpg\nSaving to: dataset/masks/0313-2/33380\nReading from: dataset/clips/0313-2/800/20.jpg\nSaving to: dataset/masks/0313-2/800\nReading from: dataset/clips/0313-2/31020/20.jpg\nSaving to: dataset/masks/0313-2/31020\nReading from: dataset/clips/0313-2/34720/20.jpg\nSaving to: dataset/masks/0313-2/34720\nReading from: dataset/clips/0313-2/1290/20.jpg\nSaving to: dataset/masks/0313-2/1290\nReading from: dataset/clips/0313-2/19560/20.jpg\nSaving to: dataset/masks/0313-2/19560\nReading from: dataset/clips/0313-2/35920/20.jpg\nSaving to: dataset/masks/0313-2/35920\nReading from: dataset/clips/0313-2/17100/20.jpg\nSaving to: dataset/masks/0313-2/17100\nReading from: dataset/clips/0313-2/31140/20.jpg\nSaving to: dataset/masks/0313-2/31140\nReading from: dataset/clips/0313-2/40960/20.jpg\nSaving to: dataset/masks/0313-2/40960\nReading from: dataset/clips/0313-2/16380/20.jpg\nSaving to: dataset/masks/0313-2/16380\nReading from: dataset/clips/0313-2/690/20.jpg\nSaving to: dataset/masks/0313-2/690\nReading from: dataset/clips/0313-2/1310/20.jpg\nSaving to: dataset/masks/0313-2/1310\nReading from: dataset/clips/0313-2/42860/20.jpg\nSaving to: dataset/masks/0313-2/42860\nReading from: dataset/clips/0313-2/1155/20.jpg\nSaving to: dataset/masks/0313-2/1155\nReading from: dataset/clips/0313-2/41440/20.jpg\nSaving to: dataset/masks/0313-2/41440\nReading from: dataset/clips/0313-2/1020/20.jpg\nSaving to: dataset/masks/0313-2/1020\nReading from: dataset/clips/0313-2/62520/20.jpg\nSaving to: dataset/masks/0313-2/62520\nReading from: dataset/clips/0313-2/32280/20.jpg\nSaving to: dataset/masks/0313-2/32280\nReading from: dataset/clips/0313-2/42280/20.jpg\nSaving to: dataset/masks/0313-2/42280\nReading from: dataset/clips/0313-2/62940/20.jpg\nSaving to: dataset/masks/0313-2/62940\nReading from: dataset/clips/0313-2/33060/20.jpg\nSaving to: dataset/masks/0313-2/33060\nReading from: dataset/clips/0313-2/200/20.jpg\nSaving to: dataset/masks/0313-2/200\nReading from: dataset/clips/0313-2/36020/20.jpg\nSaving to: dataset/masks/0313-2/36020\nReading from: dataset/clips/0313-2/13260/20.jpg\nSaving to: dataset/masks/0313-2/13260\nReading from: dataset/clips/0313-2/33780/20.jpg\nSaving to: dataset/masks/0313-2/33780\nReading from: dataset/clips/0313-2/64380/20.jpg\nSaving to: dataset/masks/0313-2/64380\nReading from: dataset/clips/0313-2/32680/20.jpg\nSaving to: dataset/masks/0313-2/32680\nReading from: dataset/clips/0313-2/1630/20.jpg\nSaving to: dataset/masks/0313-2/1630\nReading from: dataset/clips/0313-2/26220/20.jpg\nSaving to: dataset/masks/0313-2/26220\nReading from: dataset/clips/0313-2/1375/20.jpg\nSaving to: dataset/masks/0313-2/1375\nReading from: dataset/clips/0313-2/37040/20.jpg\nSaving to: dataset/masks/0313-2/37040\nReading from: dataset/clips/0313-2/1470/20.jpg\nSaving to: dataset/masks/0313-2/1470\nReading from: dataset/clips/0313-2/26280/20.jpg\nSaving to: dataset/masks/0313-2/26280\nReading from: dataset/clips/0313-2/29940/20.jpg\nSaving to: dataset/masks/0313-2/29940\nReading from: dataset/clips/0313-2/34500/20.jpg\nSaving to: dataset/masks/0313-2/34500\nReading from: dataset/clips/0313-2/42140/20.jpg\nSaving to: dataset/masks/0313-2/42140\nReading from: dataset/clips/0313-2/32760/20.jpg\nSaving to: dataset/masks/0313-2/32760\nReading from: dataset/clips/0313-2/32040/20.jpg\nSaving to: dataset/masks/0313-2/32040\nReading from: dataset/clips/0313-2/8340/20.jpg\nSaving to: dataset/masks/0313-2/8340\nReading from: dataset/clips/0313-2/31620/20.jpg\nSaving to: dataset/masks/0313-2/31620\nReading from: dataset/clips/0313-2/25500/20.jpg\nSaving to: dataset/masks/0313-2/25500\nReading from: dataset/clips/0313-2/40740/20.jpg\nSaving to: dataset/masks/0313-2/40740\nReading from: dataset/clips/0313-2/39420/20.jpg\nSaving to: dataset/masks/0313-2/39420\nReading from: dataset/clips/0313-2/61800/20.jpg\nSaving to: dataset/masks/0313-2/61800\nReading from: dataset/clips/0313-2/920/20.jpg\nSaving to: dataset/masks/0313-2/920\nReading from: dataset/clips/0313-2/12480/20.jpg\nSaving to: dataset/masks/0313-2/12480\nReading from: dataset/clips/0313-2/31740/20.jpg\nSaving to: dataset/masks/0313-2/31740\nReading from: dataset/clips/0313-2/16860/20.jpg\nSaving to: dataset/masks/0313-2/16860\nReading from: dataset/clips/0313-2/62820/20.jpg\nSaving to: dataset/masks/0313-2/62820\nReading from: dataset/clips/0313-2/33640/20.jpg\nSaving to: dataset/masks/0313-2/33640\nReading from: dataset/clips/0313-2/41400/20.jpg\nSaving to: dataset/masks/0313-2/41400\nReading from: dataset/clips/0313-2/40400/20.jpg\nSaving to: dataset/masks/0313-2/40400\nReading from: dataset/clips/0313-2/865/20.jpg\nSaving to: dataset/masks/0313-2/865\nReading from: dataset/clips/0313-2/37200/20.jpg\nSaving to: dataset/masks/0313-2/37200\nReading from: dataset/clips/0313-2/38280/20.jpg\nSaving to: dataset/masks/0313-2/38280\nReading from: dataset/clips/0313-2/26460/20.jpg\nSaving to: dataset/masks/0313-2/26460\nReading from: dataset/clips/0313-2/26820/20.jpg\nSaving to: dataset/masks/0313-2/26820\nReading from: dataset/clips/0313-2/36560/20.jpg\nSaving to: dataset/masks/0313-2/36560\nReading from: dataset/clips/0313-2/8100/20.jpg\nSaving to: dataset/masks/0313-2/8100\nReading from: dataset/clips/0313-2/61260/20.jpg\nSaving to: dataset/masks/0313-2/61260\nReading from: dataset/clips/0313-2/17760/20.jpg\nSaving to: dataset/masks/0313-2/17760\nReading from: dataset/clips/0313-2/1005/20.jpg\nSaving to: dataset/masks/0313-2/1005\nReading from: dataset/clips/0313-2/27600/20.jpg\nSaving to: dataset/masks/0313-2/27600\nReading from: dataset/clips/0313-2/770/20.jpg\nSaving to: dataset/masks/0313-2/770\nReading from: dataset/clips/0313-2/1725/20.jpg\nSaving to: dataset/masks/0313-2/1725\nReading from: dataset/clips/0313-2/1525/20.jpg\nSaving to: dataset/masks/0313-2/1525\nReading from: dataset/clips/0313-2/34980/20.jpg\nSaving to: dataset/masks/0313-2/34980\nReading from: dataset/clips/0313-2/13380/20.jpg\nSaving to: dataset/masks/0313-2/13380\nReading from: dataset/clips/0313-2/37840/20.jpg\nSaving to: dataset/masks/0313-2/37840\nReading from: dataset/clips/0313-2/42320/20.jpg\nSaving to: dataset/masks/0313-2/42320\nReading from: dataset/clips/0313-2/1335/20.jpg\nSaving to: dataset/masks/0313-2/1335\nReading from: dataset/clips/0313-2/38940/20.jpg\nSaving to: dataset/masks/0313-2/38940\nReading from: dataset/clips/0313-2/37680/20.jpg\nSaving to: dataset/masks/0313-2/37680\nReading from: dataset/clips/0313-2/30540/20.jpg\nSaving to: dataset/masks/0313-2/30540\nReading from: dataset/clips/0313-2/500/20.jpg\nSaving to: dataset/masks/0313-2/500\nReading from: dataset/clips/0313-2/41540/20.jpg\nSaving to: dataset/masks/0313-2/41540\nReading from: dataset/clips/0313-2/21660/20.jpg\nSaving to: dataset/masks/0313-2/21660\nReading from: dataset/clips/0313-2/13620/20.jpg\nSaving to: dataset/masks/0313-2/13620\nReading from: dataset/clips/0313-2/29760/20.jpg\nSaving to: dataset/masks/0313-2/29760\nReading from: dataset/clips/0313-2/35/20.jpg\nSaving to: dataset/masks/0313-2/35\nReading from: dataset/clips/0313-2/13680/20.jpg\nSaving to: dataset/masks/0313-2/13680\nReading from: dataset/clips/0313-2/535/20.jpg\nSaving to: dataset/masks/0313-2/535\nReading from: dataset/clips/0313-2/41360/20.jpg\nSaving to: dataset/masks/0313-2/41360\nReading from: dataset/clips/0313-2/37540/20.jpg\nSaving to: dataset/masks/0313-2/37540\nReading from: dataset/clips/0313-2/33400/20.jpg\nSaving to: dataset/masks/0313-2/33400\nReading from: dataset/clips/0313-2/40620/20.jpg\nSaving to: dataset/masks/0313-2/40620\nReading from: dataset/clips/0313-2/21000/20.jpg\nSaving to: dataset/masks/0313-2/21000\nReading from: dataset/clips/0313-2/14160/20.jpg\nSaving to: dataset/masks/0313-2/14160\nReading from: dataset/clips/0313-2/30780/20.jpg\nSaving to: dataset/masks/0313-2/30780\nReading from: dataset/clips/0313-2/36280/20.jpg\nSaving to: dataset/masks/0313-2/36280\nReading from: dataset/clips/0313-2/10680/20.jpg\nSaving to: dataset/masks/0313-2/10680\nReading from: dataset/clips/0313-2/100/20.jpg\nSaving to: dataset/masks/0313-2/100\nReading from: dataset/clips/0313-2/58620/20.jpg\nSaving to: dataset/masks/0313-2/58620\nReading from: dataset/clips/0313-2/1765/20.jpg\nSaving to: dataset/masks/0313-2/1765\nReading from: dataset/clips/0313-2/465/20.jpg\nSaving to: dataset/masks/0313-2/465\nReading from: dataset/clips/0313-2/42480/20.jpg\nSaving to: dataset/masks/0313-2/42480\nReading from: dataset/clips/0313-2/42360/20.jpg\nSaving to: dataset/masks/0313-2/42360\nReading from: dataset/clips/0313-2/35000/20.jpg\nSaving to: dataset/masks/0313-2/35000\nReading from: dataset/clips/0313-2/39280/20.jpg\nSaving to: dataset/masks/0313-2/39280\nReading from: dataset/clips/0313-2/1745/20.jpg\nSaving to: dataset/masks/0313-2/1745\nReading from: dataset/clips/0313-2/1075/20.jpg\nSaving to: dataset/masks/0313-2/1075\nReading from: dataset/clips/0313-2/38140/20.jpg\nSaving to: dataset/masks/0313-2/38140\nReading from: dataset/clips/0313-2/14520/20.jpg\nSaving to: dataset/masks/0313-2/14520\nReading from: dataset/clips/0313-2/31080/20.jpg\nSaving to: dataset/masks/0313-2/31080\nReading from: dataset/clips/0313-2/21540/20.jpg\nSaving to: dataset/masks/0313-2/21540\nReading from: dataset/clips/0313-2/38220/20.jpg\nSaving to: dataset/masks/0313-2/38220\nReading from: dataset/clips/0313-2/42720/20.jpg\nSaving to: dataset/masks/0313-2/42720\nReading from: dataset/clips/0313-2/16800/20.jpg\nSaving to: dataset/masks/0313-2/16800\nReading from: dataset/clips/0313-2/38360/20.jpg\nSaving to: dataset/masks/0313-2/38360\nReading from: dataset/clips/0313-2/12660/20.jpg\nSaving to: dataset/masks/0313-2/12660\nReading from: dataset/clips/0313-2/18960/20.jpg\nSaving to: dataset/masks/0313-2/18960\nReading from: dataset/clips/0313-2/37720/20.jpg\nSaving to: dataset/masks/0313-2/37720\nReading from: dataset/clips/0313-2/19680/20.jpg\nSaving to: dataset/masks/0313-2/19680\nReading from: dataset/clips/0313-2/32920/20.jpg\nSaving to: dataset/masks/0313-2/32920\nReading from: dataset/clips/0313-2/39600/20.jpg\nSaving to: dataset/masks/0313-2/39600\nReading from: dataset/clips/0313-2/34540/20.jpg\nSaving to: dataset/masks/0313-2/34540\nReading from: dataset/clips/0313-2/4440/20.jpg\nSaving to: dataset/masks/0313-2/4440\nReading from: dataset/clips/0313-2/3720/20.jpg\nSaving to: dataset/masks/0313-2/3720\nReading from: dataset/clips/0313-2/42260/20.jpg\nSaving to: dataset/masks/0313-2/42260\nReading from: dataset/clips/0313-2/33460/20.jpg\nSaving to: dataset/masks/0313-2/33460\nReading from: dataset/clips/0313-2/9360/20.jpg\nSaving to: dataset/masks/0313-2/9360\nReading from: dataset/clips/0313-2/19800/20.jpg\nSaving to: dataset/masks/0313-2/19800\nReading from: dataset/clips/0313-2/210/20.jpg\nSaving to: dataset/masks/0313-2/210\nReading from: dataset/clips/0313-2/20520/20.jpg\nSaving to: dataset/masks/0313-2/20520\nReading from: dataset/clips/0313-2/36920/20.jpg\nSaving to: dataset/masks/0313-2/36920\nReading from: dataset/clips/0313-2/62160/20.jpg\nSaving to: dataset/masks/0313-2/62160\nReading from: dataset/clips/0313-2/990/20.jpg\nSaving to: dataset/masks/0313-2/990\nReading from: dataset/clips/0313-2/63060/20.jpg\nSaving to: dataset/masks/0313-2/63060\nReading from: dataset/clips/0313-2/5040/20.jpg\nSaving to: dataset/masks/0313-2/5040\nReading from: dataset/clips/0313-2/26400/20.jpg\nSaving to: dataset/masks/0313-2/26400\nReading from: dataset/clips/0313-2/25080/20.jpg\nSaving to: dataset/masks/0313-2/25080\nReading from: dataset/clips/0313-2/12300/20.jpg\nSaving to: dataset/masks/0313-2/12300\nReading from: dataset/clips/0313-2/1735/20.jpg\nSaving to: dataset/masks/0313-2/1735\nReading from: dataset/clips/0313-2/42820/20.jpg\nSaving to: dataset/masks/0313-2/42820\nReading from: dataset/clips/0313-2/475/20.jpg\nSaving to: dataset/masks/0313-2/475\nReading from: dataset/clips/0313-2/61620/20.jpg\nSaving to: dataset/masks/0313-2/61620\nReading from: dataset/clips/0313-2/42800/20.jpg\nSaving to: dataset/masks/0313-2/42800\nReading from: dataset/clips/0313-2/63360/20.jpg\nSaving to: dataset/masks/0313-2/63360\nReading from: dataset/clips/0313-2/27420/20.jpg\nSaving to: dataset/masks/0313-2/27420\nReading from: dataset/clips/0313-2/35220/20.jpg\nSaving to: dataset/masks/0313-2/35220\nReading from: dataset/clips/0313-2/285/20.jpg\nSaving to: dataset/masks/0313-2/285\nReading from: dataset/clips/0313-2/1785/20.jpg\nSaving to: dataset/masks/0313-2/1785\nReading from: dataset/clips/0313-2/42380/20.jpg\nSaving to: dataset/masks/0313-2/42380\nReading from: dataset/clips/0313-2/42460/20.jpg\nSaving to: dataset/masks/0313-2/42460\nReading from: dataset/clips/0313-2/33860/20.jpg\nSaving to: dataset/masks/0313-2/33860\nReading from: dataset/clips/0313-2/23580/20.jpg\nSaving to: dataset/masks/0313-2/23580\nReading from: dataset/clips/0313-2/26160/20.jpg\nSaving to: dataset/masks/0313-2/26160\nReading from: dataset/clips/0313-2/60720/20.jpg\nSaving to: dataset/masks/0313-2/60720\nReading from: dataset/clips/0313-2/34960/20.jpg\nSaving to: dataset/masks/0313-2/34960\nReading from: dataset/clips/0313-2/1450/20.jpg\nSaving to: dataset/masks/0313-2/1450\nReading from: dataset/clips/0313-2/860/20.jpg\nSaving to: dataset/masks/0313-2/860\nReading from: dataset/clips/0313-2/935/20.jpg\nSaving to: dataset/masks/0313-2/935\nReading from: dataset/clips/0313-2/58800/20.jpg\nSaving to: dataset/masks/0313-2/58800\nReading from: dataset/clips/0313-2/42960/20.jpg\nSaving to: dataset/masks/0313-2/42960\nReading from: dataset/clips/0313-2/59040/20.jpg\nSaving to: dataset/masks/0313-2/59040\nReading from: dataset/clips/0313-2/7620/20.jpg\nSaving to: dataset/masks/0313-2/7620\nReading from: dataset/clips/0313-2/375/20.jpg\nSaving to: dataset/masks/0313-2/375\nReading from: dataset/clips/0313-2/240/20.jpg\nSaving to: dataset/masks/0313-2/240\nReading from: dataset/clips/0313-2/37320/20.jpg\nSaving to: dataset/masks/0313-2/37320\nReading from: dataset/clips/0313-2/885/20.jpg\nSaving to: dataset/masks/0313-2/885\nReading from: dataset/clips/0313-2/29880/20.jpg\nSaving to: dataset/masks/0313-2/29880\nReading from: dataset/clips/0313-2/36740/20.jpg\nSaving to: dataset/masks/0313-2/36740\nReading from: dataset/clips/0313-2/1100/20.jpg\nSaving to: dataset/masks/0313-2/1100\nReading from: dataset/clips/0313-2/1210/20.jpg\nSaving to: dataset/masks/0313-2/1210\nReading from: dataset/clips/0313-2/1040/20.jpg\nSaving to: dataset/masks/0313-2/1040\nReading from: dataset/clips/0313-2/1540/20.jpg\nSaving to: dataset/masks/0313-2/1540\nReading from: dataset/clips/0313-2/62040/20.jpg\nSaving to: dataset/masks/0313-2/62040\nReading from: dataset/clips/0313-2/1475/20.jpg\nSaving to: dataset/masks/0313-2/1475\nReading from: dataset/clips/0313-2/36040/20.jpg\nSaving to: dataset/masks/0313-2/36040\nReading from: dataset/clips/0313-2/41580/20.jpg\nSaving to: dataset/masks/0313-2/41580\nReading from: dataset/clips/0313-2/38000/20.jpg\nSaving to: dataset/masks/0313-2/38000\nReading from: dataset/clips/0313-2/37660/20.jpg\nSaving to: dataset/masks/0313-2/37660\nReading from: dataset/clips/0313-2/40540/20.jpg\nSaving to: dataset/masks/0313-2/40540\nReading from: dataset/clips/0313-2/945/20.jpg\nSaving to: dataset/masks/0313-2/945\nReading from: dataset/clips/0313-2/29520/20.jpg\nSaving to: dataset/masks/0313-2/29520\nReading from: dataset/clips/0313-2/195/20.jpg\nSaving to: dataset/masks/0313-2/195\nReading from: dataset/clips/0313-2/40060/20.jpg\nSaving to: dataset/masks/0313-2/40060\nReading from: dataset/clips/0313-2/15480/20.jpg\nSaving to: dataset/masks/0313-2/15480\nReading from: dataset/clips/0313-2/815/20.jpg\nSaving to: dataset/masks/0313-2/815\nReading from: dataset/clips/0313-2/38020/20.jpg\nSaving to: dataset/masks/0313-2/38020\nReading from: dataset/clips/0313-2/17220/20.jpg\nSaving to: dataset/masks/0313-2/17220\nReading from: dataset/clips/0313-2/1590/20.jpg\nSaving to: dataset/masks/0313-2/1590\nReading from: dataset/clips/0313-2/22380/20.jpg\nSaving to: dataset/masks/0313-2/22380\nReading from: dataset/clips/0313-2/33020/20.jpg\nSaving to: dataset/masks/0313-2/33020\nReading from: dataset/clips/0313-2/40860/20.jpg\nSaving to: dataset/masks/0313-2/40860\nReading from: dataset/clips/0313-2/9720/20.jpg\nSaving to: dataset/masks/0313-2/9720\nReading from: dataset/clips/0313-2/58560/20.jpg\nSaving to: dataset/masks/0313-2/58560\nReading from: dataset/clips/0313-2/16440/20.jpg\nSaving to: dataset/masks/0313-2/16440\nReading from: dataset/clips/0313-2/33080/20.jpg\nSaving to: dataset/masks/0313-2/33080\nReading from: dataset/clips/0313-2/28200/20.jpg\nSaving to: dataset/masks/0313-2/28200\nReading from: dataset/clips/0313-2/9780/20.jpg\nSaving to: dataset/masks/0313-2/9780\nReading from: dataset/clips/0313-2/1165/20.jpg\nSaving to: dataset/masks/0313-2/1165\nReading from: dataset/clips/0313-2/25560/20.jpg\nSaving to: dataset/masks/0313-2/25560\nReading from: dataset/clips/0313-2/34700/20.jpg\nSaving to: dataset/masks/0313-2/34700\nReading from: dataset/clips/0313-2/1390/20.jpg\nSaving to: dataset/masks/0313-2/1390\nReading from: dataset/clips/0313-2/27960/20.jpg\nSaving to: dataset/masks/0313-2/27960\nReading from: dataset/clips/0313-2/27060/20.jpg\nSaving to: dataset/masks/0313-2/27060\nReading from: dataset/clips/0313-2/37940/20.jpg\nSaving to: dataset/masks/0313-2/37940\nReading from: dataset/clips/0313-2/37780/20.jpg\nSaving to: dataset/masks/0313-2/37780\nReading from: dataset/clips/0313-2/7920/20.jpg\nSaving to: dataset/masks/0313-2/7920\nReading from: dataset/clips/0313-2/37060/20.jpg\nSaving to: dataset/masks/0313-2/37060\nReading from: dataset/clips/0313-2/19080/20.jpg\nSaving to: dataset/masks/0313-2/19080\nReading from: dataset/clips/0313-2/22620/20.jpg\nSaving to: dataset/masks/0313-2/22620\nReading from: dataset/clips/0313-2/480/20.jpg\nSaving to: dataset/masks/0313-2/480\nReading from: dataset/clips/0313-2/18000/20.jpg\nSaving to: dataset/masks/0313-2/18000\nReading from: dataset/clips/0313-2/390/20.jpg\nSaving to: dataset/masks/0313-2/390\nReading from: dataset/clips/0313-2/13560/20.jpg\nSaving to: dataset/masks/0313-2/13560\nReading from: dataset/clips/0313-2/59580/20.jpg\nSaving to: dataset/masks/0313-2/59580\nReading from: dataset/clips/0313-2/21360/20.jpg\nSaving to: dataset/masks/0313-2/21360\nReading from: dataset/clips/0313-2/33900/20.jpg\nSaving to: dataset/masks/0313-2/33900\nReading from: dataset/clips/0313-2/525/20.jpg\nSaving to: dataset/masks/0313-2/525\nReading from: dataset/clips/0313-2/41500/20.jpg\nSaving to: dataset/masks/0313-2/41500\nReading from: dataset/clips/0313-2/12120/20.jpg\nSaving to: dataset/masks/0313-2/12120\nReading from: dataset/clips/0313-2/28440/20.jpg\nSaving to: dataset/masks/0313-2/28440\nReading from: dataset/clips/0313-2/27540/20.jpg\nSaving to: dataset/masks/0313-2/27540\nReading from: dataset/clips/0313-2/40240/20.jpg\nSaving to: dataset/masks/0313-2/40240\nReading from: dataset/clips/0313-2/675/20.jpg\nSaving to: dataset/masks/0313-2/675\nReading from: dataset/clips/0313-2/740/20.jpg\nSaving to: dataset/masks/0313-2/740\nReading from: dataset/clips/0313-2/36540/20.jpg\nSaving to: dataset/masks/0313-2/36540\nReading from: dataset/clips/0313-2/41020/20.jpg\nSaving to: dataset/masks/0313-2/41020\nReading from: dataset/clips/0313-2/125/20.jpg\nSaving to: dataset/masks/0313-2/125\nReading from: dataset/clips/0313-2/7200/20.jpg\nSaving to: dataset/masks/0313-2/7200\nReading from: dataset/clips/0313-2/1315/20.jpg\nSaving to: dataset/masks/0313-2/1315\nReading from: dataset/clips/0313-2/19380/20.jpg\nSaving to: dataset/masks/0313-2/19380\nReading from: dataset/clips/0313-2/18360/20.jpg\nSaving to: dataset/masks/0313-2/18360\nReading from: dataset/clips/0313-2/11880/20.jpg\nSaving to: dataset/masks/0313-2/11880\nReading from: dataset/clips/0313-2/35460/20.jpg\nSaving to: dataset/masks/0313-2/35460\nReading from: dataset/clips/0313-2/11640/20.jpg\nSaving to: dataset/masks/0313-2/11640\nReading from: dataset/clips/0313-2/1300/20.jpg\nSaving to: dataset/masks/0313-2/1300\nReading from: dataset/clips/0313-2/33820/20.jpg\nSaving to: dataset/masks/0313-2/33820\nReading from: dataset/clips/0313-2/62760/20.jpg\nSaving to: dataset/masks/0313-2/62760\nReading from: dataset/clips/0313-2/21240/20.jpg\nSaving to: dataset/masks/0313-2/21240\nReading from: dataset/clips/0313-2/2220/20.jpg\nSaving to: dataset/masks/0313-2/2220\nReading from: dataset/clips/0313-2/15720/20.jpg\nSaving to: dataset/masks/0313-2/15720\nReading from: dataset/clips/0313-2/32840/20.jpg\nSaving to: dataset/masks/0313-2/32840\nReading from: dataset/clips/0313-2/40640/20.jpg\nSaving to: dataset/masks/0313-2/40640\nReading from: dataset/clips/0313-2/520/20.jpg\nSaving to: dataset/masks/0313-2/520\nReading from: dataset/clips/0313-2/61860/20.jpg\nSaving to: dataset/masks/0313-2/61860\nReading from: dataset/clips/0313-2/62880/20.jpg\nSaving to: dataset/masks/0313-2/62880\nReading from: dataset/clips/0313-2/1230/20.jpg\nSaving to: dataset/masks/0313-2/1230\nReading from: dataset/clips/0313-2/85/20.jpg\nSaving to: dataset/masks/0313-2/85\nReading from: dataset/clips/0313-2/60780/20.jpg\nSaving to: dataset/masks/0313-2/60780\nReading from: dataset/clips/0313-2/12600/20.jpg\nSaving to: dataset/masks/0313-2/12600\nReading from: dataset/clips/0313-2/38340/20.jpg\nSaving to: dataset/masks/0313-2/38340\nReading from: dataset/clips/0313-2/40420/20.jpg\nSaving to: dataset/masks/0313-2/40420\nReading from: dataset/clips/0313-2/11100/20.jpg\nSaving to: dataset/masks/0313-2/11100\nReading from: dataset/clips/0313-2/38740/20.jpg\nSaving to: dataset/masks/0313-2/38740\nReading from: dataset/clips/0313-2/38760/20.jpg\nSaving to: dataset/masks/0313-2/38760\nReading from: dataset/clips/0313-2/735/20.jpg\nSaving to: dataset/masks/0313-2/735\nReading from: dataset/clips/0313-2/1085/20.jpg\nSaving to: dataset/masks/0313-2/1085\nReading from: dataset/clips/0313-2/42220/20.jpg\nSaving to: dataset/masks/0313-2/42220\nReading from: dataset/clips/0313-2/36720/20.jpg\nSaving to: dataset/masks/0313-2/36720\nReading from: dataset/clips/0313-2/15840/20.jpg\nSaving to: dataset/masks/0313-2/15840\nReading from: dataset/clips/0313-2/35640/20.jpg\nSaving to: dataset/masks/0313-2/35640\nReading from: dataset/clips/0313-2/30120/20.jpg\nSaving to: dataset/masks/0313-2/30120\nReading from: dataset/clips/0313-2/35560/20.jpg\nSaving to: dataset/masks/0313-2/35560\nReading from: dataset/clips/0313-2/41800/20.jpg\nSaving to: dataset/masks/0313-2/41800\nReading from: dataset/clips/0313-2/295/20.jpg\nSaving to: dataset/masks/0313-2/295\nReading from: dataset/clips/0313-2/18720/20.jpg\nSaving to: dataset/masks/0313-2/18720\nReading from: dataset/clips/0313-2/41040/20.jpg\nSaving to: dataset/masks/0313-2/41040\nReading from: dataset/clips/0313-2/3540/20.jpg\nSaving to: dataset/masks/0313-2/3540\nReading from: dataset/clips/0313-2/560/20.jpg\nSaving to: dataset/masks/0313-2/560\nReading from: dataset/clips/0313-2/1260/20.jpg\nSaving to: dataset/masks/0313-2/1260\nReading from: dataset/clips/0313-2/1190/20.jpg\nSaving to: dataset/masks/0313-2/1190\nReading from: dataset/clips/0313-2/40560/20.jpg\nSaving to: dataset/masks/0313-2/40560\nReading from: dataset/clips/0313-2/63000/20.jpg\nSaving to: dataset/masks/0313-2/63000\nReading from: dataset/clips/0313-2/24060/20.jpg\nSaving to: dataset/masks/0313-2/24060\nReading from: dataset/clips/0313-2/1660/20.jpg\nSaving to: dataset/masks/0313-2/1660\nReading from: dataset/clips/0313-2/35960/20.jpg\nSaving to: dataset/masks/0313-2/35960\nReading from: dataset/clips/0313-2/280/20.jpg\nSaving to: dataset/masks/0313-2/280\nReading from: dataset/clips/0313-2/24660/20.jpg\nSaving to: dataset/masks/0313-2/24660\nReading from: dataset/clips/0313-2/1600/20.jpg\nSaving to: dataset/masks/0313-2/1600\nReading from: dataset/clips/0313-2/34860/20.jpg\nSaving to: dataset/masks/0313-2/34860\nReading from: dataset/clips/0313-2/31860/20.jpg\nSaving to: dataset/masks/0313-2/31860\nReading from: dataset/clips/0313-2/2760/20.jpg\nSaving to: dataset/masks/0313-2/2760\nReading from: dataset/clips/0313-2/40800/20.jpg\nSaving to: dataset/masks/0313-2/40800\nReading from: dataset/clips/0313-2/39320/20.jpg\nSaving to: dataset/masks/0313-2/39320\nReading from: dataset/clips/0313-2/5100/20.jpg\nSaving to: dataset/masks/0313-2/5100\nReading from: dataset/clips/0313-2/34280/20.jpg\nSaving to: dataset/masks/0313-2/34280\nReading from: dataset/clips/0313-2/42580/20.jpg\nSaving to: dataset/masks/0313-2/42580\nReading from: dataset/clips/0313-2/5520/20.jpg\nSaving to: dataset/masks/0313-2/5520\nReading from: dataset/clips/0313-2/900/20.jpg\nSaving to: dataset/masks/0313-2/900\nReading from: dataset/clips/0313-2/1115/20.jpg\nSaving to: dataset/masks/0313-2/1115\nReading from: dataset/clips/0313-2/42420/20.jpg\nSaving to: dataset/masks/0313-2/42420\nReading from: dataset/clips/0313-2/42500/20.jpg\nSaving to: dataset/masks/0313-2/42500\nReading from: dataset/clips/0313-2/120/20.jpg\nSaving to: dataset/masks/0313-2/120\nReading from: dataset/clips/0313-2/35800/20.jpg\nSaving to: dataset/masks/0313-2/35800\nReading from: dataset/clips/0313-2/42680/20.jpg\nSaving to: dataset/masks/0313-2/42680\nReading from: dataset/clips/0313-2/7140/20.jpg\nSaving to: dataset/masks/0313-2/7140\nReading from: dataset/clips/0313-2/35240/20.jpg\nSaving to: dataset/masks/0313-2/35240\nReading from: dataset/clips/0313-2/39940/20.jpg\nSaving to: dataset/masks/0313-2/39940\nReading from: dataset/clips/0313-2/24900/20.jpg\nSaving to: dataset/masks/0313-2/24900\nReading from: dataset/clips/0313-2/14100/20.jpg\nSaving to: dataset/masks/0313-2/14100\nReading from: dataset/clips/0313-2/1240/20.jpg\nSaving to: dataset/masks/0313-2/1240\nReading from: dataset/clips/0313-2/39480/20.jpg\nSaving to: dataset/masks/0313-2/39480\nReading from: dataset/clips/0313-2/25920/20.jpg\nSaving to: dataset/masks/0313-2/25920\nReading from: dataset/clips/0313-2/23940/20.jpg\nSaving to: dataset/masks/0313-2/23940\nReading from: dataset/clips/0313-2/15000/20.jpg\nSaving to: dataset/masks/0313-2/15000\nReading from: dataset/clips/0313-2/42020/20.jpg\nSaving to: dataset/masks/0313-2/42020\nReading from: dataset/clips/0313-2/32400/20.jpg\nSaving to: dataset/masks/0313-2/32400\nReading from: dataset/clips/0313-2/32860/20.jpg\nSaving to: dataset/masks/0313-2/32860\nReading from: dataset/clips/0313-2/460/20.jpg\nSaving to: dataset/masks/0313-2/460\nReading from: dataset/clips/0313-2/36240/20.jpg\nSaving to: dataset/masks/0313-2/36240\nReading from: dataset/clips/0313-2/38580/20.jpg\nSaving to: dataset/masks/0313-2/38580\nReading from: dataset/clips/0313-2/36000/20.jpg\nSaving to: dataset/masks/0313-2/36000\nReading from: dataset/clips/0313-2/38540/20.jpg\nSaving to: dataset/masks/0313-2/38540\nReading from: dataset/clips/0313-2/36320/20.jpg\nSaving to: dataset/masks/0313-2/36320\nReading from: dataset/clips/0313-2/1365/20.jpg\nSaving to: dataset/masks/0313-2/1365\nReading from: dataset/clips/0313-2/21480/20.jpg\nSaving to: dataset/masks/0313-2/21480\nReading from: dataset/clips/0313-2/37400/20.jpg\nSaving to: dataset/masks/0313-2/37400\nReading from: dataset/clips/0313-2/985/20.jpg\nSaving to: dataset/masks/0313-2/985\nReading from: dataset/clips/0313-2/32440/20.jpg\nSaving to: dataset/masks/0313-2/32440\nReading from: dataset/clips/0313-2/41060/20.jpg\nSaving to: dataset/masks/0313-2/41060\nReading from: dataset/clips/0313-2/1515/20.jpg\nSaving to: dataset/masks/0313-2/1515\nReading from: dataset/clips/0313-2/37620/20.jpg\nSaving to: dataset/masks/0313-2/37620\nReading from: dataset/clips/0313-2/8400/20.jpg\nSaving to: dataset/masks/0313-2/8400\nReading from: dataset/clips/0313-2/22200/20.jpg\nSaving to: dataset/masks/0313-2/22200\nReading from: dataset/clips/0313-2/34360/20.jpg\nSaving to: dataset/masks/0313-2/34360\nReading from: dataset/clips/0313-2/32100/20.jpg\nSaving to: dataset/masks/0313-2/32100\nReading from: dataset/clips/0313-2/41340/20.jpg\nSaving to: dataset/masks/0313-2/41340\nReading from: dataset/clips/0313-2/39680/20.jpg\nSaving to: dataset/masks/0313-2/39680\nReading from: dataset/clips/0313-2/3780/20.jpg\nSaving to: dataset/masks/0313-2/3780\nReading from: dataset/clips/0313-2/38860/20.jpg\nSaving to: dataset/masks/0313-2/38860\nReading from: dataset/clips/0313-2/29040/20.jpg\nSaving to: dataset/masks/0313-2/29040\nReading from: dataset/clips/0313-2/38440/20.jpg\nSaving to: dataset/masks/0313-2/38440\nReading from: dataset/clips/0313-2/260/20.jpg\nSaving to: dataset/masks/0313-2/260\nReading from: dataset/clips/0313-2/26940/20.jpg\nSaving to: dataset/masks/0313-2/26940\nReading from: dataset/clips/0313-2/63960/20.jpg\nSaving to: dataset/masks/0313-2/63960\nReading from: dataset/clips/0313-2/11460/20.jpg\nSaving to: dataset/masks/0313-2/11460\nReading from: dataset/clips/0313-2/40460/20.jpg\nSaving to: dataset/masks/0313-2/40460\nReading from: dataset/clips/0313-2/34040/20.jpg\nSaving to: dataset/masks/0313-2/34040\nReading from: dataset/clips/0313-2/8640/20.jpg\nSaving to: dataset/masks/0313-2/8640\nReading from: dataset/clips/0313-2/1060/20.jpg\nSaving to: dataset/masks/0313-2/1060\nReading from: dataset/clips/0313-2/1015/20.jpg\nSaving to: dataset/masks/0313-2/1015\nReading from: dataset/clips/0313-2/37180/20.jpg\nSaving to: dataset/masks/0313-2/37180\nReading from: dataset/clips/0313-2/64140/20.jpg\nSaving to: dataset/masks/0313-2/64140\nReading from: dataset/clips/0313-2/18480/20.jpg\nSaving to: dataset/masks/0313-2/18480\nReading from: dataset/clips/0313-2/1265/20.jpg\nSaving to: dataset/masks/0313-2/1265\nReading from: dataset/clips/0313-2/37020/20.jpg\nSaving to: dataset/masks/0313-2/37020\nReading from: dataset/clips/0313-2/41660/20.jpg\nSaving to: dataset/masks/0313-2/41660\nReading from: dataset/clips/0313-2/64200/20.jpg\nSaving to: dataset/masks/0313-2/64200\nReading from: dataset/clips/0313-2/10380/20.jpg\nSaving to: dataset/masks/0313-2/10380\nReading from: dataset/clips/0313-2/40300/20.jpg\nSaving to: dataset/masks/0313-2/40300\nReading from: dataset/clips/0313-2/37280/20.jpg\nSaving to: dataset/masks/0313-2/37280\nReading from: dataset/clips/0313-2/34760/20.jpg\nSaving to: dataset/masks/0313-2/34760\nReading from: dataset/clips/0313-2/1800/20.jpg\nSaving to: dataset/masks/0313-2/1800\nReading from: dataset/clips/0313-2/230/20.jpg\nSaving to: dataset/masks/0313-2/230\nReading from: dataset/clips/0313-2/39580/20.jpg\nSaving to: dataset/masks/0313-2/39580\nReading from: dataset/clips/0313-2/1555/20.jpg\nSaving to: dataset/masks/0313-2/1555\nReading from: dataset/clips/0313-2/20340/20.jpg\nSaving to: dataset/masks/0313-2/20340\nReading from: dataset/clips/0313-2/450/20.jpg\nSaving to: dataset/masks/0313-2/450\nReading from: dataset/clips/0313-2/33560/20.jpg\nSaving to: dataset/masks/0313-2/33560\nReading from: dataset/clips/0313-2/10800/20.jpg\nSaving to: dataset/masks/0313-2/10800\nReading from: dataset/clips/0313-2/35600/20.jpg\nSaving to: dataset/masks/0313-2/35600\nReading from: dataset/clips/0313-2/37420/20.jpg\nSaving to: dataset/masks/0313-2/37420\nReading from: dataset/clips/0313-2/2460/20.jpg\nSaving to: dataset/masks/0313-2/2460\nReading from: dataset/clips/0313-2/13500/20.jpg\nSaving to: dataset/masks/0313-2/13500\nReading from: dataset/clips/0313-2/145/20.jpg\nSaving to: dataset/masks/0313-2/145\nReading from: dataset/clips/0313-2/18060/20.jpg\nSaving to: dataset/masks/0313-2/18060\nReading from: dataset/clips/0313-2/34940/20.jpg\nSaving to: dataset/masks/0313-2/34940\nReading from: dataset/clips/0313-2/35020/20.jpg\nSaving to: dataset/masks/0313-2/35020\nReading from: dataset/clips/0313-2/61920/20.jpg\nSaving to: dataset/masks/0313-2/61920\nReading from: dataset/clips/0313-2/38160/20.jpg\nSaving to: dataset/masks/0313-2/38160\nReading from: dataset/clips/0313-2/32220/20.jpg\nSaving to: dataset/masks/0313-2/32220\nReading from: dataset/clips/0313-2/33960/20.jpg\nSaving to: dataset/masks/0313-2/33960\nReading from: dataset/clips/0313-2/915/20.jpg\nSaving to: dataset/masks/0313-2/915\nReading from: dataset/clips/0313-2/60840/20.jpg\nSaving to: dataset/masks/0313-2/60840\nReading from: dataset/clips/0313-2/58860/20.jpg\nSaving to: dataset/masks/0313-2/58860\nReading from: dataset/clips/0313-2/5700/20.jpg\nSaving to: dataset/masks/0313-2/5700\nReading from: dataset/clips/0313-2/64440/20.jpg\nSaving to: dataset/masks/0313-2/64440\nReading from: dataset/clips/0313-2/13860/20.jpg\nSaving to: dataset/masks/0313-2/13860\nReading from: dataset/clips/0313-2/35160/20.jpg\nSaving to: dataset/masks/0313-2/35160\nReading from: dataset/clips/0313-2/30000/20.jpg\nSaving to: dataset/masks/0313-2/30000\nReading from: dataset/clips/0313-2/32460/20.jpg\nSaving to: dataset/masks/0313-2/32460\nReading from: dataset/clips/0313-2/395/20.jpg\nSaving to: dataset/masks/0313-2/395\nReading from: dataset/clips/0313-2/40040/20.jpg\nSaving to: dataset/masks/0313-2/40040\nReading from: dataset/clips/0313-2/58920/20.jpg\nSaving to: dataset/masks/0313-2/58920\nReading from: dataset/clips/0313-2/40/20.jpg\nSaving to: dataset/masks/0313-2/40\nReading from: dataset/clips/0313-2/1455/20.jpg\nSaving to: dataset/masks/0313-2/1455\nReading from: dataset/clips/0313-2/34840/20.jpg\nSaving to: dataset/masks/0313-2/34840\nReading from: dataset/clips/0313-2/41520/20.jpg\nSaving to: dataset/masks/0313-2/41520\nReading from: dataset/clips/0313-2/1675/20.jpg\nSaving to: dataset/masks/0313-2/1675\nReading from: dataset/clips/0313-2/38500/20.jpg\nSaving to: dataset/masks/0313-2/38500\nReading from: dataset/clips/0313-2/10620/20.jpg\nSaving to: dataset/masks/0313-2/10620\nReading from: dataset/clips/0313-2/42180/20.jpg\nSaving to: dataset/masks/0313-2/42180\nReading from: dataset/clips/0313-2/32740/20.jpg\nSaving to: dataset/masks/0313-2/32740\nReading from: dataset/clips/0313-2/110/20.jpg\nSaving to: dataset/masks/0313-2/110\nReading from: dataset/clips/0313-2/6240/20.jpg\nSaving to: dataset/masks/0313-2/6240\nReading from: dataset/clips/0313-2/1215/20.jpg\nSaving to: dataset/masks/0313-2/1215\nReading from: dataset/clips/0313-2/42240/20.jpg\nSaving to: dataset/masks/0313-2/42240\nReading from: dataset/clips/0313-2/38720/20.jpg\nSaving