Upload image_dataset.ipynb
Browse files- image_dataset.ipynb +56 -81
image_dataset.ipynb
CHANGED
@@ -581,7 +581,7 @@
|
|
581 |
},
|
582 |
{
|
583 |
"cell_type": "code",
|
584 |
-
"execution_count":
|
585 |
"metadata": {},
|
586 |
"outputs": [
|
587 |
{
|
@@ -589,14 +589,8 @@
|
|
589 |
"output_type": "stream",
|
590 |
"text": [
|
591 |
"Reading metadata from /fsx/avijit/projects/datacommonsMA/massgis_2023_aerial_imagery/imagefolder-dataset/train/metadata.csv\n",
|
592 |
-
"Selected 50 samples\n"
|
593 |
-
|
594 |
-
},
|
595 |
-
{
|
596 |
-
"name": "stderr",
|
597 |
-
"output_type": "stream",
|
598 |
-
"text": [
|
599 |
-
"Processing images: 0%| | 0/50 [00:00<?, ?it/s]"
|
600 |
]
|
601 |
},
|
602 |
{
|
@@ -604,28 +598,20 @@
|
|
604 |
"output_type": "stream",
|
605 |
"text": [
|
606 |
"/fsx/avijit/anaconda3/envs/py312/lib/python3.12/site-packages/PIL/Image.py:3402: DecompressionBombWarning: Image size (100000000 pixels) exceeds limit of 89478485 pixels, could be decompression bomb DOS attack.\n",
|
607 |
-
" warnings.warn(\n"
|
608 |
-
"Processing images: 100%|βββββββββββββββββββββββββββββββββββββββββ| 50/50 [00:48<00:00, 1.04it/s]"
|
609 |
]
|
610 |
},
|
611 |
{
|
612 |
"name": "stdout",
|
613 |
"output_type": "stream",
|
614 |
"text": [
|
|
|
|
|
615 |
"Saved sample dataset to /fsx/avijit/projects/datacommonsMA/massgis_2023_aerial_imagery/imagefolder-dataset/data/sample_dataset_256x256.parquet\n",
|
616 |
-
"
|
617 |
-
"
|
618 |
-
"
|
619 |
-
"
|
620 |
-
"tilename: string\n",
|
621 |
-
"zone: int64\n"
|
622 |
-
]
|
623 |
-
},
|
624 |
-
{
|
625 |
-
"name": "stderr",
|
626 |
-
"output_type": "stream",
|
627 |
-
"text": [
|
628 |
-
"\n"
|
629 |
]
|
630 |
}
|
631 |
],
|
@@ -637,10 +623,17 @@
|
|
637 |
"from PIL import Image as PILImage\n",
|
638 |
"import numpy as np\n",
|
639 |
"from tqdm import tqdm\n",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
640 |
"\n",
|
641 |
"base_dir = \"/fsx/avijit/projects/datacommonsMA/massgis_2023_aerial_imagery/imagefolder-dataset\"\n",
|
642 |
"train_dir = os.path.join(base_dir, \"train\")\n",
|
643 |
-
"output_dir = os.path.join(base_dir, \"data\")
|
644 |
"output_path = os.path.join(output_dir, \"sample_dataset_256x256.parquet\")\n",
|
645 |
"\n",
|
646 |
"# Create the output directory if it doesn't exist\n",
|
@@ -659,84 +652,54 @@
|
|
659 |
"\n",
|
660 |
"print(f\"Selected {len(metadata_df)} samples\")\n",
|
661 |
"\n",
|
662 |
-
"#
|
663 |
-
"
|
664 |
-
"
|
665 |
-
"
|
666 |
-
"\n",
|
667 |
-
"
|
668 |
-
"os.makedirs(processed_images_dir, exist_ok=True)\n",
|
669 |
-
"\n",
|
670 |
-
"# Process each image\n",
|
671 |
-
"for _, row in tqdm(metadata_df.iterrows(), total=len(metadata_df), desc=\"Processing images\"):\n",
|
672 |
-
" try:\n",
|
673 |
-
" # Get image path\n",
|
674 |
-
" img_path = os.path.join(train_dir, row['file_name'])\n",
|
675 |
-
" \n",
|
676 |
-
" # Open, resize, and save the processed image\n",
|
677 |
-
" with PILImage.open(img_path) as img:\n",
|
678 |
-
" # Convert to RGB if needed\n",
|
679 |
-
" if img.mode != 'RGB':\n",
|
680 |
-
" img = img.convert('RGB')\n",
|
681 |
-
" \n",
|
682 |
-
" # Resize to target size\n",
|
683 |
-
" img_resized = img.resize(target_size)\n",
|
684 |
-
" \n",
|
685 |
-
" # Save the resized image in a separate folder\n",
|
686 |
-
" output_image_path = os.path.join(processed_images_dir, row['file_name'])\n",
|
687 |
-
