Upload 9 files

v1/README.md

@@ -0,0 +1,169 @@
+# SigLIP2 Gardner Grading Project
+
+This project implements an AI-based embryo grading system using the SigLIP2 vision-language model for automated Gardner grading in IVF procedures. The system fine-tunes a pre-trained SigLIP2 model on embryo images to classify them into different grade categories.
+
+## Project Structure
+
+```
+siglip2/
+├── filtered_data.csv          # Dataset with image filenames and grades
+├── finetune.py               # Script for fine-tuning the SigLIP2 model
+├── infer.py                  # GUI application for single image classification
+├── inference.py              # Alternative GUI application for image classification
+├── inspect_modules.py        # Script to inspect model architecture modules
+├── Images/                   # Directory containing embryo images
+├── model/                    # Pre-trained SigLIP2 model files
+├── Organized_Images[raw]/    # Raw organized embryo images by experiment
+└── siglip2_finetuned/        # Directory where fine-tuned model is saved
+```
+## the main files of model are 
+- config.json - Model configuration
+- model.safetensors - Model weights
+- preprocessor_config.json - Image preprocessing settings
+- training_args.bin - Training arguments
+
+## Prerequisites
+
+- Python 3.8+
+- PyTorch
+- Transformers library
+- Datasets library
+- PIL (Pillow)
+- scikit-learn
+- imbalanced-learn
+- evaluate
+- tkinter (usually included with Python)
+
+## Installation
+
+1. Clone or download this repository
+2. Install required packages:
+   ```bash
+   pip install torch torchvision transformers datasets pillow scikit-learn imbalanced-learn evaluate
+   ```
+
+## Data Description
+
+### filtered_data.csv
+- **Purpose**: Contains the training dataset with image filenames and corresponding grades
+- **Columns**:
+  - `Image`: Filename of the embryo image (e.g., "0001_03.png")
+  - `grade`: Gardner grade classification (e.g., "4BA", "3AB", "5AB")
+- **Size**: 1654 entries (after oversampling for class balance)
+
+### Images/
+- Contains PNG images of embryos used for training and inference
+- Images are named with format: `{experiment}_{image_number}.png`
+
+### Organized_Images[raw]/
+- Raw embryo images organized by experiment folders
+- Includes subfolders like EXP3/, EXP4/, EXP5/, EXP6/
+- Also contains web-sourced embryo pictures for reference
+
+## Scripts Description
+
+### finetune.py
+**Purpose**: Fine-tunes the pre-trained SigLIP2 model on the embryo grading dataset.
+
+**Key Features**:
+- Loads and preprocesses the dataset from `filtered_data.csv`
+- Applies oversampling to balance classes using RandomOverSampler
+- Defines data augmentations: rotation, sharpness adjustment, resizing, normalization
+- Uses Hugging Face Trainer for training with mixed precision (FP16)
+- Saves the fine-tuned model to `siglip2_finetuned/`
+
+**Usage**:
+```bash
+python finetune.py
+```
+
+**Configuration**:
+- Learning rate: 2e-5
+- Batch size: 32
+- Epochs: 3
+- Weight decay: 0.02
+- Warmup steps: 50
+
+### infer.py
+**Purpose**: GUI application for classifying individual embryo images using the fine-tuned model.
+
+**Features**:
+- Tkinter-based graphical interface
+- File dialog to select PNG/JPG images
+- Displays selected image and predicted grade
+- Explicit preprocessing with size 224x224
+- Error handling for image processing
+
+**Usage**:
+```bash
+python infer.py
+```
+
+**Requirements**: Fine-tuned model in `siglip2_finetuned/` directory
+
+### inference.py
+**Purpose**: Simplified GUI application for embryo image classification.
+
+**Features**:
+- Similar to `infer.py` but with minimal preprocessing
+- Tkinter interface for image selection and display
+- Shows predicted Gardner grade
+
+**Usage**:
+```bash
+python inference.py
+```
+
+**Note**: This is a lighter version of `infer.py` with less explicit preprocessing settings.
+
+### inspect_modules.py
+**Purpose**: Utility script to inspect the architecture of the SigLIP2 model.
+
+**Features**:
+- Loads the pre-trained SigLIP2 model from `model/` directory
+- Prints vision model modules containing 'proj' or 'attn' (projection/attention layers)
+- Prints text model modules containing 'proj' or 'attn'
+
+**Usage**:
+```bash
+python inspect_modules.py
+```
+
+**Output**: Lists of module names for vision and text components useful for understanding model architecture.
+
+## Model Information
+
+### Pre-trained Model (model/)
+- **Type**: SigLIP2 (SigLIP version 2)
+- **Source**: Hugging Face transformers
+- **Components**: Vision encoder, text encoder, and classification head
+- **Files**: config.json, model.safetensors, preprocessor_config.json, tokenizer files
+
+### Fine-tuned Model (siglip2_finetuned/)
+- **Base**: Pre-trained SigLIP2 adapted for image classification
+- **Classes**: Gardner grading categories (e.g., 3AA, 3AB, 3BA, 4AA, etc.)
+- **Training**: Fine-tuned on embryo images with data augmentation
+- **Output**: Classification logits for grade prediction
+
+## Usage Workflow
+
+1. **Prepare Data**: Ensure `filtered_data.csv` and `Images/` are in place
+2. **Inspect Model** (optional): Run `inspect_modules.py` to understand architecture
+3. **Fine-tune Model**: Execute `finetune.py` to train on the dataset
+4. **Inference**: Use `infer.py` or `inference.py` to classify new embryo images
+
+## Notes
+
+- The project uses oversampling to handle class imbalance in the dataset
+- Training includes data augmentations for better generalization
+- GUI applications require a display environment (not suitable for headless servers)
+- Model expects RGB images resized to 224x224 pixels
+- Fine-tuning requires GPU for reasonable training time (FP16 enabled)
+
+## License
+
+[Add license information if applicable]
+
+## Contributing
+
+[Add contribution guidelines if applicable]