to: dataset/masks/0313-2/38720\nReading from: dataset/clips/0313-2/38040/20.jpg\nSaving to: dataset/masks/0313-2/38040\nReading from: dataset/clips/0313-2/5640/20.jpg\nSaving to: dataset/masks/0313-2/5640\nReading from: dataset/clips/0313-2/62580/20.jpg\nSaving to: dataset/masks/0313-2/62580\nReading from: dataset/clips/0313-2/31380/20.jpg\nSaving to: dataset/masks/0313-2/31380\nReading from: dataset/clips/0313-2/4560/20.jpg\nSaving to: dataset/masks/0313-2/4560\nReading from: dataset/clips/0313-2/38980/20.jpg\nSaving to: dataset/masks/0313-2/38980\nReading from: dataset/clips/0313-2/35700/20.jpg\nSaving to: dataset/masks/0313-2/35700\nReading from: dataset/clips/0313-2/80/20.jpg\nSaving to: dataset/masks/0313-2/80\nReading from: dataset/clips/0313-2/365/20.jpg\nSaving to: dataset/masks/0313-2/365\nReading from: dataset/clips/0313-2/595/20.jpg\nSaving to: dataset/masks/0313-2/595\nReading from: dataset/clips/0313-2/39300/20.jpg\nSaving to: dataset/masks/0313-2/39300\nReading from: dataset/clips/0313-2/730/20.jpg\nSaving to: dataset/masks/0313-2/730\nReading from: dataset/clips/0313-2/39060/20.jpg\nSaving to: dataset/masks/0313-2/39060\nReading from: dataset/clips/0313-2/41380/20.jpg\nSaving to: dataset/masks/0313-2/41380\nReading from: dataset/clips/0313-2/38200/20.jpg\nSaving to: dataset/masks/0313-2/38200\nReading from: dataset/clips/0313-2/34320/20.jpg\nSaving to: dataset/masks/0313-2/34320\nReading from: dataset/clips/0313-2/40200/20.jpg\nSaving to: dataset/masks/0313-2/40200\nReading from: dataset/clips/0313-2/12420/20.jpg\nSaving to: dataset/masks/0313-2/12420\nReading from: dataset/clips/0313-2/33200/20.jpg\nSaving to: dataset/masks/0313-2/33200\nReading from: dataset/clips/0313-2/470/20.jpg\nSaving to: dataset/masks/0313-2/470\nReading from: dataset/clips/0313-2/28920/20.jpg\nSaving to: dataset/masks/0313-2/28920\nReading from: dataset/clips/0313-2/26100/20.jpg\nSaving to: dataset/masks/0313-2/26100\nReading from: dataset/clips/0313-2/6420/20.jpg\nSaving to: dataset/masks/0313-2/6420\nReading from: dataset/clips/0313-2/25260/20.jpg\nSaving to: dataset/masks/0313-2/25260\nReading from: dataset/clips/0313-2/17460/20.jpg\nSaving to: dataset/masks/0313-2/17460\nReading from: dataset/clips/0313-2/40440/20.jpg\nSaving to: dataset/masks/0313-2/40440\nReading from: dataset/clips/0313-2/61080/20.jpg\nSaving to: dataset/masks/0313-2/61080\nReading from: dataset/clips/0313-2/4140/20.jpg\nSaving to: dataset/masks/0313-2/4140\nReading from: dataset/clips/0313-2/35440/20.jpg\nSaving to: dataset/masks/0313-2/35440\nReading from: dataset/clips/0313-2/61560/20.jpg\nSaving to: dataset/masks/0313-2/61560\nReading from: dataset/clips/0313-2/1090/20.jpg\nSaving to: dataset/masks/0313-2/1090\nReading from: dataset/clips/0313-2/61680/20.jpg\nSaving to: dataset/masks/0313-2/61680\nReading from: dataset/clips/0313-2/5340/20.jpg\nSaving to: dataset/masks/0313-2/5340\nReading from: dataset/clips/0313-2/9000/20.jpg\nSaving to: dataset/masks/0313-2/9000\nReading from: dataset/clips/0313-2/1490/20.jpg\nSaving to: dataset/masks/0313-2/1490\nReading from: dataset/clips/0313-2/665/20.jpg\nSaving to: dataset/masks/0313-2/665\nReading from: dataset/clips/0313-2/1775/20.jpg\nSaving to: dataset/masks/0313-2/1775\nReading from: dataset/clips/0313-2/42200/20.jpg\nSaving to: dataset/masks/0313-2/42200\nReading from: dataset/clips/0313-2/27840/20.jpg\nSaving to: dataset/masks/0313-2/27840\nReading from: dataset/clips/0313-2/5880/20.jpg\nSaving to: dataset/masks/0313-2/5880\nReading from: dataset/clips/0313-2/58740/20.jpg\nSaving to: dataset/masks/0313-2/58740\nReading from: dataset/clips/0313-2/34900/20.jpg\nSaving to: dataset/masks/0313-2/34900\nReading from: dataset/clips/0313-2/37580/20.jpg\nSaving to: dataset/masks/0313-2/37580\nReading from: dataset/clips/0313-2/34780/20.jpg\nSaving to: dataset/masks/0313-2/34780\nReading from: dataset/clips/0313-2/7560/20.jpg\nSaving to: dataset/masks/0313-2/7560\nReading from: dataset/clips/0313-2/18660/20.jpg\nSaving to: dataset/masks/0313-2/18660\nReading from: dataset/clips/0313-2/17280/20.jpg\nSaving to: dataset/masks/0313-2/17280\nReading from: dataset/clips/0313-2/37960/20.jpg\nSaving to: dataset/masks/0313-2/37960\nReading from: dataset/clips/0313-2/22920/20.jpg\nSaving to: dataset/masks/0313-2/22920\nReading from: dataset/clips/0313-2/22680/20.jpg\nSaving to: dataset/masks/0313-2/22680\nReading from: dataset/clips/0313-2/420/20.jpg\nSaving to: dataset/masks/0313-2/420\nReading from: dataset/clips/0313-2/8760/20.jpg\nSaving to: dataset/masks/0313-2/8760\nReading from: dataset/clips/0313-2/37140/20.jpg\nSaving to: dataset/masks/0313-2/37140\nReading from: dataset/clips/0313-2/7740/20.jpg\nSaving to: dataset/masks/0313-2/7740\nReading from: dataset/clips/0313-2/34580/20.jpg\nSaving to: dataset/masks/0313-2/34580\nReading from: dataset/clips/0313-2/34260/20.jpg\nSaving to: dataset/masks/0313-2/34260\nReading from: dataset/clips/0313-2/21600/20.jpg\nSaving to: dataset/masks/0313-2/21600\nReading from: dataset/clips/0313-2/1220/20.jpg\nSaving to: dataset/masks/0313-2/1220\nReading from: dataset/clips/0313-2/33320/20.jpg\nSaving to: dataset/masks/0313-2/33320\nReading from: dataset/clips/0313-2/9420/20.jpg\nSaving to: dataset/masks/0313-2/9420\nReading from: dataset/clips/0313-2/1695/20.jpg\nSaving to: dataset/masks/0313-2/1695\nReading from: dataset/clips/0313-2/1030/20.jpg\nSaving to: dataset/masks/0313-2/1030\nReading from: dataset/clips/0313-2/39260/20.jpg\nSaving to: dataset/masks/0313-2/39260\nReading from: dataset/clips/0313-2/33300/20.jpg\nSaving to: dataset/masks/0313-2/33300\nReading from: dataset/clips/0313-2/60480/20.jpg\nSaving to: dataset/masks/0313-2/60480\nReading from: dataset/clips/0313-2/34120/20.jpg\nSaving to: dataset/masks/0313-2/34120\nReading from: dataset/clips/0313-2/6600/20.jpg\nSaving to: dataset/masks/0313-2/6600\nReading from: dataset/clips/0313-2/40720/20.jpg\nSaving to: dataset/masks/0313-2/40720\nReading from: dataset/clips/0313-2/24780/20.jpg\nSaving to: dataset/masks/0313-2/24780\nReading from: dataset/clips/0313-2/38680/20.jpg\nSaving to: dataset/masks/0313-2/38680\nReading from: dataset/clips/0313-2/1730/20.jpg\nSaving to: dataset/masks/0313-2/1730\nReading from: dataset/clips/0313-2/380/20.jpg\nSaving to: dataset/masks/0313-2/380\nReading from: dataset/clips/0313-2/220/20.jpg\nSaving to: dataset/masks/0313-2/220\nReading from: dataset/clips/0313-2/22440/20.jpg\nSaving to: dataset/masks/0313-2/22440\nReading from: dataset/clips/0313-2/61740/20.jpg\nSaving to: dataset/masks/0313-2/61740\nReading from: dataset/clips/0313-2/34880/20.jpg\nSaving to: dataset/masks/0313-2/34880\nReading from: dataset/clips/0313-2/845/20.jpg\nSaving to: dataset/masks/0313-2/845\nReading from: dataset/clips/0313-2/33340/20.jpg\nSaving to: dataset/masks/0313-2/33340\nReading from: dataset/clips/0313-2/1395/20.jpg\nSaving to: dataset/masks/0313-2/1395\nReading from: dataset/clips/0313-2/30240/20.jpg\nSaving to: dataset/masks/0313-2/30240\nReading from: dataset/clips/0313-2/37600/20.jpg\nSaving to: dataset/masks/0313-2/37600\nReading from: dataset/clips/0313-2/2280/20.jpg\nSaving to: dataset/masks/0313-2/2280\nReading from: dataset/clips/0313-2/60660/20.jpg\nSaving to: dataset/masks/0313-2/60660\nReading from: dataset/clips/0313-2/42120/20.jpg\nSaving to: dataset/masks/0313-2/42120\nReading from: dataset/clips/0313-2/620/20.jpg\nSaving to: dataset/masks/0313-2/620\nReading from: dataset/clips/0313-2/10320/20.jpg\nSaving to: dataset/masks/0313-2/10320\nReading from: dataset/clips/0313-2/1645/20.jpg\nSaving to: dataset/masks/0313-2/1645\nReading from: dataset/clips/0313-2/565/20.jpg\nSaving to: dataset/masks/0313-2/565\nReading from: dataset/clips/0313-2/12240/20.jpg\nSaving to: dataset/masks/0313-2/12240\nReading from: dataset/clips/0313-2/1035/20.jpg\nSaving to: dataset/masks/0313-2/1035\nReading from: dataset/clips/0313-2/33720/20.jpg\nSaving to: dataset/masks/0313-2/33720\nReading from: dataset/clips/0313-2/3960/20.jpg\nSaving to: dataset/masks/0313-2/3960\nReading from: dataset/clips/0313-2/33880/20.jpg\nSaving to: dataset/masks/0313-2/33880\nReading from: dataset/clips/0313-2/2700/20.jpg\nSaving to: dataset/masks/0313-2/2700\nReading from: dataset/clips/0313-2/245/20.jpg\nSaving to: dataset/masks/0313-2/245\nReading from: dataset/clips/0313-2/1195/20.jpg\nSaving to: dataset/masks/0313-2/1195\nReading from: dataset/clips/0313-2/1640/20.jpg\nSaving to: dataset/masks/0313-2/1640\nReading from: dataset/clips/0313-2/39560/20.jpg\nSaving to: dataset/masks/0313-2/39560\nReading from: dataset/clips/0313-2/36760/20.jpg\nSaving to: dataset/masks/0313-2/36760\nReading from: dataset/clips/0313-2/16620/20.jpg\nSaving to: dataset/masks/0313-2/16620\nReading from: dataset/clips/0313-2/1635/20.jpg\nSaving to: dataset/masks/0313-2/1635\nReading from: dataset/clips/0313-2/24840/20.jpg\nSaving to: dataset/masks/0313-2/24840\nReading from: dataset/clips/0313-2/8580/20.jpg\nSaving to: dataset/masks/0313-2/8580\nReading from: dataset/clips/0313-2/41680/20.jpg\nSaving to: dataset/masks/0313-2/41680\nReading from: dataset/clips/0313-2/345/20.jpg\nSaving to: dataset/masks/0313-2/345\nReading from: dataset/clips/0313-2/1650/20.jpg\nSaving to: dataset/masks/0313-2/1650\nReading from: dataset/clips/0313-2/1575/20.jpg\nSaving to: dataset/masks/0313-2/1575\nReading from: dataset/clips/0313-2/39540/20.jpg\nSaving to: dataset/masks/0313-2/39540\nReading from: dataset/clips/0313-2/63540/20.jpg\nSaving to: dataset/masks/0313-2/63540\nReading from: dataset/clips/0313-2/29400/20.jpg\nSaving to: dataset/masks/0313-2/29400\nReading from: dataset/clips/0313-2/42000/20.jpg\nSaving to: dataset/masks/0313-2/42000\nReading from: dataset/clips/0313-2/1065/20.jpg\nSaving to: dataset/masks/0313-2/1065\nReading from: dataset/clips/0313-2/27720/20.jpg\nSaving to: dataset/masks/0313-2/27720\nReading from: dataset/clips/0313-2/35320/20.jpg\nSaving to: dataset/masks/0313-2/35320\nReading from: dataset/clips/0313-2/25020/20.jpg\nSaving to: dataset/masks/0313-2/25020\nReading from: dataset/clips/0313-2/1560/20.jpg\nSaving to: dataset/masks/0313-2/1560\nReading from: dataset/clips/0313-2/895/20.jpg\nSaving to: dataset/masks/0313-2/895\nReading from: dataset/clips/0313-2/35940/20.jpg\nSaving to: dataset/masks/0313-2/35940\nReading from: dataset/clips/0313-2/41560/20.jpg\nSaving to: dataset/masks/0313-2/41560\nReading from: dataset/clips/0313-2/20580/20.jpg\nSaving to: dataset/masks/0313-2/20580\nReading from: dataset/clips/0313-2/405/20.jpg\nSaving to: dataset/masks/0313-2/405\nReading from: dataset/clips/0313-2/40020/20.jpg\nSaving to: dataset/masks/0313-2/40020\nReading from: dataset/clips/0313-2/41220/20.jpg\nSaving to: dataset/masks/0313-2/41220\nReading from: dataset/clips/0313-2/7980/20.jpg\nSaving to: dataset/masks/0313-2/7980\nReading from: dataset/clips/0313-2/29280/20.jpg\nSaving to: dataset/masks/0313-2/29280\nReading from: dataset/clips/0313-2/60180/20.jpg\nSaving to: dataset/masks/0313-2/60180\nReading from: dataset/clips/0313-2/11820/20.jpg\nSaving to: dataset/masks/0313-2/11820\nReading from: dataset/clips/0313-2/29340/20.jpg\nSaving to: dataset/masks/0313-2/29340\nReading from: dataset/clips/0313-2/31200/20.jpg\nSaving to: dataset/masks/0313-2/31200\nReading from: dataset/clips/0313-2/25380/20.jpg\nSaving to: dataset/masks/0313-2/25380\nReading from: dataset/clips/0313-2/37520/20.jpg\nSaving to: dataset/masks/0313-2/37520\nReading from: dataset/clips/0313-2/15600/20.jpg\nSaving to: dataset/masks/0313-2/15600\nReading from: dataset/clips/0313-2/38920/20.jpg\nSaving to: dataset/masks/0313-2/38920\nReading from: dataset/clips/0313-2/17940/20.jpg\nSaving to: dataset/masks/0313-2/17940\nReading from: dataset/clips/0313-2/36440/20.jpg\nSaving to: dataset/masks/0313-2/36440\nReading from: dataset/clips/0313-2/40680/20.jpg\nSaving to: dataset/masks/0313-2/40680\nReading from: dataset/clips/0313-2/7380/20.jpg\nSaving to: dataset/masks/0313-2/7380\nReading from: dataset/clips/0313-2/580/20.jpg\nSaving to: dataset/masks/0313-2/580\nReading from: dataset/clips/0313-2/59760/20.jpg\nSaving to: dataset/masks/0313-2/59760\nReading from: dataset/clips/0313-2/36980/20.jpg\nSaving to: dataset/masks/0313-2/36980\nReading from: dataset/clips/0313-2/34620/20.jpg\nSaving to: dataset/masks/0313-2/34620\nReading from: dataset/clips/0313-2/1070/20.jpg\nSaving to: dataset/masks/0313-2/1070\nReading from: dataset/clips/0313-2/42080/20.jpg\nSaving to: dataset/masks/0313-2/42080\nReading from: dataset/clips/0313-2/165/20.jpg\nSaving to: dataset/masks/0313-2/165\nReading from: dataset/clips/0313-2/1285/20.jpg\nSaving to: dataset/masks/0313-2/1285\nReading from: dataset/clips/0313-2/42700/20.jpg\nSaving to: dataset/masks/0313-2/42700\nReading from: dataset/clips/0313-2/590/20.jpg\nSaving to: dataset/masks/0313-2/590\nReading from: dataset/clips/0313-2/18840/20.jpg\nSaving to: dataset/masks/0313-2/18840\nReading from: dataset/clips/0313-2/32820/20.jpg\nSaving to: dataset/masks/0313-2/32820\nReading from: dataset/clips/0313-2/950/20.jpg\nSaving to: dataset/masks/0313-2/950\nReading from: dataset/clips/0313-2/26760/20.jpg\nSaving to: dataset/masks/0313-2/26760\nReading from: dataset/clips/0313-2/1330/20.jpg\nSaving to: dataset/masks/0313-2/1330\nReading from: dataset/clips/0313-2/40700/20.jpg\nSaving to: dataset/masks/0313-2/40700\nReading from: dataset/clips/0313-2/27900/20.jpg\nSaving to: dataset/masks/0313-2/27900\nReading from: dataset/clips/0313-2/60960/20.jpg\nSaving to: dataset/masks/0313-2/60960\nReading from: dataset/clips/0313-2/30360/20.jpg\nSaving to: dataset/masks/0313-2/30360\nReading from: dataset/clips/0313-2/19980/20.jpg\nSaving to: dataset/masks/0313-2/19980\nReading from: dataset/clips/0313-2/17880/20.jpg\nSaving to: dataset/masks/0313-2/17880\nReading from: dataset/clips/0313-2/34200/20.jpg\nSaving to: dataset/masks/0313-2/34200\nReading from: dataset/clips/0313-2/780/20.jpg\nSaving to: dataset/masks/0313-2/780\nReading from: dataset/clips/0313-2/1415/20.jpg\nSaving to: dataset/masks/0313-2/1415\nReading from: dataset/clips/0313-2/32580/20.jpg\nSaving to: dataset/masks/0313-2/32580\nReading from: dataset/clips/0313-2/19140/20.jpg\nSaving to: dataset/masks/0313-2/19140\nReading from: dataset/clips/0313-2/40340/20.jpg\nSaving to: dataset/masks/0313-2/40340\nReading from: dataset/clips/0313-2/32660/20.jpg\nSaving to: dataset/masks/0313-2/32660\nReading from: dataset/clips/0313-2/34060/20.jpg\nSaving to: dataset/masks/0313-2/34060\nReading from: dataset/clips/0313-2/6840/20.jpg\nSaving to: dataset/masks/0313-2/6840\nReading from: dataset/clips/0313-2/33180/20.jpg\nSaving to: dataset/masks/0313-2/33180\nReading from: dataset/clips/0313-2/62700/20.jpg\nSaving to: dataset/masks/0313-2/62700\nReading from: dataset/clips/0313-2/4320/20.jpg\nSaving to: dataset/masks/0313-2/4320\nReading from: dataset/clips/0313-2/340/20.jpg\nSaving to: dataset/masks/0313-2/340\nReading from: dataset/clips/0313-2/1080/20.jpg\nSaving to: dataset/masks/0313-2/1080\nReading from: dataset/clips/0313-2/90/20.jpg\nSaving to: dataset/masks/0313-2/90\nReading from: dataset/clips/0313-2/130/20.jpg\nSaving to: dataset/masks/0313-2/130\nReading from: dataset/clips/0313-2/15660/20.jpg\nSaving to: dataset/masks/0313-2/15660\nReading from: dataset/clips/0313-2/38300/20.jpg\nSaving to: dataset/masks/0313-2/38300\nReading from: dataset/clips/0313-2/14880/20.jpg\nSaving to: dataset/masks/0313-2/14880\nReading from: dataset/clips/0313-2/39620/20.jpg\nSaving to: dataset/masks/0313-2/39620\nReading from: dataset/clips/0313-2/64320/20.jpg\nSaving to: dataset/masks/0313-2/64320\nReading from: dataset/clips/0313-2/41940/20.jpg\nSaving to: dataset/masks/0313-2/41940\nReading from: dataset/clips/0313-2/34520/20.jpg\nSaving to: dataset/masks/0313-2/34520\nReading from: dataset/clips/0313-2/265/20.jpg\nSaving to: dataset/masks/0313-2/265\nReading from: dataset/clips/0313-2/36260/20.jpg\nSaving to: dataset/masks/0313-2/36260\nReading from: dataset/clips/0313-2/1625/20.jpg\nSaving to: dataset/masks/0313-2/1625\nReading from: dataset/clips/0313-2/35280/20.jpg\nSaving to: dataset/masks/0313-2/35280\nReading from: dataset/clips/0313-2/30/20.jpg\nSaving to: dataset/masks/0313-2/30\nReading from: dataset/clips/0313-2/1505/20.jpg\nSaving to: dataset/masks/0313-2/1505\nReading from: dataset/clips/0313-2/23640/20.jpg\nSaving to: dataset/masks/0313-2/23640\nReading from: dataset/clips/0313-2/59940/20.jpg\nSaving to: dataset/masks/0313-2/59940\nReading from: dataset/clips/0313-2/755/20.jpg\nSaving to: dataset/masks/0313-2/755\nReading from: dataset/clips/0313-2/37480/20.jpg\nSaving to: dataset/masks/0313-2/37480\nReading from: dataset/clips/0313-2/305/20.jpg\nSaving to: dataset/masks/0313-2/305\nReading from: dataset/clips/0313-2/29220/20.jpg\nSaving to: dataset/masks/0313-2/29220\nReading from: dataset/clips/0313-2/58980/20.jpg\nSaving to: dataset/masks/0313-2/58980\nReading from: dataset/clips/0313-2/40360/20.jpg\nSaving to: dataset/masks/0313-2/40360\nReading from: dataset/clips/0313-2/37300/20.jpg\nSaving to: dataset/masks/0313-2/37300\nReading from: dataset/clips/0313-2/39240/20.jpg\nSaving to: dataset/masks/0313-2/39240\nReading from: dataset/clips/0313-2/13080/20.jpg\nSaving to: dataset/masks/0313-2/13080\nReading from: dataset/clips/0313-2/26520/20.jpg\nSaving to: dataset/masks/0313-2/26520\nReading from: dataset/clips/0313-2/30960/20.jpg\nSaving to: dataset/masks/0313-2/30960\nReading from: dataset/clips/0313-2/23220/20.jpg\nSaving to: dataset/masks/0313-2/23220\nReading from: dataset/clips/0313-2/35300/20.jpg\nSaving to: dataset/masks/0313-2/35300\nReading from: dataset/clips/0313-2/400/20.jpg\nSaving to: dataset/masks/0313-2/400\n" ], [ "cv2.imwrite('/Users/srinivas/Projects/Lane_Detection/datasets/LaneDetection/train/masks/0313-1/300/1.tiff', mask)", "_____no_output_____" ], [ "mask_img = cv2.imread('20.tiff', cv2.IMREAD_GRAYSCALE)\nmask_img.shape", "_____no_output_____" ], [ "plt.imshow(mask_img)", "_____no_output_____" ], [ "print(np.unique(mask_img))\nprint(np.unique(mask))", "[0 1 2 3 4]\n[0 1 2 3 4]\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "from pykoreaqi import AirKorea", "_____no_output_____" ], [ "aqi = AirKorea()\n\ndata = aqi.get_all_realtime()", "_____no_output_____" ], [ "data", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Navigation\n\n---\n\nIn this notebook, you will learn how to use the Unity ML-Agents environment for the first project of the [Deep Reinforcement Learning Nanodegree](https://www.udacity.com/course/deep-reinforcement-learning-nanodegree--nd893).\n\n### 1. Start the Environment\n\nWe begin by importing some necessary packages. If the code cell below returns an error, please revisit the project instructions to double-check that you have installed [Unity ML-Agents](https://github.com/Unity-Technologies/ml-agents/blob/master/docs/Installation.md) and [NumPy](http://www.numpy.org/).", "_____no_output_____" ] ], [ [ "from unityagents import UnityEnvironment\nimport numpy as np\nfrom collections import deque\nimport torch\nimport matplotlib.pyplot as plt\n%matplotlib inline", "_____no_output_____" ] ], [ [ "Next, we will start the environment! **_Before running the code cell below_**, change the `file_name` parameter to match the location of the Unity environment that you downloaded.\n\n- **Mac**: `\"path/to/Banana.app\"`\n- **Windows** (x86): `\"path/to/Banana_Windows_x86/Banana.exe\"`\n- **Windows** (x86_64): `\"path/to/Banana_Windows_x86_64/Banana.exe\"`\n- **Linux** (x86): `\"path/to/Banana_Linux/Banana.x86\"`\n- **Linux** (x86_64): `\"path/to/Banana_Linux/Banana.x86_64\"`\n- **Linux** (x86, headless): `\"path/to/Banana_Linux_NoVis/Banana.x86\"`\n- **Linux** (x86_64, headless): `\"path/to/Banana_Linux_NoVis/Banana.x86_64\"`\n\nFor instance, if you are using a Mac, then you downloaded `Banana.app`. If this file is in the same folder as the notebook, then the line below should appear as follows:\n```\nenv = UnityEnvironment(file_name=\"Banana.app\")\n```", "_____no_output_____" ] ], [ [ "env = UnityEnvironment(file_name=\"Banana.app\")", "INFO:unityagents:\n'Academy' started successfully!\nUnity Academy name: Academy\n Number of Brains: 1\n Number of External Brains : 1\n Lesson number : 0\n Reset Parameters :\n\t\t\nUnity brain name: BananaBrain\n Number of Visual Observations (per agent): 0\n Vector Observation space type: continuous\n Vector Observation space size (per agent): 37\n Number of stacked Vector Observation: 1\n Vector Action space type: discrete\n Vector Action space size (per agent): 4\n Vector Action descriptions: , , , \n" ] ], [ [ "Environments contain **_brains_** which are responsible for deciding the actions of their associated agents. Here we check for the first brain available, and set it as the default brain we will be controlling from Python.", "_____no_output_____" ] ], [ [ "# get the default brain\nbrain_name = env.brain_names[0]\nbrain = env.brains[brain_name]", "_____no_output_____" ] ], [ [ "### 2. Examine the State and Action Spaces\n\nThe simulation contains a single agent that navigates a large environment. At each time step, it has four actions at its disposal:\n- `0` - walk forward \n- `1` - walk backward\n- `2` - turn left\n- `3` - turn right\n\nThe state space has `37` dimensions and contains the agent's velocity, along with ray-based perception of objects around agent's forward direction. A reward of `+1` is provided for collecting a yellow banana, and a reward of `-1` is provided for collecting a blue banana. \n\nRun the code cell below to print some information about the environment.", "_____no_output_____" ] ], [ [ "# reset the environment\nenv_info = env.reset(train_mode=True)[brain_name]\n\n# number of agents in the environment\nprint('Number of agents:', len(env_info.agents))\n\n# number of actions\naction_size = brain.vector_action_space_size\nprint('Number of actions:', action_size)\n\n# examine the state space \nstate = env_info.vector_observations[0]\nprint('States look like:', state)\nstate_size = len(state)\nprint('States have length:', state_size)", "Number of agents: 1\nNumber of actions: 4\nStates look like: [1. 0. 0. 0. 0.84408134 0.\n 0. 1. 0. 0.0748472 0. 1.\n 0. 0. 0.25755 1. 0. 0.\n 0. 0.74177343 0. 1. 0. 0.\n 0.25854847 0. 0. 1. 0. 0.09355672\n 0. 1. 0. 0. 0.31969345 0.\n 0. ]\nStates have length: 37\n" ] ], [ [ "### 3. Take Random Actions in the Environment\n\nIn the next code cell, you will learn how to use the Python API to control the agent and receive feedback from the environment.\n\nOnce this cell is executed, you will watch the agent's performance, if it selects an action (uniformly) at random with each time step. A window should pop up that allows you to observe the agent, as it moves through the environment. \n\nOf course, as part of the project, you'll have to change the code so that the agent is able to use its experience to gradually choose better actions when interacting with the environment!", "_____no_output_____" ] ], [ [ "env_info = env.reset(train_mode=False)[brain_name] # reset the environment\nstate = env_info.vector_observations[0] # get the current state\nscore = 0 # initialize the score\nwhile True:\n action = np.random.randint(action_size) # select an action\n env_info = env.step(action)[brain_name] # send the action to the environment\n next_state = env_info.vector_observations[0] # get the next state\n reward = env_info.rewards[0] # get the reward\n done = env_info.local_done[0] # see if episode has finished\n score += reward # update the score\n state = next_state # roll over the state to next time step\n if done: # exit loop if episode finished\n break\n \nprint(\"Score: {}\".format(score))", "Score: 0.0\n" ] ], [ [ "When finished, you can close the environment.", "_____no_output_____" ] ], [ [ "#env.close()", "_____no_output_____" ] ], [ [ "### 4. It's Your Turn!\n\nNow it's your turn to train your own agent to solve the environment! When training the environment, set `train_mode=True`, so that the line for resetting the environment looks like the following:\n```python\nenv_info = env.reset(train_mode=True)[brain_name]\n```", "_____no_output_____" ], [ "### Training using DQN based off the previous assignments", "_____no_output_____" ] ], [ [ "from dqn_agent import Agent\nagent = Agent(state_size=37, action_size=4, seed=42)\n\ndef dqn(n_episodes=2000, max_t=1000, eps_start=1.0, eps_end=0.01, eps_decay=0.99):\n \"\"\"Deep Q-Learning.\n \n Params\n ======\n n_episodes (int): maximum number of training episodes\n max_t (int): maximum number of timesteps per episode\n eps_start (float): starting value of epsilon, for epsilon-greedy action selection\n eps_end (float): minimum value of epsilon\n eps_decay (float): multiplicative factor (per episode) for decreasing epsilon\n \"\"\"\n scores = [] # list containing scores from each episode\n scores_window = deque(maxlen=100) # last 100 scores\n eps = eps_start # initialize epsilon\n for i_episode in range(1, n_episodes+1):\n env_info = env.reset(train_mode=True)[brain_name]\n state = env_info.vector_observations[0]\n score = 0\n for t in range(max_t):\n action = agent.act(state, eps)\n env_info = env.step(action)[brain_name]\n next_state = env_info.vector_observations[0]\n reward = env_info.rewards[0]\n done = env_info.local_done[0]\n agent.step(state, action, reward, next_state, done)\n state = next_state\n score += reward\n if done:\n break \n scores_window.append(score) # save most recent score\n scores.append(score) # save most recent score\n eps = max(eps_end, eps_decay*eps) # decrease epsilon\n print('\\rEpisode {}\\tAverage Score: {:.2f}'.format(i_episode, np.mean(scores_window)), end=\"\")\n if i_episode % 100 == 0:\n print('\\rEpisode {}\\tAverage Score: {:.2f}'.format(i_episode, np.mean(scores_window)))\n if np.mean(scores_window)>=13.0:\n print('\\nEnvironment solved in {:d} episodes!\\tAverage Score: {:.2f}'.format(i_episode-100, np.mean(scores_window)))\n torch.save(agent.qnetwork_local.state_dict(), 'checkpoint_311220201515.pth')\n break\n return scores\n\nscores = dqn()\n\n# plot the scores\nfig = plt.figure()\nax = fig.add_subplot(111)\nplt.plot(np.arange(len(scores)), scores)\nplt.ylabel('Score')\nplt.xlabel('Episode #')\nplt.show()", "State Size: 37\nAction_size: 4\nEpisode 100\tAverage Score: 1.40\nEpisode 200\tAverage Score: 7.13\nEpisode 300\tAverage Score: 11.09\nEpisode 383\tAverage Score: 13.03\nEnvironment solved in 283 episodes!\tAverage Score: 13.03\n" ], [ "# plot the scores\nfig = plt.figure()\nax = fig.add_subplot(111)\nplt.plot(np.arange(len(scores)), scores)\nplt.ylabel('Score')\nplt.xlabel('Episode #')\nplt.show()", "_____no_output_____" ], [ "# closes the environment\nenv.close()", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Data Download - *read description before running*", "_____no_output_____" ], [ "Term project for ESS 490/590\n\nGrad: Erik Fredrickson\n\nUndergrad: Ashika Capirala", "_____no_output_____" ], [ "*This notebook demonstrates how the open access datasets can be downloaded, but these data are provided at significantly higher temporal resolution than needed for the purposes of our study, so for the sake of this project we recommend that the user use the provided reduced data, which will save significant time, computing power, and disc space.*", "_____no_output_____" ] ], [ [ "# Imports\nimport obspy\nimport obspy.clients.fdsn.client as fdsn\nfrom obspy import UTCDateTime", "_____no_output_____" ] ], [ [ "## APG pressures and temperatures\n\nGet bottom temperature and bottom pressure from IRIS (https://www.iris.edu/hq/; https://doi.org/10.7914/SN/XO_2018)\n\n<img src=\"AACSE.png\" width=\"600\">\n<center>Alaska Amphibious Community Seismic Experiment</center>", "_____no_output_____" ] ], [ [ "# Pull pressure and temperature data from IRIS\nnetwork = 'XO'\nstaNames = ['LA21', 'LA34', 'LA33', 'LA23', 'LA25', 'LA22', 'LA28', 'LA39', 'LA32', 'LA30', 'LT07', 'LT06', \\\n 'LT13', 'LT03', 'LT11', 'LT04', 'LT01', 'LT20', 'LT14', 'LT16', 'LT10', 'LT12']\nstaCodes = 'LA21,LA34,LA33,LA23,LA25,LA22,LA28,LA39,LA32,LA30,LT07,LT06,LT13,LT03,LT11,LT04,LT01,LT20,LT14,LT16,LT10,LT12'\nchaNames = ['HDH', 'HKO']\nchaCodes='HDH,HKO'\nTstart = UTCDateTime(2018, 06, 01)\nTend = UTCDateTime(2019, 06, 20)\n\nfdsn_client = fdsn.Client('IRIS')\n\n# DO NOT RUN AS WRITTEN -- way too much data, so we'll need to make a loop to parse it by station and by day\nDtmp = fdsn_client.get_waveforms(network=network, station=staCodes, location='--', channel=chaCodes, starttime=Tstart, \\\n endtime=Tend, attach_response=False)", "_____no_output_____" ] ], [ [ "## Satellite altimetry\n\nGet altimetry data from Copernicus Marine (https://marine.copernicus.eu/; https://resources.marine.copernicus.eu/?option=com_csw&view=details&product_id=SEALEVEL_GLO_PHY_CLIMATE_L4_REP_OBSERVATIONS_008_057)\n\n<img src=\"jason-2-altimeter.jpg\" width=\"600\">\n<center>Jason-2 Satellite</center>", "_____no_output_____" ] ], [ [ "# installer to handle data download\n!pip install motuclient --upgrade", "Collecting motuclient\n Downloading motuclient-1.8.8.tar.gz (30 kB)\nBuilding wheels for collected packages: motuclient\n Building wheel for motuclient (setup.py) ... \u001b[?25ldone\n\u001b[?25h Created wheel for motuclient: filename=motuclient-1.8.8-py3-none-any.whl size=34205 sha256=83d62167dbfe284e1de30763a39c7b71a92ab49edd6ff972d4a9b09e6cc6a208\n Stored in directory: /Users/erikfred/Library/Caches/pip/wheels/e6/16/94/3e40d579a5a03c17f2123e2dd13c5389c2e7515f49487fd88e\nSuccessfully built motuclient\nInstalling collected packages: motuclient\nSuccessfully installed motuclient-1.8.8\n" ], [ "# Get desired data (would need to change directory, user, and password fields)\n!python -m motuclient --motu https://my.cmems-du.eu/motu-web/Motu --service-id \\\n SEALEVEL_GLO_PHY_CLIMATE_L4_REP_OBSERVATIONS_008_057-TDS --product-id \\\n dataset-duacs-rep-global-merged-twosat-phy-l4 --longitude-min 198 \\\n --longitude-max 210 --latitude-min 53 --latitude-max 60 \\\n --date-min \"2018-06-01 00:00:00\" --date-max \"2019-06-20 23:59:59\" \\\n --variable adt --variable err --variable sla --variable ugos --variable ugosa \\\n --variable vgos --variable vgosa --out-dir <OUTPUT_DIRECTORY> --out-name \\\n <OUTPUT_FILENAME> --user <USERNAME> --pwd <PASSWORD>", "_____no_output_____" ] ], [ [ "## Oceanographic model\n\nModel data not currently publicly available :(\nLoad from netcdf", "_____no_output_____" ], [ "## Eddy catalog", "_____no_output_____" ], [ "Labeled dataset hosted by AVISO. Requires registration, but free for academics (https://www.aviso.altimetry.fr/en/home.html; https://doi.org/10.24400/527896/a01-2021.001)\n\nFull code is available on GitHub! (https://github.com/AntSimi/py-eddy-tracker)\n\n<img src=\"eddy_field.jpg\" width=\"600\">\n<center>https://doi.org/10.1175/JTECH-D-14-00019.1</center>", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\n\nfrom sklearn.model_selection import train_test_split\n\nimport keras\nfrom keras.models import Sequential\nfrom keras.layers import Dense, Dropout, ReLU, Flatten\nfrom keras.layers import Conv2D, MaxPooling2D\nfrom keras.losses import categorical_crossentropy\nfrom keras.optimizers import Adam\nfrom keras.preprocessing.image import ImageDataGenerator\nfrom keras import backend as K\n\n%matplotlib inline", "_____no_output_____" ], [ "# load data\ntrain_df = pd.read_csv('data/train.csv')\ntest_df = pd.read_csv('data/test.csv')", "_____no_output_____" ], [ "train_df.head(3)", "_____no_output_____" ], [ "test_df.head(3)", "_____no_output_____" ], [ "# Convert data to numpy arrays\nimg_rows, img_cols = 28, 28\nnum_classes = train_df['label'].nunique()\n\n# Split train + val data\nX = train_df.drop(columns=['label']).values\ny = train_df['label'].values\n\n# one-hot encode the labels\ny = keras.utils.to_categorical(y, num_classes)\n\nif K.image_data_format() == 'channels_first':\n X = X.reshape(-1, 1, img_rows, img_cols)\n X_test = test_df.values.reshape(-1, 1, img_rows, img_cols)\n input_shape = (1, img_rows, img_cols)\nelse:\n X = X.reshape(-1, img_rows, img_cols, 1)\n X_test = test_df.values.reshape(-1, img_rows, img_cols, 1)\n input_shape = (img_rows, img_cols, 1)", "_____no_output_____" ], [ "# normalize data\nX = X.astype('float32')\nX_test = X_test.astype('float32')\nX /= 255\nX_test /= 255", "_____no_output_____" ], [ "# Split train + val data\nX_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.33, random_state=42)", "_____no_output_____" ], [ "# train array shape\nX_train.shape, y_train.shape", "_____no_output_____" ], [ "# val array shape\nX_val.shape, y_val.shape", "_____no_output_____" ], [ "# plot first image to check if it is in the correct format\nplt.imshow(X_train[0,:,:, 0])", "_____no_output_____" ], [ "# create model\ndef get_model():\n model = Sequential()\n model.add(Conv2D(filters=32, \n kernel_size=(3,3),\n activation=\"relu\",\n input_shape=input_shape))\n model.add(Conv2D(filters=64, \n kernel_size=(3,3),\n activation=\"relu\"))\n model.add(MaxPooling2D(pool_size=(2,2)))\n model.add(Conv2D(filters=128, \n kernel_size=(3,3),\n activation=\"relu\"))\n model.add(MaxPooling2D(pool_size=(2,2)))\n model.add(Dropout(0.25))\n model.add(Flatten())\n model.add(Dense(128, activation='relu'))\n model.add(Dropout(0.5))\n model.add(Dense(num_classes, activation='softmax'))\n\n model.compile(loss=categorical_crossentropy,\n optimizer=Adam(),\n metrics=['accuracy'])\n return model", "_____no_output_____" ], [ "# Model summary\nmodel = get_model()\nmodel.summary()", "_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\nconv2d_7 (Conv2D) (None, 26, 26, 32) 320 \n_________________________________________________________________\nconv2d_8 (Conv2D) (None, 24, 24, 64) 18496 \n_________________________________________________________________\nmax_pooling2d_5 (MaxPooling2 (None, 12, 12, 64) 0 \n_________________________________________________________________\nconv2d_9 (Conv2D) (None, 10, 10, 128) 73856 \n_________________________________________________________________\nmax_pooling2d_6 (MaxPooling2 (None, 5, 5, 128) 0 \n_________________________________________________________________\ndropout_5 (Dropout) (None, 5, 5, 128) 0 \n_________________________________________________________________\nflatten_3 (Flatten) (None, 3200) 0 \n_________________________________________________________________\ndense_5 (Dense) (None, 128) 409728 \n_________________________________________________________________\ndropout_6 (Dropout) (None, 128) 0 \n_________________________________________________________________\ndense_6 (Dense) (None, 10) 1290 \n=================================================================\nTotal params: 503,690\nTrainable params: 503,690\nNon-trainable params: 0\n_________________________________________________________________\n" ], [ "# Setup train data transformer\ntrain_datagen = ImageDataGenerator(featurewise_center=True,\n featurewise_std_normalization=True,\n width_shift_range=0.2,\n height_shift_range=0.2,\n horizontal_flip=True)\n\ntrain_datagen.fit(X_train)", "_____no_output_____" ], [ "# Setup validation data transformer\nval_datagen = ImageDataGenerator(featurewise_center=True,\n featurewise_std_normalization=True,\n horizontal_flip=False)\n\nval_datagen.fit(X_train)", "_____no_output_____" ], [ "batch_size = 128\nepochs = 10\nmodel_gen = get_model()\nmodel_gen.fit_generator(train_datagen.flow(X_train, y_train, batch_size=batch_size),\n steps_per_epoch=len(X_train) / batch_size,\n epochs=epochs,\n shuffle=True,\n validation_data=val_datagen.flow(X_val, y_val, batch_size=batch_size),\n validation_steps=len(X_val) / batch_size,\n verbose=1)", "Epoch 1/10\n220/219 [==============================] - 89s 405ms/step - loss: 1.2205 - acc: 0.5704 - val_loss: 0.3468 - val_acc: 0.8898\nEpoch 2/10\n220/219 [==============================] - 97s 442ms/step - loss: 0.5769 - acc: 0.8096 - val_loss: 0.1883 - val_acc: 0.9426\nEpoch 3/10\n220/219 [==============================] - 88s 400ms/step - loss: 0.4190 - acc: 0.8655 - val_loss: 0.1466 - val_acc: 0.9542\nEpoch 4/10\n220/219 [==============================] - 98s 443ms/step - loss: 0.3380 - acc: 0.8945 - val_loss: 0.1110 - val_acc: 0.9655\nEpoch 5/10\n220/219 [==============================] - 97s 440ms/step - loss: 0.2889 - acc: 0.9113 - val_loss: 0.1315 - val_acc: 0.9595\nEpoch 6/10\n220/219 [==============================] - 99s 452ms/step - loss: 0.2588 - acc: 0.9198 - val_loss: 0.1108 - val_acc: 0.9652\nEpoch 7/10\n220/219 [==============================] - 95s 434ms/step - loss: 0.2430 - acc: 0.9252 - val_loss: 0.0798 - val_acc: 0.9752\nEpoch 8/10\n220/219 [==============================] - 97s 442ms/step - loss: 0.2179 - acc: 0.9340 - val_loss: 0.0775 - val_acc: 0.9764\nEpoch 9/10\n220/219 [==============================] - 97s 441ms/step - loss: 0.2033 - acc: 0.9359 - val_loss: 0.0773 - val_acc: 0.9760\nEpoch 10/10\n220/219 [==============================] - 97s 440ms/step - loss: 0.1960 - acc: 0.9414 - val_loss: 0.0694 - val_acc: 0.9781\n" ], [ "score = model_gen.evaluate_generator(val_datagen.flow(X_val, y_val, batch_size=batch_size), \n steps=len(X_val) / batch_size,\n verbose=1)\nprint('Test loss:', score[0])\nprint('Test accuracy:', score[1])", "109/108 [==============================] - 8s 74ms/step\nTest loss: 0.06943384749548775\nTest accuracy: 0.9781385280697205\n" ], [ "batch_size = 128\nepochs = 10\nmodel.fit(X_train, y_train,\n batch_size=batch_size,\n epochs=epochs,\n verbose=1,\n validation_data=(X_val, y_val))", "Train on 28140 samples, validate on 13860 samples\nEpoch 1/10\n28140/28140 [==============================] - 96s 3ms/step - loss: 0.3474 - acc: 0.8908 - val_loss: 0.0738 - val_acc: 0.9776\nEpoch 2/10\n28140/28140 [==============================] - 92s 3ms/step - loss: 0.1053 - acc: 0.9687 - val_loss: 0.0515 - val_acc: 0.9838\nEpoch 3/10\n28140/28140 [==============================] - 94s 3ms/step - loss: 0.0763 - acc: 0.9769 - val_loss: 0.0438 - val_acc: 0.9854\nEpoch 4/10\n28140/28140 [==============================] - 93s 3ms/step - loss: 0.0623 - acc: 0.9813 - val_loss: 0.0413 - val_acc: 0.9874\nEpoch 5/10\n28140/28140 [==============================] - 93s 3ms/step - loss: 0.0513 - acc: 0.9857 - val_loss: 0.0342 - val_acc: 0.9891\nEpoch 6/10\n28140/28140 [==============================] - 91s 3ms/step - loss: 0.0467 - acc: 0.9856 - val_loss: 0.0303 - val_acc: 0.9903\nEpoch 7/10\n28140/28140 [==============================] - 93s 3ms/step - loss: 0.0397 - acc: 0.9882 - val_loss: 0.0320 - val_acc: 0.9911\nEpoch 8/10\n28140/28140 [==============================] - 98s 3ms/step - loss: 0.0348 - acc: 0.9885 - val_loss: 0.0288 - val_acc: 0.9918\nEpoch 9/10\n28140/28140 [==============================] - 91s 3ms/step - loss: 0.0334 - acc: 0.9897 - val_loss: 0.0295 - val_acc: 0.9916\nEpoch 10/10\n28140/28140 [==============================] - 94s 3ms/step - loss: 0.0297 - acc: 0.9906 - val_loss: 0.0311 - val_acc: 0.9908\n" ], [ "score = model.evaluate(X_val, y_val, batch_size=256, verbose=1)\nprint('Test loss:', score[0])\nprint('Test accuracy:', score[1])", "13860/13860 [==============================] - 11s 810us/step\nTest loss: 0.031108998779266598\nTest accuracy: 0.9907647907303869\n" ], [ "# train model on the full train data\nbatch_size = 128\nepochs = 10\nmodel_final = get_model()\nmodel_final.fit(X, y,\n batch_size=batch_size,\n epochs=epochs,\n verbose=1)", "Epoch 1/10\n42000/42000 [==============================] - 111s 3ms/step - loss: 0.2927 - acc: 0.9086\nEpoch 2/10\n42000/42000 [==============================] - 122s 3ms/step - loss: 0.0909 - acc: 0.9726\nEpoch 3/10\n42000/42000 [==============================] - 119s 3ms/step - loss: 0.0647 - acc: 0.9797\nEpoch 4/10\n42000/42000 [==============================] - 121s 3ms/step - loss: 0.0507 - acc: 0.9853\nEpoch 5/10\n42000/42000 [==============================] - 121s 3ms/step - loss: 0.0455 - acc: 0.9857\nEpoch 6/10\n42000/42000 [==============================] - 122s 3ms/step - loss: 0.0370 - acc: 0.9887\nEpoch 7/10\n42000/42000 [==============================] - 122s 3ms/step - loss: 0.0356 - acc: 0.9888\nEpoch 8/10\n42000/42000 [==============================] - 122s 3ms/step - loss: 0.0305 - acc: 0.9906\nEpoch 9/10\n42000/42000 [==============================] - 121s 3ms/step - loss: 0.0268 - acc: 0.9917\nEpoch 10/10\n42000/42000 [==============================] - 125s 3ms/step - loss: 0.0248 - acc: 0.9918\n" ], [ "# create submission predictions\npredictions = model_final.predict(X_test, batch_size=256, verbose=1)", "28000/28000 [==============================] - 25s 879us/step\n" ], [ "# save predictions\nout_df = pd.DataFrame({\"ImageId\": list(range(1, len(predictions) + 1)),\n \"Label\": np.argmax(predictions, axis=1)})\nout_df.to_csv('keras_submission.csv', index=False)", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "##### Copyright 2018 The TensorFlow Authors.", "_____no_output_____" ] ], [ [ "#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# https://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n", "_____no_output_____" ] ], [ [ "# tf.dataを使って画像をロードする", "_____no_output_____" ], [ "<table class=\"tfo-notebook-buttons\" align=\"left\">\n <td>\n <a target=\"_blank\" href=\"https://www.tensorflow.org/tutorials/load_data/images\"><img src=\"https://www.tensorflow.org/images/tf_logo_32px.png\" />View on TensorFlow.org</a>\n </td>\n <td>\n <a target=\"_blank\" href=\"https://colab.research.google.com/github/tensorflow/docs-l10n/blob/master/site/ja/tutorials/load_data/images.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />Run in Google Colab</a>\n </td>\n <td>\n <a target=\"_blank\" href=\"https://github.com/tensorflow/docs-l10n/blob/master/site/ja/tutorials/load_data/images.ipynb\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" />View source on GitHub</a>\n </td>\n</table>", "_____no_output_____" ], [ "Note: これらのドキュメントは私たちTensorFlowコミュニティが翻訳したものです。コミュニティによる 翻訳は**ベストエフォート**であるため、この翻訳が正確であることや[英語の公式ドキュメント](https://www.tensorflow.org/?hl=en)の 最新の状態を反映したものであることを保証することはできません。 この翻訳の品質を向上させるためのご意見をお持ちの方は、GitHubリポジトリ[tensorflow/docs](https://github.com/tensorflow/docs)にプルリクエストをお送りください。 コミュニティによる翻訳やレビューに参加していただける方は、 [[email protected] メーリングリスト](https://groups.google.com/a/tensorflow.org/forum/#!forum/docs-ja)にご連絡ください。", "_____no_output_____" ], [ "このチュートリアルでは、'tf.data' を使って画像データセットをロードする簡単な例を示します。\n\nこのチュートリアルで使用するデータセットは、クラスごとに別々のディレクトリに別れた形で配布されています。", "_____no_output_____" ], [ "## 設定", "_____no_output_____" ] ], [ [ "from __future__ import absolute_import, division, print_function, unicode_literals\n\ntry:\n # Colab only\n %tensorflow_version 2.x\nexcept Exception:\n pass\nimport tensorflow as tf", "_____no_output_____" ], [ "AUTOTUNE = tf.data.experimental.AUTOTUNE", "_____no_output_____" ] ], [ [ "## データセットのダウンロードと検査", "_____no_output_____" ], [ "### 画像の取得\n\n訓練を始める前に、ネットワークに認識すべき新しいクラスを教えるために画像のセットが必要です。最初に使うためのクリエイティブ・コモンズでライセンスされた花の画像のアーカイブを作成してあります。", "_____no_output_____" ] ], [ [ "import pathlib\ndata_root_orig = tf.keras.utils.get_file(origin='https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz',\n fname='flower_photos', untar=True)\ndata_root = pathlib.Path(data_root_orig)\nprint(data_root)", "_____no_output_____" ] ], [ [ "218MB をダウンロードすると、花の画像のコピーが使えるようになっているはずです。", "_____no_output_____" ] ], [ [ "for item in data_root.iterdir():\n print(item)", "_____no_output_____" ], [ "import random\nall_image_paths = list(data_root.glob('*/*'))\nall_image_paths = [str(path) for path in all_image_paths]\nrandom.shuffle(all_image_paths)\n\nimage_count = len(all_image_paths)\nimage_count", "_____no_output_____" ], [ "all_image_paths[:10]", "_____no_output_____" ] ], [ [ "### 画像の検査\n\n扱っている画像について知るために、画像のいくつかを見てみましょう。", "_____no_output_____" ] ], [ [ "import os\nattributions = (data_root/\"LICENSE.txt\").open(encoding='utf-8').readlines()[4:]\nattributions = [line.split(' CC-BY') for line in attributions]\nattributions = dict(attributions)", "_____no_output_____" ], [ "import IPython.display as display\n\ndef caption_image(image_path):\n image_rel = pathlib.Path(image_path).relative_to(data_root)\n return \"Image (CC BY 2.0) \" + ' - '.join(attributions[str(image_rel)].split(' - ')[:-1])\n ", "_____no_output_____" ], [ "for n in range(3):\n image_path = random.choice(all_image_paths)\n display.display(display.Image(image_path))\n print(caption_image(image_path))\n print()", "_____no_output_____" ] ], [ [ "### 各画像のラベルの決定", "_____no_output_____" ], [ "ラベルを一覧してみます。", "_____no_output_____" ] ], [ [ "label_names = sorted(item.name for item in data_root.glob('*/') if item.is_dir())\nlabel_names", "_____no_output_____" ] ], [ [ "ラベルにインデックスを割り当てます。", "_____no_output_____" ] ], [ [ "label_to_index = dict((name, index) for index,name in enumerate(label_names))\nlabel_to_index", "_____no_output_____" ] ], [ [ "ファイルとラベルのインデックスの一覧を作成します。", "_____no_output_____" ] ], [ [ "all_image_labels = [label_to_index[pathlib.Path(path).parent.name]\n for path in all_image_paths]\n\nprint(\"First 10 labels indices: \", all_image_labels[:10])", "_____no_output_____" ] ], [ [ "### 画像の読み込みと整形", "_____no_output_____" ], [ "TensorFlow には画像を読み込んで処理するために必要なツールが備わっています。", "_____no_output_____" ] ], [ [ "img_path = all_image_paths[0]\nimg_path", "_____no_output_____" ] ], [ [ "以下は生のデータです。", "_____no_output_____" ] ], [ [ "img_raw = tf.io.read_file(img_path)\nprint(repr(img_raw)[:100]+\"...\")", "_____no_output_____" ] ], [ [ "画像のテンソルにデコードします。", "_____no_output_____" ] ], [ [ "img_tensor = tf.image.decode_image(img_raw)\n\nprint(img_tensor.shape)\nprint(img_tensor.dtype)", "_____no_output_____" ] ], [ [ "モデルに合わせてリサイズします。", "_____no_output_____" ] ], [ [ "img_final = tf.image.resize(img_tensor, [192, 192])\nimg_final = img_final/255.0\nprint(img_final.shape)\nprint(img_final.numpy().min())\nprint(img_final.numpy().max())\n", "_____no_output_____" ] ], [ [ "このあと使用するために、簡単な関数にまとめます。", "_____no_output_____" ] ], [ [ "def preprocess_image(image):\n image = tf.image.decode_jpeg(image, channels=3)\n image = tf.image.resize(image, [192, 192])\n image /= 255.0 # normalize to [0,1] range\n\n return image", "_____no_output_____" ], [ "def load_and_preprocess_image(path):\n image = tf.io.read_file(path)\n return preprocess_image(image)", "_____no_output_____" ], [ "import matplotlib.pyplot as plt\n\nimage_path = all_image_paths[0]\nlabel = all_image_labels[0]\n\nplt.imshow(load_and_preprocess_image(img_path))\nplt.grid(False)\nplt.xlabel(caption_image(img_path))\nplt.title(label_names[label].title())\nprint()", "_____no_output_____" ] ], [ [ "## `tf.data.Dataset`の構築", "_____no_output_____" ], [ "### 画像のデータセット", "_____no_output_____" ], [ "`tf.data.Dataset` を構築するもっとも簡単な方法は、`from_tensor_slices` メソッドを使うことです。\n\n文字列の配列をスライスすると、文字列のデータセットが出来上がります。", "_____no_output_____" ] ], [ [ "path_ds = tf.data.Dataset.from_tensor_slices(all_image_paths)", "_____no_output_____" ] ], [ [ "`shapes` と `types` は、データセット中のそれぞれのアイテムの内容を示しています。この場合には、バイナリ文字列のスカラーのセットです。 ", "_____no_output_____" ] ], [ [ "print(path_ds)", "_____no_output_____" ] ], [ [ "`preprocess_image` をファイルパスのデータセットにマップすることで、画像を実行時にロードし整形する新しいデータセットを作成します。", "_____no_output_____" ] ], [ [ "image_ds = path_ds.map(load_and_preprocess_image, num_parallel_calls=AUTOTUNE)", "_____no_output_____" ], [ "import matplotlib.pyplot as plt\n\nplt.figure(figsize=(8,8))\nfor n,image in enumerate(image_ds.take(4)):\n plt.subplot(2,2,n+1)\n plt.imshow(image)\n plt.grid(False)\n plt.xticks([])\n plt.yticks([])\n plt.xlabel(caption_image(all_image_paths[n]))\n plt.show()", "_____no_output_____" ] ], [ [ "### `(image, label)`のペアのデータセット", "_____no_output_____" ], [ "おなじ `from_tensor_slices` メソッドを使ってラベルのデータセットを作ることができます。", "_____no_output_____" ] ], [ [ "label_ds = tf.data.Dataset.from_tensor_slices(tf.cast(all_image_labels, tf.int64))", "_____no_output_____" ], [ "for label in label_ds.take(10):\n print(label_names[label.numpy()])", "_____no_output_____" ] ], [ [ "これらのデータセットはおなじ順番なので、zip することで `(image, label)` というペアのデータセットができます。", "_____no_output_____" ] ], [ [ "image_label_ds = tf.data.Dataset.zip((image_ds, label_ds))", "_____no_output_____" ] ], [ [ "新しいデータセットの `shapes` と `types` は、それぞれのフィールドを示すシェイプと型のタプルです。", "_____no_output_____" ] ], [ [ "print(image_label_ds)", "_____no_output_____" ] ], [ [ "注： `all_image_labels` や `all_image_paths` のような配列がある場合、 `tf.data.dataset.Dataset.zip` メソッドの代わりとなるのは、配列のペアをスライスすることです。", "_____no_output_____" ] ], [ [ "ds = tf.data.Dataset.from_tensor_slices((all_image_paths, all_image_labels))\n\n# The tuples are unpacked into the positional arguments of the mapped function\n# タプルは展開され、マップ関数の位置引数に割り当てられます\ndef load_and_preprocess_from_path_label(path, label):\n return load_and_preprocess_image(path), label\n\nimage_label_ds = ds.map(load_and_preprocess_from_path_label)\nimage_label_ds", "_____no_output_____" ] ], [ [ "### 基本的な訓練手法", "_____no_output_____" ], [ "このデータセットを使ってモデルの訓練を行うには、データが\n\n* よくシャッフルされ\n* バッチ化され\n* 限りなく繰り返され\n* バッチが出来るだけ早く利用できる\n\nことが必要です。\n\nこれらの特性は `tf.data` APIを使えば簡単に付け加えることができます。", "_____no_output_____" ] ], [ [ "BATCH_SIZE = 32\n\n# シャッフルバッファのサイズをデータセットとおなじに設定することで、データが完全にシャッフルされる\n# ようにできます。\nds = image_label_ds.shuffle(buffer_size=image_count)\nds = ds.repeat()\nds = ds.batch(BATCH_SIZE)\n# `prefetch`を使うことで、モデルの訓練中にバックグラウンドでデータセットがバッチを取得できます。\nds = ds.prefetch(buffer_size=AUTOTUNE)\nds", "_____no_output_____" ] ], [ [ "注意すべきことがいくつかあります。\n\n1. 順番が重要です。\n\n * `.repeat` の前に `.shuffle` すると、エポックの境界を越えて要素がシャッフルされます。（ほかの要素がすべて出現する前に2回出現する要素があるかもしれません）\n * `.batch` の後に `.shuffle` すると、バッチの順番がシャッフルされますが、要素がバッチを越えてシャッフルされることはありません。\n\n1. 完全なシャッフルのため、 `buffer_size` をデータセットとおなじサイズに設定しています。データセットのサイズ未満の場合、値が大きいほど良くランダム化されますが、より多くのメモリーを使用します。\n\n1. シャッフルバッファがいっぱいになってから要素が取り出されます。そのため、大きな `buffer_size` が `Dataset` を使い始める際の遅延の原因になります。\n\n1. シャッフルされたデータセットは、シャッフルバッファが完全に空になるまでデータセットが終わりであることを伝えません。 `.repeat` によって `Dataset` が再起動されると、シャッフルバッファが一杯になるまでもう一つの待ち時間が発生します。\n\n最後の問題は、 `tf.data.Dataset.apply` メソッドを、融合された `tf.data.experimental.shuffle_and_repeat` 関数と組み合わせることで対処できます。", "_____no_output_____" ] ], [ [ "ds = image_label_ds.apply(\n tf.data.experimental.shuffle_and_repeat(buffer_size=image_count))\nds = ds.batch(BATCH_SIZE)\nds = ds.prefetch(buffer_size=AUTOTUNE)\nds", "_____no_output_____" ] ], [ [ "### データセットをモデルにつなぐ\n\n`tf.keras.applications`からMobileNet v2のコピーを取得します。\n\nこれを簡単な転移学習のサンプルに使用します。\n\nMobileNetの重みを訓練不可に設定します。", "_____no_output_____" ] ], [ [ "mobile_net = tf.keras.applications.MobileNetV2(input_shape=(192, 192, 3), include_top=False)\nmobile_net.trainable=False", "_____no_output_____" ] ], [ [ "このモデルは、入力が `[-1,1]` の範囲に正規化されていることを想定しています。\n\n```\nhelp(keras_applications.mobilenet_v2.preprocess_input)\n```\n\n<pre>\n...\nThis function applies the \"Inception\" preprocessing which converts\nthe RGB values from [0, 255] to [-1, 1] \n...\n</pre>", "_____no_output_____" ], [ "このため、データをMobileNetモデルに渡す前に、入力を`[0,1]`の範囲から`[-1,1]`の範囲に変換する必要があります。", "_____no_output_____" ] ], [ [ "def change_range(image,label):\n return 2*image-1, label\n\nkeras_ds = ds.map(change_range)", "_____no_output_____" ] ], [ [ "MobileNetは画像ごとに `6x6` の特徴量の空間を返します。\n\nバッチを１つ渡してみましょう。", "_____no_output_____" ] ], [ [ "# シャッフルバッファがいっぱいになるまで、データセットは何秒かかかります。\nimage_batch, label_batch = next(iter(keras_ds))", "_____no_output_____" ], [ "feature_map_batch = mobile_net(image_batch)\nprint(feature_map_batch.shape)", "_____no_output_____" ] ], [ [ "MobileNet をラップしたモデルを作り、出力層である `tf.keras.layers.Dense` の前に、`tf.keras.layers.GlobalAveragePooling2D` で空間の軸にそって平均値を求めます。", "_____no_output_____" ] ], [ [ "model = tf.keras.Sequential([\n mobile_net,\n tf.keras.layers.GlobalAveragePooling2D(),\n tf.keras.layers.Dense(len(label_names))])", "_____no_output_____" ] ], [ [ "期待したとおりの形状の出力が得られます。", "_____no_output_____" ] ], [ [ "logit_batch = model(image_batch).numpy()\n\nprint(\"min logit:\", logit_batch.min())\nprint(\"max logit:\", logit_batch.max())\nprint()\n\nprint(\"Shape:\", logit_batch.shape)", "_____no_output_____" ] ], [ [ "訓練手法を記述するためにモデルをコンパイルします。", "_____no_output_____" ] ], [ [ "model.compile(optimizer=tf.keras.optimizers.Adam(), \n loss='sparse_categorical_crossentropy',\n metrics=[\"accuracy\"])", "_____no_output_____" ] ], [ [ "訓練可能な変数は2つ、全結合層の `weights` と `bias` です。", "_____no_output_____" ] ], [ [ "len(model.trainable_variables) ", "_____no_output_____" ], [ "model.summary()", "_____no_output_____" ] ], [ [ "モデルを訓練します。\n\n普通は、エポックごとの本当のステップ数を指定しますが、ここではデモの目的なので3ステップだけとします。", "_____no_output_____" ] ], [ [ "steps_per_epoch=tf.math.ceil(len(all_image_paths)/BATCH_SIZE).numpy()\nsteps_per_epoch", "_____no_output_____" ], [ "model.fit(ds, epochs=1, steps_per_epoch=3)", "_____no_output_____" ] ], [ [ "## 性能\n\n注：このセクションでは性能の向上に役立ちそうな簡単なトリックをいくつか紹介します。詳しくは、[Input Pipeline Performance](https://www.tensorflow.org/guide/performance/datasets) を参照してください。\n\n上記の単純なパイプラインは、エポックごとにそれぞれのファイルを一つずつ読み込みます。これは、CPU を使ったローカルでの訓練では問題になりませんが、GPU を使った訓練では十分ではなく、いかなる分散訓練でも使うべきではありません。", "_____no_output_____" ], [ "調査のため、まず、データセットの性能をチェックする簡単な関数を定義します。", "_____no_output_____" ] ], [ [ "import time\ndefault_timeit_steps = 2*steps_per_epoch+1\n\ndef timeit(ds, steps=default_timeit_steps):\n overall_start = time.time()\n # Fetch a single batch to prime the pipeline (fill the shuffle buffer),\n # before starting the timer\n it = iter(ds.take(steps+1))\n next(it)\n\n start = time.time()\n for i,(images,labels) in enumerate(it):\n if i%10 == 0:\n print('.',end='')\n print()\n end = time.time()\n\n duration = end-start\n print(\"{} batches: {} s\".format(steps, duration))\n print(\"{:0.5f} Images/s\".format(BATCH_SIZE*steps/duration))\n print(\"Total time: {}s\".format(end-overall_start))", "_____no_output_____" ] ], [ [ "現在のデータセットの性能は次のとおりです。", "_____no_output_____" ] ], [ [ "ds = image_label_ds.apply(\n tf.data.experimental.shuffle_and_repeat(buffer_size=image_count))\nds = ds.batch(BATCH_SIZE).prefetch(buffer_size=AUTOTUNE)\nds", "_____no_output_____" ], [ "timeit(ds)", "_____no_output_____" ] ], [ [ "### キャッシュ", "_____no_output_____" ], [ "`tf.data.Dataset.cache` を使うと、エポックを越えて計算結果を簡単にキャッシュできます。特に、データがメモリに収まるときには効果的です。\n\nここでは、画像が前処理（デコードとリサイズ）された後でキャッシュされます。", "_____no_output_____" ] ], [ [ "ds = image_label_ds.cache()\nds = ds.apply(\n tf.data.experimental.shuffle_and_repeat(buffer_size=image_count))\nds = ds.batch(BATCH_SIZE).prefetch(buffer_size=AUTOTUNE)\nds", "_____no_output_____" ], [ "timeit(ds)", "_____no_output_____" ] ], [ [ "メモリキャッシュを使う際の欠点のひとつは、実行の都度キャッシュを再構築しなければならないことです。このため、データセットがスタートするたびにおなじだけ起動のための遅延が発生します。", "_____no_output_____" ] ], [ [ "timeit(ds)", "_____no_output_____" ] ], [ [ "データがメモリに収まらない場合には、キャッシュファイルを使用します。", "_____no_output_____" ] ], [ [ "ds = image_label_ds.cache(filename='./cache.tf-data')\nds = ds.apply(\n tf.data.experimental.shuffle_and_repeat(buffer_size=image_count))\nds = ds.batch(BATCH_SIZE).prefetch(1)\nds", "_____no_output_____" ], [ "timeit(ds)", "_____no_output_____" ] ], [ [ "キャッシュファイルには、キャッシュを再構築することなくデータセットを再起動できるという利点もあります。2回めがどれほど早いか見てみましょう。", "_____no_output_____" ] ], [ [ "timeit(ds)", "_____no_output_____" ] ], [ [ "### TFRecord ファイル", "_____no_output_____" ], [ "#### 生の画像データ\n\nTFRecord ファイルは、バイナリの大きなオブジェクトのシーケンスを保存するための単純なフォーマットです。複数のサンプルをおなじファイルに詰め込むことで、TensorFlow は複数のサンプルを一度に読み込むことができます。これは、特に GCS のようなリモートストレージサービスを使用する際の性能にとって重要です。\n\n最初に、生の画像データから TFRecord ファイルを構築します。", "_____no_output_____" ] ], [ [ "image_ds = tf.data.Dataset.from_tensor_slices(all_image_paths).map(tf.io.read_file)\ntfrec = tf.data.experimental.TFRecordWriter('images.tfrec')\ntfrec.write(image_ds)", "_____no_output_____" ] ], [ [ "次に、TFRecord ファイルを読み込み、以前定義した `preprocess_image` 関数を使って画像のデコード/リフォーマットを行うデータセットを構築します。", "_____no_output_____" ] ], [ [ "image_ds = tf.data.TFRecordDataset('images.tfrec').map(preprocess_image)", "_____no_output_____" ] ], [ [ "これを、前に定義済みのラベルデータセットと zip し、期待どおりの `(image,label)` のペアを得ます。", "_____no_output_____" ] ], [ [ "ds = tf.data.Dataset.zip((image_ds, label_ds))\nds = ds.apply(\n tf.data.experimental.shuffle_and_repeat(buffer_size=image_count))\nds=ds.batch(BATCH_SIZE).prefetch(AUTOTUNE)\nds", "_____no_output_____" ], [ "timeit(ds)", "_____no_output_____" ] ], [ [ "これは、`cache` バージョンよりも低速です。前処理をキャッシュしていないからです。", "_____no_output_____" ], [ "#### シリアライズしたテンソル", "_____no_output_____" ], [ "前処理を TFRecord ファイルに保存するには、前やったように前処理した画像のデータセットを作ります。", "_____no_output_____" ] ], [ [ "paths_ds = tf.data.Dataset.from_tensor_slices(all_image_paths)\nimage_ds = paths_ds.map(load_and_preprocess_image)\nimage_ds", "_____no_output_____" ] ], [ [ "`.jpeg` 文字列のデータセットではなく、これはテンソルのデータセットです。\n\nこれを TFRecord ファイルにシリアライズするには、まず、テンソルのデータセットを文字列のデータセットに変換します。", "_____no_output_____" ] ], [ [ "ds = image_ds.map(tf.io.serialize_tensor)\nds", "_____no_output_____" ], [ "tfrec = tf.data.experimental.TFRecordWriter('images.tfrec')\ntfrec.write(ds)", "_____no_output_____" ] ], [ [ "前処理をキャッシュしたことにより、データは TFRecord ファイルから非常に効率的にロードできます。テンソルを使用する前にデシリアライズすることを忘れないでください。", "_____no_output_____" ] ], [ [ "ds = tf.data.TFRecordDataset('images.tfrec')\n\ndef parse(x):\n result = tf.io.parse_tensor(x, out_type=tf.float32)\n result = tf.reshape(result, [192, 192, 3])\n return result\n\nds = ds.map(parse, num_parallel_calls=AUTOTUNE)\nds", "_____no_output_____" ] ], [ [ "次にラベルを追加し、以前とおなじような標準的な処理を適用します。", "_____no_output_____" ] ], [ [ "ds = tf.data.Dataset.zip((ds, label_ds))\nds = ds.apply(\n tf.data.experimental.shuffle_and_repeat(buffer_size=image_count))\nds=ds.batch(BATCH_SIZE).prefetch(AUTOTUNE)\nds", "_____no_output_____" ], [ "timeit(ds)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Character level language model - Dinosaurus Island\n\nWelcome to Dinosaurus Island! 65 million years ago, dinosaurs existed, and in this assignment they are back. You are in charge of a special task. Leading biology researchers are creating new breeds of dinosaurs and bringing them to life on earth, and your job is to give names to these dinosaurs. If a dinosaur does not like its name, it might go berserk, so choose wisely! \n\n<table>\n<td>\n<img src=\"images/dino.jpg\" style=\"width:250;height:300px;\">\n\n</td>\n\n</table>\n\nLuckily you have learned some deep learning and you will use it to save the day. Your assistant has collected a list of all the dinosaur names they could find, and compiled them into this [dataset](dinos.txt). (Feel free to take a look by clicking the previous link.) To create new dinosaur names, you will build a character level language model to generate new names. Your algorithm will learn the different name patterns, and randomly generate new names. Hopefully this algorithm will keep you and your team safe from the dinosaurs' wrath! \n\nBy completing this assignment you will learn:\n\n- How to store text data for processing using an RNN \n- How to synthesize data, by sampling predictions at each time step and passing it to the next RNN-cell unit\n- How to build a character-level text generation recurrent neural network\n- Why clipping the gradients is important\n\nWe will begin by loading in some functions that we have provided for you in `rnn_utils`. Specifically, you have access to functions such as `rnn_forward` and `rnn_backward` which are equivalent to those you've implemented in the previous assignment. ", "_____no_output_____" ], [ "## <font color='darkblue'>Updates</font>\n\n#### If you were working on the notebook before this update...\n* The current notebook is version \"3a\".\n* You can find your original work saved in the notebook with the previous version name (\"v3\") \n* To view the file directory, go to the menu \"File->Open\", and this will open a new tab that shows the file directory.\n\n#### List of updates\n* Sort and print `chars` list of characters.\n* Import and use pretty print\n* `clip`: \n - Additional details on why we need to use the \"out\" parameter.\n - Modified for loop to have students fill in the correct items to loop through.\n - Added a test case to check for hard-coding error.\n* `sample`\n - additional hints added to steps 1,2,3,4.\n - \"Using 2D arrays instead of 1D arrays\".\n - explanation of numpy.ravel().\n - fixed expected output.\n - clarified comments in the code.\n* \"training the model\"\n - Replaced the sample code with explanations for how to set the index, X and Y (for a better learning experience).\n* Spelling, grammar and wording corrections.", "_____no_output_____" ] ], [ [ "import numpy as np\nfrom utils import *\nimport random\nimport pprint", "_____no_output_____" ] ], [ [ "## 1 - Problem Statement\n\n### 1.1 - Dataset and Preprocessing\n\nRun the following cell to read the dataset of dinosaur names, create a list of unique characters (such as a-z), and compute the dataset and vocabulary size. ", "_____no_output_____" ] ], [ [ "data = open('dinos.txt', 'r').read()\ndata= data.lower()\nchars = list(set(data))\ndata_size, vocab_size = len(data), len(chars)\nprint('There are %d total characters and %d unique characters in your data.' % (data_size, vocab_size))", "There are 19909 total characters and 27 unique characters in your data.\n" ] ], [ [ "\n* The characters are a-z (26 characters) plus the \"\\n\" (or newline character).\n* In this assignment, the newline character \"\\n\" plays a role similar to the `<EOS>` (or \"End of sentence\") token we had discussed in lecture. \n - Here, \"\\n\" indicates the end of the dinosaur name rather than the end of a sentence. \n* `char_to_ix`: In the cell below, we create a python dictionary (i.e., a hash table) to map each character to an index from 0-26.\n* `ix_to_char`: We also create a second python dictionary that maps each index back to the corresponding character. \n - This will help you figure out what index corresponds to what character in the probability distribution output of the softmax layer. ", "_____no_output_____" ] ], [ [ "chars = sorted(chars)\nprint(chars)", "['\\n', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z']\n" ], [ "char_to_ix = { ch:i for i,ch in enumerate(chars) }\nix_to_char = { i:ch for i,ch in enumerate(chars) }\npp = pprint.PrettyPrinter(indent=4)\npp.pprint(ix_to_char)", "{ 0: '\\n',\n 1: 'a',\n 2: 'b',\n 3: 'c',\n 4: 'd',\n 5: 'e',\n 6: 'f',\n 7: 'g',\n 8: 'h',\n 9: 'i',\n 10: 'j',\n 11: 'k',\n 12: 'l',\n 13: 'm',\n 14: 'n',\n 15: 'o',\n 16: 'p',\n 17: 'q',\n 18: 'r',\n 19: 's',\n 20: 't',\n 21: 'u',\n 22: 'v',\n 23: 'w',\n 24: 'x',\n 25: 'y',\n 26: 'z'}\n" ] ], [ [ "### 1.2 - Overview of the model\n\nYour model will have the following structure: \n\n- Initialize parameters \n- Run the optimization loop\n - Forward propagation to compute the loss function\n - Backward propagation to compute the gradients with respect to the loss function\n - Clip the gradients to avoid exploding gradients\n - Using the gradients, update your parameters with the gradient descent update rule.\n- Return the learned parameters \n \n<img src=\"images/rnn.png\" style=\"width:450;height:300px;\">\n<caption><center> **Figure 1**: Recurrent Neural Network, similar to what you had built in the previous notebook \"Building a Recurrent Neural Network - Step by Step\". </center></caption>\n\n* At each time-step, the RNN tries to predict what is the next character given the previous characters. \n* The dataset $\\mathbf{X} = (x^{\\langle 1 \\rangle}, x^{\\langle 2 \\rangle}, ..., x^{\\langle T_x \\rangle})$ is a list of characters in the training set.\n* $\\mathbf{Y} = (y^{\\langle 1 \\rangle}, y^{\\langle 2 \\rangle}, ..., y^{\\langle T_x \\rangle})$ is the same list of characters but shifted one character forward. \n* At every time-step $t$, $y^{\\langle t \\rangle} = x^{\\langle t+1 \\rangle}$. The prediction at time $t$ is the same as the input at time $t + 1$.", "_____no_output_____" ], [ "## 2 - Building blocks of the model\n\nIn this part, you will build two important blocks of the overall model:\n- Gradient clipping: to avoid exploding gradients\n- Sampling: a technique used to generate characters\n\nYou will then apply these two functions to build the model.", "_____no_output_____" ], [ "### 2.1 - Clipping the gradients in the optimization loop\n\nIn this section you will implement the `clip` function that you will call inside of your optimization loop. \n\n#### Exploding gradients\n* When gradients are very large, they're called \"exploding gradients.\" \n* Exploding gradients make the training process more difficult, because the updates may be so large that they \"overshoot\" the optimal values during back propagation.\n\nRecall that your overall loop structure usually consists of:\n* forward pass, \n* cost computation, \n* backward pass, \n* parameter update. \n\nBefore updating the parameters, you will perform gradient clipping to make sure that your gradients are not \"exploding.\"\n\n#### gradient clipping\nIn the exercise below, you will implement a function `clip` that takes in a dictionary of gradients and returns a clipped version of gradients if needed. \n* There are different ways to clip gradients.\n* We will use a simple element-wise clipping procedure, in which every element of the gradient vector is clipped to lie between some range [-N, N]. \n* For example, if the N=10\n - The range is [-10, 10]\n - If any component of the gradient vector is greater than 10, it is set to 10.\n - If any component of the gradient vector is less than -10, it is set to -10. \n - If any components are between -10 and 10, they keep their original values.\n\n<img src=\"images/clip.png\" style=\"width:400;height:150px;\">\n<caption><center> **Figure 2**: Visualization of gradient descent with and without gradient clipping, in a case where the network is running into \"exploding gradient\" problems. </center></caption>\n\n**Exercise**: \nImplement the function below to return the clipped gradients of your dictionary `gradients`. \n* Your function takes in a maximum threshold and returns the clipped versions of the gradients. \n* You can check out [numpy.clip](https://docs.scipy.org/doc/numpy-1.13.0/reference/generated/numpy.clip.html). \n - You will need to use the argument \"`out = ...`\".\n - Using the \"`out`\" parameter allows you to update a variable \"in-place\".\n - If you don't use \"`out`\" argument, the clipped variable is stored in the variable \"gradient\" but does not update the gradient variables `dWax`, `dWaa`, `dWya`, `db`, `dby`.", "_____no_output_____" ] ], [ [ "### GRADED FUNCTION: clip\n\ndef clip(gradients, maxValue):\n '''\n Clips the gradients' values between minimum and maximum.\n \n Arguments:\n gradients -- a dictionary containing the gradients \"dWaa\", \"dWax\", \"dWya\", \"db\", \"dby\"\n maxValue -- everything above this number is set to this number, and everything less than -maxValue is set to -maxValue\n \n Returns: \n gradients -- a dictionary with the clipped gradients.\n '''\n \n dWaa, dWax, dWya, db, dby = gradients['dWaa'], gradients['dWax'], gradients['dWya'], gradients['db'], gradients['dby']\n \n ### START CODE HERE ###\n # clip to mitigate exploding gradients, loop over [dWax, dWaa, dWya, db, dby]. (≈2 lines)\n for gradient in [dWaa, dWax, dWya, db, dby]:\n np.clip(gradient,a_min=-maxValue,a_max=maxValue,out=gradient)\n ### END CODE HERE ###\n \n gradients = {\"dWaa\": dWaa, \"dWax\": dWax, \"dWya\": dWya, \"db\": db, \"dby\": dby}\n \n return gradients", "_____no_output_____" ], [ "# Test with a maxvalue of 10\nmaxValue = 10\nnp.random.seed(3)\ndWax = np.random.randn(5,3)*10\ndWaa = np.random.randn(5,5)*10\ndWya = np.random.randn(2,5)*10\ndb = np.random.randn(5,1)*10\ndby = np.random.randn(2,1)*10\ngradients = {\"dWax\": dWax, \"dWaa\": dWaa, \"dWya\": dWya, \"db\": db, \"dby\": dby}\ngradients = clip(gradients, maxValue)\nprint(\"gradients[\\\"dWaa\\\"][1][2] =\", gradients[\"dWaa\"][1][2])\nprint(\"gradients[\\\"dWax\\\"][3][1] =\", gradients[\"dWax\"][3][1])\nprint(\"gradients[\\\"dWya\\\"][1][2] =\", gradients[\"dWya\"][1][2])\nprint(\"gradients[\\\"db\\\"][4] =\", gradients[\"db\"][4])\nprint(\"gradients[\\\"dby\\\"][1] =\", gradients[\"dby\"][1])", "gradients[\"dWaa\"][1][2] = 10.0\ngradients[\"dWax\"][3][1] = -10.0\ngradients[\"dWya\"][1][2] = 0.29713815361\ngradients[\"db\"][4] = [ 10.]