" img_resized.save(output_image_path, format='JPEG', quality=90)\n",
|
688 |
-
" \n",
|
689 |
-
" # Append data\n",
|
690 |
-
" image_paths.append(output_image_path) # Store file path instead of bytes\n",
|
691 |
-
" tilenames.append(row['tilename'])\n",
|
692 |
-
" zones.append(int(row['zone'])) # Ensure zone is an integer\n",
|
693 |
-
" except Exception as e:\n",
|
694 |
-
" print(f\"Error processing {row['file_name']}: {e}\")\n",
|
695 |
"\n",
|
696 |
-
"#
|
697 |
-
"
|
698 |
-
"
|
699 |
-
"zone_array = pa.array(zones, type=pa.int64())\n",
|
700 |
"\n",
|
701 |
-
"#
|
702 |
-
"
|
703 |
-
"
|
704 |
-
" ('tilename', pa.string()),\n",
|
705 |
-
" ('zone', pa.int64())\n",
|
706 |
-
"])\n",
|
707 |
"\n",
|
708 |
-
"#
|
709 |
-
"
|
710 |
"\n",
|
711 |
-
"#
|
712 |
-
"
|
|
|
713 |
"\n",
|
714 |
"print(f\"Saved sample dataset to {output_path}\")\n",
|
715 |
-
"print(f\"Processed images saved in {processed_images_dir}\")\n",
|
716 |
"print(f\"File size: {os.path.getsize(output_path) / (1024 * 1024):.2f} MB\")\n",
|
717 |
"\n",
|
718 |
-
"#
|
719 |
-
"print(\"
|
720 |
-
"
|
721 |
-
"print(
|
|
|
722 |
]
|
723 |
},
|
724 |
{
|
725 |
"cell_type": "code",
|
726 |
-
"execution_count":
|
727 |
"metadata": {},
|
728 |
"outputs": [
|
729 |
{
|
730 |
"name": "stdout",
|
731 |
"output_type": "stream",
|
732 |
"text": [
|
733 |
-
"<pyarrow._parquet.FileMetaData object at
|
734 |
" created_by: parquet-cpp-arrow version 19.0.0\n",
|
735 |
-
" num_columns:
|
736 |
" num_rows: 50\n",
|
737 |
" num_row_groups: 1\n",
|
738 |
" format_version: 2.6\n",
|
739 |
-
" serialized_size:
|
740 |
]
|
741 |
}
|
742 |
],
|
@@ -758,6 +721,18 @@
|
|
758 |
"display_name": "py312",
|
759 |
"language": "python",
|
760 |
"name": "py312"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
761 |
}
|
762 |
},
|
763 |
"nbformat": 4,
|
|
|
581 |
},
|
582 |
{
|
583 |
"cell_type": "code",
|
584 |
+
"execution_count": 22,
|
585 |
"metadata": {},
|
586 |
"outputs": [
|
587 |
{
|
|
|
589 |
"output_type": "stream",
|
590 |
"text": [
|
591 |
"Reading metadata from /fsx/avijit/projects/datacommonsMA/massgis_2023_aerial_imagery/imagefolder-dataset/train/metadata.csv\n",
|
592 |
+
"Selected 50 samples\n",
|
593 |
+
"Loading and resizing images...\n"
|
|
|
|
|
|
|
|
|
|
|
|
|
594 |
]
|
595 |
},
|
596 |
{
|
|
|
598 |
"output_type": "stream",
|
599 |
"text": [
|
600 |
"/fsx/avijit/anaconda3/envs/py312/lib/python3.12/site-packages/PIL/Image.py:3402: DecompressionBombWarning: Image size (100000000 pixels) exceeds limit of 89478485 pixels, could be decompression bomb DOS attack.\n",
|
601 |
+
" warnings.warn(\n"
|
|
|
602 |
]
|
603 |
},
|
604 |
{
|
605 |
"name": "stdout",
|
606 |
"output_type": "stream",
|
607 |
"text": [
|
608 |
+
"Resizing images...\n",
|
609 |
+
"Saving to /fsx/avijit/projects/datacommonsMA/massgis_2023_aerial_imagery/imagefolder-dataset/data/sample_dataset_256x256.parquet...\n",
|
610 |
"Saved sample dataset to /fsx/avijit/projects/datacommonsMA/massgis_2023_aerial_imagery/imagefolder-dataset/data/sample_dataset_256x256.parquet\n",
|
611 |
+
"File size: 6.28 MB\n",
|
612 |
+
"Verifying saved file...\n",
|
613 |
+
"Columns in saved file: ['image', 'tilename', 'zone']\n",
|
614 |
+
"Number of rows: 50\n"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
615 |
]
|
616 |
}
|
617 |
],
|
|
|
623 |
"from PIL import Image as PILImage\n",
|
624 |
"import numpy as np\n",
|
625 |
"from tqdm import tqdm\n",
