File size: 6,545 Bytes
901dbf6 3c8c115 901dbf6 73bc32e 901dbf6 545abd8 901dbf6 ab1449b a7eb438 ab1449b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 |
import streamlit as st
from paddleocr import PaddleOCR
from PIL import ImageDraw, ImageFont
import torch
from transformers import AutoProcessor,LayoutLMv3ForTokenClassification
import numpy as np
model_Hugging_path = "Noureddinesa/Output_LayoutLMv3_v5"
#############################################################################
#############################################################################
def Labels():
labels = ['InvNum', 'InvDate', 'Fourni', 'TTC', 'TVA', 'TT', 'Autre']
id2label = {v: k for v, k in enumerate(labels)}
label2id = {k: v for v, k in enumerate(labels)}
return id2label, label2id
#############################################################################
#############################################################################
def Paddle():
ocr = PaddleOCR(use_angle_cls=False,lang='fr',rec=False)
return ocr
def processbbox(BBOX, width, height):
bbox = []
bbox.append(BBOX[0][0])
bbox.append(BBOX[0][1])
bbox.append(BBOX[2][0])
bbox.append(BBOX[2][1])
#Scaling
bbox[0]= 1000*bbox[0]/width # X1
bbox[1]= 1000*bbox[1]/height # Y1
bbox[2]= 1000*bbox[2]/width # X2
bbox[3]= 1000*bbox[3]/height # Y2
for i in range(4):
bbox[i] = int(bbox[i])
return bbox
def Preprocess(image):
image_array = np.array(image)
ocr = Paddle()
width, height = image.size
results = ocr.ocr(image_array, cls=True)
results = results[0]
test_dict = {'image': image ,'tokens':[], "bboxes":[]}
for item in results :
bbox = processbbox(item[0], width, height)
test_dict['tokens'].append(item[1][0])
test_dict['bboxes'].append(bbox)
print(test_dict['bboxes'])
print(test_dict['tokens'])
return test_dict
#############################################################################
#############################################################################
def Encode(image):
example = Preprocess(image)
image = example["image"]
words = example["tokens"]
boxes = example["bboxes"]
processor = AutoProcessor.from_pretrained(model_Hugging_path, apply_ocr=False)
encoding = processor(image, words, boxes=boxes,return_offsets_mapping=True,truncation=True, max_length=512, padding="max_length", return_tensors="pt")
offset_mapping = encoding.pop('offset_mapping')
return encoding, offset_mapping,words
def unnormalize_box(bbox, width, height):
return [
width * (bbox[0] / 1000),
height * (bbox[1] / 1000),
width * (bbox[2] / 1000),
height * (bbox[3] / 1000),
]
def Run_model(image):
encoding,offset_mapping,words = Encode(image)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# load the fine-tuned model from the hub
model = LayoutLMv3ForTokenClassification.from_pretrained(model_Hugging_path)
model.to(device)
# forward pass
outputs = model(**encoding)
predictions = outputs.logits.argmax(-1).squeeze().tolist()
token_boxes = encoding.bbox.squeeze().tolist()
width, height = image.size
id2label, _ = Labels()
is_subword = np.array(offset_mapping.squeeze().tolist())[:,0] != 0
true_predictions = [id2label[pred] for idx, pred in enumerate(predictions) if not is_subword[idx]]
true_boxes = [unnormalize_box(box, width, height) for idx, box in enumerate(token_boxes) if not is_subword[idx]]
return true_predictions,true_boxes,words
def Get_Json(true_predictions,words):
Results = {}
i = 0
for prd in true_predictions:
if prd in ['InvNum','Fourni', 'InvDate','TT','TTC','TVA']:
#print(i,prd,words[i-1])
Results[prd] = words[i-1]
i+=1
key_mapping = {'InvNum':'Numéro de facture','Fourni':'Fournisseur', 'InvDate':'Date Facture','TT':'Total HT','TTC':'Total TTC','TVA':'TVA'}
Results = {key_mapping.get(key, key): value for key, value in Results.items()}
return Results
def Draw(image):
true_predictions, true_boxes,words = Run_model(image)
draw = ImageDraw.Draw(image)
label2color = {
'InvNum': 'blue',
'InvDate': 'green',
'Fourni': 'orange',
'TTC':'purple',
'TVA': 'magenta',
'TT': 'red',
'Autre': 'black'
}
# Adjust the thickness of the rectangle outline and label text position
rectangle_thickness = 4
label_x_offset = 20
label_y_offset = -30
# Custom font size
custom_font_size = 25
# Load a font with the custom size
font_path = "arial.ttf" # Specify the path to your font file
custom_font = ImageFont.truetype(font_path, custom_font_size)
for prediction, box in zip(true_predictions, true_boxes):
predicted_label = prediction
# Check if the predicted label exists in the label2color dictionary
if predicted_label in label2color:
color = label2color[predicted_label]
else:
color = 'black' # Default color if label is not found
if predicted_label != "Autre":
draw.rectangle(box, outline=color, width=rectangle_thickness)
# Draw text using the custom font and size
draw.rectangle((box[0], box[1]+ label_y_offset,box[2],box[3]+ label_y_offset), fill=color)
draw.text((box[0] + label_x_offset, box[1] + label_y_offset), text=predicted_label, fill='white', font=custom_font)
# Get the Results Json File
Results = Get_Json(true_predictions,words)
return image,Results
def Add_Results(data):
# Render the table
for key, value in data.items():
data[key] = st.sidebar.text_input(key, value)
#############################################################################
#############################################################################
def Change_Image(image1,image2):
# Initialize session state
if 'current_image' not in st.session_state:
st.session_state.current_image = 'image1'
# Button to switch between images
if st.sidebar.button('Remove'):
if st.session_state.current_image == 'image1':
st.session_state.current_image = 'image2'
else:
st.session_state.current_image = 'image1'
# Display the selected image
if st.session_state.current_image == 'image1':
st.image(image1, caption='Output', use_column_width=True)
else:
st.image(image2, caption='Image initiale', use_column_width=True) |