\ngradients[\"dby\"][1] = [ 8.45833407]\n" ] ], [ [ "** Expected output:**\n\n```Python\ngradients[\"dWaa\"][1][2] = 10.0\ngradients[\"dWax\"][3][1] = -10.0\ngradients[\"dWya\"][1][2] = 0.29713815361\ngradients[\"db\"][4] = [ 10.]\ngradients[\"dby\"][1] = [ 8.45833407]\n```", "_____no_output_____" ] ], [ [ "# Test with a maxValue of 5\nmaxValue = 5\nnp.random.seed(3)\ndWax = np.random.randn(5,3)*10\ndWaa = np.random.randn(5,5)*10\ndWya = np.random.randn(2,5)*10\ndb = np.random.randn(5,1)*10\ndby = np.random.randn(2,1)*10\ngradients = {\"dWax\": dWax, \"dWaa\": dWaa, \"dWya\": dWya, \"db\": db, \"dby\": dby}\ngradients = clip(gradients, maxValue)\nprint(\"gradients[\\\"dWaa\\\"][1][2] =\", gradients[\"dWaa\"][1][2])\nprint(\"gradients[\\\"dWax\\\"][3][1] =\", gradients[\"dWax\"][3][1])\nprint(\"gradients[\\\"dWya\\\"][1][2] =\", gradients[\"dWya\"][1][2])\nprint(\"gradients[\\\"db\\\"][4] =\", gradients[\"db\"][4])\nprint(\"gradients[\\\"dby\\\"][1] =\", gradients[\"dby\"][1])", "gradients[\"dWaa\"][1][2] = 5.0\ngradients[\"dWax\"][3][1] = -5.0\ngradients[\"dWya\"][1][2] = 0.29713815361\ngradients[\"db\"][4] = [ 5.]\ngradients[\"dby\"][1] = [ 5.]\n" ] ], [ [ "** Expected Output: **\n```Python\ngradients[\"dWaa\"][1][2] = 5.0\ngradients[\"dWax\"][3][1] = -5.0\ngradients[\"dWya\"][1][2] = 0.29713815361\ngradients[\"db\"][4] = [ 5.]\ngradients[\"dby\"][1] = [ 5.]\n```", "_____no_output_____" ], [ "### 2.2 - Sampling\n\nNow assume that your model is trained. You would like to generate new text (characters). The process of generation is explained in the picture below:\n\n<img src=\"images/dinos3.png\" style=\"width:500;height:300px;\">\n<caption><center> **Figure 3**: In this picture, we assume the model is already trained. We pass in $x^{\\langle 1\\rangle} = \\vec{0}$ at the first time step, and have the network sample one character at a time. </center></caption>", "_____no_output_____" ], [ "**Exercise**: Implement the `sample` function below to sample characters. You need to carry out 4 steps:\n\n- **Step 1**: Input the \"dummy\" vector of zeros $x^{\\langle 1 \\rangle} = \\vec{0}$. \n - This is the default input before we've generated any characters. \n We also set $a^{\\langle 0 \\rangle} = \\vec{0}$", "_____no_output_____" ], [ "- **Step 2**: Run one step of forward propagation to get $a^{\\langle 1 \\rangle}$ and $\\hat{y}^{\\langle 1 \\rangle}$. Here are the equations:\n\nhidden state: \n$$ a^{\\langle t+1 \\rangle} = \\tanh(W_{ax} x^{\\langle t+1 \\rangle } + W_{aa} a^{\\langle t \\rangle } + b)\\tag{1}$$\n\nactivation:\n$$ z^{\\langle t + 1 \\rangle } = W_{ya} a^{\\langle t + 1 \\rangle } + b_y \\tag{2}$$\n\nprediction:\n$$ \\hat{y}^{\\langle t+1 \\rangle } = softmax(z^{\\langle t + 1 \\rangle })\\tag{3}$$\n\n- Details about $\\hat{y}^{\\langle t+1 \\rangle }$:\n - Note that $\\hat{y}^{\\langle t+1 \\rangle }$ is a (softmax) probability vector (its entries are between 0 and 1 and sum to 1). \n - $\\hat{y}^{\\langle t+1 \\rangle}_i$ represents the probability that the character indexed by \"i\" is the next character. \n - We have provided a `softmax()` function that you can use.", "_____no_output_____" ], [ "#### Additional Hints\n\n- $x^{\\langle 1 \\rangle}$ is `x` in the code. When creating the one-hot vector, make a numpy array of zeros, with the number of rows equal to the number of unique characters, and the number of columns equal to one. It's a 2D and not a 1D array.\n- $a^{\\langle 0 \\rangle}$ is `a_prev` in the code. It is a numpy array of zeros, where the number of rows is $n_{a}$, and number of columns is 1. It is a 2D array as well. $n_{a}$ is retrieved by getting the number of columns in $W_{aa}$ (the numbers need to match in order for the matrix multiplication $W_{aa}a^{\\langle t \\rangle}$ to work.\n- [numpy.dot](https://docs.scipy.org/doc/numpy/reference/generated/numpy.dot.html)\n- [numpy.tanh](https://docs.scipy.org/doc/numpy/reference/generated/numpy.tanh.html)", "_____no_output_____" ], [ "#### Using 2D arrays instead of 1D arrays\n* You may be wondering why we emphasize that $x^{\\langle 1 \\rangle}$ and $a^{\\langle 0 \\rangle}$ are 2D arrays and not 1D vectors.\n* For matrix multiplication in numpy, if we multiply a 2D matrix with a 1D vector, we end up with with a 1D array.\n* This becomes a problem when we add two arrays where we expected them to have the same shape.\n* When two arrays with a different number of dimensions are added together, Python \"broadcasts\" one across the other.\n* Here is some sample code that shows the difference between using a 1D and 2D array.", "_____no_output_____" ] ], [ [ "import numpy as np", "_____no_output_____" ], [ "matrix1 = np.array([[1,1],[2,2],[3,3]]) # (3,2)\nmatrix2 = np.array([[0],[0],[0]]) # (3,1) \nvector1D = np.array([1,1]) # (2,) \nvector2D = np.array([[1],[1]]) # (2,1)\nprint(\"matrix1 \\n\", matrix1,\"\\n\")\nprint(\"matrix2 \\n\", matrix2,\"\\n\")\nprint(\"vector1D \\n\", vector1D,\"\\n\")\nprint(\"vector2D \\n\", vector2D)", "matrix1 \n [[1 1]\n [2 2]\n [3 3]] \n\nmatrix2 \n [[0]\n [0]\n [0]] \n\nvector1D \n [1 1] \n\nvector2D \n [[1]\n [1]]\n" ], [ "print(\"Multiply 2D and 1D arrays: result is a 1D array\\n\", \n np.dot(matrix1,vector1D))\nprint(\"Multiply 2D and 2D arrays: result is a 2D array\\n\", \n np.dot(matrix1,vector2D))", "Multiply 2D and 1D arrays: result is a 1D array\n [2 4 6]\nMultiply 2D and 2D arrays: result is a 2D array\n [[2]\n [4]\n [6]]\n" ], [ "print(\"Adding (3 x 1) vector to a (3 x 1) vector is a (3 x 1) vector\\n\",\n \"This is what we want here!\\n\", \n np.dot(matrix1,vector2D) + matrix2)", "Adding (3 x 1) vector to a (3 x 1) vector is a (3 x 1) vector\n This is what we want here!\n [[2]\n [4]\n [6]]\n" ], [ "print(\"Adding a (3,) vector to a (3 x 1) vector\\n\",\n \"broadcasts the 1D array across the second dimension\\n\",\n \"Not what we want here!\\n\",\n np.dot(matrix1,vector1D) + matrix2\n )", "Adding a (3,) vector to a (3 x 1) vector\n broadcasts the 1D array across the second dimension\n Not what we want here!\n [[2 4 6]\n [2 4 6]\n [2 4 6]]\n" ] ], [ [ "- **Step 3**: Sampling: \n - Now that we have $y^{\\langle t+1 \\rangle}$, we want to select the next letter in the dinosaur name. If we select the most probable, the model will always generate the same result given a starting letter. \n - To make the results more interesting, we will use np.random.choice to select a next letter that is likely, but not always the same.\n - Sampling is the selection of a value from a group of values, where each value has a probability of being picked. \n - Sampling allows us to generate random sequences of values.\n - Pick the next character's index according to the probability distribution specified by $\\hat{y}^{\\langle t+1 \\rangle }$. \n - This means that if $\\hat{y}^{\\langle t+1 \\rangle }_i = 0.16$, you will pick the index \"i\" with 16% probability. \n - You can use [np.random.choice](https://docs.scipy.org/doc/numpy-1.13.0/reference/generated/numpy.random.choice.html).\n\n Example of how to use `np.random.choice()`:\n ```python\n np.random.seed(0)\n probs = np.array([0.1, 0.0, 0.7, 0.2])\n idx = np.random.choice([0, 1, 2, 3] p = probs)\n ```\n - This means that you will pick the index (`idx`) according to the distribution: \n\n $P(index = 0) = 0.1, P(index = 1) = 0.0, P(index = 2) = 0.7, P(index = 3) = 0.2$.\n\n - Note that the value that's set to `p` should be set to a 1D vector.\n - Also notice that $\\hat{y}^{\\langle t+1 \\rangle}$, which is `y` in the code, is a 2D array.", "_____no_output_____" ], [ "##### Additional Hints\n- [range](https://docs.python.org/3/library/functions.html#func-range)\n- [numpy.ravel](https://docs.scipy.org/doc/numpy/reference/generated/numpy.ravel.html) takes a multi-dimensional array and returns its contents inside of a 1D vector.\n```Python\narr = np.array([[1,2],[3,4]])\nprint(\"arr\")\nprint(arr)\nprint(\"arr.ravel()\")\nprint(arr.ravel())\n```\nOutput:\n```Python\narr\n[[1 2]\n [3 4]]\narr.ravel()\n[1 2 3 4]\n```\n\n- Note that `append` is an \"in-place\" operation. In other words, don't do this:\n```Python\nfun_hobbies = fun_hobbies.append('learning') ## Doesn't give you what you want\n```", "_____no_output_____" ], [ "- **Step 4**: Update to $x^{\\langle t \\rangle }$ \n - The last step to implement in `sample()` is to update the variable `x`, which currently stores $x^{\\langle t \\rangle }$, with the value of $x^{\\langle t + 1 \\rangle }$. \n - You will represent $x^{\\langle t + 1 \\rangle }$ by creating a one-hot vector corresponding to the character that you have chosen as your prediction. \n - You will then forward propagate $x^{\\langle t + 1 \\rangle }$ in Step 1 and keep repeating the process until you get a \"\\n\" character, indicating that you have reached the end of the dinosaur name. ", "_____no_output_____" ], [ "##### Additional Hints\n- In order to reset `x` before setting it to the new one-hot vector, you'll want to set all the values to zero.\n - You can either create a new numpy array: [numpy.zeros](https://docs.scipy.org/doc/numpy/reference/generated/numpy.zeros.html)\n - Or fill all values with a single number: [numpy.ndarray.fill](https://docs.scipy.org/doc/numpy/reference/generated/numpy.ndarray.fill.html)", "_____no_output_____" ] ], [ [ "# GRADED FUNCTION: sample\n\ndef sample(parameters, char_to_ix, seed):\n \"\"\"\n Sample a sequence of characters according to a sequence of probability distributions output of the RNN\n\n Arguments:\n parameters -- python dictionary containing the parameters Waa, Wax, Wya, by, and b. \n char_to_ix -- python dictionary mapping each character to an index.\n seed -- used for grading purposes. Do not worry about it.\n\n Returns:\n indices -- a list of length n containing the indices of the sampled characters.\n \"\"\"\n \n # Retrieve parameters and relevant shapes from \"parameters\" dictionary\n Waa, Wax, Wya, by, b = parameters['Waa'], parameters['Wax'], parameters['Wya'], parameters['by'], parameters['b']\n vocab_size = by.shape[0]\n n_a = Waa.shape[1]\n \n ### START CODE HERE ###\n # Step 1: Create the a zero vector x that can be used as the one-hot vector \n # representing the first character (initializing the sequence generation). (≈1 line)\n x = np.zeros((vocab_size,1))\n # Step 1': Initialize a_prev as zeros (≈1 line)\n a_prev = np.zeros((n_a,1))\n \n # Create an empty list of indices, this is the list which will contain the list of indices of the characters to generate (≈1 line)\n indices = []\n \n # idx is the index of the one-hot vector x that is set to 1\n # All other positions in x are zero.\n # We will initialize idx to -1\n idx = -1 \n \n # Loop over time-steps t. At each time-step:\n # sample a character from a probability distribution \n # and append its index (`idx`) to the list \"indices\". \n # We'll stop if we reach 50 characters \n # (which should be very unlikely with a well trained model).\n # Setting the maximum number of characters helps with debugging and prevents infinite loops. \n counter = 0\n newline_character = char_to_ix['\\n']\n \n while (idx != newline_character and counter != 50):\n \n # Step 2: Forward propagate x using the equations (1), (2) and (3)\n a = np.tanh(np.dot(Wax,x)+np.dot(Waa,a_prev)+b)\n z = np.dot(Wya,a)+by\n y = softmax(z)\n \n # for grading purposes\n np.random.seed(counter+seed) \n \n # Step 3: Sample the index of a character within the vocabulary from the probability distribution y\n # (see additional hints above)\n idx = np.random.choice(list(range(0,vocab_size)),p=y.ravel())\n\n # Append the index to \"indices\"\n indices.append(idx)\n \n # Step 4: Overwrite the input x with one that corresponds to the sampled index `idx`.\n # (see additional hints above)\n x = np.zeros((vocab_size,1))\n x[idx] = 1\n \n # Update \"a_prev\" to be \"a\"\n a_prev = a\n \n # for grading purposes\n seed += 1\n counter +=1\n \n ### END CODE HERE ###\n\n if (counter == 50):\n indices.append(char_to_ix['\\n'])\n \n return indices", "_____no_output_____" ], [ "np.random.seed(2)\n_, n_a = 20, 100\nWax, Waa, Wya = np.random.randn(n_a, vocab_size), np.random.randn(n_a, n_a), np.random.randn(vocab_size, n_a)\nb, by = np.random.randn(n_a, 1), np.random.randn(vocab_size, 1)\nparameters = {\"Wax\": Wax, \"Waa\": Waa, \"Wya\": Wya, \"b\": b, \"by\": by}\n\n\nindices = sample(parameters, char_to_ix, 0)\nprint(\"Sampling:\")\nprint(\"list of sampled indices:\\n\", indices)\nprint(\"size indices:\\n\", len(indices))\nprint(\"list of sampled characters:\\n\", [ix_to_char[i] for i in indices])", "Sampling:\nlist of sampled indices:\n [12, 17, 24, 14, 13, 9, 10, 22, 24, 6, 13, 11, 12, 6, 21, 15, 21, 14, 3, 2, 1, 21, 18, 24, 7, 25, 6, 25, 18, 10, 16, 2, 3, 8, 15, 12, 11, 7, 1, 12, 10, 2, 7, 7, 11, 17, 24, 12, 13, 24, 0]\nsize indices:\n 51\nlist of sampled characters:\n ['l', 'q', 'x', 'n', 'm', 'i', 'j', 'v', 'x', 'f', 'm', 'k', 'l', 'f', 'u', 'o', 'u', 'n', 'c', 'b', 'a', 'u', 'r', 'x', 'g', 'y', 'f', 'y', 'r', 'j', 'p', 'b', 'c', 'h', 'o', 'l', 'k', 'g', 'a', 'l', 'j', 'b', 'g', 'g', 'k', 'q', 'x', 'l', 'm', 'x', '\\n']\n" ] ], [ [ "** Expected output:**\n\n```Python\nSampling:\nlist of sampled indices:\n [12, 17, 24, 14, 13, 9, 10, 22, 24, 6, 13, 11, 12, 6, 21, 15, 21, 14, 3, 2, 1, 21, 18, 24, 7, 25, 6, 25, 18, 10, 16, 2, 3, 8, 15, 12, 11, 7, 1, 12, 10, 2, 7, 7, 11, 17, 24, 12, 13, 24, 0]\nlist of sampled characters:\n ['l', 'q', 'x', 'n', 'm', 'i', 'j', 'v', 'x', 'f', 'm', 'k', 'l', 'f', 'u', 'o', 'u', 'n', 'c', 'b', 'a', 'u', 'r', 'x', 'g', 'y', 'f', 'y', 'r', 'j', 'p', 'b', 'c', 'h', 'o', 'l', 'k', 'g', 'a', 'l', 'j', 'b', 'g', 'g', 'k', 'q', 'x', 'l', 'm', 'x', '\\n']\n```\n\n* Please note that over time, if there are updates to the back-end of the Coursera platform (that may update the version of numpy), the actual list of sampled indices and sampled characters may change. \n* If you follow the instructions given above and get an output without errors, it's possible the routine is correct even if your output doesn't match the expected output. Submit your assignment to the grader to verify its correctness.", "_____no_output_____" ], [ "## 3 - Building the language model \n\nIt is time to build the character-level language model for text generation. \n\n\n### 3.1 - Gradient descent \n\n* In this section you will implement a function performing one step of stochastic gradient descent (with clipped gradients). \n* You will go through the training examples one at a time, so the optimization algorithm will be stochastic gradient descent. \n\nAs a reminder, here are the steps of a common optimization loop for an RNN:\n\n- Forward propagate through the RNN to compute the loss\n- Backward propagate through time to compute the gradients of the loss with respect to the parameters\n- Clip the gradients\n- Update the parameters using gradient descent \n\n**Exercise**: Implement the optimization process (one step of stochastic gradient descent). \n\nThe following functions are provided:\n\n```python\ndef rnn_forward(X, Y, a_prev, parameters):\n \"\"\" Performs the forward propagation through the RNN and computes the cross-entropy loss.\n It returns the loss' value as well as a \"cache\" storing values to be used in backpropagation.\"\"\"\n ....\n return loss, cache\n \ndef rnn_backward(X, Y, parameters, cache):\n \"\"\" Performs the backward propagation through time to compute the gradients of the loss with respect\n to the parameters. It returns also all the hidden states.\"\"\"\n ...\n return gradients, a\n\ndef update_parameters(parameters, gradients, learning_rate):\n \"\"\" Updates parameters using the Gradient Descent Update Rule.\"\"\"\n ...\n return parameters\n```\n\nRecall that you previously implemented the `clip` function:\n\n```Python\ndef clip(gradients, maxValue)\n \"\"\"Clips the gradients' values between minimum and maximum.\"\"\"\n ...\n return gradients\n```", "_____no_output_____" ], [ "#### parameters\n\n* Note that the weights and biases inside the `parameters` dictionary are being updated by the optimization, even though `parameters` is not one of the returned values of the `optimize` function. The `parameters` dictionary is passed by reference into the function, so changes to this dictionary are making changes to the `parameters` dictionary even when accessed outside of the function.\n* Python dictionaries and lists are \"pass by reference\", which means that if you pass a dictionary into a function and modify the dictionary within the function, this changes that same dictionary (it's not a copy of the dictionary).", "_____no_output_____" ] ], [ [ "# GRADED FUNCTION: optimize\n\ndef optimize(X, Y, a_prev, parameters, learning_rate = 0.01):\n \"\"\"\n Execute one step of the optimization to train the model.\n \n Arguments:\n X -- list of integers, where each integer is a number that maps to a character in the vocabulary.\n Y -- list of integers, exactly the same as X but shifted one index to the left.\n a_prev -- previous hidden state.\n parameters -- python dictionary containing:\n Wax -- Weight matrix multiplying the input, numpy array of shape (n_a, n_x)\n Waa -- Weight matrix multiplying the hidden state, numpy array of shape (n_a, n_a)\n Wya -- Weight matrix relating the hidden-state to the output, numpy array of shape (n_y, n_a)\n b -- Bias, numpy array of shape (n_a, 1)\n by -- Bias relating the hidden-state to the output, numpy array of shape (n_y, 1)\n learning_rate -- learning rate for the model.\n \n Returns:\n loss -- value of the loss function (cross-entropy)\n gradients -- python dictionary containing:\n dWax -- Gradients of input-to-hidden weights, of shape (n_a, n_x)\n dWaa -- Gradients of hidden-to-hidden weights, of shape (n_a, n_a)\n dWya -- Gradients of hidden-to-output weights, of shape (n_y, n_a)\n db -- Gradients of bias vector, of shape (n_a, 1)\n dby -- Gradients of output bias vector, of shape (n_y, 1)\n a[len(X)-1] -- the last hidden state, of shape (n_a, 1)\n \"\"\"\n \n ### START CODE HERE ###\n \n # Forward propagate through time (≈1 line)\n loss, cache = rnn_forward(X, Y, a_prev, parameters)\n \n # Backpropagate through time (≈1 line)\n gradients, a = rnn_backward(X, Y, parameters, cache)\n \n # Clip your gradients between -5 (min) and 5 (max) (≈1 line)\n gradients = clip(gradients, 5)\n \n # Update parameters (≈1 line)\n parameters = update_parameters(parameters, gradients, learning_rate)\n \n ### END CODE HERE ###\n \n return loss, gradients, a[len(X)-1]", "_____no_output_____" ], [ "np.random.seed(1)\nvocab_size, n_a = 27, 100\na_prev = np.random.randn(n_a, 1)\nWax, Waa, Wya = np.random.randn(n_a, vocab_size), np.random.randn(n_a, n_a), np.random.randn(vocab_size, n_a)\nb, by = np.random.randn(n_a, 1), np.random.randn(vocab_size, 1)\nparameters = {\"Wax\": Wax, \"Waa\": Waa, \"Wya\": Wya, \"b\": b, \"by\": by}\nX = [12,3,5,11,22,3]\nY = [4,14,11,22,25, 26]\n\nloss, gradients, a_last = optimize(X, Y, a_prev, parameters, learning_rate = 0.01)\nprint(\"Loss =\", loss)\nprint(\"gradients[\\\"dWaa\\\"][1][2] =\", gradients[\"dWaa\"][1][2])\nprint(\"np.argmax(gradients[\\\"dWax\\\"]) =\", np.argmax(gradients[\"dWax\"]))\nprint(\"gradients[\\\"dWya\\\"][1][2] =\", gradients[\"dWya\"][1][2])\nprint(\"gradients[\\\"db\\\"][4] =\", gradients[\"db\"][4])\nprint(\"gradients[\\\"dby\\\"][1] =\", gradients[\"dby\"][1])\nprint(\"a_last[4] =\", a_last[4])", "Loss = 126.503975722\ngradients[\"dWaa\"][1][2] = 0.194709315347\nnp.argmax(gradients[\"dWax\"]) = 93\ngradients[\"dWya\"][1][2] = -0.007773876032\ngradients[\"db\"][4] = [-0.06809825]\ngradients[\"dby\"][1] = [ 0.01538192]\na_last[4] = [-1.]\n" ] ], [ [ "** Expected output:**\n\n```Python\nLoss = 126.503975722\ngradients[\"dWaa\"][1][2] = 0.194709315347\nnp.argmax(gradients[\"dWax\"]) = 93\ngradients[\"dWya\"][1][2] = -0.007773876032\ngradients[\"db\"][4] = [-0.06809825]\ngradients[\"dby\"][1] = [ 0.01538192]\na_last[4] = [-1.]\n```", "_____no_output_____" ], [ "### 3.2 - Training the model ", "_____no_output_____" ], [ "* Given the dataset of dinosaur names, we use each line of the dataset (one name) as one training example. \n* Every 100 steps of stochastic gradient descent, you will sample 10 randomly chosen names to see how the algorithm is doing. \n* Remember to shuffle the dataset, so that stochastic gradient descent visits the examples in random order. \n\n**Exercise**: Follow the instructions and implement `model()`. When `examples[index]` contains one dinosaur name (string), to create an example (X, Y), you can use this:\n\n##### Set the index `idx` into the list of examples\n* Using the for-loop, walk through the shuffled list of dinosaur names in the list \"examples\".\n* If there are 100 examples, and the for-loop increments the index to 100 onwards, think of how you would make the index cycle back to 0, so that we can continue feeding the examples into the model when j is 100, 101, etc.\n* Hint: 101 divided by 100 is zero with a remainder of 1.\n* `%` is the modulus operator in python.\n\n##### Extract a single example from the list of examples\n* `single_example`: use the `idx` index that you set previously to get one word from the list of examples.", "_____no_output_____" ], [ "##### Convert a string into a list of characters: `single_example_chars`\n* `single_example_chars`: A string is a list of characters.\n* You can use a list comprehension (recommended over for-loops) to generate a list of characters.\n```Python\nstr = 'I love learning'\nlist_of_chars = [c for c in str]\nprint(list_of_chars)\n```\n\n```\n['I', ' ', 'l', 'o', 'v', 'e', ' ', 'l', 'e', 'a', 'r', 'n', 'i', 'n', 'g']\n```", "_____no_output_____" ], [ "##### Convert list of characters to a list of integers: `single_example_ix`\n* Create a list that contains the index numbers associated with each character.\n* Use the dictionary `char_to_ix`\n* You can combine this with the list comprehension that is used to get a list of characters from a string.\n* This is a separate line of code below, to help learners clarify each step in the function.", "_____no_output_____" ], [ "##### Create the list of input characters: `X`\n* `rnn_forward` uses the `None` value as a flag to set the input vector as a zero-vector.\n* Prepend the `None` value in front of the list of input characters.\n* There is more than one way to prepend a value to a list. One way is to add two lists together: `['a'] + ['b']`", "_____no_output_____" ], [ "##### Get the integer representation of the newline character `ix_newline`\n* `ix_newline`: The newline character signals the end of the dinosaur name.\n - get the integer representation of the newline character `'\\n'`.\n - Use `char_to_ix`", "_____no_output_____" ], [ "##### Set the list of labels (integer representation of the characters): `Y`\n* The goal is to train the RNN to predict the next letter in the name, so the labels are the list of characters that are one time step ahead of the characters in the input `X`.\n - For example, `Y[0]` contains the same value as `X[1]` \n* The RNN should predict a newline at the last letter so add ix_newline to the end of the labels. \n - Append the integer representation of the newline character to the end of `Y`.\n - Note that `append` is an in-place operation.\n - It might be easier for you to add two lists together.", "_____no_output_____" ] ], [ [ "# GRADED FUNCTION: model\n\ndef model(data, ix_to_char, char_to_ix, num_iterations = 35000, n_a = 50, dino_names = 7, vocab_size = 27):\n \"\"\"\n Trains the model and generates dinosaur names. \n \n Arguments:\n data -- text corpus\n ix_to_char -- dictionary that maps the index to a character\n char_to_ix -- dictionary that maps a character to an index\n num_iterations -- number of iterations to train the model for\n n_a -- number of units of the RNN cell\n dino_names -- number of dinosaur names you want to sample at each iteration. \n vocab_size -- number of unique characters found in the text (size of the vocabulary)\n \n Returns:\n parameters -- learned parameters\n \"\"\"\n \n # Retrieve n_x and n_y from vocab_size\n n_x, n_y = vocab_size, vocab_size\n \n # Initialize parameters\n parameters = initialize_parameters(n_a, n_x, n_y)\n \n # Initialize loss (this is required because we want to smooth our loss)\n loss = get_initial_loss(vocab_size, dino_names)\n \n # Build list of all dinosaur names (training examples).\n with open(\"dinos.txt\") as f:\n examples = f.readlines()\n examples = [x.lower().strip() for x in examples]\n \n # Shuffle list of all dinosaur names\n np.random.seed(0)\n np.random.shuffle(examples)\n \n # Initialize the hidden state of your LSTM\n a_prev = np.zeros((n_a, 1))\n \n # Optimization loop\n for j in range(num_iterations):\n \n ### START CODE HERE ###\n \n # Set the index `idx` (see instructions above)\n idx = j%len(examples)\n \n # Set the input X (see instructions above)\n single_example = examples[idx]\n single_example_chars = [c for c in single_example]\n single_example_ix = [char_to_ix[c] for c in single_example_chars]\n X = [None]+single_example_ix\n \n # Set the labels Y (see instructions above)\n ix_newline = char_to_ix[\"\\n\"]\n Y = X[1:]+[ix_newline]\n \n # Perform one optimization step: Forward-prop -> Backward-prop -> Clip -> Update parameters\n # Choose a learning rate of 0.01\n curr_loss, gradients, a_prev = optimize(X, Y, a_prev, parameters, learning_rate = 0.01)\n \n ### END CODE HERE ###\n \n # Use a latency trick to keep the loss smooth. It happens here to accelerate the training.\n loss = smooth(loss, curr_loss)\n\n # Every 2000 Iteration, generate \"n\" characters thanks to sample() to check if the model is learning properly\n if j % 2000 == 0:\n \n print('Iteration: %d, Loss: %f' % (j, loss) + '\\n')\n \n # The number of dinosaur names to print\n seed = 0\n for name in range(dino_names):\n \n # Sample indices and print them\n sampled_indices = sample(parameters, char_to_ix, seed)\n print_sample(sampled_indices, ix_to_char)\n \n seed += 1 # To get the same result (for grading purposes), increment the seed by one. \n \n print('\\n')\n \n return parameters", "_____no_output_____" ] ], [ [ "Run the following cell, you should observe your model outputting random-looking characters at the first iteration. After a few thousand iterations, your model should learn to generate reasonable-looking names. ", "_____no_output_____" ] ], [ [ "parameters = model(data, ix_to_char, char_to_ix)", "Iteration: 0, Loss: 23.087336\n\nNkzxwtdmfqoeyhsqwasjkjvu\nKneb\nKzxwtdmfqoeyhsqwasjkjvu\nNeb\nZxwtdmfqoeyhsqwasjkjvu\nEb\nXwtdmfqoeyhsqwasjkjvu\n\n\nIteration: 2000, Loss: 27.884160\n\nLiusskeomnolxeros\nHmdaairus\nHytroligoraurus\nLecalosapaus\nXusicikoraurus\nAbalpsamantisaurus\nTpraneronxeros\n\n\nIteration: 4000, Loss: 25.901815\n\nMivrosaurus\nInee\nIvtroplisaurus\nMbaaisaurus\nWusichisaurus\nCabaselachus\nToraperlethosdarenitochusthiamamumamaon\n\n\nIteration: 6000, Loss: 24.608779\n\nOnwusceomosaurus\nLieeaerosaurus\nLxussaurus\nOma\nXusteonosaurus\nEeahosaurus\nToreonosaurus\n\n\nIteration: 8000, Loss: 24.070350\n\nOnxusichepriuon\nKilabersaurus\nLutrodon\nOmaaerosaurus\nXutrcheps\nEdaksoje\nTrodiktonus\n\n\nIteration: 10000, Loss: 23.844446\n\nOnyusaurus\nKlecalosaurus\nLustodon\nOla\nXusodonia\nEeaeosaurus\nTroceosaurus\n\n\nIteration: 12000, Loss: 23.291971\n\nOnyxosaurus\nKica\nLustrepiosaurus\nOlaagrraiansaurus\nYuspangosaurus\nEealosaurus\nTrognesaurus\n\n\nIteration: 14000, Loss: 23.382338\n\nMeutromodromurus\nInda\nIutroinatorsaurus\nMaca\nYusteratoptititan\nCa\nTroclosaurus\n\n\nIteration: 16000, Loss: 23.255630\n\nMeustolkanolus\nIndabestacarospceryradwalosaurus\nJustolopinaveraterasauracoptelalenyden\nMaca\nYusocles\nDaahosaurus\nTrodon\n\n\nIteration: 18000, Loss: 22.905483\n\nPhytronn\nMeicanstolanthus\nMustrisaurus\nPegalosaurus\nYuskercis\nEgalosaurus\nTromelosaurus\n\n\nIteration: 20000, Loss: 22.873854\n\nNlyushanerohyisaurus\nLoga\nLustrhigosaurus\nNedalosaurus\nYuslangosaurus\nElagosaurus\nTrrangosaurus\n\n\nIteration: 22000, Loss: 22.710545\n\nOnyxromicoraurospareiosatrus\nLiga\nMustoffankeugoptardoros\nOla\nYusodogongterosaurus\nEhaerona\nTrododongxernochenhus\n\n\nIteration: 24000, Loss: 22.604827\n\nMeustognathiterhucoplithaloptha\nJigaadosaurus\nKurrodon\nMecaistheansaurus\nYuromelosaurus\nEiaeropeeton\nTroenathiteritaus\n\n\nIteration: 26000, Loss: 22.714486\n\nNhyxosaurus\nKola\nLvrosaurus\nNecalosaurus\nYurolonlus\nEjakosaurus\nTroindronykus\n\n\nIteration: 28000, Loss: 22.647640\n\nOnyxosaurus\nLoceahosaurus\nLustleonlonx\nOlabasicachudrakhurgawamosaurus\nYtrojianiisaurus\nEladon\nTromacimathoshargicitan\n\n\nIteration: 30000, Loss: 22.598485\n\nOryuton\nLocaaesaurus\nLustoendosaurus\nOlaahus\nYusaurus\nEhadopldarshuellus\nTroia\n\n\nIteration: 32000, Loss: 22.211861\n\nMeutronlapsaurus\nKracallthcaps\nLustrathus\nMacairugeanosaurus\nYusidoneraverataus\nEialosaurus\nTroimaniathonsaurus\n\n\nIteration: 34000, Loss: 22.447230\n\nOnyxipaledisons\nKiabaeropa\nLussiamang\nPacaeptabalsaurus\nXosalong\nEiacoteg\nTroia\n\n\n" ] ], [ [ "** Expected Output**\n\nThe output of your model may look different, but it will look something like this:\n\n```Python\nIteration: 34000, Loss: 22.447230\n\nOnyxipaledisons\nKiabaeropa\nLussiamang\nPacaeptabalsaurus\nXosalong\nEiacoteg\nTroia\n```", "_____no_output_____" ], [ "## Conclusion\n\nYou can see that your algorithm has started to generate plausible dinosaur names towards the end of the training. At first, it was generating random characters, but towards the end you could see dinosaur names with cool endings. Feel free to run the algorithm even longer and play with hyperparameters to see if you can get even better results. Our implementation generated some really cool names like `maconucon`, `marloralus` and `macingsersaurus`. Your model hopefully also learned that dinosaur names tend to end in `saurus`, `don`, `aura`, `tor`, etc.\n\nIf your model generates some non-cool names, don't blame the model entirely--not all actual dinosaur names sound cool. (For example, `dromaeosauroides` is an actual dinosaur name and is in the training set.) But this model should give you a set of candidates from which you can pick the coolest! \n\nThis assignment had used a relatively small dataset, so that you could train an RNN quickly on a CPU. Training a model of the english language requires a much bigger dataset, and usually needs much more computation, and could run for many hours on GPUs. We ran our dinosaur name for quite some time, and so far our favorite name is the great, undefeatable, and fierce: Mangosaurus!\n\n<img src=\"images/mangosaurus.jpeg\" style=\"width:250;height:300px;\">", "_____no_output_____" ], [ "## 4 - Writing like Shakespeare\n\nThe rest of this notebook is optional and is not graded, but we hope you'll do it anyway since it's quite fun and informative. \n\nA similar (but more complicated) task is to generate Shakespeare poems. Instead of learning from a dataset of Dinosaur names you can use a collection of Shakespearian poems. Using LSTM cells, you can learn longer term dependencies that span many characters in the text--e.g., where a character appearing somewhere a sequence can influence what should be a different character much much later in the sequence. These long term dependencies were less important with dinosaur names, since the names were quite short. \n\n\n<img src=\"images/shakespeare.jpg\" style=\"width:500;height:400px;\">\n<caption><center> Let's become poets! </center></caption>\n\nWe have implemented a Shakespeare poem generator with Keras. Run the following cell to load the required packages and models. This may take a few minutes. ", "_____no_output_____" ] ], [ [ "from __future__ import print_function\nfrom keras.callbacks import LambdaCallback\nfrom keras.models import Model, load_model, Sequential\nfrom keras.layers import Dense, Activation, Dropout, Input, Masking\nfrom keras.layers import LSTM\nfrom keras.utils.data_utils import get_file\nfrom keras.preprocessing.sequence import pad_sequences\nfrom shakespeare_utils import *\nimport sys\nimport io", "Using TensorFlow backend.\n" ] ], [ [ "To save you some time, we have already trained a model for ~1000 epochs on a collection of Shakespearian poems called [*\"The Sonnets\"*](shakespeare.txt). ", "_____no_output_____" ], [ "Let's train the model for one more epoch. When it finishes training for an epoch---this will also take a few minutes---you can run `generate_output`, which will prompt asking you for an input (`<`40 characters). The poem will start with your sentence, and our RNN-Shakespeare will complete the rest of the poem for you! For example, try \"Forsooth this maketh no sense \" (don't enter the quotation marks). Depending on whether you include the space at the end, your results might also differ--try it both ways, and try other inputs as well. \n", "_____no_output_____" ] ], [ [ "print_callback = LambdaCallback(on_epoch_end=on_epoch_end)\n\nmodel.fit(x, y, batch_size=128, epochs=1, callbacks=[print_callback])", "Epoch 1/1\n31412/31412 [==============================] - 263s - loss: 2.5635 \n" ], [ "# Run this cell to try with different inputs without having to re-train the model \ngenerate_output()", "Write the beginning of your poem, the Shakespeare machine will complete it. Your input is: Forsooth this maketh no sense \n\n\nHere is your poem: \n\nForsooth this maketh no sense renping.\nbut a did sind make hil ons dede is men,\nwithou's inus will o decanot,\nlek o whle thou debert should every forted,\nwhice muse whe bow way i hath, wom's (leccanny,\nthat minge in adited and forso caned doass,\nwith this of ares mote di have no lade peres,\nin live mater for worre la cinse with will when thy comserd srobld,\nnow, boroor, my ho thate thought dewhice hevinging,\nnow not his wrecio" ] ], [ [ "The RNN-Shakespeare model is very similar to the one you have built for dinosaur names. The only major differences are:\n- LSTMs instead of the basic RNN to capture longer-range dependencies\n- The model is a deeper, stacked LSTM model (2 layer)\n- Using Keras instead of python to simplify the code \n\nIf you want to learn more, you can also check out the Keras Team's text generation implementation on GitHub: https://github.com/keras-team/keras/blob/master/examples/lstm_text_generation.py.