|
626 |
+
"import io\n",
|
627 |
+
"\n",
|
628 |
+
"# Import the pandas_image_methods library\n",
|
629 |
+
"from pandas_image_methods import PILMethods\n",
|
630 |
+
"\n",
|
631 |
+
"# Register the PIL methods accessor\n",
|
632 |
+
"pd.api.extensions.register_series_accessor(\"pil\")(PILMethods)\n",
|
633 |
"\n",
|
634 |
"base_dir = \"/fsx/avijit/projects/datacommonsMA/massgis_2023_aerial_imagery/imagefolder-dataset\"\n",
|
635 |
"train_dir = os.path.join(base_dir, \"train\")\n",
|
636 |
+
"output_dir = os.path.join(base_dir, \"data\")\n",
|
637 |
"output_path = os.path.join(output_dir, \"sample_dataset_256x256.parquet\")\n",
|
638 |
"\n",
|
639 |
"# Create the output directory if it doesn't exist\n",
|
|
|
652 |
"\n",
|
653 |
"print(f\"Selected {len(metadata_df)} samples\")\n",
|
654 |
"\n",
|
655 |
+
"# Create DataFrame with just the paths first\n",
|
656 |
+
"df = pd.DataFrame({\n",
|
657 |
+
" 'file_path': [os.path.join(train_dir, row['file_name']) for _, row in metadata_df.iterrows()],\n",
|
658 |
+
" 'tilename': metadata_df['tilename'].tolist(),\n",
|
659 |
+
" 'zone': metadata_df['zone'].astype('int64').tolist()\n",
|
660 |
+
"})\n",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
661 |
"\n",
|
662 |
+
"# Load images using the pil accessor\n",
|
663 |
+
"print(\"Loading and resizing images...\")\n",
|
664 |
+
"df['image'] = df['file_path'].pil.open()\n",
|
|
|
665 |
"\n",
|
666 |
+
"# Resize the images\n",
|
667 |
+
"print(\"Resizing images...\")\n",
|
668 |
+
"df['image'] = df['image'].pil.resize(target_size)\n",
|
|
|
|
|
|
|
669 |
"\n",
|
670 |
+
"# Keep only the required columns for the preview\n",
|
671 |
+
"df = df[['image', 'tilename', 'zone']]\n",
|
672 |
"\n",
|
673 |
+
"# Save to Parquet (the library will handle the PIL images correctly)\n",
|
674 |
+
"print(f\"Saving to {output_path}...\")\n",
|
675 |
+
"df.to_parquet(output_path)\n",
|
676 |
"\n",
|
677 |
"print(f\"Saved sample dataset to {output_path}\")\n",
|
|
|
678 |
"print(f\"File size: {os.path.getsize(output_path) / (1024 * 1024):.2f} MB\")\n",
|
679 |
"\n",
|
680 |
+
"# Verify the saved file\n",
|
681 |
+
"print(\"Verifying saved file...\")\n",
|
682 |
+
"df_check = pd.read_parquet(output_path)\n",
|
683 |
+
"print(\"Columns in saved file:\", df_check.columns.tolist())\n",
|
684 |
+
"print(\"Number of rows:\", len(df_check))"
|
685 |
]
|
686 |
},
|
687 |
{
|
688 |
"cell_type": "code",
|
689 |
+
"execution_count": 24,
|
690 |
"metadata": {},
|
691 |
"outputs": [
|
692 |
{
|
693 |
"name": "stdout",
|
694 |
"output_type": "stream",
|
695 |
"text": [
|
696 |
+
"<pyarrow._parquet.FileMetaData object at 0x7fa0fc694f90>\n",
|
697 |
" created_by: parquet-cpp-arrow version 19.0.0\n",
|
698 |
+
" num_columns: 4\n",
|
699 |
" num_rows: 50\n",
|
700 |
" num_row_groups: 1\n",
|
701 |
" format_version: 2.6\n",
|
702 |
+
" serialized_size: 2731\n"
|
703 |
]
|
704 |
}
|
705 |
],
|
|
|
721 |
"display_name": "py312",
|
722 |
"language": "python",
|
723 |
"name": "py312"
|
724 |
+
},
|
725 |
+
"language_info": {
|
726 |
+
"codemirror_mode": {
|
727 |
+
"name": "ipython",
|
728 |
+
"version": 3
|
729 |
+
},
|
730 |
+
"file_extension": ".py",
|
731 |
+
"mimetype": "text/x-python",
|
732 |
+
"name": "python",
|
733 |
+
"nbconvert_exporter": "python",
|
734 |
+
"pygments_lexer": "ipython3",
|
735 |
+
"version": "3.12.9"
|
736 |
}
|
737 |
},
|
738 |
"nbformat": 4,
|