\n\nCongratulations on finishing this notebook! ", "_____no_output_____" ], [ "**References**:\n- This exercise took inspiration from Andrej Karpathy's implementation: https://gist.github.com/karpathy/d4dee566867f8291f086. To learn more about text generation, also check out Karpathy's [blog post](http://karpathy.github.io/2015/05/21/rnn-effectiveness/).\n- For the Shakespearian poem generator, our implementation was based on the implementation of an LSTM text generator by the Keras team: https://github.com/keras-team/keras/blob/master/examples/lstm_text_generation.py ", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "import cv2 as cv\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport argparse", "_____no_output_____" ], [ "net = cv.dnn.readNetFromTensorflow(\"graph_opt.pb\")", "_____no_output_____" ], [ "inWidth = 368\ninHeight = 368\nthr = 0.2", "_____no_output_____" ], [ "BODY_PARTS = { \"Nose\": 0, \"Neck\": 1, \"RShoulder\": 2, \"RElbow\": 3, \"RWrist\": 4,\n \"LShoulder\": 5, \"LElbow\": 6, \"LWrist\": 7, \"RHip\": 8, \"RKnee\": 9,\n \"RAnkle\": 10, \"LHip\": 11, \"LKnee\": 12, \"LAnkle\": 13, \"REye\": 14,\n \"LEye\": 15, \"REar\": 16, \"LEar\": 17, \"Background\": 18 }\n\nPOSE_PAIRS = [ [\"Neck\", \"RShoulder\"], [\"Neck\", \"LShoulder\"], [\"RShoulder\", \"RElbow\"],\n [\"RElbow\", \"RWrist\"], [\"LShoulder\", \"LElbow\"], [\"LElbow\", \"LWrist\"],\n [\"Neck\", \"RHip\"], [\"RHip\", \"RKnee\"], [\"RKnee\", \"RAnkle\"], [\"Neck\", \"LHip\"],\n [\"LHip\", \"LKnee\"], [\"LKnee\", \"LAnkle\"], [\"Neck\", \"Nose\"], [\"Nose\", \"REye\"],\n [\"REye\", \"REar\"], [\"Nose\", \"LEye\"], [\"LEye\", \"LEar\"] ]\n", "_____no_output_____" ], [ "img = cv.imread(\"image.jpg\")", "_____no_output_____" ], [ "plt.imshow(img)", "_____no_output_____" ], [ "plt.imshow(cv.cvtColor(img,cv.COLOR_BGR2RGB))", "_____no_output_____" ], [ "def pose_estiamtion(frame):\n frameWidth = frame.shape[1]\n frameHeight = frame.shape[0]\n net.setInput(cv.dnn.blobFromImage(frame, 1.0, (inWidth, inHeight), (127.5, 127.5, 127.5), swapRB=True, crop=False))\n out = net.forward()\n out = out[:, :19, :, :] # MobileNet output [1, 57, -1, -1], we only need the first 19 elements\n assert(len(BODY_PARTS) == out.shape[1])\n points = []\n for i in range(len(BODY_PARTS)):\n # Slice heatmap of corresponging body's part.\n heatMap = out[0, i, :, :]\n\n # Originally, we try to find all the local maximums. To simplify a sample\n # we just find a global one. However only a single pose at the same time\n # could be detected this way.\n _, conf, _, point = cv.minMaxLoc(heatMap)\n x = (frameWidth * point[0]) / out.shape[3]\n y = (frameHeight * point[1]) / out.shape[2]\n # Add a point if it's confidence is higher than threshold.\n points.append((int(x), int(y)) if conf > thr else None)\n \n for pair in POSE_PAIRS:\n partFrom = pair[0]\n partTo = pair[1]\n assert(partFrom in BODY_PARTS)\n assert(partTo in BODY_PARTS)\n\n idFrom = BODY_PARTS[partFrom]\n idTo = BODY_PARTS[partTo]\n\n if points[idFrom] and points[idTo]:\n cv.line(frame, points[idFrom], points[idTo], (0, 255, 0), 3)\n cv.ellipse(frame, points[idFrom], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED)\n cv.ellipse(frame, points[idTo], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED)\n \n t, _ = net.getPerfProfile()\n freq = cv.getTickFrequency() / 1000\n cv.putText(frame, '%.2fms' % (t / freq), (10, 20), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0))\n return frame", "_____no_output_____" ], [ "estimated_image = pose_estiamtion(img)", "_____no_output_____" ], [ "plt.imshow(cv.cvtColor(estimated_image,cv.COLOR_BGR2RGB))", "_____no_output_____" ], [ "cap = cv.VideoCapture(1)\ncap.set(cv.CAP_PROP_FPS,10)\ncap.set(3,800)\ncap.set(4,800)\n\nif not cap.isOpened():\n cap = cv.VideoCapture(0)\nif not cap.isOpened():\n raise IOError(\"Cannot open the Webcame\")\n\nwhile cv.waitKey(1) < 0 :\n hasFrame, frame = cap.read()\n if not hasFrame:\n cv.waitKey()\n break\n frameWidth = frame.shape[1]\n frameHeight = frame.shape[0]\n net.setInput(cv.dnn.blobFromImage(frame, 1.0, (inWidth, inHeight), (127.5, 127.5, 127.5), swapRB=True, crop=False))\n out = net.forward()\n out = out[:, :19, :, :] # MobileNet output [1, 57, -1, -1], we only need the first 19 elements\n assert(len(BODY_PARTS) == out.shape[1])\n points = []\n for i in range(len(BODY_PARTS)):\n # Slice heatmap of corresponging body's part.\n heatMap = out[0, i, :, :]\n\n # Originally, we try to find all the local maximums. To simplify a sample\n # we just find a global one. However only a single pose at the same time\n # could be detected this way.\n _, conf, _, point = cv.minMaxLoc(heatMap)\n x = (frameWidth * point[0]) / out.shape[3]\n y = (frameHeight * point[1]) / out.shape[2]\n # Add a point if it's confidence is higher than threshold.\n points.append((int(x), int(y)) if conf > thr else None)\n \n for pair in POSE_PAIRS:\n partFrom = pair[0]\n partTo = pair[1]\n assert(partFrom in BODY_PARTS)\n assert(partTo in BODY_PARTS)\n\n idFrom = BODY_PARTS[partFrom]\n idTo = BODY_PARTS[partTo]\n\n if points[idFrom] and points[idTo]:\n cv.line(frame, points[idFrom], points[idTo], (0, 255, 0), 3)\n cv.ellipse(frame, points[idFrom], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED)\n cv.ellipse(frame, points[idTo], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED)\n \n t, _ = net.getPerfProfile()\n freq = cv.getTickFrequency() / 1000\n cv.putText(frame, '%.2fms' % (t / freq), (10, 20), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0))\n \n cv.imshow('Pose Estimtion Tutorial',frame)", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Moodle Database: Educational Data Log Analysis \nThe Moodle LMS is a free and open-source learning management system written in PHP and distributed under the GNU General Public License. It is used for blended learning, distance education, flipped classroom and other e-learning projects in schools, universities, workplaces and other sectors. With customizable management features, it is used to create private websites with online courses for educators and trainers to achieve learning goals. Moodle allows for extending and tailoring learning environments using community-sourced plugins .\n\nIn this notebokk we are going to explore the 10 Academy Moodle logs stored in the database together with many other relevant tables. ", "_____no_output_____" ], [ "# Table of content\n1. Installing the required libraries \n2. Importing the required libraries \n3. Moodle database understanding \n4. Data Extraction Transformation and Loading (ETL)", "_____no_output_____" ], [ "### Installing the necessary libraries", "_____no_output_____" ] ], [ [ "#!pip install ipython-sql\n#!pip install sqlalchemy\n#!pip install psycopg2", "_____no_output_____" ] ], [ [ "### Importing necessary libraries", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport numpy as np\nfrom sqlalchemy import create_engine\nimport psycopg2\nimport logging\nfrom IPython.display import display", "_____no_output_____" ], [ "#allowing connection to the database\n%load_ext sql", "The sql extension is already loaded. To reload it, use:\n %reload_ext sql\n" ], [ "#ipython-sql\n%sql postgresql://bessy:Streetdance53@localhost/moodle", "_____no_output_____" ], [ "#sqlalchemy\nengine = create_engine('postgresql://bessy:Streetdance53@localhost/moodle')", "_____no_output_____" ] ], [ [ "### Moodle database Understanding.\nNow, let's have a glance of how some of the tables look like.We will consider the following tables;\n`mdl_logstore_standard_log`,\n`mdl_context`,\n`mdl_user`,\n`mdl_course`,\n`mdl_modules `,\n`mdl_course_modules`,\n`mdl_course_modules_completion`, \n`mdl_grade_items`,\n`mdl_grade_grades`,\n`mdl_grade_categories`,\n`mdl_grade_items_history`,\n`mdl_grade_grades_history`,\n`mdl_grade_categories_history`,\n`mdl_forum`,\n`mdl_forum_discussions`,\n`mdl_forum_posts`.", "_____no_output_____" ], [ "`Table:mdl_logstore_standard_log`", "_____no_output_____" ] ], [ [ "%%sql\nSELECT *FROM mdl_logstore_standard_log LIMIT 3;", " * postgresql://bessy:***@localhost/moodle\n3 rows affected.\n" ] ], [ [ "`Table: mdl_context`", "_____no_output_____" ] ], [ [ "%%sql\nSELECT * FROM mdl_context LIMIT 3;", " * postgresql://bessy:***@localhost/moodle\n3 rows affected.\n" ] ], [ [ "`mdl_course`", "_____no_output_____" ] ], [ [ "%%sql\nSELECT * FROM mdl_course LIMIT 3;", " * postgresql://bessy:***@localhost/moodle\n3 rows affected.\n" ] ], [ [ "`mdl_user`", "_____no_output_____" ] ], [ [ "%%sql\nSELECT * FROM mdl_user LIMIT 3;", " * postgresql://bessy:***@localhost/moodle\n3 rows affected.\n" ] ], [ [ "`mdl_modules`", "_____no_output_____" ] ], [ [ "%%sql\nSELECT * FROM mdl_modules LIMIT 3;", " * postgresql://bessy:***@localhost/moodle\n3 rows affected.\n" ] ], [ [ "`mdl_course_modules`", "_____no_output_____" ] ], [ [ "%%sql\nSELECT * FROM mdl_course_modules LIMIT 3;", " * postgresql://bessy:***@localhost/moodle\n3 rows affected.\n" ] ], [ [ "`mdl_course_modules_completion`", "_____no_output_____" ] ], [ [ "%%sql\nSELECT * FROM mdl_course_modules_completion LIMIT 3", " * postgresql://bessy:***@localhost/moodle\n3 rows affected.\n" ] ], [ [ "`mdl_grade_grades`", "_____no_output_____" ] ], [ [ "%%sql\nSELECT * FROM mdl_grade_grades LIMIT 3", " * postgresql://bessy:***@localhost/moodle\n3 rows affected.\n" ] ], [ [ "### Number of tables in the database;", "_____no_output_____" ] ], [ [ "%%sql\nSELECT COUNT(*) FROM information_schema.tables", " * postgresql://bessy:***@localhost/moodle\n1 rows affected.\n" ] ], [ [ "### Number of records in the following tables; \n", "_____no_output_____" ] ], [ [ "mit = ['mdl_logstore_standard_log', 'mdl_context', 'mdl_user', 'mdl_course', 'mdl_modules' , 'mdl_course_modules', 'mdl_course_modules_completion',\n 'mdl_grade_items', 'mdl_grade_grades', 'mdl_grade_categories', 'mdl_grade_items_history', 'mdl_grade_grades_history', \n 'mdl_grade_categories_history', 'mdl_forum', 'mdl_forum_discussions', 'mdl_forum_posts']\n\n# fetches and returns number of records of a given table in a moodle database\ndef table_count(table):\n count = %sql SELECT COUNT(*) as {table}_count from {table}\n return count\n\nfor table in mit:\n display(table_count(table))", " * postgresql://bessy:***@localhost/moodle\n1 rows affected.\n" ] ], [ [ "### Number of quiz submission by time", "_____no_output_____" ] ], [ [ "%%sql\nselect date_part('hour', timestamp with time zone 'epoch' + timefinish * interval '1 second') as hour, count(1)\nfrom mdl_quiz_attempts qa\nwhere qa.preview = 0 and qa.timefinish <> 0\ngroup by date_part('hour', timestamp with time zone 'epoch' + timefinish * interval '1 second')\norder by hour", " * postgresql://bessy:***@localhost/moodle\n24 rows affected.\n" ], [ "%%sql\nSELECT COUNT(id), EXTRACT(HOUR FROM to_timestamp(timecreated)) FROM mdl_logstore_standard_log WHERE action ='submitted' AND component='mod_quiz'\ngroup by EXTRACT(HOUR FROM to_timestamp(timecreated));", " * postgresql://bessy:***@localhost/moodle\n24 rows affected.\n" ] ], [ [ "## Monthly usage time of learners who have confirmed and are not deleted", "_____no_output_____" ] ], [ [ "%%sql\nselect extract(month from to_timestamp(mdl_stats_user_monthly.timeend)) as calendar_month,\n count(distinct mdl_stats_user_monthly.userid) as total_users\nfrom mdl_stats_user_monthly\n inner join mdl_role_assignments on mdl_stats_user_monthly.userid = mdl_role_assignments.userid\n inner join mdl_context on mdl_role_assignments.contextid = mdl_context.id\nwhere mdl_stats_user_monthly.stattype = 'activity'\n and mdl_stats_user_monthly.courseid <>1\ngroup by extract(month from to_timestamp(mdl_stats_user_monthly.timeend))\norder by extract(month from to_timestamp(mdl_stats_user_monthly.timeend))", " * postgresql://bessy:***@localhost/moodle\n2 rows affected.\n" ], [ "%%sql \nSELECT COUNT(lastaccess - firstaccess) AS usagetime, EXTRACT (MONTH FROM to_timestamp(firstaccess)) AS month \nFROM mdl_user WHERE confirmed = 1 AND deleted = 0 GROUP BY EXTRACT (MONTH FROM to_timestamp(firstaccess))", " * postgresql://bessy:***@localhost/moodle\n7 rows affected.\n" ] ], [ [ "## Count of log events per user", "_____no_output_____" ] ], [ [ "actions = ['loggedin', 'viewed', 'started', 'submitted', 'uploaded', 'updated', 'searched', \n 'answered', 'attempted', 'abandoned']\n\n# fetch and return count of log events of an action per user\ndef event_count(action):\n count = %sql SELECT userid, COUNT(action) AS {action}_count FROM mdl_logstore_standard_log WHERE action='{action}' GROUP BY userid limit 5\n return count\n\nfor action in actions:\n display(event_count(action))", " * postgresql://bessy:***@localhost/moodle\n5 rows affected.\n" ] ], [ [ "### python class to pull \n* Overall grade of learners \n* Number of forum posts\n", "_____no_output_____" ] ], [ [ "class PullGrade():\n def __init__(self):\n pass\n \n def open_db(self, **kwargs):\n # extract args, if they are not provided assign a default value\n user = kwargs.get('user', 'briodev')\n password = kwargs.get('password', '14ConnectPsq')\n db = kwargs.get('db', 'moodle')\n \n # make a connection to PostgreSQL\n # use exception to show error message if failed to connect\n try:\n params = dict(user=user, \n password=password,\n host=\"127.0.0.1\",\n port = \"5432\",\n database = db)\n proot = 'postgresql://{user}@{host}:5432/{database}'.format(**params)\n logging.info('Connecting to the PostgreSQL database... using sqlalchemy engine')\n engine = create_engine(proot)\n except (Exception, psycopg2.Error) as error:\n logging.error(r\"Error while connecting to PostgreSQL {error}\")\n \n return engine\n \n # fetch and return number of forum posts\n def forum_posts(self):\n count = %sql SELECT COUNT(*) from mdl_forum_posts\n return count\n \n # fetch and return overall grade of learners\n def overall_grade(self):\n overall = %sql SELECT userid, round(SUM(finalgrade)/count(*), 2) as overall_grade from mdl_grade_grades WHERE finalgrade is not null group by userid LIMIT 10\n return overall", "_____no_output_____" ], [ "db = PullGrade()\ndb.open_db()", "_____no_output_____" ], [ "#Forum_posts\ndb.forum_posts()", " * postgresql://bessy:***@localhost/moodle\n1 rows affected.\n" ], [ "#Overall grade.\ndb.overall_grade()", " * postgresql://bessy:***@localhost/moodle\n10 rows affected.\n" ] ], [ [ "### Data Extraction Transformation and Loading (ETL)", "_____no_output_____" ] ], [ [ "#reading the mdl_logstore_standard_log\nlog_df = pd.read_sql(\"select * from mdl_logstore_standard_log\", engine)", "_____no_output_____" ], [ "def top_x(df, percent):\n total_len = df.shape[0]\n top = int((total_len * percent)/100)\n return df.iloc[:top,]", "_____no_output_____" ] ], [ [ "### Login count", "_____no_output_____" ] ], [ [ "log_df_logged_in = log_df[log_df.action == 'loggedin'][['userid', 'action']]\nlogin_by_user = log_df_logged_in.groupby('userid').count().sort_values('action', ascending=False)", "_____no_output_____" ], [ "login_by_user.columns = [\"login_count\"]\ntop_x(login_by_user, 1)", "_____no_output_____" ] ], [ [ "### Activity count", "_____no_output_____" ] ], [ [ "activity_log = log_df[['userid', 'action']]\nactivity_log_by_user = activity_log.groupby('userid').count().sort_values('action', ascending=False)", "_____no_output_____" ], [ "activity_log_by_user.columns = ['activity_count']\ntop_x(activity_log_by_user, 1)", "_____no_output_____" ], [ "log_in_out = log_df[(log_df.action == \"loggedin\") | (log_df.action == \"loggedout\")]", "_____no_output_____" ], [ "user_id = log_df.userid.unique()\n\nd_times = {}\n\nfor user in user_id:\n log_user = log_df[log_df.userid == user].sort_values('timecreated')\n \n d_time = 0 \n isLoggedIn = 0\n loggedIn_timecreated = 0\n \n for i in range(len(log_user)): \n row = log_user.iloc[i,]\n \n row_next = log_user.iloc[i+1,] if i+1 < len(log_user) else row\n \n if(row.action == \"loggedin\"): \n isLoggedIn = 1\n loggedIn_timecreated = row.timecreated\n\n if( (i+1 == len(log_user)) | ( (row_next.action == \"loggedin\") & (isLoggedIn == 1) ) ):\n d_time += row.timecreated - loggedIn_timecreated\n isLoggedIn = 0\n\n d_times[user] = d_time\n", "_____no_output_____" ], [ "\ndedication_time_df = pd.DataFrame({'userid':list(d_times.keys()),\n 'dedication_time':list(d_times.values())})", "_____no_output_____" ], [ "dedication_time_df", "_____no_output_____" ], [ "top_x(dedication_time_df.sort_values('dedication_time', ascending=False), 35)", "_____no_output_____" ] ], [ [ "### References\n*\thttps://docs.moodle.org/39/en/Custom_SQL_queries_report\n*\thttps://docs.moodle.org/39/en/ad-hoc_contributed_reports\n*\thttps://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.331.667&rep=rep1&type=pdf\n*\thttp://informatics.ue-varna.bg/conference19/Conf.proceedings_Informatics-50.years%20177-187.pdf\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Lecture 4\n\n* Optidigits notebook in day 2 directory should be looked at.\n* Slide \"Other tools\" feature index vs row index, `x.value_counts()`, `x.isnull()` should be used in report\n* `pd.scatter_matrix(df)`, `plt.matshow(df.corr(df))`. Group features.\n* Use test set without labels in exploratory data analysis.\n* NA values should be replaced with the `MICE` R-package, subshelling out.", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Format DataFrame", "_____no_output_____" ] ], [ [ "import pandas as pd\nfrom sklearn.datasets import make_regression\n\ndata = make_regression(n_samples=600, n_features=50, noise=0.1, random_state=42)\ntrain_df = pd.DataFrame(data[0], columns=[\"x_{}\".format(_) for _ in range(data[0].shape[1])])\ntrain_df[\"target\"] = data[1]\n\nprint(train_df.shape)\ntrain_df.head()", "(600, 51)\n" ] ], [ [ "# Set Up Environment", "_____no_output_____" ] ], [ [ "from hyperparameter_hunter import Environment, CVExperiment\nfrom sklearn.metrics import explained_variance_score\n\nenv = Environment(\n train_dataset=train_df,\n results_path=\"HyperparameterHunterAssets\",\n metrics=dict(evs=explained_variance_score),\n cv_type=\"KFold\",\n cv_params=dict(n_splits=3, shuffle=True, random_state=1337),\n runs=2,\n)", "Cross-Experiment Key: 'Sxkz36nLbTyi4QJM7w6wCWkbIucXm0RbzMsirLPuYmw='\n" ] ], [ [ "Now that HyperparameterHunter has an active `Environment`, we can do two things:\n\n# 1. Perform Experiments\n\n*Note: If this is your first HyperparameterHunter example, the CatBoost classification example may be a better starting point.*\n\nIn this Experiment, we're also going to use `model_extra_params` to provide arguments to `CatBoostRegressor`'s `fit` method, just like we would if we weren't using HyperparameterHunter.\n\nWe'll be using the `verbose` argument to print evaluations of our `CatBoostRegressor` every 50 iterations, and we'll also be using the dataset sentinels offered by `Environment`. You can read more about the exciting thing you can do with the `Environment` sentinels in the documentation and in the example dedicated to them. For now, though, we'll be using them to provide each fold's `env.validation_input`, and `env.validation_target` to `CatBoostRegressor.fit` via its `eval_set` argument.\n\nYou could also easily add `CatBoostRegressor.fit`'s `early_stopping_rounds` argument to `model_extra_params[\"fit\"]` to use early stopping, but doing so here with only `iterations=100` doesn't make much sense.", "_____no_output_____" ] ], [ [ "from catboost import CatBoostRegressor\n\nexperiment = CVExperiment(\n model_initializer=CatBoostRegressor,\n model_init_params=dict(\n iterations=100,\n learning_rate=0.05,\n depth=5,\n bootstrap_type=\"Bayesian\",\n save_snapshot=False,\n allow_writing_files=False,\n ),\n model_extra_params=dict(\n fit=dict(\n verbose=50,\n eval_set=[(env.validation_input, env.validation_target)],\n ),\n ),\n)", "<22:13:22> Validated Environment: 'Sxkz36nLbTyi4QJM7w6wCWkbIucXm0RbzMsirLPuYmw='\n<22:13:22> Initialized Experiment: 'df2982f6-ab54-4c74-80b6-ad1c1e49fe45'\n<22:13:22> Hyperparameter Key: '913_iDDPY_PMp5ulOyK251mgLCZ2qau7kjOvDdy1rz8='\n<22:13:22> \n0:\tlearn: 181.0383680\ttest: 179.9132919\tbest: 179.9132919 (0)\ttotal: 63ms\tremaining: 6.24s\n50:\tlearn: 97.8034574\ttest: 110.2078251\tbest: 110.2078251 (50)\ttotal: 427ms\tremaining: 410ms\n99:\tlearn: 65.3572502\ttest: 86.2235253\tbest: 86.2235253 (99)\ttotal: 783ms\tremaining: 0us\n\nbestTest = 86.22352531\nbestIteration = 99\n\n<22:13:22> F0/R0 | OOF(evs=0.77530) | Time Elapsed: 0.81737 s\n0:\tlearn: 180.3273426\ttest: 178.7632887\tbest: 178.7632887 (0)\ttotal: 7.5ms\tremaining: 742ms\n50:\tlearn: 95.3065291\ttest: 108.1124182\tbest: 108.1124182 (50)\ttotal: 373ms\tremaining: 358ms\n99:\tlearn: 63.7322170\ttest: 85.4546004\tbest: 85.4546004 (99)\ttotal: 724ms\tremaining: 0us\n\nbestTest = 85.45460041\nbestIteration = 99\n\n<22:13:23> F0/R1 | OOF(evs=0.77929) | Time Elapsed: 0.73847 s\n<22:13:23> F0.0 AVG: OOF(evs=0.77998) | Time Elapsed: 1.56024 s\n0:\tlearn: 183.5077544\ttest: 173.1917854\tbest: 173.1917854 (0)\ttotal: 7.24ms\tremaining: 717ms\n50:\tlearn: 100.5844263\ttest: 108.1710145\tbest: 108.1710145 (50)\ttotal: 375ms\tremaining: 360ms\n99:\tlearn: 67.5414394\ttest: 85.5053407\tbest: 85.5053407 (99)\ttotal: 734ms\tremaining: 0us\n\nbestTest = 85.5053407\nbestIteration = 99\n\n<22:13:24> F1/R0 | OOF(evs=0.76335) | Time Elapsed: 0.7484 s\n0:\tlearn: 183.4412848\ttest: 172.7748236\tbest: 172.7748236 (0)\ttotal: 7.37ms\tremaining: 730ms\n50:\tlearn: 98.9207169\ttest: 107.3348632\tbest: 107.3348632 (50)\ttotal: 372ms\tremaining: 357ms\n99:\tlearn: 67.1194369\ttest: 84.1135630\tbest: 84.1135630 (99)\ttotal: 728ms\tremaining: 0us\n\nbestTest = 84.11356298\nbestIteration = 99\n\n<22:13:25> F1/R1 | OOF(evs=0.77169) | Time Elapsed: 0.74292 s\n<22:13:25> F0.1 AVG: OOF(evs=0.77061) | Time Elapsed: 1.49607 s\n0:\tlearn: 176.0961712\ttest: 188.8149185\tbest: 188.8149185 (0)\ttotal: 6.9ms\tremaining: 683ms\n50:\tlearn: 94.2771704\ttest: 120.4426175\tbest: 120.4426175 (50)\ttotal: 375ms\tremaining: 360ms\n99:\tlearn: 67.9729670\ttest: 99.9417843\tbest: 99.9417843 (99)\ttotal: 730ms\tremaining: 0us\n\nbestTest = 99.94178427\nbestIteration = 99\n\n<22:13:25> F2/R0 | OOF(evs=0.72638) | Time Elapsed: 0.74454 s\n0:\tlearn: 177.0392612\ttest: 189.8521136\tbest: 189.8521136 (0)\ttotal: 6.98ms\tremaining: 691ms\n50:\tlearn: 95.5295037\ttest: 121.3560567\tbest: 121.3560567 (50)\ttotal: 377ms\tremaining: 362ms\n99:\tlearn: 66.3592641\ttest: 99.0826473\tbest: 99.0826473 (99)\ttotal: 736ms\tremaining: 0us\n\nbestTest = 99.08264731\nbestIteration = 99\n\n<22:13:26> F2/R1 | OOF(evs=0.73186) | Time Elapsed: 0.75069 s\n<22:13:26> F0.2 AVG: OOF(evs=0.73123) | Time Elapsed: 1.49972 s\n<22:13:26> \n<22:13:26> FINAL: OOF(evs=0.75971) | Time Elapsed: 4.56196 s\n<22:13:26> \n<22:13:26> Saving results for Experiment: 'df2982f6-ab54-4c74-80b6-ad1c1e49fe45'\n" ] ], [ [ "Notice above that CatBoost printed scores for our `eval_set` every 50 iterations just like we said in `model_extra_params[\"fit\"]`; although, it made our results rather difficult to read, so we'll switch back to `verbose=False` during optimization.\n\n# 2. Hyperparameter Optimization\n\nNotice below that `optimizer` still recognizes the results of `experiment` as valid learning material even though their `verbose` values differ. This is because it knows that `verbose` has no effect on actual results.", "_____no_output_____" ] ], [ [ "from hyperparameter_hunter import DummyOptPro, Real, Integer, Categorical\n\noptimizer = DummyOptPro(iterations=10, random_state=777)\n\noptimizer.forge_experiment(\n model_initializer=CatBoostRegressor,\n model_init_params=dict(\n iterations=100,\n learning_rate=Real(0.001, 0.2),\n depth=Integer(3, 7),\n bootstrap_type=Categorical([\"Bayesian\", \"Bernoulli\"]),\n save_snapshot=False,\n allow_writing_files=False,\n ),\n model_extra_params=dict(\n fit=dict(\n verbose=False,\n eval_set=[(env.validation_input, env.validation_target)],\n ),\n ),\n)\n\noptimizer.go()", "Validated Environment with key: \"Sxkz36nLbTyi4QJM7w6wCWkbIucXm0RbzMsirLPuYmw=\"\n\u001b[31mSaved Result Files\u001b[0m\n\u001b[31m_________________________________________________________________________________________\u001b[0m\n Step | ID | Time | Value | bootstrap_type | depth | learning_rate | \nExperiments matching cross-experiment key/algorithm: 1\nExperiments fitting in the given space: 1\nExperiments matching current guidelines: 1\n 0 | df2982f6 | 00m00s | \u001b[35m 0.75971\u001b[0m | \u001b[32m Bayesian\u001b[0m | \u001b[32m 5\u001b[0m | \u001b[32m 0.0500\u001b[0m | \n\u001b[31mHyperparameter Optimization\u001b[0m\n\u001b[31m_________________________________________________________________________________________\u001b[0m\n Step | ID | Time | Value | bootstrap_type | depth | learning_rate | \n 1 | e21de31a | 00m19s | \u001b[35m 0.86756\u001b[0m | \u001b[32m Bayesian\u001b[0m | \u001b[32m 7\u001b[0m | \u001b[32m 0.1719\u001b[0m | \n 2 | ea45250b | 00m02s | \u001b[35m 0.89994\u001b[0m | \u001b[32m Bayesian\u001b[0m | \u001b[32m 4\u001b[0m | \u001b[32m 0.1510\u001b[0m | \n 3 | dc911fbd | 00m19s | 0.86487 | Bernoulli | 7 | 0.1858 | \n 4 | a344b5f7 | 00m04s | 0.89580 | Bayesian | 5 | 0.1656 | \n 5 | 186a787e | 00m04s | 0.89318 | Bernoulli | 5 | 0.1329 | \n 6 | 30d43916 | 00m02s | \u001b[35m 0.90540\u001b[0m | \u001b[32m Bernoulli\u001b[0m | \u001b[32m 4\u001b[0m | \u001b[32m 0.1422\u001b[0m | \n 7 | bfe06f05 | 00m08s | 0.40699 | Bayesian | 6 | 0.0109 | \n 8 | 514bf91a | 00m19s | 0.87122 | Bernoulli | 7 | 0.1991 | \n 9 | 2109f8fb | 00m08s | 0.84663 | Bayesian | 6 | 0.0910 | \n 10 | ebd045eb | 00m02s | \u001b[35m 0.92369\u001b[0m | \u001b[32m Bernoulli\u001b[0m | \u001b[32m 4\u001b[0m | \u001b[32m 0.1999\u001b[0m | \nOptimization loop completed in 0:01:34.529025\nBest score was 0.9236872318041063 from Experiment \"ebd045eb-1224-49f7-89f3-9ff37bd0fdb7\"\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "#!/usr/bin/env python\n\"\"\"File format conversion\ncategory: vtk, file conversion, tomb\"\"\"\nimport os, sys\nimport vtk\n\ninvtkfile='./tubes.vtk'\noutvtpfile ='tubes.vtp'\n\ndef vtk2vtp(invtkfile, outvtpfile, binary=False):\n \"\"\"What it says on the label\"\"\"\n reader = vtk.vtkPolyDataReader()\n reader.SetFileName(invtkfile)\n\n writer = vtk.vtkXMLPolyDataWriter()\n writer.SetFileName(outvtpfile)\n if binary:\n writer.SetFileTypeToBinary()\n writer.SetInput(reader.GetOutput())\n writer.Update()\n\nif __name__ == '__main__':\n args = sys.argv\n binary = False\n if '-b' in args:\n args.remove('-b')\n binary = True\n if len(args) < 2:\n print('Batch converts vtk files to vtp files.\\nUsage:\\n vtk2vtp.py model1.vtk model2.vtk ...')\n print(' [-b] causes output to be in binary format, much smaller vtp file size, if it happens to work')\n sys.exit()\n infiles = args[1:]\n for vtkfile in infiles:\n if vtkfile[-4:] != '.vtk':\n print (vtkfile, \"doesn't look like a vtk file, won't convert\")\n continue\n vtk2vtp(vtkfile, vtkfile[:-4]+'.vtp', binary=binary)", "-f doesn't look like a vtk file, won't convert\n/home/user/.local/share/jupyter/runtime/kernel-3ceff57c-4ef5-4097-bce5-9b9b56b9707c.json doesn't look like a vtk file, won't convert\n" ], [ "invtkfile='./tubes.vtk'\noutvtpfile ='./tubes.vtp'\nreader = vtk.vtkPolyDataReader()\nreader.SetFileName(invtkfile)", "_____no_output_____" ], [ "writer = vtk.vtkXMLPolyDataWriter()\nwriter.SetFileName(outvtpfile)\nwriter.SetInputData(reader.GetOutput())\nwriter.Update()", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Durables vs Non Durables At Low And High Frequencies", "_____no_output_____" ] ], [ [ "!pip install numpy\n!pip install matplotlib \n!pip install pandas\n!pip install pandas_datareader\n!pip install datetime\n!pip install seaborn\n\n# Some initial setup\nfrom matplotlib import pyplot as plt\nimport numpy as np\nplt.style.use('seaborn-darkgrid')\nimport pandas as pd\nimport pandas_datareader.data as web\nimport datetime\nimport seaborn as sns", "Requirement already satisfied: numpy in c:\\users\\mateo\\anaconda3\\lib\\site-packages (1.18.5)\nRequirement already satisfied: matplotlib in c:\\users\\mateo\\anaconda3\\lib\\site-packages (3.2.2)\nRequirement already satisfied: numpy>=1.11 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib) (1.18.5)\nRequirement already satisfied: kiwisolver>=1.0.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib) (1.2.0)\nRequirement already satisfied: cycler>=0.10 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib) (0.10.0)\nRequirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib) (2.4.7)\nRequirement already satisfied: python-dateutil>=2.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib) (2.8.1)\nRequirement already satisfied: six in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from cycler>=0.10->matplotlib) (1.15.0)\nRequirement already satisfied: pandas in c:\\users\\mateo\\anaconda3\\lib\\site-packages (1.0.5)\nRequirement already satisfied: pytz>=2017.2 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas) (2020.1)\nRequirement already satisfied: numpy>=1.13.3 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas) (1.18.5)\nRequirement already satisfied: python-dateutil>=2.6.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas) (2.8.1)\nRequirement already satisfied: six>=1.5 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from python-dateutil>=2.6.1->pandas) (1.15.0)\nRequirement already satisfied: pandas_datareader in c:\\users\\mateo\\anaconda3\\lib\\site-packages (0.9.0)\nRequirement already satisfied: pandas>=0.23 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas_datareader) (1.0.5)\nRequirement already satisfied: requests>=2.19.0 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas_datareader) (2.24.0)\nRequirement already satisfied: lxml in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas_datareader) (4.5.2)\nRequirement already satisfied: python-dateutil>=2.6.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas>=0.23->pandas_datareader) (2.8.1)\nRequirement already satisfied: numpy>=1.13.3 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas>=0.23->pandas_datareader) (1.18.5)\nRequirement already satisfied: pytz>=2017.2 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas>=0.23->pandas_datareader) (2020.1)\nRequirement already satisfied: chardet<4,>=3.0.2 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from requests>=2.19.0->pandas_datareader) (3.0.4)\nRequirement already satisfied: idna<3,>=2.5 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from requests>=2.19.0->pandas_datareader) (2.10)\nRequirement already satisfied: certifi>=2017.4.17 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from requests>=2.19.0->pandas_datareader) (2020.6.20)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from requests>=2.19.0->pandas_datareader) (1.25.9)\nRequirement already satisfied: six>=1.5 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from python-dateutil>=2.6.1->pandas>=0.23->pandas_datareader) (1.15.0)\nRequirement already satisfied: datetime in c:\\users\\mateo\\anaconda3\\lib\\site-packages (4.3)\nRequirement already satisfied: pytz in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from datetime) (2020.1)\nRequirement already satisfied: zope.interface in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from datetime) (4.7.1)\nRequirement already satisfied: setuptools in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from zope.interface->datetime) (49.2.0.post20200714)\nRequirement already satisfied: seaborn in c:\\users\\mateo\\anaconda3\\lib\\site-packages (0.10.1)\nRequirement already satisfied: scipy>=1.0.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from seaborn) (1.5.0)\nRequirement already satisfied: matplotlib>=2.1.2 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from seaborn) (3.2.2)\nRequirement already satisfied: numpy>=1.13.3 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from seaborn) (1.18.5)\nRequirement already satisfied: pandas>=0.22.0 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from seaborn) (1.0.5)\nRequirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib>=2.1.2->seaborn) (2.4.7)\nRequirement already satisfied: python-dateutil>=2.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib>=2.1.2->seaborn) (2.8.1)\nRequirement already satisfied: kiwisolver>=1.0.1 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib>=2.1.2->seaborn) (1.2.0)\nRequirement already satisfied: cycler>=0.10 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from matplotlib>=2.1.2->seaborn) (0.10.0)\nRequirement already satisfied: pytz>=2017.2 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from pandas>=0.22.0->seaborn) (2020.1)\nRequirement already satisfied: six>=1.5 in c:\\users\\mateo\\anaconda3\\lib\\site-packages (from python-dateutil>=2.1->matplotlib>=2.1.2->seaborn) (1.15.0)\n" ], [ "# Import Quarterly data from Fred using Data Reader\n\nstart = datetime.datetime(1947, 1, 1) #beginning of series\nstart1 = datetime.datetime(1956, 10, 1) #beginning of series\n\nend = datetime.datetime(2018, 4, 1) #end of series\n\nPCDG = web.DataReader('PCDG', 'fred', start, end) #loads your durable goods quarterly series data\nPCND= web.DataReader('PCND', 'fred', start, end) #Loads your non durable goods quarterly series data\nPCDG1 = web.DataReader('PCDG', 'fred', start1, end) #loads your durable goods quarterly series data, helps in having time series of identical length\nPCND1= web.DataReader('PCND', 'fred', start1, end) #Loads your non durable goods quarterly series data, , helps in having time series of identical length", "_____no_output_____" ], [ "# Constructing PCDG and PCND growth series () \nz1=PCDG.pct_change(periods=40)# 10*4\nz2=PCND.pct_change(periods=40)#10*4\nz3=PCDG1.pct_change(periods=1)# \nz4=PCND1.pct_change(periods=1)#\ns1=z1*100 #(In percentage terms)\ns2=z2*100 #(In percentage terms)\ns3=z3*100 #(In percentage terms)\ns4=z4*100 #(In percentage terms)", "_____no_output_____" ], [ "# Plotting the growth rates\nplt.figure(figsize=((14,8))) # set the plot size\nplt.title('Durables vs Non Durables Growth 10 year vs Quarterly')\nplt.xlabel('Year')\nplt.ylabel(' Growth (Percentage Terms)')\nplt.plot(s1,label=\"PCDG 10 year growth\")\nplt.plot(s2,label=\"PCND 10 year growth\")\nplt.plot(s3,label=\"PCDG quarterly growth\")\nplt.plot(s4,label=\"PCND quarterly growth\")\nplt.legend()\nplt.show()", "_____no_output_____" ], [ "# Drops the missing NAN observations\n\na1=s1.dropna()#Drops the missing values from s1 series\na2=s2.dropna()#Drops the missing values from s2 series\na3=s3.dropna()#Drops the missing values from s3 series\na4=s4.dropna()#Drops the missing values from s4 series", "_____no_output_____" ], [ "# concatate (merge) the two series\nc1=pd.concat([a1, a2], axis=1)\nc2=pd.concat([a3, a4], axis=1)", "_____no_output_____" ], [ "#Pairwise Plotting for the 10 year growth series\n\nsns.pairplot(c1)\nplt.suptitle('10 Year Growth Rates')\nplt.show()", "_____no_output_____" ], [ "#Pairwise Plotting for the quarterly growth series\n\nsns.pairplot(c2)\nplt.suptitle('1 Quarter Growth Rates')\nplt.show()", "_____no_output_____" ] ], [ [ "For each frequency [quarterly|10-year] each moment of time would correspond to a single point (x=nondurables growth, y=durables growth). Such a plot shows that at the 10 year frequency, there is a very strong relationship between the two growth rates, and at the 1 quarter frequency, much much less.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ] ]

[ "ECL-2.0", "Apache-2.0" ]

[ "ECL-2.0", "Apache-2.0" ]

[ "ECL-2.0", "Apache-2.0" ]

[ [ [ "[source](../../api/alibi_detect.od.isolationforest.rst)", "_____no_output_____" ], [ "# Isolation Forest", "_____no_output_____" ], [ "## Overview\n\n[Isolation forests](https://cs.nju.edu.cn/zhouzh/zhouzh.files/publication/icdm08b.pdf) (IF) are tree based models specifically used for outlier detection. The IF isolates observations by randomly selecting a feature and then randomly selecting a split value between the maximum and minimum values of the selected feature. The number of splittings required to isolate a sample is equivalent to the path length from the root node to the terminating node. This path length, averaged over a forest of random trees, is a measure of normality and is used to define an anomaly score. Outliers can typically be isolated quicker, leading to shorter paths. The algorithm is suitable for low to medium dimensional tabular data.", "_____no_output_____" ], [ "## Usage\n\n### Initialize\n\nParameters:\n\n* `threshold`: threshold value for the outlier score above which the instance is flagged as an outlier.\n\n* `n_estimators`: number of base estimators in the ensemble. Defaults to 100.\n\n* `max_samples`: number of samples to draw from the training data to train each base estimator. If *int*, draw `max_samples` samples. If *float*, draw `max_samples` *times number of features* samples. If *'auto'*, `max_samples` = min(256, number of samples).\n\n* `max_features`: number of features to draw from the training data to train each base estimator. If *int*, draw `max_features` features. If float, draw `max_features` *times number of features* features.\n\n* `bootstrap`: whether to fit individual trees on random subsets of the training data, sampled with replacement.\n\n* `n_jobs`: number of jobs to run in parallel for `fit` and `predict`.\n\n* `data_type`: can specify data type added to metadata. E.g. *'tabular'* or *'image'*.\n\nInitialized outlier detector example:\n\n```python\nfrom alibi_detect.od import IForest\n\nod = IForest(\n threshold=0.,\n n_estimators=100\n)\n```", "_____no_output_____" ], [ "### Fit\n\nWe then need to train the outlier detector. The following parameters can be specified:\n\n* `X`: training batch as a numpy array.\n\n* `sample_weight`: array with shape *(batch size,)* used to assign different weights to each instance during training. Defaults to *None*.\n\n```python\nod.fit(\n X_train\n)\n```\n\nIt is often hard to find a good threshold value. If we have a batch of normal and outlier data and we know approximately the percentage of normal data in the batch, we can infer a suitable threshold:\n\n```python\nod.infer_threshold(\n X, \n threshold_perc=95\n)\n```", "_____no_output_____" ], [ "### Detect\n\nWe detect outliers by simply calling `predict` on a batch of instances `X` to compute the instance level outlier scores. We can also return the instance level outlier score by setting `return_instance_score` to True.\n\nThe prediction takes the form of a dictionary with `meta` and `data` keys. `meta` contains the detector's metadata while `data` is also a dictionary which contains the actual predictions stored in the following keys:\n\n* `is_outlier`: boolean whether instances are above the threshold and therefore outlier instances. The array is of shape *(batch size,)*.\n\n* `instance_score`: contains instance level scores if `return_instance_score` equals True.\n\n\n```python\npreds = od.predict(\n X,\n return_instance_score=True\n)\n```", "_____no_output_____" ], [ "## Examples\n\n### Tabular\n\n[Outlier detection on KDD Cup 99](../../examples/od_if_kddcup.nblink)", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "from google.colab import drive\ndrive.mount('/content/drive')\nimport warnings\nwarnings.filterwarnings('ignore')", "Mounted at /content/drive\n" ], [ "# !pip install tensorflow_text \n!pip install transformers emoji\n# !pip install ktrain", "Collecting transformers\n Downloading transformers-4.10.0-py3-none-any.whl (2.8 MB)\n\u001b[K |████████████████████████████████| 2.8 MB 5.4 MB/s \n\u001b[?25hCollecting emoji\n Downloading emoji-1.4.2.tar.gz (184 kB)\n\u001b[K |████████████████████████████████| 184 kB 49.7 MB/s \n\u001b[?25hRequirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.7/dist-packages (from transformers) (4.62.0)\nRequirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers) (3.0.12)\nRequirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.7/dist-packages (from transformers) (2019.12.20)\nRequirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from transformers) (4.6.4)\nRequirement already satisfied: packaging in /usr/local/lib/python3.7/dist-packages (from transformers) (21.0)\nCollecting huggingface-hub>=0.0.12\n Downloading huggingface_hub-0.0.16-py3-none-any.whl (50 kB)\n\u001b[K |████████████████████████████████| 50 kB 5.7 MB/s \n\u001b[?25hRequirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers) (2.23.0)\nCollecting tokenizers<0.11,>=0.10.1\n Downloading tokenizers-0.10.3-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (3.3 MB)\n\u001b[K |████████████████████████████████| 3.3 MB 31.8 MB/s \n\u001b[?25hCollecting sacremoses\n Downloading sacremoses-0.0.45-py3-none-any.whl (895 kB)\n\u001b[K |████████████████████████████████| 895 kB 33.4 MB/s \n\u001b[?25hRequirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.7/dist-packages (from transformers) (1.19.5)\nCollecting pyyaml>=5.1\n Downloading PyYAML-5.4.1-cp37-cp37m-manylinux1_x86_64.whl (636 kB)\n\u001b[K |████████████████████████████████| 636 kB 42.3 MB/s \n\u001b[?25hRequirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from huggingface-hub>=0.0.12->transformers) (3.7.4.3)\nRequirement already satisfied: pyparsing>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging->transformers) (2.4.7)\nRequirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->transformers) (3.5.0)\nRequirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (3.0.4)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (1.24.3)\nRequirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (2021.5.30)\nRequirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers) (2.10)\nRequirement already satisfied: joblib in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (1.0.1)\nRequirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (1.15.0)\nRequirement already satisfied: click in /usr/local/lib/python3.7/dist-packages (from sacremoses->transformers) (7.1.2)\nBuilding wheels for collected packages: emoji\n Building wheel for emoji (setup.py) ... \u001b[?25l\u001b[?25hdone\n Created wheel for emoji: filename=emoji-1.4.2-py3-none-any.whl size=186469 sha256=8ba5b5c09c979a1b3981450a8e7eea569942b19925ff33996b1b652da84cd70b\n Stored in directory: /root/.cache/pip/wheels/e4/61/e7/2fc1ac8f306848fc66c6c013ab511f0a39ef4b1825b11363b2\nSuccessfully built emoji\nInstalling collected packages: tokenizers, sacremoses, pyyaml, huggingface-hub, transformers, emoji\n Attempting uninstall: pyyaml\n Found existing installation: PyYAML 3.13\n Uninstalling PyYAML-3.13:\n Successfully uninstalled PyYAML-3.13\nSuccessfully installed emoji-1.4.2 huggingface-hub-0.0.16 pyyaml-5.4.1 sacremoses-0.0.45 tokenizers-0.10.3 transformers-4.10.0\n" ], [ "from transformers import AutoTokenizer\n", "_____no_output_____" ], [ "import pandas as pd\ndataset = pd.read_excel(\"/content/drive/MyDrive/English Category Transformer Model/en_category_model_data.xlsx\")\ndataset = dataset.sample(frac=1, axis=1).sample(frac=1).reset_index(drop=True) \ndataset = dataset[['p_message','Category']]\ndataset.head()", "_____no_output_____" ], [ "dataset.groupby('Category').size()", "_____no_output_____" ], [ "def to_int_sentiment(label):\n if label == \"business\":\n return 0\n elif label == \"education\":\n return 1\n elif label == 'entertainment':\n return 2\n elif label == 'fashion':\n return 3\n elif label == 'food':\n return 4\n elif label == 'health':\n return 5\n elif label == 'politics':\n return 6\n elif label == 'sports':\n return 7\n elif label == 'technology':\n return 8\n elif label == 'telecom':\n return 9\n elif label == 'tourism':\n return 10\n elif label == 'transport':\n return 11\n elif label == 'weather':\n return 12\n\ndataset['Category'] = dataset.Category.apply(to_int_sentiment)\ndataset = dataset.dropna()\ndataset['Category'] = dataset.Category.apply(int)", "_____no_output_____" ], [ "dataset.head()", "_____no_output_____" ], [ "from preprocessing import CleaningText\nen_text_clean = CleaningText()\n\n", "_____no_output_____" ], [ "dataset['p_message'] = dataset['p_message'].apply(str)\ndataset['p_message'] = dataset['p_message'].apply(en_text_clean.text_preprocessing)\ndataset.head()\n", "_____no_output_____" ], [ "dataset.groupby('Category').size()", "_____no_output_____" ], [ "\nimport numpy as np\nimport pandas as pd\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.metrics import accuracy_score, recall_score, precision_score, f1_score\nimport torch\nfrom transformers import TrainingArguments, Trainer\nfrom transformers import BertTokenizer, BertForSequenceClassification\nfrom transformers import EarlyStoppingCallback\n\n\n# Read data\n# dataset = pd.read_excel(\"/content/drive/MyDrive/ar_general_sentiment_data.xlsx\")\n# dataset = dataset[['text','Sentiment']]\n\n# Define pretrained tokenizer and model\nmodel_name = \"prajjwal1/bert-small\"\ntokenizer = BertTokenizer.from_pretrained(model_name)\nmodel = BertForSequenceClassification.from_pretrained(model_name, num_labels=13)\n\n# ----- 1. Preprocess data -----#\n# Preprocess data\nX = list(dataset[\"p_message\"])\ny = list(dataset[\"Category\"])\nX_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2)\nX_train_tokenized = tokenizer(X_train, padding=True, truncation=True, max_length=512)\nX_val_tokenized = tokenizer(X_val, padding=True, truncation=True, max_length=512)\n\n# Create torch dataset\nclass Dataset(torch.utils.data.Dataset):\n def __init__(self, encodings, labels=None):\n self.encodings = encodings\n self.labels = labels\n\n def __getitem__(self, idx):\n item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}\n if self.labels:\n item[\"labels\"] = torch.tensor(self.labels[idx])\n return item\n\n def __len__(self):\n return len(self.encodings[\"input_ids\"])\n\ntrain_dataset = Dataset(X_train_tokenized, y_train)\nval_dataset = Dataset(X_val_tokenized, y_val)\n\n\n\n", "_____no_output_____" ], [ "# ----- 2. Fine-tune pretrained model -----#\n# Define Trainer parameters\ndef compute_metrics(p):\n pred, labels = p\n pred = np.argmax(pred, axis=1)\n\n accuracy = accuracy_score(y_true=labels, y_pred=pred)\n recall = recall_score(y_true=labels, y_pred=pred,average='micro')\n precision = precision_score(y_true=labels, y_pred=pred,average='micro')\n f1 = f1_score(y_true=labels, y_pred=pred,average='micro')\n\n return {\"accuracy\": accuracy, \"precision\": precision, \"recall\": recall, \"f1\": f1}\n\n# Define Trainer\nargs = TrainingArguments(\n output_dir=\"output\",\n evaluation_strategy=\"steps\",\n eval_steps=500,\n per_device_train_batch_size=8,\n per_device_eval_batch_size=8,\n num_train_epochs=5,\n seed=0,\n load_best_model_at_end=True,\n)\ntrainer = Trainer(\n model=model,\n args=args,\n train_dataset=train_dataset,\n eval_dataset=val_dataset,\n compute_metrics=compute_metrics,\n # callbacks=[EarlyStoppingCallback(early_stopping_patience=3)],\n)\n\n# Train pre-trained model\ntrainer.train()\n", "***** Running training *****\n Num examples = 25391\n Num Epochs = 5\n Instantaneous batch size per device = 8\n Total train batch size (w. parallel, distributed & accumulation) = 8\n Gradient Accumulation steps = 1\n Total optimization steps = 15870\n" ], [ "model_path = \"/content/drive/MyDrive/English Category Transformer Model/en_category_model\"\ntrainer.save_model(model_path)", "Saving model checkpoint to /content/drive/MyDrive/English Category Transformer Model/en_category_model\nConfiguration saved in /content/drive/MyDrive/English Category Transformer Model/en_category_model/config.json\nModel weights saved in /content/drive/MyDrive/English Category Transformer Model/en_category_model/pytorch_model.bin\n" ], [ "\nX_test_tokenized = tokenizer(X_val, padding=True, truncation=True, max_length=512) \n# Create torch dataset\ntest_dataset = Dataset(X_test_tokenized) \n# Load trained model\n# model_path = \"sentiment_model\"\nmodel = BertForSequenceClassification.from_pretrained(model_path, num_labels=13) \n# Define test trainer\ntest_trainer = Trainer(model) \n# Make prediction\nraw_pred, _, _ = test_trainer.predict(test_dataset) \n# Preprocess raw predictions\ny_pred = np.argmax(raw_pred, axis=1)", "loading configuration file /content/drive/MyDrive/English Category Transformer Model/en_category_model/config.json\nModel config BertConfig {\n \"_name_or_path\": \"prajjwal1/bert-small\",\n \"architectures\": [\n \"BertForSequenceClassification\"\n ],\n \"attention_probs_dropout_prob\": 0.1,\n \"classifier_dropout\": null,\n \"gradient_checkpointing\": false,\n \"hidden_act\": \"gelu\",\n \"hidden_dropout_prob\": 0.1,\n \"hidden_size\": 512,\n \"id2label\": {\n \"0\": \"LABEL_0\",\n \"1\": \"LABEL_1\",\n \"2\": \"LABEL_2\",\n \"3\": \"LABEL_3\",\n \"4\": \"LABEL_4\",\n \"5\": \"LABEL_5\",\n \"6\": \"LABEL_6\",\n \"7\": \"LABEL_7\",\n \"8\": \"LABEL_8\",\n \"9\": \"LABEL_9\",\n \"10\": \"LABEL_10\",\n \"11\": \"LABEL_11\",\n \"12\": \"LABEL_12\"\n },\n \"initializer_range\": 0.02,\n \"intermediate_size\": 2048,\n \"label2id\": {\n \"LABEL_0\": 0,\n \"LABEL_1\": 1,\n \"LABEL_10\": 10,\n \"LABEL_11\": 11,\n \"LABEL_12\": 12,\n \"LABEL_2\": 2,\n \"LABEL_3\": 3,\n \"LABEL_4\": 4,\n \"LABEL_5\": 5,\n \"LABEL_6\": 6,\n \"LABEL_7\": 7,\n \"LABEL_8\": 8,\n \"LABEL_9\": 9\n },\n \"layer_norm_eps\": 1e-12,\n \"max_position_embeddings\": 512,\n \"model_type\": \"bert\",\n \"num_attention_heads\": 8,\n \"num_hidden_layers\": 4,\n \"pad_token_id\": 0,\n \"position_embedding_type\": \"absolute\",\n \"problem_type\": \"single_label_classification\",\n \"torch_dtype\": \"float32\",\n \"transformers_version\": \"4.10.0\",\n \"type_vocab_size\": 2,\n \"use_cache\": true,\n \"vocab_size\": 30522\n}\n\nloading weights file /content/drive/MyDrive/English Category Transformer Model/en_category_model/pytorch_model.bin\nAll model checkpoint weights were used when initializing BertForSequenceClassification.\n\nAll the weights of BertForSequenceClassification were initialized from the model checkpoint at /content/drive/MyDrive/English Category Transformer Model/en_category_model.\nIf your task is similar to the task the model of the checkpoint was trained on, you can already use BertForSequenceClassification for predictions without further training.\nNo `TrainingArguments` passed, using `output_dir=tmp_trainer`.\nPyTorch: setting up devices\nThe default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).\n***** Running Prediction *****\n Num examples = 6348\n Batch size = 8\n" ], [ "y_pred", "_____no_output_____" ], [ "from sklearn.metrics import classification_report, confusion_matrix\nclass_names = [\"0\",\"1\",\"2\",\"3\",\"4\",\"5\",\"6\",\"7\",\"8\",\"9\",\"10\",\"11\",\"12\"]\nprint(classification_report(y_val, y_pred, target_names=class_names))", " precision recall f1-score support\n\n 0 0.91 0.91 0.91 527\n 1 0.94 0.93 0.94 495\n 2 0.96 0.93 0.95 526\n 3 0.98 0.99 0.98 438\n 4 0.98 0.94 0.96 491\n 5 0.91 0.95 0.93 522\n 6 0.96 0.92 0.94 554\n 7 0.97 0.97 0.97 449\n 8 0.90 0.92 0.91 537\n 9 0.95 0.97 0.96 578\n 10 0.96 0.96 0.96 292\n 11 0.94 0.95 0.95 523\n 12 0.96 0.97 0.96 416\n\n accuracy 0.95 6348\n macro avg 0.95 0.95 0.95 6348\nweighted avg 0.95 0.95 0.95 6348\n\n" ], [ " print(confusion_matrix(y_val, y_pred))", "[[480 5 2 1 3 6 11 0 8 3 2 4 2]\n [ 5 459 5 1 1 7 1 1 8 5 1 1 0]\n [ 2 5 490 3 1 6 2 2 5 3 0 5 2]\n [ 0 1 0 434 0 0 0 1 0 0 0 1 1]\n [ 3 2 2 0 462 10 0 1 0 2 1 5 3]\n [ 1 3 4 3 2 496 1 1 7 1 1 1 1]\n [ 16 4 0 0 1 2 509 2 10 2 2 2 4]\n [ 0 1 3 1 0 3 1 436 0 0 0 3 1]\n [ 9 5 1 0 1 8 1 1 492 13 0 4 2]\n [ 2 0 0 0 1 1 0 0 13 560 0 0 1]\n [ 8 0 0 0 0 0 0 0 0 0 279 5 0]\n [ 3 1 2 0 0 5 0 3 5 1 3 499 1]\n [ 1 0 1 1 0 0 4 2 1 1 1 0 404]]\n" ], [ "df = pd.DataFrame(X_val,columns =['p_message'])\ndf.head()", "_____no_output_____" ], [ "df['prediction'] = y_pred", "_____no_output_____" ], [ "df.head()", "_____no_output_____" ], [ "df.groupby('prediction').size()", "_____no_output_____" ], [ "from transformers import BertTokenizer, BertForSequenceClassification\nfrom transformers import EarlyStoppingCallback\nfrom transformers import AutoTokenizer\nfrom preprocessing import CleaningText\n\nmodel_name = \"prajjwal1/bert-small\"\nmodel_path = \"en_category\"", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Self-Driving Car Engineer Nanodegree\n\n\n## Project: **Finding Lane Lines on the Road** \n***\nIn this project, you will use the tools you learned about in the lesson to identify lane lines on the road. You can develop your pipeline on a series of individual images, and later apply the result to a video stream (really just a series of images). Check out the video clip \"raw-lines-example.mp4\" (also contained in this repository) to see what the output should look like after using the helper functions below. \n\nOnce you have a result that looks roughly like \"raw-lines-example.mp4\", you'll need to get creative and try to average and/or extrapolate the line segments you've detected to map out the full extent of the lane lines. You can see an example of the result you're going for in the video \"P1_example.mp4\". Ultimately, you would like to draw just one line for the left side of the lane, and one for the right.\n\nIn addition to implementing code, there is a brief writeup to complete. The writeup should be completed in a separate file, which can be either a markdown file or a pdf document. There is a [write up template](https://github.com/udacity/CarND-LaneLines-P1/blob/master/writeup_template.md) that can be used to guide the writing process. Completing both the code in the Ipython notebook and the writeup template will cover all of the [rubric points](https://review.udacity.com/#!/rubrics/322/view) for this project.\n\n---\nLet's have a look at our first image called 'test_images/solidWhiteRight.jpg'. Run the 2 cells below (hit Shift-Enter or the \"play\" button above) to display the image.\n\n**Note: If, at any point, you encounter frozen display windows or other confounding issues, you can always start again with a clean slate by going to the \"Kernel\" menu above and selecting \"Restart & Clear Output\".**\n\n---", "_____no_output_____" ], [ "**The tools you have are color selection, region of interest selection, grayscaling, Gaussian smoothing, Canny Edge Detection and Hough Tranform line detection. You are also free to explore and try other techniques that were not presented in the lesson. Your goal is piece together a pipeline to detect the line segments in the image, then average/extrapolate them and draw them onto the image for display (as below). Once you have a working pipeline, try it out on the video stream below.**\n\n---\n\n<figure>\n <img src=\"examples/line-segments-example.jpg\" width=\"380\" alt=\"Combined Image\" />\n <figcaption>\n <p></p> \n <p style=\"text-align: center;\"> Your output should look something like this (above) after detecting line segments using the helper functions below </p> \n </figcaption>\n</figure>\n <p></p> \n<figure>\n <img src=\"examples/laneLines_thirdPass.jpg\" width=\"380\" alt=\"Combined Image\" />\n <figcaption>\n <p></p> \n <p style=\"text-align: center;\"> Your goal is to connect/average/extrapolate line segments to get output like this</p> \n </figcaption>\n</figure>", "_____no_output_____" ], [ "**Run the cell below to import some packages. If you get an `import error` for a package you've already installed, try changing your kernel (select the Kernel menu above --> Change Kernel). Still have problems? Try relaunching Jupyter Notebook from the terminal prompt. Also, consult the forums for more troubleshooting tips.** ", "_____no_output_____" ], [ "## Import Packages", "_____no_output_____" ] ], [ [ "#importing some useful packages\nimport matplotlib.pyplot as plt\nimport matplotlib.image as mpimg\nimport numpy as np\nimport cv2\n%matplotlib inline", "_____no_output_____" ] ], [ [ "## Read in an Image", "_____no_output_____" ] ], [ [ "#reading in an image\nimage = mpimg.imread('test_images/solidWhiteRight.jpg')\n\n#printing out some stats and plotting\nprint('This image is:', type(image), 'with dimensions:', image.shape)\nplt.imshow(image) # if you wanted to show a single color channel image called 'gray', for example, call as plt.imshow(gray, cmap='gray')", "This image is: <class 'numpy.ndarray'> with dimensions: (540, 960, 3)\n" ] ], [ [ "## Ideas for Lane Detection Pipeline", "_____no_output_____" ], [ "**Some OpenCV functions (beyond those introduced in the lesson) that might be useful for this project are:**\n\n`cv2.inRange()` for color selection \n`cv2.fillPoly()` for regions selection \n`cv2.line()` to draw lines on an image given endpoints \n`cv2.addWeighted()` to coadd / overlay two images \n`cv2.cvtColor()` to grayscale or change color \n`cv2.imwrite()` to output images to file \n`cv2.bitwise_and()` to apply a mask to an image\n\n**Check out the OpenCV documentation to learn about these and discover even more awesome functionality!**", "_____no_output_____" ], [ "## Helper Functions", "_____no_output_____" ], [ "Below are some helper functions to help get you started. They should look familiar from the lesson!", "_____no_output_____" ] ], [ [ "import math\n\ndef grayscale(img):\n \"\"\"Applies the Grayscale transform\n This will return an image with only one color channel\n but NOTE: to see the returned image as grayscale\n (assuming your grayscaled image is called 'gray')\n you should call plt.imshow(gray, cmap='gray')\"\"\"\n return cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)\n # Or use BGR2GRAY if you read an image with cv2.imread()\n # return cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)\n \ndef canny(img, low_threshold, high_threshold):\n \"\"\"Applies the Canny transform\"\"\"\n return cv2.Canny(img, low_threshold, high_threshold)\n\ndef gaussian_blur(img, kernel_size):\n \"\"\"Applies a Gaussian Noise kernel\"\"\"\n return cv2.GaussianBlur(img, (kernel_size, kernel_size), 0)\n\ndef region_of_interest(img, vertices):\n \"\"\"\n Applies an image mask.\n \n Only keeps the region of the image defined by the polygon\n formed from `vertices`. The rest of the image is set to black.\n `vertices` should be a numpy array of integer points.\n \"\"\"\n #defining a blank mask to start with\n mask = np.zeros_like(img) \n \n #defining a 3 channel or 1 channel color to fill the mask with depending on the input image\n if len(img.shape) > 2:\n channel_count = img.shape[2] # i.e. 3 or 4 depending on your image\n ignore_mask_color = (255,) * channel_count\n else:\n ignore_mask_color = 255\n \n #filling pixels inside the polygon defined by \"vertices\" with the fill color \n cv2.fillPoly(mask, vertices, ignore_mask_color)\n \n #returning the image only where mask pixels are nonzero\n masked_image = cv2.bitwise_and(img, mask)\n return masked_image\n\n\n# def draw_lines(img, lines, color=[255, 0, 0], thickness=10):\n# \"\"\"\n# NOTE: this is the function you might want to use as a starting point once you want to \n# average/extrapolate the line segments you detect to map out the full\n# extent of the lane (going from the result shown in raw-lines-example.mp4\n# to that shown in P1_example.mp4). \n \n# Think about things like separating line segments by their \n# slope ((y2-y1)/(x2-x1)) to decide which segments are part of the left\n# line vs. the right line. Then, you can average the position of each of \n# the lines and extrapolate to the top and bottom of the lane.\n \n# This function draws `lines` with `color` and `thickness`. \n# Lines are drawn on the image inplace (mutates the image).\n# If you want to make the lines semi-transparent, think about combining\n# this function with the weighted_img() function below\n# \"\"\"\n# for line in lines:\n# for x1,y1,x2,y2 in line:\n# cv2.line(img, (x1, y1), (x2, y2), color, thickness)\n \n\ndef hough_lines(img, rho, theta, threshold, min_line_len, max_line_gap):\n \"\"\"\n `img` should be the output of a Canny transform.\n \n Returns an image with hough lines drawn.\n \"\"\"\n lines = cv2.HoughLinesP(img, rho, theta, threshold, np.array([]), minLineLength=min_line_len, maxLineGap=max_line_gap)\n line_img = np.zeros((img.shape[0], img.shape[1], 3), dtype=np.uint8)\n draw_lines(line_img, lines)\n return line_img\n\n# Python 3 has support for cool math symbols.\n\ndef weighted_img(img, initial_img, α=0.8, β=1., γ=0.):\n \"\"\"\n `img` is the output of the hough_lines(), An image with lines drawn on it.\n Should be a blank image (all black) with lines drawn on it.\n \n `initial_img` should be the image before any processing.\n \n The result image is computed as follows:\n \n initial_img * α + img * β + γ\n NOTE: initial_img and img must be the same shape!\n \"\"\"\n return cv2.addWeighted(initial_img, α, img, β, γ)", "_____no_output_____" ] ], [ [ "New draw_lines function to extrapolate the line<br/>\nHere fraction of shape of image i.e **fraction*img.shape[0]** is taken to make it work with optoional problem as its image has different dimensions.", "_____no_output_____" ] ], [ [ "def draw_lines(img, lines, color=[255, 0, 0], thickness=10):\n left_slope = 0\n right_slope = 0\n x_left = 0\n y_left = 0\n x_right = 0\n y_right = 0\n count_left = 0 # count of left linesegments for average calculation\n count_right = 0 # count of right linesegments for average calculation\n \n for line in lines:\n for x1,y1,x2,y2 in line:\n slope = (y2-y1)/(x2-x1)\n # slope boundary is taken almost tan(30) as from camera point of view it has slope arount the same\n if slope>0.5: #Left lane\n # Adding all values of slope and average positions of a line\n left_slope += slope\n x_left += (x1+x2)/2\n y_left += (y1+y2)/2\n count_left += 1\n \n elif slope<-0.5: # right lane\n # Adding all values of slope and average positions of a line\n right_slope += slope\n x_right += (x1+x2)/2\n y_right += (y1+y2)/2\n count_right += 1\n \n # Left lane - averaging all slopes, x co-ordinates and y co-ordinates\n if count_left>0: # if left lane has been detected\n avg_left_slope = left_slope/count_left\n avg_left_x = x_left/count_left\n avg_left_y = y_left/count_left\n # Calculate bottom x and top x assuming fixed positions for corresponding y\n # It has been calculated based on slope formula y = mx+c then x = (y-c)/m\n bottom_x_left = int(((int(img.shape[0])-avg_left_y)/avg_left_slope) + avg_left_x)\n top_x_left = int(((int(0.60*img.shape[0])-avg_left_y)/avg_left_slope)+ avg_left_x)\n \n else: # If Left lane is not detected - best guess positions of bottom x and top x\n bottom_x_left = int(0.21*img.shape[1])\n top_x_left = int(0.43*img.shape[1])\n \n # Draw a line\n cv2.line(img, (top_x_left, int(0.60*img.shape[0])), (bottom_x_left, int(img.shape[0])), color, thickness)\n \n #Right lane - Average across all slope and intercepts\n if count_right>0: # If right lane is detected\n avg_right_slope = right_slope/count_right\n avg_right_x = x_right/count_right\n avg_right_y = y_right/count_right\n # Calculate bottom x and top x assuming fixed positions for corresponding y\n # It has been calculated based on slope formula y = mx+c then x = (y-c)/m\n bottom_x_right = int(((int(img.shape[0])-avg_right_y)/avg_right_slope) + avg_right_x)\n top_x_right = int(((int(0.60*img.shape[0])-avg_right_y)/avg_right_slope)+ avg_right_x)\n \n else: # If right lane is not detected - best guess positions of bottom x and top x\n bottom_x_right = int(0.89*img.shape[1])\n top_x_right = int(0.53*img.shape[1])\n \n # Draw a line \n cv2.line(img, (top_x_right, int(0.60*img.shape[0])), (bottom_x_right, int(img.shape[0])), color, thickness)", "_____no_output_____" ] ], [ [ "## Test Images\n\nBuild your pipeline to work on the images in the directory \"test_images\" \n**You should make sure your pipeline works well on these images before you try the videos.**", "_____no_output_____" ] ], [ [ "import os\nos.listdir(\"test_images/\")", "_____no_output_____" ], [ "image_file = ['test_images/whiteCarLaneSwitch.jpg','test_images/solidWhiteCurve.jpg', 'test_images/solidWhiteRight.jpg', 'test_images/solidYellowCurve.jpg', 'test_images/solidYellowCurve2.jpg', 'test_images/solidYellowLeft.jpg']", "_____no_output_____" ] ], [ [ "## Build a Lane Finding Pipeline\n\n", "_____no_output_____" ], [ "Build the pipeline and run your solution on all test_images. Make copies into the `test_images_output` directory, and you can use the images in your writeup report.\n\nTry tuning the various parameters, especially the low and high Canny thresholds as well as the Hough lines parameters.", "_____no_output_____" ] ], [ [ "# TODO: Build your pipeline that will draw lane lines on the test_images\n# then save them to the test_images_output directory.\ndef lane_finding_pipeline(img, vertices):\n# image = mpimg.imread(img)\n gray = grayscale(img)\n kernel_size = 5\n blur_gray = gaussian_blur(gray, kernel_size)\n low_threshold = 50\n high_threshold = 150\n edges = canny(blur_gray, low_threshold, high_threshold)\n masked_edges = region_of_interest(edges, vertices)\n rho = 1 # distance resolution in pixels of the Hough grid\n theta = np.pi/180 # angular resolution in radians of the Hough grid\n threshold = 10 # minimum number of votes (intersections in Hough grid cell)\n min_line_len = 60 #minimum number of pixels making up a line\n max_line_gap = 30 # maximum gap in pixels between connectable line segments\n final = hough_lines(masked_edges, rho, theta, threshold, min_line_len, max_line_gap)\n # color_edges = np.dstack((edges, edges, edges))\n final_image = weighted_img(final, img)\n return final_image", "_____no_output_____" ], [ "img = mpimg.imread(image_file[0])\nvertices = np.array([[(150,img.shape[0]),(450, 323), (500, 323), (img.shape[1],img.shape[0])]], dtype=np.int32)\ncv2.imwrite('test_images_output/whiteCarLaneSwitch.png', cv2.cvtColor(lane_finding_pipeline(img, vertices), cv2.COLOR_BGR2RGB))\nplt.imshow(lane_finding_pipeline(img, vertices))", "_____no_output_____" ], [ "img = mpimg.imread(image_file[1])\nvertices = np.array([[(150,img.shape[0]),(450, 323), (500, 323), (img.shape[1],img.shape[0])]], dtype=np.int32)\ncv2.imwrite('test_images_output/solidWhiteCurve.png', cv2.cvtColor(lane_finding_pipeline(img, vertices), cv2.COLOR_BGR2RGB))\nplt.imshow(lane_finding_pipeline(img, vertices))", "_____no_output_____" ], [ "img = mpimg.imread(image_file[2])\nvertices = np.array([[(150,img.shape[0]),(450, 323), (500, 323), (img.shape[1],img.shape[0])]], dtype=np.int32)\ncv2.imwrite('test_images_output/solidWhiteRight.png', cv2.cvtColor(lane_finding_pipeline(img, vertices), cv2.COLOR_BGR2RGB))\nplt.imshow(lane_finding_pipeline(img, vertices))", "_____no_output_____" ], [ "img = mpimg.imread(image_file[3])\nvertices = np.array([[(150,img.shape[0]),(450, 323), (500, 323), (img.shape[1],img.shape[0])]], dtype=np.int32)\ncv2.imwrite('test_images_output/solidYellowCurve.png', cv2.cvtColor(lane_finding_pipeline(img, vertices), cv2.COLOR_BGR2RGB))\nplt.imshow(lane_finding_pipeline(img, vertices))", "_____no_output_____" ], [ "img = mpimg.imread(image_file[4])\nvertices = np.array([[(150,img.shape[0]),(450, 323), (500, 323), (img.shape[1],img.shape[0])]], dtype=np.int32)\ncv2.imwrite('test_images_output/solidYellowCurve2.png', cv2.cvtColor(lane_finding_pipeline(img, vertices), cv2.COLOR_BGR2RGB))\nplt.imshow(lane_finding_pipeline(img, vertices))", "_____no_output_____" ], [ "img = mpimg.imread(image_file[5])\nvertices = np.array([[(150,img.shape[0]),(450, 305), (500, 305), (img.shape[1],img.shape[0])]], dtype=np.int32)\ncv2.imwrite('test_images_output/solidYellowLeft.png', cv2.cvtColor(lane_finding_pipeline(img, vertices), cv2.COLOR_BGR2RGB))\nplt.imshow(lane_finding_pipeline(img, vertices))", "_____no_output_____" ] ], [ [ "## Test on Videos\n\nYou know what's cooler than drawing lanes over images? Drawing lanes over video!\n\nWe can test our solution on two provided videos:\n\n`solidWhiteRight.mp4`\n\n`solidYellowLeft.mp4`\n\n**Note: if you get an import error when you run the next cell, try changing your kernel (select the Kernel menu above --> Change Kernel). Still have problems? Try relaunching Jupyter Notebook from the terminal prompt. Also, consult the forums for more troubleshooting tips.**\n\n**If you get an error that looks like this:**\n```\nNeedDownloadError: Need ffmpeg exe. \nYou can download it by calling: \nimageio.plugins.ffmpeg.download()\n```\n**Follow the instructions in the error message and check out [this forum post](https://discussions.udacity.com/t/project-error-of-test-on-videos/274082) for more troubleshooting tips across operating systems.**", "_____no_output_____" ] ], [ [ "# Import everything needed to edit/save/watch video clips\nfrom moviepy.editor import VideoFileClip\nfrom IPython.display import HTML", "_____no_output_____" ], [ "def process_image(image):\n # NOTE: The output you return should be a color image (3 channel) for processing video below\n # TODO: put your pipeline here,\n # you should return the final output (image where lines are drawn on lanes)\n gray = grayscale(image)\n kernel_size = 5\n blur_gray = gaussian_blur(gray, kernel_size)\n low_threshold = 50\n high_threshold = 150\n edges = canny(blur_gray, low_threshold, high_threshold)\n vertices = np.array([[(150,image.shape[0]),(445, 320), (500, 320), (image.shape[1],image.shape[0])]], dtype=np.int32)\n masked_edges = region_of_interest(edges, vertices)\n rho = 1 # distance resolution in pixels of the Hough grid\n theta = np.pi/180 # angular resolution in radians of the Hough grid\n threshold = 10 # minimum number of votes (intersections in Hough grid cell)\n min_line_len = 60 #minimum number of pixels making up a line\n max_line_gap = 30 # maximum gap in pixels between connectable line segments\n final = hough_lines(masked_edges, rho, theta, threshold, min_line_len, max_line_gap)\n # color_edges = np.dstack((edges, edges, edges))\n result = weighted_img(final, image)\n# plt.imshow(result)\n return result", "_____no_output_____" ] ], [ [ "Let's try the one with the solid white lane on the right first ...", "_____no_output_____" ] ], [ [ "white_output = 'test_videos_output/solidWhiteRight.mp4'\n## To speed up the testing process you may want to try your pipeline on a shorter subclip of the video\n## To do so add .subclip(start_second,end_second) to the end of the line below\n## Where start_second and end_second are integer values representing the start and end of the subclip\n## You may also uncomment the following line for a subclip of the first 5 seconds\n##clip1 = VideoFileClip(\"test_videos/solidWhiteRight.mp4\").subclip(0,5)\nclip1 = VideoFileClip(\"test_videos/solidWhiteRight.mp4\")\nwhite_clip = clip1.fl_image(process_image) #NOTE: this function expects color images!!\n%time white_clip.write_videofile(white_output, audio=False)", "[MoviePy] >>>> Building video test_videos_output/solidWhiteRight.mp4\n[MoviePy] Writing video test_videos_output/solidWhiteRight.mp4\n" ] ], [ [ "Play the video inline, or if you prefer find the video in your filesystem (should be in the same directory) and play it in your video player of choice.", "_____no_output_____" ] ], [ [ "HTML(\"\"\"\n<video width=\"960\" height=\"540\" controls>\n <source src=\"{0}\">\n</video>\n\"\"\".format(white_output))", "_____no_output_____" ] ], [ [ "## Improve the draw_lines() function\n\n**At this point, if you were successful with making the pipeline and tuning parameters, you probably have the Hough line segments drawn onto the road, but what about identifying the full extent of the lane and marking it clearly as in the example video (P1_example.mp4)? Think about defining a line to run the full length of the visible lane based on the line segments you identified with the Hough Transform. As mentioned previously, try to average and/or extrapolate the line segments you've detected to map out the full extent of the lane lines. You can see an example of the result you're going for in the video \"P1_example.mp4\".**\n\n**Go back and modify your draw_lines function accordingly and try re-running your pipeline. The new output should draw a single, solid line over the left lane line and a single, solid line over the right lane line. The lines should start from the bottom of the image and extend out to the top of the region of interest.**", "_____no_output_____" ], [ "Now for the one with the solid yellow lane on the left. This one's more tricky!", "_____no_output_____" ] ], [ [ "yellow_output = 'test_videos_output/solidYellowLeft.mp4'\n## To speed up the testing process you may want to try your pipeline on a shorter subclip of the video\n## To do so add .subclip(start_second,end_second) to the end of the line below\n## Where start_second and end_second are integer values representing the start and end of the subclip\n## You may also uncomment the following line for a subclip of the first 5 seconds\n##clip2 = VideoFileClip('test_videos/solidYellowLeft.mp4').subclip(0,5)\nclip2 = VideoFileClip('test_videos/solidYellowLeft.mp4')\nyellow_clip = clip2.fl_image(process_image)\n%time yellow_clip.write_videofile(yellow_output, audio=False)", "[MoviePy] >>>> Building video test_videos_output/solidYellowLeft.mp4\n[MoviePy] Writing video test_videos_output/solidYellowLeft.mp4\n" ], [ "HTML(\"\"\"\n<video width=\"960\" height=\"540\" controls>\n <source src=\"{0}\">\n</video>\n\"\"\".format(yellow_output))", "_____no_output_____" ] ], [ [ "## Writeup and Submission\n\nIf you're satisfied with your video outputs, it's time to make the report writeup in a pdf or markdown file. Once you have this Ipython notebook ready along with the writeup, it's time to submit for review! Here is a [link](https://github.com/udacity/CarND-LaneLines-P1/blob/master/writeup_template.md) to the writeup template file.\n", "_____no_output_____" ], [ "## Optional Challenge\n\nTry your lane finding pipeline on the video below. Does it still work? Can you figure out a way to make it more robust? If you're up for the challenge, modify your pipeline so it works with this video and submit it along with the rest of your project!", "_____no_output_____" ] ], [ [ "def process_image_1(image):\n # NOTE: The output you return should be a color image (3 channel) for processing video below\n # TODO: put your pipeline here,\n # you should return the final output (image where lines are drawn on lanes)\n gray = grayscale(image)\n kernel_size = 5\n blur_gray = gaussian_blur(gray, kernel_size)\n low_threshold = 50\n high_threshold = 150\n edges = canny(blur_gray, low_threshold, high_threshold)\n# vertices = np.array([[(220,690),(550, 450), (740, 450), (1150,690), (1000, 650),(750, 500),(600, 500),(400, 650),(700, 550)]], dtype=np.int32)\n vertices = np.array([[(220,690),(550, 450), (740, 450), (1150,690)]], dtype=np.int32)\n masked_edges = region_of_interest(edges, vertices)\n rho = 1 # distance resolution in pixels of the Hough grid\n theta = np.pi/180 # angular resolution in radians of the Hough grid\n threshold = 10 # minimum number of votes (intersections in Hough grid cell)\n min_line_len = 20 #minimum number of pixels making up a line\n max_line_gap = 7 # maximum gap in pixels between connectable line segments\n final = hough_lines(masked_edges, rho, theta, threshold, min_line_len, max_line_gap)\n # color_edges = np.dstack((edges, edges, edges))\n result = weighted_img(final, image)\n# plt.imshow(image)\n return result", "_____no_output_____" ], [ "challenge_output = 'test_videos_output/challenge.mp4'\n## To speed up the testing process you may want to try your pipeline on a shorter subclip of the video\n## To do so add .subclip(start_second,end_second) to the end of the line below\n## Where start_second and end_second are integer values representing the start and end of the subclip\n## You may also uncomment the following line for a subclip of the first 5 seconds\n##clip3 = VideoFileClip('test_videos/challenge.mp4').subclip(0,5)\nclip3 = VideoFileClip('test_videos/challenge.mp4')\nchallenge_clip = clip3.fl_image(process_image_1)\n%time challenge_clip.write_videofile(challenge_output, audio=False)", "[MoviePy] >>>> Building video test_videos_output/challenge.mp4\n[MoviePy] Writing video test_videos_output/challenge.mp4\n" ], [ "HTML(\"\"\"\n<video width=\"960\" height=\"540\" controls>\n <source src=\"{0}\">\n</video>\n\"\"\".format(challenge_output))", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Classification models using python and scikit-learn\n\nThere are many users of online trading platforms and these companies would like to run analytics on and predict churn based on user activity on the platform. Keeping customers happy so they do not move their investments elsewhere is key to maintaining profitability.\n\nIn this notebook, we'll use scikit-learn to predict classes. scikit-learn provides implementations of many classification algorithms. In here, we have chosen the random forest classification algorithm to walk through all the different steps.\n\n\n<a id=\"top\"></a>\n## Table of Contents\n\n1. [Load libraries](#load_libraries)\n2. [Data exploration](#explore_data)\n3. [Prepare data for building classification model](#prepare_data)\n4. [Split data into train and test sets](#split_data)\n5. [Helper methods for graph generation](#helper_methods)\n6. [Prepare Random Forest classification model](#prepare_model)\n7. [Train Random Forest classification model](#train_model)\n8. [Test Random Forest classification model](#test_model)\n9. [Evaluate Random Forest classification model](#evaluate_model)\n10.[Build K-Nearest classification model](#model_knn)\n11. [Comparative study of both classification algorithms](#compare_classification)", "_____no_output_____" ], [ "### Quick set of instructions to work through the notebook\n\nIf you are new to Notebooks, here's a quick overview of how to work in this environment.\n\n1. The notebook has 2 types of cells - markdown (text) such as this and code such as the one below. \n2. Each cell with code can be executed independently or together (see options under the Cell menu). When working in this notebook, we will be running one cell at a time.\n3. To run the cell, position cursor in the code cell and click the Run (arrow) icon. The cell is running when you see the * next to it. Some cells have printable output.\n4. Work through this notebook by reading the instructions and executing code cell by cell. Some cells will require modifications before you run them. ", "_____no_output_____" ], [ "<a id=\"load_libraries\"></a>\n## 1. Load libraries\n[Top](#top)\n\n\nInstall python modules\nNOTE! Some pip installs require a kernel restart.\nThe shell command pip install is used to install Python modules. Some installs require a kernel restart to complete. To avoid confusing errors, run the following cell once and then use the Kernel menu to restart the kernel before proceeding.", "_____no_output_____" ] ], [ [ "!pip install pandas==0.24.2\n!pip install --user pandas_ml==0.6.1\n#downgrade matplotlib to bypass issue with confusion matrix being chopped out\n!pip install matplotlib==3.1.0\n!pip install --user scikit-learn==0.21.3\n!pip install -q scikit-plot", "_____no_output_____" ], [ "from sklearn.preprocessing import LabelEncoder, OneHotEncoder\nfrom sklearn.impute import SimpleImputer\nfrom sklearn.preprocessing import MinMaxScaler\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.compose import ColumnTransformer, make_column_transformer\nfrom sklearn.pipeline import Pipeline\n\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.metrics import accuracy_score, classification_report\n\nimport pandas as pd, numpy as np\n\nimport matplotlib.pyplot as plt\nfrom mpl_toolkits.mplot3d import Axes3D\nimport matplotlib.colors as mcolors\nimport matplotlib.patches as mpatches\nimport scikitplot as skplt", "_____no_output_____" ] ], [ [ "<a id=\"explore_data\"></a>\n## 2. Data exploration\n[Top](#top)\n\nIn this tutorial, we use a data set that contains information about customers of an online trading platform to classify whether a given customer’s probability of churn will be high, medium, or low. This provides a good example to learn how a classification model is built from start to end.", "_____no_output_____" ] ], [ [ "df_churn_pd = pd.read_csv(\"https://raw.githubusercontent.com/IBM/ml-learning-path-assets/master/data/mergedcustomers_missing_values_GENDER.csv\")\ndf_churn_pd.head()", "_____no_output_____" ] ], [ [ "We use numpy and matplotlib to get some statistics and visualize data.", "_____no_output_____" ], [ "print(\"The dataset contains columns of the following data types : \\n\" +str(df_churn_pd.dtypes))", "_____no_output_____" ], [ "Notice below that Gender has three missing values. This will be handled in one of the preprocessing steps that is to follow. ", "_____no_output_____" ] ], [ [ "print(\"The dataset contains following number of records for each of the columns : \\n\" +str(df_churn_pd.count()))", "_____no_output_____" ] ], [ [ "If we are not satisfied with the representational data, now is the time to get more data to be used for training and testing.", "_____no_output_____" ] ], [ [ "print( \"Each category within the churnrisk column has the following count : \")\nprint(df_churn_pd.groupby(['CHURNRISK']).size())\n#bar chart to show split of data\nindex = ['High','Medium','Low']\nchurn_plot = df_churn_pd['CHURNRISK'].value_counts(sort=True, ascending=False).plot(kind='bar',\n figsize=(4,4),title=\"Total number for occurences of churn risk \" \n + str(df_churn_pd['CHURNRISK'].count()), color=['#BB6B5A','#8CCB9B','#E5E88B'])\nchurn_plot.set_xlabel(\"Churn Risk\")\nchurn_plot.set_ylabel(\"Frequency\")", "_____no_output_____" ] ], [ [ "<a id=\"prepare_data\"></a>\n## 3. Data preparation\n[Top](#top)\n\nData preparation is a very important step in machine learning model building. This is because the model can perform well only when the data it is trained on is good and well prepared. Hence, this step consumes the bulk of a data scientist's time spent building models.\n\nDuring this process, we identify categorical columns in the dataset. Categories need to be indexed, which means the string labels are converted to label indices. These label indices are encoded using One-hot encoding to a binary vector with at most a single value indicating the presence of a specific feature value from among the set of all feature values. This encoding allows algorithms which expect continuous features to use categorical features.\n", "_____no_output_____" ] ], [ [ "\n#remove columns that are not required\ndf_churn_pd = df_churn_pd.drop(['ID'], axis=1)\n\ndf_churn_pd.head()\n", "_____no_output_____" ] ], [ [ "\n### [Preprocessing Data](https://scikit-learn.org/stable/modules/preprocessing.html)", "_____no_output_____" ], [ "Scikit-learn provides a method to fill empty values with something that would be applicable in its context. We used the <i><b> SimpleImputer <b></i> class that is provided by Sklearn and filled the missing values with the most frequent value in the column.", "_____no_output_____" ], [ "### [One Hot Encoder](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.OneHotEncoder.html)", "_____no_output_____" ] ], [ [ "# Defining the categorical columns \ncategoricalColumns = ['GENDER', 'STATUS', 'HOMEOWNER']\n\nprint(\"Categorical columns : \" )\nprint(categoricalColumns)\n\nimpute_categorical = SimpleImputer(strategy=\"most_frequent\")\n\nonehot_categorical = OneHotEncoder(handle_unknown='ignore')\n\ncategorical_transformer = Pipeline(steps=[('impute',impute_categorical),('onehot',onehot_categorical)])\n", "_____no_output_____" ] ], [ [ "The numerical columns from the data set are identified, and StandardScaler is applied to each of the columns. This way, each value is subtracted with the mean of its column and divided by its standard deviation.<br>\n### [Standard Scaler](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html)", "_____no_output_____" ] ], [ [ "# Defining the numerical columns \nnumericalColumns = df_churn_pd.select_dtypes(include=[np.float,np.int]).columns\n\nprint(\"Numerical columns : \" )\nprint(numericalColumns)\n\nscaler_numerical = StandardScaler()\n\nnumerical_transformer = Pipeline(steps=[('scale',scaler_numerical)])\n", "_____no_output_____" ] ], [ [ "The preprocessing techniques that are applied must be customized for each of the columns. Sklearn provides a library called the [ColumnTransformer](https://scikit-learn.org/stable/modules/generated/sklearn.compose.ColumnTransformer.html?highlight=columntransformer#sklearn.compose.ColumnTransformer), which allows a sequence of these techniques to be applied to selective columns using a pipeline.\n\nOnly the specified columns in transformers are transformed and combined in the output, and the non-specified columns are dropped. By specifying remainder='passthrough', all remaining columns that were not specified in transformers will be automatically passed through", "_____no_output_____" ] ], [ [ "preprocessorForCategoricalColumns = ColumnTransformer(transformers=[('cat', categorical_transformer, \n categoricalColumns)],\n remainder=\"passthrough\")\npreprocessorForAllColumns = ColumnTransformer(transformers=[('cat', categorical_transformer, categoricalColumns),\n ('num',numerical_transformer,numericalColumns)],\n remainder=\"passthrough\")\n\n", "_____no_output_____" ] ], [ [ "Machine learning algorithms cannot use simple text. We must convert the data from text to a number. Therefore, for each string that is a class we assign a label that is a number. For example, in the customer churn data set, the CHURNRISK output label is classified as high, medium, or low and is assigned labels 0, 1, or 2. We use the [LabelEncoder](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.LabelEncoder.html?highlight=labelencoder#sklearn.preprocessing.LabelEncoder) class provided by Sklearn for this.", "_____no_output_____" ] ], [ [ "# prepare data frame for splitting data into train and test datasets\n\nfeatures = []\nfeatures = df_churn_pd.drop(['CHURNRISK'], axis=1)\n\nlabel_churn = pd.DataFrame(df_churn_pd, columns = ['CHURNRISK']) \nlabel_encoder = LabelEncoder()\nlabel = df_churn_pd['CHURNRISK']\n\nlabel = label_encoder.fit_transform(label)\nprint(\"Encoded value of Churnrisk after applying label encoder : \" + str(label))\n", "_____no_output_____" ] ], [ [ "### These are some of the popular preprocessing steps that are applied on the data sets. Look at [Data Processing in detail](https://developer.ibm.com/articles/data-preprocessing-in-detail/) for more information", "_____no_output_____" ] ], [ [ "area = 75\nx = df_churn_pd['ESTINCOME']\ny = df_churn_pd['DAYSSINCELASTTRADE']\nz = df_churn_pd['TOTALDOLLARVALUETRADED']\n\npop_a = mpatches.Patch(color='#BB6B5A', label='High')\npop_b = mpatches.Patch(color='#E5E88B', label='Medium')\npop_c = mpatches.Patch(color='#8CCB9B', label='Low')\ndef colormap(risk_list):\n cols=[]\n for l in risk_list:\n if l==0:\n cols.append('#BB6B5A')\n elif l==2:\n cols.append('#E5E88B')\n elif l==1:\n cols.append('#8CCB9B')\n return cols\n\nfig = plt.figure(figsize=(12,6))\nfig.suptitle('2D and 3D view of churnrisk data')\n\n# First subplot\nax = fig.add_subplot(1, 2,1)\n\nax.scatter(x, y, alpha=0.8, c=colormap(label), s= area)\nax.set_ylabel('DAYS SINCE LAST TRADE')\nax.set_xlabel('ESTIMATED INCOME')\n\nplt.legend(handles=[pop_a,pop_b,pop_c])\n\n# Second subplot\nax = fig.add_subplot(1,2,2, projection='3d')\n\nax.scatter(z, x, y, c=colormap(label), marker='o')\n\nax.set_xlabel('TOTAL DOLLAR VALUE TRADED')\nax.set_ylabel('ESTIMATED INCOME')\nax.set_zlabel('DAYS SINCE LAST TRADE')\n\nplt.legend(handles=[pop_a,pop_b,pop_c])\n\nplt.show()", "_____no_output_____" ] ], [ [ "\n<a id=\"split_data\"></a>\n## 4. Split data into test and train\n[Top](#top)\n\nScikit-learn provides in built API to split the original dataset into train and test datasets. random_state is set to a number to be able to reproduce the same data split combination through multiple runs. \n\n[Split arrays or matrices into random train and test subsets](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html)", "_____no_output_____" ] ], [ [ "\nX_train, X_test, y_train, y_test = train_test_split(features,label , random_state=0)\nprint(\"Dimensions of datasets that will be used for training : Input features\"+str(X_train.shape)+ \n \" Output label\" + str(y_train.shape))\nprint(\"Dimensions of datasets that will be used for testing : Input features\"+str(X_test.shape)+ \n \" Output label\" + str(y_test.shape))\n", "_____no_output_____" ] ], [ [ "<a id=\"helper_methods\"></a>\n## 5. Helper methods for graph generation\n[Top](#top)\n", "_____no_output_____" ] ], [ [ "def colormap(risk_list):\n cols=[]\n for l in risk_list:\n if l==0:\n cols.append('#BB6B5A')\n elif l==2:\n cols.append('#E5E88B')\n elif l==1:\n cols.append('#8CCB9B')\n return cols\n\ndef two_d_compare(y_test,y_pred,model_name):\n #y_pred = label_encoder.fit_transform(y_pred)\n #y_test = label_encoder.fit_transform(y_test)\n area = (12 * np.random.rand(40))**2 \n plt.subplots(ncols=2, figsize=(10,4))\n plt.suptitle('Actual vs Predicted data : ' +model_name + '. Accuracy : %.2f' % accuracy_score(y_test, y_pred))\n\n plt.subplot(121)\n plt.scatter(X_test['ESTINCOME'], X_test['DAYSSINCELASTTRADE'], alpha=0.8, c=colormap(y_test))\n plt.title('Actual')\n plt.legend(handles=[pop_a,pop_b,pop_c])\n\n plt.subplot(122)\n plt.scatter(X_test['ESTINCOME'], X_test['DAYSSINCELASTTRADE'],alpha=0.8, c=colormap(y_pred))\n plt.title('Predicted')\n plt.legend(handles=[pop_a,pop_b,pop_c])\n\n plt.show()\n \nx = X_test['TOTALDOLLARVALUETRADED']\ny = X_test['ESTINCOME']\nz = X_test['DAYSSINCELASTTRADE']\n\npop_a = mpatches.Patch(color='#BB6B5A', label='High')\npop_b = mpatches.Patch(color='#E5E88B', label='Medium')\npop_c = mpatches.Patch(color='#8CCB9B', label='Low')\n\ndef three_d_compare(y_test,y_pred,model_name):\n fig = plt.figure(figsize=(12,10))\n fig.suptitle('Actual vs Predicted (3D) data : ' +model_name + '. Accuracy : %.2f' % accuracy_score(y_test, y_pred))\n \n ax = fig.add_subplot(121, projection='3d')\n ax.scatter(x, y, z, c=colormap(y_test), marker='o')\n ax.set_xlabel('TOTAL DOLLAR VALUE TRADED')\n ax.set_ylabel('ESTIMATED INCOME')\n ax.set_zlabel('DAYS SINCE LAST TRADE')\n plt.legend(handles=[pop_a,pop_b,pop_c])\n plt.title('Actual')\n\n ax = fig.add_subplot(122, projection='3d')\n ax.scatter(x, y, z, c=colormap(y_pred), marker='o')\n ax.set_xlabel('TOTAL DOLLAR VALUE TRADED')\n ax.set_ylabel('ESTIMATED INCOME')\n ax.set_zlabel('DAYS SINCE LAST TRADE')\n plt.legend(handles=[pop_a,pop_b,pop_c])\n plt.title('Predicted')\n\n plt.show()\n \n\ndef model_metrics(y_test,y_pred):\n print(\"Decoded values of Churnrisk after applying inverse of label encoder : \" + str(np.unique(y_pred)))\n\n skplt.metrics.plot_confusion_matrix(y_test,y_pred,text_fontsize=\"small\",cmap='Greens',figsize=(6,4))\n plt.show()\n \n print(\"The classification report for the model : \\n\\n\"+ classification_report(y_test, y_pred))\n", "_____no_output_____" ] ], [ [ "<a id=\"prepare_model\"></a>\n## 6. Prepare Random Forest classification model\n[Top](#top)", "_____no_output_____" ], [ "We instantiate a decision-tree based classification algorithm, namely, RandomForestClassifier. Next we define a pipeline to chain together the various transformers and estimators defined during the data preparation step before. \nScikit-learn provides APIs that make it easier to combine multiple algorithms into a single pipeline.\n\nWe fit the pipeline to training data and apply the trained model to transform test data and generate churn risk class prediction.", "_____no_output_____" ], [ "[Understanding Random Forest Classifier](https://towardsdatascience.com/understanding-random-forest-58381e0602d2)", "_____no_output_____" ] ], [ [ "from sklearn.ensemble import RandomForestClassifier\n\nmodel_name = \"Random Forest Classifier\"\n\nrandomForestClassifier = RandomForestClassifier(n_estimators=100, max_depth=2,random_state=0)\n", "_____no_output_____" ] ], [ [ "Pipelines are a convenient way of designing your data processing in a machine learning flow. The following code example shows how pipelines are set up using sklearn.\n\nRead more [Here](https://scikit-learn.org/stable/modules/classes.html?highlight=pipeline#module-sklearn.pipeline)", "_____no_output_____" ] ], [ [ "rfc_model = Pipeline(steps=[('preprocessorAll',preprocessorForAllColumns),('classifier', randomForestClassifier)])", "_____no_output_____" ] ], [ [ "<a id=\"train_model\"></a>\n## 7. Train Random Forest classification model\n[Top](#top)", "_____no_output_____" ] ], [ [ "# Build models\n\nrfc_model.fit(X_train,y_train)\n", "_____no_output_____" ] ], [ [ "<a id=\"test_model\"></a>\n## 8. Test Random Forest classification model\n[Top](#top)", "_____no_output_____" ] ], [ [ "\ny_pred_rfc = rfc_model.predict(X_test)\n", "_____no_output_____" ] ], [ [ "<a id=\"evaluate_model\"></a>\n## 9. Evaluate Random Forest classification model\n[Top](#top)", "_____no_output_____" ], [ "### Model results\n\nIn a supervised classification problem such as churn risk classification, we have a true output and a model-generated predicted output for each data point. For this reason, the results for each data point can be assigned to one of four categories:\n\n1. True Positive (TP) - label is positive and prediction is also positive\n2. True Negative (TN) - label is negative and prediction is also negative\n3. False Positive (FP) - label is negative but prediction is positive\n4. False Negative (FN) - label is positive but prediction is negative\n\nThese four numbers are the building blocks for most classifier evaluation metrics. A fundamental point when considering classifier evaluation is that pure accuracy (i.e. was the prediction correct or incorrect) is not generally a good metric. The reason for this is because a dataset may be highly unbalanced. For example, if a model is designed to predict fraud from a dataset where 95% of the data points are not fraud and 5% of the data points are fraud, then a naive classifier that predicts not fraud, regardless of input, will be 95% accurate. For this reason, metrics like precision and recall are typically used because they take into account the type of error. In most applications there is some desired balance between precision and recall, which can be captured by combining the two into a single metric, called the F-measure.\n\n", "_____no_output_____" ] ], [ [ "two_d_compare(y_test,y_pred_rfc,model_name)\n\n#three_d_compare(y_test,y_pred_rfc,model_name)", "_____no_output_____" ] ], [ [ "### Confusion matrix \n\nIn the graph below we have printed a confusion matrix and a self-explanotary classification report.\n\nThe confusion matrix shows that, 42 mediums were wrongly predicted as high, 2 mediums were wrongly predicted as low and 52 mediums were accurately predicted as mediums.", "_____no_output_____" ] ], [ [ "y_test = label_encoder.inverse_transform(y_test)\ny_pred_rfc = label_encoder.inverse_transform(y_pred_rfc)\nmodel_metrics(y_test,y_pred_rfc)", "_____no_output_____" ] ], [ [ "[Precision Recall Fscore support](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.precision_recall_fscore_support.html)", "_____no_output_____" ], [ "[Understanding the Confusion Matrix](https://towardsdatascience.com/confusion-matrix-for-your-multi-class-machine-learning-model-ff9aa3bf7826)", "_____no_output_____" ], [ "### Comparative study\nIn the bar chart below, we have compared the random forest classification algorithm output classes against the actual values. ", "_____no_output_____" ] ], [ [ "uniqueValues, occurCount = np.unique(y_test, return_counts=True)\nfrequency_actual = (occurCount[0],occurCount[2],occurCount[1])\n\nuniqueValues, occurCount = np.unique(y_pred_rfc, return_counts=True)\nfrequency_predicted_rfc = (occurCount[0],occurCount[2],occurCount[1])\n\nn_groups = 3\nfig, ax = plt.subplots(figsize=(10,5))\nindex = np.arange(n_groups)\nbar_width = 0.1\nopacity = 0.8\n\nrects1 = plt.bar(index, frequency_actual, bar_width,\nalpha=opacity,\ncolor='g',\nlabel='Actual')\n\nrects6 = plt.bar(index + bar_width, frequency_predicted_rfc, bar_width,\nalpha=opacity,\ncolor='purple',\nlabel='Random Forest - Predicted')\n\nplt.xlabel('Churn Risk')\nplt.ylabel('Frequency')\nplt.title('Actual vs Predicted frequency.')\nplt.xticks(index + bar_width, ('High', 'Medium', 'Low'))\nplt.legend()\n\nplt.tight_layout()\nplt.show()", "_____no_output_____" ] ], [ [ "<a id=\"model_knn\"></a>\n## 10. Build K-Nearest classification model\n[Top](#top)\n\nK number of nearest points around the data point to be predicted are taken into consideration. These K points at this time, already belong to a class. The data point under consideration, is said to belong to the class with which most number of points from these k points belong to. ", "_____no_output_____" ] ], [ [ "from sklearn.neighbors import KNeighborsClassifier\n\nmodel_name = \"K-Nearest Neighbor Classifier\"\n\nknnClassifier = KNeighborsClassifier(n_neighbors = 5, metric='minkowski', p=2)\n\nknn_model = Pipeline(steps=[('preprocessorAll',preprocessorForAllColumns),('classifier', knnClassifier)]) \n\nknn_model.fit(X_train,y_train)\n\ny_pred_knn = knn_model.predict(X_test)\n", "_____no_output_____" ], [ "y_test = label_encoder.transform(y_test)\ntwo_d_compare(y_test,y_pred_knn,model_name)", "_____no_output_____" ], [ "y_test = label_encoder.inverse_transform(y_test)\ny_pred_knn = label_encoder.inverse_transform(y_pred_knn)\nmodel_metrics(y_test,y_pred_knn)", "_____no_output_____" ] ], [ [ "<a id=\"compare_classification\"></a>\n## 11. Comparative study of both classification algorithms. \n[Top](#top)\n\n\n ", "_____no_output_____" ] ], [ [ "uniqueValues, occurCount = np.unique(y_test, return_counts=True)\nfrequency_actual = (occurCount[0],occurCount[2],occurCount[1])\n\nuniqueValues, occurCount = np.unique(y_pred_rfc, return_counts=True)\nfrequency_predicted_rfc = (occurCount[0],occurCount[2],occurCount[1])\n\nuniqueValues, occurCount = np.unique(y_pred_knn, return_counts=True)\nfrequency_predicted_knn = (occurCount[0],occurCount[2],occurCount[1])\n\nn_groups = 3\nfig, ax = plt.subplots(figsize=(10,5))\nindex = np.arange(n_groups)\nbar_width = 0.1\nopacity = 0.8\n\nrects1 = plt.bar(index, frequency_actual, bar_width,\nalpha=opacity,\ncolor='g',\nlabel='Actual')\n\n\nrects6 = plt.bar(index + bar_width*2, frequency_predicted_rfc, bar_width,\nalpha=opacity,\ncolor='purple',\nlabel='Random Forest - Predicted')\n\nrects4 = plt.bar(index + bar_width*4, frequency_predicted_knn, bar_width,\nalpha=opacity,\ncolor='b',\nlabel='K-Nearest Neighbor - Predicted')\n\nplt.xlabel('Churn Risk')\nplt.ylabel('Frequency')\nplt.title('Actual vs Predicted frequency.')\nplt.xticks(index + bar_width, ('High', 'Medium', 'Low'))\nplt.legend()\n\nplt.tight_layout()\nplt.show()", "_____no_output_____" ] ], [ [ "Until evaluation provides satisfactory scores, you would repeat the data preprocessing through evaluating steps by tuning what are called the hyperparameters.", "_____no_output_____" ], [ "[Choosing the right estimator](https://scikit-learn.org/stable/tutorial/machine_learning_map/index.html)", "_____no_output_____" ], [ "### For a comparative study of some of the current most popular algorithms Please refer to this [tutorial](https://developer.ibm.com/tutorials/learn-classification-algorithms-using-python-and-scikit-learn/)", "_____no_output_____" ], [ "<p><font size=-1 color=gray>\n&copy; Copyright 2019 IBM Corp. All Rights Reserved.\n<p>\nLicensed under the Apache License, Version 2.0 (the \"License\"); you may not use this file\nexcept in compliance with the License. You may obtain a copy of the License at\nhttps://www.apache.org/licenses/LICENSE-2.0\nUnless required by applicable law or agreed to in writing, software distributed under the\nLicense is distributed on an \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either\nexpress or implied. See the License for the specific language governing permissions and\nlimitations under the License.\n</font></p>", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ] ]

[ "ADSL" ]

[ "ADSL" ]

[ "ADSL" ]

[ [ [ "from bs4 import BeautifulSoup as bs\nfrom splinter import Browser\nimport pandas as pd", "_____no_output_____" ], [ "with Browser(\"chrome\") as browser:\n # Visit URL\n url = \"https://mars.nasa.gov/news/\"\n browser.visit(url)\n browser.fill('search', 'splinter - python acceptance testing for web applications')\n # Find and click the 'search' button\n button = browser.find_by_css('input.search_submit')\n print(button[0])\n # Interact with elements\n button[0].click()\n if browser.is_text_present('There are no items matching these criteria.'):\n print(\"Yes, the official website was found!\")\n else:\n print(\"No, it wasn't found... We need to improve our SEO techniques\")", "<splinter.driver.webdriver.WebDriverElement object at 0x0000027A80EFAF88>\nNo, it wasn't found... We need to improve our SEO techniques\n" ], [ "#Scrape the [NASA Mars News Site](https://mars.nasa.gov/news/) and collect the latest News Title and Paragraph Text. Assign the text to variables that you can reference later.\n\n#```python\n# Example:\nnews_title = \"NASA's Next Mars Mission to Investigate Interior of Red Planet\"\n\nnews_p = \"Preparation of NASA's next spacecraft to Mars, InSight, has ramped up this summer, on course for launch next May from Vandenberg Air Force Base in central California -- the first interplanetary launch in history from America's West Coast.\"", "_____no_output_____" ], [ "with Browser(\"chrome\") as browser:\n url = \"https://mars.nasa.gov/news/\"\n browser.visit(url)\n html = bs(browser.html, 'html.parser')\n \n body = html.body\n# body.strippedtext\n# print(body.a)\n\n print(body.find_all(\"div\", class_='content_title')[1].getText())\n# titles = body.find_all(\"div\", class_='content_title')\n# first = titles[1].getText()\n# print(first)\n ", "\n\nAlabama High School Student Names NASA's Mars Helicopter\n\n\n" ], [ "# Use splinter to navigate the site and find the image url for the current Featured Mars Image and \n#assign the url string to a variable called `featured_image_url`.\n\n# * Make sure to find the image url to the full size `.jpg` image.\n\n\n# * Make sure to save a complete url string for this image.\n\n# ```python\n# # Example:\n# featured_image_url = 'https://www.jpl.nasa.gov/spaceimages/images/largesize/PIA16225_hires.jpg'\n\nbase_url = \"https://www.jpl.nasa.gov\"\nsearch_url = '/spaceimages/?search=&category=Mars'\nurl = base_url + search_url\nfeatured_image_url = None\nwith Browser(\"chrome\") as browser:\n browser.visit(url)\n image = browser.find_by_id('full_image')[0]\n featured_image_url = base_url + image['data-fancybox-href']\n print(featured_image_url)\n", "https://www.jpl.nasa.gov/spaceimages/images/mediumsize/PIA17838_ip.jpg\n" ], [ "#Visit the Mars Weather twitter account [here](https://twitter.com/marswxreport?lang=en) and \n#scrape the latest Mars weather tweet from the page. Save the tweet text for the weather \n#report as a variable called `mars_weather`.\n#Note: Be sure you are not signed in to twitter, or scraping may become more difficult.**\nimport time\nwith Browser(\"chrome\") as browser:\n url_weather = \"https://twitter.com/marswxreport?lang=en\"\n browser.visit(url_weather) \n html_weather = browser.html\n time.sleep(5)\n soup = bs(html_weather, \"html.parser\")\n main = soup.main\n temp = main.find_all('section', attrs={\"aria-labelledby\": \"accessible-list-0\"})\n elements = soup.find_all(\"section\", class_=\"css-1dbjc4n\")\n for e in elements:\n print(e)\n print(temp.text)\n print(temp[0])\n# url= \"https://astrogeology.usgs.gov/search/results?q=hemisphere+enhanced&k1=target&v1=Mars\"\n# with Browser(\"chrome\") as browser:\n\n# browser.visit(url) \n# html = browser.html\n# soup = bs(html, 'html.parser')\n# xpath = \"//div[@class='description']//a[@class='itemLink product-item']/h3\"\n# results = browser.find_by_xpath(xpath)\n# hemisphere_image_urls = []\n# for i in range(4):\n# html = browser.html\n# soup = bs(html, 'html.parser')\n# # find the new Splinter elements\n# results = browser.find_by_xpath(xpath)\n# # save name of the hemisphere\n# header = results[i].html\n# # go to hemisphere details page \n# details_link = results[i]\n# details_link.click()\n\n# html = browser.html\n# soup = bs(html, 'html.parser')\n# # Save the image url\n# hemisphere_image_urls.append({\"title\": header, \"image_url\": soup.find(\"div\", class_=\"downloads\").a[\"href\"]})\n# # Go back to the original page\n# browser.back()\n\n# print(hemisphere_image_urls)\n", "_____no_output_____" ], [ "#Visit the Mars Facts webpage [here](https://space-facts.com/mars/) and use Pandas to scrape the table \n#containing facts about the planet including Diameter, Mass, etc.\nurl = \"https://space-facts.com/mars/\"\ndata = pd.read_html(url)\ndf = data[0]\n\ndf.columns = ['Description', 'Value']\ndf.set_index('Description', inplace=True)\nmars_html_table = df.to_html()\nprint(df)\n\n\n#Use Pandas to convert the data to a HTML table string.\n", "_____no_output_____" ], [ "# * Visit the USGS Astrogeology site [here](https://astrogeology.usgs.gov/search/results?q=hemisphere+enhanced&k1=target&v1=Mars)\n# to obtain high resolution images for each of Mar's hemispheres.\n\n# * You will need to click each of the links to the hemispheres in order to find \n# the image url to the full resolution image.\n\n# * Save both the image url string for the full resolution hemisphere image, and \n# the Hemisphere title containing the hemisphere name. Use a Python dictionary to store \n#the data using the keys `img_url` and `title`.\n\n# * Append the dictionary with the image url string and the hemisphere title to a list.\n# This list will contain one dictionary for each hemisphere.\n\n# ```python\n# # Example:\n# hemisphere_image_urls = [\n# {\"title\": \"Valles Marineris Hemisphere\", \"img_url\": \"...\"},\n# {\"title\": \"Cerberus Hemisphere\", \"img_url\": \"...\"},\n# {\"title\": \"Schiaparelli Hemisphere\", \"img_url\": \"...\"},\n# {\"title\": \"Syrtis Major Hemisphere\", \"img_url\": \"...\"},\n# ]\n# ```\n\nurl= \"https://astrogeology.usgs.gov/search/results?q=hemisphere+enhanced&k1=target&v1=Mars\"\nwith Browser(\"chrome\") as browser:\n\n browser.visit(url) \n html = browser.html\n soup = bs(html, 'html.parser')\n xpath = \"//div[@class='description']//a[@class='itemLink product-item']/h3\"\n results = browser.find_by_xpath(xpath)\n hemisphere_image_urls = []\n for i in range(4):\n html = browser.html\n soup = bs(html, 'html.parser')\n # find the new Splinter elements\n results = browser.find_by_xpath(xpath)\n # save name of the hemisphere\n header = results[i].html\n # go to hemisphere details page \n details_link = results[i]\n details_link.click()\n\n html = browser.html\n soup = bs(html, 'html.parser')\n # Save the image url\n hemisphere_image_urls.append({\"title\": header, \"image_url\": soup.find(\"div\", class_=\"downloads\").a[\"href\"]})\n # Go back to the original page\n browser.back()\n\n print(hemisphere_image_urls)\n #to make it prettier\n for i in hemisphere_image_urls: print(i)", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "CC-BY-3.0" ]

[ "CC-BY-3.0" ]

[ "CC-BY-3.0" ]

[ [ [ "<small><small><i>\nIntroduction to Python - available from https://gitlab.erc.monash.edu.au/andrease/Python4Maths.git\n\nThe original version was written by Rajath Kumar and is available at https://github.com/rajathkumarmp/Python-Lectures.\nThe notes have been updated for Python 3 and amended for use in Monash University mathematics courses by [Andreas Ernst](http://users.monash.edu.au/~andreas) \n</small></small></i>", "_____no_output_____" ], [ "# Python-Lectures", "_____no_output_____" ], [ "## Introduction\n\nPython is a modern, robust, high level programming language. It is very easy to pick up even if you are completely new to programming. \n\nPython, similar to other languages like matlab or R, is interpreted hence runs slowly compared to C++, Fortran or Java. However writing programs in Python is very quick. Python has a very large collection of libraries for everything from scientific computing to web services. It caters for object oriented and functional programming with module system that allows large and complex applications to be developed in Python. \n\nThese lectures are using jupyter notebooks which mix Python code with documentation. The python notebooks can be run on a webserver or stand-alone on a computer.\n\nTo give an indication of what Python code looks like, here is a simple bit of code that defines a set $N=\\{1,3,4,5,7\\}$ and calculates the sum of the squared elements of this set: $$\\sum_{i\\in N} i^2=100$$", "_____no_output_____" ] ], [ [ "N={1,3,4,5,7,8}\nprint('The sum of ∑_i∈N i*i =',sum( i**2 for i in N ) )", "The sum of ∑_i∈N i*i = 164\n" ] ], [ [ "## Contents\n\nThis course is broken up into a number of notebooks (chapters).\n\n* [00](00.ipynb) This introduction with additional information below on how to get started in running python\n* [01](01.ipynb) Basic data types and operations (numbers, strings) \n* [02](02.ipynb) String manipulation \n* [03](03.ipynb) Data structures: Lists and Tuples\n* [04](04.ipynb) Data structures (continued): dictionaries\n* [05](05.ipynb) Control statements: if, for, while, try statements\n* [06](06.ipynb) Functions\n* [07](07.ipynb) Classes and basic object oriented programming\n* [08](08.ipynb) Scipy: libraries for arrays (matrices) and plotting\n* [09](09.ipynb) Mixed Integer Linear Programming using the mymip library.\n* [10](10.ipynb) Networks and graphs under python - a very brief introduction\n* [11](11.ipynb) Using the numba library for fast numerical computing.\n\nThis is a tutorial style introduction to Python. For a quick reminder / summary of Python syntax the following [Quick Reference Card](http://www.cs.put.poznan.pl/csobaniec/software/python/py-qrc.html) may be useful. A longer and more detailed tutorial style introduction to python is available from the python site at: https://docs.python.org/3/tutorial/\n", "_____no_output_____" ], [ "## Installation\n\n### Loging into the web server\nThe easiest way to run this and other notebooks for staff and students at Monash University is to log into the Jupyter server at [https://sci-web17-v01.ocio.monash.edu.au/hub](https://sci-web17-v01.ocio.monash.edu.au/hub). The steps for running notebooks are:\n* Log in using your monash email address. The first time you log in an empty account will automatically be set up for you.\n* Press the start button (if prompted by the system)\n* Use the menu of the jupyter system to upload a .ipynb python notebook file or to start a new notebook.\n\n### Installing \n\nPython runs on windows, linux, mac and other environments. There are many python distributions available. However the recommended way to install python under Microsoft Windows or Linux is to use the Anaconda distribution available at [https://www.continuum.io/downloads](https://www.continuum.io/downloads). Make sure to get the Python *3.6* version, not 2.7. This distribution comes with the [SciPy](https://www.scipy.org/) collection of scientific python tools as well as the iron python notebook. For developing python code without notebooks consider using [spyder](https://github.com/spyder-ide/spyder) (also included with Anaconda)\n\nTo open a notebook with anaconda installed, from the terminal run:\n\n ipython notebook\n\nNote that for the Monash University optimisation course additional modules relating to the commercial optimisation library [CPLEX](http://www-01.ibm.com/software/commerce/optimization/cplex-optimizer/index.html) and possibly [Gurobi](http://www.gurobi.com/) will be used. These libraries are not available as part of any standard distribution but are available under academic licence. Cplex is included on the [Monash server](https://sci-web17-v01.ocio.monash.edu.au/hub).", "_____no_output_____" ], [ "## How to learn from this resource?\n\nDownload all the notebooks from Moodle or https://gitlab.erc.monash.edu.au/andrease/Python4Maths.git\n\nUpload them to the monash server and lauch them or launch ipython notebook from the folder which contains the notebooks. Open each one of them\n\nCell > All Output > Clear\n\nThis will clear all the outputs and now you can understand each statement and learn interactively.\n", "_____no_output_____" ], [ "## License\nThis work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "from ..src.data.get_config import get_config", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Overfitting demo\n\n## Create a dataset based on a true sinusoidal relationship\nLet's look at a synthetic dataset consisting of 30 points drawn from the sinusoid $y = \\sin(4x)$:", "_____no_output_____" ] ], [ [ "import graphlab\nimport math\nimport random\nimport numpy\nfrom matplotlib import pyplot as plt\n%matplotlib inline", "_____no_output_____" ] ], [ [ "Create random values for x in interval [0,1)", "_____no_output_____" ] ], [ [ "random.seed(98103)\nn = 30\nx = graphlab.SArray([random.random() for i in range(n)]).sort()", "2016-04-25 03:15:49,473 [INFO] graphlab.cython.cy_server, 176: GraphLab Create v1.8.5 started. Logging: C:\\Users\\PHILIP~1\\AppData\\Local\\Temp\\graphlab_server_1461579348.log.0\n" ] ], [ [ "Compute y", "_____no_output_____" ] ], [ [ "y = x.apply(lambda x: math.sin(4*x))", "_____no_output_____" ] ], [ [ "Add random Gaussian noise to y", "_____no_output_____" ] ], [ [ "random.seed(1)\ne = graphlab.SArray([random.gauss(0,1.0/3.0) for i in range(n)])\ny = y + e", "_____no_output_____" ] ], [ [ "### Put data into an SFrame to manipulate later", "_____no_output_____" ] ], [ [ "data = graphlab.SFrame({'X1':x,'Y':y})\ndata", "_____no_output_____" ] ], [ [ "### Create a function to plot the data, since we'll do it many times", "_____no_output_____" ] ], [ [ "def plot_data(data): \n plt.plot(data['X1'],data['Y'],'k.')\n plt.xlabel('x')\n plt.ylabel('y')\n\nplot_data(data)", "_____no_output_____" ] ], [ [ "## Define some useful polynomial regression functions", "_____no_output_____" ], [ "Define a function to create our features for a polynomial regression model of any degree:", "_____no_output_____" ] ], [ [ "def polynomial_features(data, deg):\n data_copy=data.copy()\n for i in range(1,deg):\n data_copy['X'+str(i+1)]=data_copy['X'+str(i)]*data_copy['X1']\n return data_copy", "_____no_output_____" ] ], [ [ "Define a function to fit a polynomial linear regression model of degree \"deg\" to the data in \"data\":", "_____no_output_____" ] ], [ [ "def polynomial_regression(data, deg):\n model = graphlab.linear_regression.create(polynomial_features(data,deg), \n target='Y', l2_penalty=0.,l1_penalty=0.,\n validation_set=None,verbose=False)\n return model", "_____no_output_____" ] ], [ [ "Define function to plot data and predictions made, since we are going to use it many times.", "_____no_output_____" ] ], [ [ "def plot_poly_predictions(data, model):\n plot_data(data)\n\n # Get the degree of the polynomial\n deg = len(model.coefficients['value'])-1\n \n # Create 200 points in the x axis and compute the predicted value for each point\n x_pred = graphlab.SFrame({'X1':[i/200.0 for i in range(200)]})\n y_pred = model.predict(polynomial_features(x_pred,deg))\n \n # plot predictions\n plt.plot(x_pred['X1'], y_pred, 'g-', label='degree ' + str(deg) + ' fit')\n plt.legend(loc='upper left')\n plt.axis([0,1,-1.5,2])", "_____no_output_____" ] ], [ [ "Create a function that prints the polynomial coefficients in a pretty way :)", "_____no_output_____" ] ], [ [ "def print_coefficients(model): \n # Get the degree of the polynomial\n deg = len(model.coefficients['value'])-1\n\n # Get learned parameters as a list\n w = list(model.coefficients['value'])\n\n # Numpy has a nifty function to print out polynomials in a pretty way\n # (We'll use it, but it needs the parameters in the reverse order)\n print 'Learned polynomial for degree ' + str(deg) + ':'\n w.reverse()\n print numpy.poly1d(w)", "_____no_output_____" ] ], [ [ "## Fit a degree-2 polynomial", "_____no_output_____" ], [ "Fit our degree-2 polynomial to the data generated above:", "_____no_output_____" ] ], [ [ "model = polynomial_regression(data, deg=2)", "_____no_output_____" ] ], [ [ "Inspect learned parameters", "_____no_output_____" ] ], [ [ "print_coefficients(model)", "Learned polynomial for degree 2:\n 2\n-5.129 x + 4.147 x + 0.07471\n" ] ], [ [ "Form and plot our predictions along a grid of x values:", "_____no_output_____" ] ], [ [ "plot_poly_predictions(data,model)", "_____no_output_____" ] ], [ [ "## Fit a degree-4 polynomial", "_____no_output_____" ] ], [ [ "model = polynomial_regression(data, deg=4)\nprint_coefficients(model)\nplot_poly_predictions(data,model)", "Learned polynomial for degree 4:\n 4 3 2\n23.87 x - 53.82 x + 35.23 x - 6.828 x + 0.7755\n" ] ], [ [ "## Fit a degree-16 polynomial", "_____no_output_____" ] ], [ [ "model = polynomial_regression(data, deg=16)\nprint_coefficients(model)", "Learned polynomial for degree 16:\n 16 15 14 13\n-4.537e+05 x + 1.129e+06 x + 4.821e+05 x - 3.81e+06 x \n 12 11 10 9\n + 3.536e+06 x + 5.753e+04 x - 1.796e+06 x + 2.178e+06 x\n 8 7 6 5 4\n - 3.662e+06 x + 4.442e+06 x - 3.13e+06 x + 1.317e+06 x - 3.356e+05 x\n 3 2\n + 5.06e+04 x - 4183 x + 160.8 x - 1.621\n" ] ], [ [ "###Woah!!!! Those coefficients are *crazy*! On the order of 10^6.", "_____no_output_____" ] ], [ [ "plot_poly_predictions(data,model)", "_____no_output_____" ] ], [ [ "### Above: Fit looks pretty wild, too. Here's a clear example of how overfitting is associated with very large magnitude estimated coefficients.", "_____no_output_____" ], [ "# ", "_____no_output_____" ], [ "# ", "_____no_output_____" ], [ " # ", "_____no_output_____" ], [ " # ", "_____no_output_____" ], [ "# Ridge Regression", "_____no_output_____" ], [ "Ridge regression aims to avoid overfitting by adding a cost to the RSS term of standard least squares that depends on the 2-norm of the coefficients $\\|w\\|$. The result is penalizing fits with large coefficients. The strength of this penalty, and thus the fit vs. model complexity balance, is controled by a parameter lambda (here called \"L2_penalty\").", "_____no_output_____" ], [ "Define our function to solve the ridge objective for a polynomial regression model of any degree:", "_____no_output_____" ] ], [ [ "def polynomial_ridge_regression(data, deg, l2_penalty):\n model = graphlab.linear_regression.create(polynomial_features(data,deg), \n target='Y', l2_penalty=l2_penalty,\n validation_set=None,verbose=False)\n return model", "_____no_output_____" ] ], [ [ "## Perform a ridge fit of a degree-16 polynomial using a *very* small penalty strength", "_____no_output_____" ] ], [ [ "model = polynomial_ridge_regression(data, deg=16, l2_penalty=1e-25)\nprint_coefficients(model)", "Learned polynomial for degree 16:\n 16 15 14 13\n-4.537e+05 x + 1.129e+06 x + 4.821e+05 x - 3.81e+06 x \n 12 11 10 9\n + 3.536e+06 x + 5.753e+04 x - 1.796e+06 x + 2.178e+06 x\n 8 7 6 5 4\n - 3.662e+06 x + 4.442e+06 x - 3.13e+06 x + 1.317e+06 x - 3.356e+05 x\n 3 2\n + 5.06e+04 x - 4183 x + 160.8 x - 1.621\n" ], [ "plot_poly_predictions(data,model)", "_____no_output_____" ] ], [ [ "## Perform a ridge fit of a degree-16 polynomial using a very large penalty strength", "_____no_output_____" ] ], [ [ "model = polynomial_ridge_regression(data, deg=16, l2_penalty=100)\nprint_coefficients(model)", "Learned polynomial for degree 16:\n 16 15 14 13 12 11\n-0.301 x - 0.2802 x - 0.2604 x - 0.2413 x - 0.2229 x - 0.205 x \n 10 9 8 7 6 5\n - 0.1874 x - 0.1699 x - 0.1524 x - 0.1344 x - 0.1156 x - 0.09534 x\n 4 3 2\n - 0.07304 x - 0.04842 x - 0.02284 x - 0.002257 x + 0.6416\n" ], [ "plot_poly_predictions(data,model)", "_____no_output_____" ] ], [ [ "## Let's look at fits for a sequence of increasing lambda values", "_____no_output_____" ] ], [ [ "for l2_penalty in [1e-25, 1e-10, 1e-6, 1e-3, 1e2]:\n model = polynomial_ridge_regression(data, deg=16, l2_penalty=l2_penalty)\n print 'lambda = %.2e' % l2_penalty\n print_coefficients(model)\n print '\\n'\n plt.figure()\n plot_poly_predictions(data,model)\n plt.title('Ridge, lambda = %.2e' % l2_penalty)", "lambda = 1.00e-25\nLearned polynomial for degree 16:\n 16 15 14 13\n-4.537e+05 x + 1.129e+06 x + 4.821e+05 x - 3.81e+06 x \n 12 11 10 9\n + 3.536e+06 x + 5.753e+04 x - 1.796e+06 x + 2.178e+06 x\n 8 7 6 5 4\n - 3.662e+06 x + 4.442e+06 x - 3.13e+06 x + 1.317e+06 x - 3.356e+05 x\n 3 2\n + 5.06e+04 x - 4183 x + 160.8 x - 1.621\n\n\nlambda = 1.00e-10\nLearned polynomial for degree 16:\n 16 15 14 13\n4.975e+04 x - 7.821e+04 x - 2.265e+04 x + 3.949e+04 x \n 12 11 10 9 8\n + 4.366e+04 x + 3074 x - 3.332e+04 x - 2.786e+04 x + 1.032e+04 x\n 7 6 5 4 3 2\n + 2.962e+04 x - 1440 x - 2.597e+04 x + 1.839e+04 x - 5596 x + 866.1 x - 65.19 x + 2.159\n\n\nlambda = 1.00e-06\nLearned polynomial for degree 16:\n 16 15 14 13 12 11\n329.1 x - 356.4 x - 264.2 x + 33.8 x + 224.7 x + 210.8 x \n 10 9 8 7 6 5 4\n + 49.62 x - 122.4 x - 178 x - 79.13 x + 84.89 x + 144.9 x + 5.123 x\n 3 2\n - 156.9 x + 88.21 x - 14.82 x + 1.059\n\n\nlambda = 1.00e-03\nLearned polynomial for degree 16:\n 16 15 14 13 12 11\n6.364 x - 1.596 x - 4.807 x - 4.778 x - 2.776 x + 0.1238 x \n 10 9 8 7 6 5\n + 2.977 x + 4.926 x + 5.203 x + 3.248 x - 0.9291 x - 6.011 x\n 4 3 2\n - 8.395 x - 2.655 x + 9.861 x - 2.225 x + 0.5636\n\n\nlambda = 1.00e+02\nLearned polynomial for degree 16:\n 16 15 14 13 12 11\n-0.301 x - 0.2802 x - 0.2604 x - 0.2413 x - 0.2229 x - 0.205 x \n 10 9 8 7 6 5\n - 0.1874 x - 0.1699 x - 0.1524 x - 0.1344 x - 0.1156 x - 0.09534 x\n 4 3 2\n - 0.07304 x - 0.04842 x - 0.02284 x - 0.002257 x + 0.6416\n\n\n" ], [ "data", "_____no_output_____" ] ], [ [ "## Perform a ridge fit of a degree-16 polynomial using a \"good\" penalty strength", "_____no_output_____" ], [ "We will learn about cross validation later in this course as a way to select a good value of the tuning parameter (penalty strength) lambda. Here, we consider \"leave one out\" (LOO) cross validation, which one can show approximates average mean square error (MSE). As a result, choosing lambda to minimize the LOO error is equivalent to choosing lambda to minimize an approximation to average MSE.", "_____no_output_____" ] ], [ [ "# LOO cross validation -- return the average MSE\ndef loo(data, deg, l2_penalty_values):\n # Create polynomial features\n data = polynomial_features(data, deg)\n \n # Create as many folds for cross validatation as number of data points\n num_folds = len(data)\n folds = graphlab.cross_validation.KFold(data,num_folds)\n \n # for each value of l2_penalty, fit a model for each fold and compute average MSE\n l2_penalty_mse = []\n min_mse = None\n best_l2_penalty = None\n for l2_penalty in l2_penalty_values:\n next_mse = 0.0\n for train_set, validation_set in folds:\n # train model\n model = graphlab.linear_regression.create(train_set,target='Y', \n l2_penalty=l2_penalty,\n validation_set=None,verbose=False)\n \n # predict on validation set \n y_test_predicted = model.predict(validation_set)\n # compute squared error\n next_mse += ((y_test_predicted-validation_set['Y'])**2).sum()\n \n # save squared error in list of MSE for each l2_penalty\n next_mse = next_mse/num_folds\n l2_penalty_mse.append(next_mse)\n if min_mse is None or next_mse < min_mse:\n min_mse = next_mse\n best_l2_penalty = l2_penalty\n \n return l2_penalty_mse,best_l2_penalty", "_____no_output_____" ] ], [ [ "Run LOO cross validation for \"num\" values of lambda, on a log scale", "_____no_output_____" ] ], [ [ "l2_penalty_values = numpy.logspace(-4, 10, num=10)\nl2_penalty_mse,best_l2_penalty = loo(data, 16, l2_penalty_values)", "_____no_output_____" ] ], [ [ "Plot results of estimating LOO for each value of lambda", "_____no_output_____" ] ], [ [ "plt.plot(l2_penalty_values,l2_penalty_mse,'k-')\nplt.xlabel('$\\ell_2$ penalty')\nplt.ylabel('LOO cross validation error')\nplt.xscale('log')\nplt.yscale('log')", "_____no_output_____" ] ], [ [ "Find the value of lambda, $\\lambda_{\\mathrm{CV}}$, that minimizes the LOO cross validation error, and plot resulting fit", "_____no_output_____" ] ], [ [ "best_l2_penalty", "_____no_output_____" ], [ "model = polynomial_ridge_regression(data, deg=16, l2_penalty=best_l2_penalty)\nprint_coefficients(model)", "Learned polynomial for degree 16:\n 16 15 14 13 12 11\n1.345 x + 1.141 x + 0.9069 x + 0.6447 x + 0.3569 x + 0.04947 x \n 10 9 8 7 6 5\n - 0.2683 x - 0.5821 x - 0.8701 x - 1.099 x - 1.216 x - 1.145 x\n 4 3 2\n - 0.7837 x - 0.07406 x + 0.7614 x + 0.7703 x + 0.3918\n" ], [ "plot_poly_predictions(data,model)", "_____no_output_____" ] ], [ [ "# ", "_____no_output_____" ], [ "# ", "_____no_output_____" ], [ "# ", "_____no_output_____" ], [ "# ", "_____no_output_____" ], [ "# Lasso Regression", "_____no_output_____" ], [ "Lasso regression jointly shrinks coefficients to avoid overfitting, and implicitly performs feature selection by setting some coefficients exactly to 0 for sufficiently large penalty strength lambda (here called \"L1_penalty\"). In particular, lasso takes the RSS term of standard least squares and adds a 1-norm cost of the coefficients $\\|w\\|$.", "_____no_output_____" ], [ "Define our function to solve the lasso objective for a polynomial regression model of any degree:", "_____no_output_____" ] ], [ [ "def polynomial_lasso_regression(data, deg, l1_penalty):\n model = graphlab.linear_regression.create(polynomial_features(data,deg), \n target='Y', l2_penalty=0.,\n l1_penalty=l1_penalty,\n validation_set=None, \n solver='fista', verbose=False,\n max_iterations=3000, convergence_threshold=1e-10)\n return model", "_____no_output_____" ] ], [ [ "## Explore the lasso solution as a function of a few different penalty strengths", "_____no_output_____" ], [ "We refer to lambda in the lasso case below as \"l1_penalty\"", "_____no_output_____" ] ], [ [ "for l1_penalty in [0.0001, 0.01, 0.1, 10]:\n model = polynomial_lasso_regression(data, deg=16, l1_penalty=l1_penalty)\n print 'l1_penalty = %e' % l1_penalty\n print 'number of nonzeros = %d' % (model.coefficients['value']).nnz()\n print_coefficients(model)\n print '\\n'\n plt.figure()\n plot_poly_predictions(data,model)\n plt.title('LASSO, lambda = %.2e, # nonzeros = %d' % (l1_penalty, (model.coefficients['value']).nnz()))", "l1_penalty = 1.000000e-04\nnumber of nonzeros = 17\nLearned polynomial for degree 16:\n 16 15 14 13 12 11\n29.02 x + 1.35 x - 12.72 x - 16.93 x - 13.82 x - 6.698 x \n 10 9 8 7 6 5\n + 1.407 x + 8.939 x + 12.88 x + 11.44 x + 3.759 x - 8.062 x\n 4 3 2\n - 16.28 x - 7.682 x + 17.86 x - 4.384 x + 0.685\n\n\nl1_penalty = 1.000000e-02\nnumber of nonzeros = 14\nLearned polynomial for degree 16:\n 16 15 11 10 9 8\n-1.18 x - 0.001318 x + 0.08745 x + 0.7389 x + 3.828 x + 0.4761 x\n 7 6 5 4 3 2\n + 0.1282 x + 0.001952 x - 0.6151 x - 10.11 x - 0.0003954 x + 6.686 x - 1.28 x + 0.5056\n\n\nl1_penalty = 1.000000e-01\nnumber of nonzeros = 5\nLearned polynomial for degree 16:\n 16 6 5\n2.21 x - 1.002 x - 2.962 x + 1.216 x + 0.3473\n\n\nl1_penalty = 1.000000e+01\nnumber of nonzeros = 2\nLearned polynomial for degree 16:\n 9\n-1.526 x + 0.5755\n\n\n" ] ], [ [ "Above: We see that as lambda increases, we get sparser and sparser solutions. However, even for our non-sparse case for lambda=0.0001, the fit of our high-order polynomial is not too wild. This is because, like in ridge, coefficients included in the lasso solution are shrunk relative to those of the least squares (unregularized) solution. This leads to better behavior even without sparsity. Of course, as lambda goes to 0, the amount of this shrinkage decreases and the lasso solution approaches the (wild) least squares solution.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Gridworld", "_____no_output_____" ], [ "\n\nThe Gridworld environment (inspired from Sutton and Barto, Reinforcement Learning: an Introduction) is represented in figure. The environment is a finite MDP in which states are represented by grid cells. The available actions are 4: left, right, up, down. Actions move the current state in the action direction and the associated reward is 0 for all actions. Exceptions are:\n\n- Border cells: if the action brings the agent outside of the grid the agent state does not change and the agent receives a reward of -1.\n\n- Good cells: $G_1$ and $G_2$ are special cells. For these cells each action brings the agent in state $G_1$' and $G_2$' respectively. The associated reward is +10 for going outside state $G_1$ and +5 for going outside state $G_2$.\n\n- Bad cells: $B_1$ and $B_2$ are bad cells. For these cells the associated reward is -1 for all actions.\n\nThe goal of the activity is to calculate and represent visually the state values for the random policy, in which the agent selects each action with equal probability (1/4) in all states. The discount factor is assumed to be equal to 0.9.", "_____no_output_____" ], [ "## Solution", "_____no_output_____" ], [ "Imports:", "_____no_output_____" ] ], [ [ "from enum import Enum, auto\nimport matplotlib.pyplot as plt\nimport numpy as np\nfrom scipy import linalg\nfrom typing import Tuple", "_____no_output_____" ] ], [ [ "Visualization function:", "_____no_output_____" ] ], [ [ "# helper function\ndef vis_matrix(M, cmap=plt.cm.Blues):\n fig, ax = plt.subplots()\n ax.matshow(M, cmap=cmap)\n for i in range(M.shape[0]):\n for j in range(M.shape[1]):\n c = M[j, i]\n ax.text(i, j, \"%.2f\" % c, va=\"center\", ha=\"center\")", "_____no_output_____" ], [ "# Define the actions\nclass Action(Enum):\n UP = auto()\n DOWN = auto()\n LEFT = auto()\n RIGHT = auto()\n\n\n# Agent Policy, random\nclass Policy:\n def __init__(self):\n self._possible_actions = [action for action in Action]\n self._action_probs = {\n a: 1 / len(self._possible_actions) for a in self._possible_actions\n }\n\n def __call__(self, state: Tuple[int, int], action: Action) -> float:\n \"\"\"\n Returns the action probability\n \"\"\"\n assert action in self._possible_actions\n # state is unused for this policy\n return self._action_probs[action]", "_____no_output_____" ], [ "class Environment:\n def __init__(self):\n self.grid_width = 5\n self.grid_height = 5\n self._good_state1 = (0, 1)\n self._good_state2 = (0, 3)\n self._to_state1 = (4, 2)\n self._to_state2 = (2, 3)\n self._bad_state1 = (1, 1)\n self._bad_state2 = (4, 4)\n self._bad_states = [self._bad_state1, self._bad_state2]\n self._good_states = [self._good_state1, self._good_state2]\n self._to_states = [self._to_state1, self._to_state2]\n self._good_rewards = [10, 5]\n\n def step(self, state, action):\n i, j = state\n for good_state, reward, to_state in zip(\n self._good_states, self._good_rewards, self._to_states\n ):\n if (i, j) == good_state:\n return (to_state, reward)\n reward = 0\n if state in self._bad_states:\n reward = -1\n if action == Action.LEFT:\n j_next = max(j - 1, 0)\n i_next = i\n if j - 1 < 0:\n reward = -1\n elif action == Action.RIGHT:\n j_next = min(j + 1, self.grid_width - 1)\n i_next = i\n if j + 1 > self.grid_width - 1:\n reward = -1\n elif action == Action.UP:\n j_next = j\n i_next = max(i - 1, 0)\n if i - 1 < 0:\n reward = -1\n elif action == Action.DOWN:\n j_next = j\n i_next = min(i + 1, self.grid_height - 1)\n if i + 1 > self.grid_height - 1:\n reward = -1\n else:\n raise ValueError(\"Invalid action\")\n return ((i_next, j_next), reward)", "_____no_output_____" ] ], [ [ "Probability and reward matrix:", "_____no_output_____" ] ], [ [ "pi = Policy()\nenv = Environment()\n\n# setup probability matrix and reward matrix\nP = np.zeros((env.grid_width * env.grid_height, env.grid_width * env.grid_height))\nR = np.zeros_like(P)\npossible_actions = [action for action in Action]\n\n# Loop for all states and fill up P and R\nfor i in range(env.grid_height):\n for j in range(env.grid_width):\n state = (i, j)\n # loop for all action and setup P and R\n for action in possible_actions:\n next_state, reward = env.step(state, action)\n (i_next, j_next) = next_state\n P[i * env.grid_width + j, i_next * env.grid_width + j_next] += pi(\n state, action\n )\n # the reward depends only on the starting state and the final state\n R[i * env.grid_width + j, i_next * env.grid_width + j_next] = reward", "_____no_output_____" ], [ "# check the correctness\nassert((np.sum(P, axis=1) == 1).all())", "_____no_output_____" ], [ "# expected reward for each state\nR_expected = np.sum(P * R, axis=1, keepdims=True)", "_____no_output_____" ], [ "# reshape the state values in a matrix\nR_square = R_expected.reshape((env.grid_height,env.grid_width))\n# Visualize\nvis_matrix(R_square, cmap=plt.cm.Reds)", "_____no_output_____" ] ], [ [ "The previous figure is a color representation of the expected reward associated to each state considering the current policy. Notice the expected reward of bad states is exactly equal to -1. The expected reward of good states is exactly equal to 10 and 5 respectively.", "_____no_output_____" ] ], [ [ "# define the discount factor\ngamma = 0.9", "_____no_output_____" ], [ "# Now it is possible to solve the Bellman Equation\nA = np.eye(env.grid_width*env.grid_height) - gamma * P\nB = R_expected", "_____no_output_____" ], [ "# solve using scipy linalg\nV = linalg.solve(A, B)", "_____no_output_____" ], [ "# reshape the state values in a matrix\nV_square = V.reshape((env.grid_height,env.grid_width))\n# visualize results\nvis_matrix(V_square, cmap=plt.cm.Reds)", "_____no_output_____" ] ], [ [ "Notice that the value of good states is less than the expected reward from those states. This is because landing states have an expected reward that is negative or because landing states are close to states for which the reward is negative. You can notice that the state with higher value is state $G_1$, followed by state $G_2$. It is also interesting to notice the high value of state in position (1, 2), being close to good states.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## evaluation episodeのrewardsの結果を見るnotebook", "_____no_output_____" ] ], [ [ "# import csv\n# def text_csv_converter(datas):\n# file_csv = datas.replace(\"txt\", \"csv\")\n# with open(datas) as rf:\n# with open(file_csv, \"w\") as wf:\n# readfile = rf.readlines()\n# for read_text in readfile:\n# read_text = read_text.split()\n# writer = csv.writer(wf, delimiter=',')\n# writer.writerow(read_text)\n\n# filename = \"/content/drive/My Drive/Gfootball/kaggle_simulations/rainbow/agent2/1-7results/scores.txt\"\n# text_csv_converter(filename)", "_____no_output_____" ], [ "import pandas as pd\ndf1 = pd.read_csv(\"/content/drive/My Drive/Gfootball/kaggle_simulations/rainbow/agent/1-6results/scores.csv\")\ndf2 = pd.read_csv(\"/content/drive/My Drive/Gfootball/kaggle_simulations/rainbow/agent/7-10results/scores.csv\")", "_____no_output_____" ], [ "df1 = df1[df1[\"episodes\"] < 207]\ndf1", "_____no_output_____" ], [ "import numpy as np\ndf2 = df2.iloc[6:]\nepi = np.arange(207, len(df2)+207)\ndf2[\"episodes\"] = epi\ndf2", "_____no_output_____" ], [ "df = pd.concat([df1, df2])\ndf.to_csv('/content/drive/My Drive/Gfootball/kaggle_simulations/rainbow/agent/scores.csv')", "_____no_output_____" ], [ "df = pd.read_csv(\"/content/drive/My Drive/Gfootball/kaggle_simulations/rainbow/agent2/scores.csv\")\ndf", "_____no_output_____" ], [ "df['roll_median'] = df['median'].rolling(5).mean()\ndf", "_____no_output_____" ], [ "import matplotlib.pyplot as plt\nimport seaborn as sns\nsns.set()\n\n# visualize reward each episodes\nfig = plt.figure(figsize=(7, 5))\nax1 = fig.add_subplot(111)\nax1.set_title(\"median reward\")\nsns.lineplot(x=\"episodes\", y=\"roll_median\", data=df, ax=ax1, color=(0, 0, 1, 1))\nsns.lineplot(x=\"episodes\", y=\"median\", data=df, ax=ax1, color=(0, 0, 1, 0.2))\nplt.tight_layout()\nplt.savefig(\"/content/drive/My Drive/Gfootball/kaggle_simulations/rainbow/agent2/result.png\")\nplt.show()", "_____no_output_____" ], [ "", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]