v1/config.json

@@ -0,0 +1,86 @@
+{
+  "architectures": [
+    "SiglipForImageClassification"
+  ],
+  "dtype": "float32",
+  "id2label": {
+    "0": "3AA",
+    "1": "3AB",
+    "2": "3AC",
+    "3": "3BA",
+    "4": "3BB",
+    "5": "3BC",
+    "6": "3CA",
+    "7": "3CB",
+    "8": "4AA",
+    "9": "4AB",
+    "10": "4AC",
+    "11": "4BA",
+    "12": "4BB",
+    "13": "4BC",
+    "14": "4CA",
+    "15": "4CB",
+    "16": "5AA",
+    "17": "5AB",
+    "18": "5AC",
+    "19": "5BA",
+    "20": "5BB",
+    "21": "5BC"
+  },
+  "initializer_factor": 1.0,
+  "label2id": {
+    "3AA": 0,
+    "3AB": 1,
+    "3AC": 2,
+    "3BA": 3,
+    "3BB": 4,
+    "3BC": 5,
+    "3CA": 6,
+    "3CB": 7,
+    "4AA": 8,
+    "4AB": 9,
+    "4AC": 10,
+    "4BA": 11,
+    "4BB": 12,
+    "4BC": 13,
+    "4CA": 14,
+    "4CB": 15,
+    "5AA": 16,
+    "5AB": 17,
+    "5AC": 18,
+    "5BA": 19,
+    "5BB": 20,
+    "5BC": 21
+  },
+  "model_type": "siglip",
+  "problem_type": "single_label_classification",
+  "text_config": {
+    "attention_dropout": 0.0,
+    "dtype": "float32",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 768,
+    "intermediate_size": 3072,
+    "layer_norm_eps": 1e-06,
+    "max_position_embeddings": 64,
+    "model_type": "siglip_text_model",
+    "num_attention_heads": 12,
+    "num_hidden_layers": 12,
+    "projection_size": 768,
+    "vocab_size": 256000
+  },
+  "transformers_version": "4.56.1",
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "float32",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 768,
+    "image_size": 224,
+    "intermediate_size": 3072,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 12,
+    "num_channels": 3,
+    "num_hidden_layers": 12,
+    "patch_size": 16
+  }
+}

v1/filtered_data.csv

@@ -0,0 +1,1653 @@
+Image,grade
+0001_03.png,4BA
+0003_01.png,3AB
+0003_02.png,4AA
+0003_03.png,4AB
+0004_01.png,3AA
+0004_02.png,5AB
+0005_01.png,4AA
+0005_02.png,3BA
+0005_04.png,4AA
+0005_05.png,4AA
+0005_06.png,4AA
+0005_07.png,4BA
+0007_01.png,4AA
+0007_02.png,3BA
+0009_01.png,3BA
+0009_02.png,3AB
+0009_03.png,4BB
+0011_01.png,3BA
+0012_03.png,4BA
+0014_01.png,4AB
+0015_01.png,3BA
+0016_01.png,3BA
+0018_02.png,3AB
+0018_03.png,4AB
+0019_02.png,3AA
+0019_03.png,4BB
+0019_04.png,3BC
+0023_01.png,4AB
+0023_02.png,3BB
+0023_03.png,3BB
+0023_04.png,3BA
+0024_01.png,3BB
+0025_01.png,3BB
+0025_05.png,4AA
+0027_01.png,3BA
+0028_02.png,4BA
+0031_01.png,3AA
+0031_03.png,3CB
+0033_02.png,3BB
+0034_01.png,4AB
+0034_02.png,3AA
+0035_01.png,3AA
+0036_01.png,3AA
+0036_04.png,3BA
+0039_01.png,3AB
+0039_02.png,4BC
+0040_01.png,3BA
+0041_01.png,3BB
+0042_01.png,4AB
+0042_05.png,4AA
+0042_06.png,4AB
+0045_01.png,3BB
+0046_02.png,3AA
+0046_04.png,4AA
+0046_05.png,4BB
+0046_06.png,4AB
+0047_03.png,4AB
+0049_01.png,3BA
+0049_03.png,4AA
+0049_04.png,4AB
+0049_05.png,3AB
+0050_01.png,4AB
+0051_01.png,3AC
+0051_04.png,3AB
+0051_05.png,4AB
+0052_01.png,3AA
+0052_02.png,3BA
+0052_04.png,4BC
+0054_02.png,3AB
+0054_03.png,4AB
+0054_04.png,5AA
+0054_05.png,4AB
+0055_01.png,3AB
+0056_01.png,4AA
+0057_03.png,3AA
+0057_05.png,4AB
+0058_01.png,4AA
+0058_02.png,3AB
+0058_03.png,4AB
+0058_04.png,5AB
+0058_05.png,4AB
+0059_01.png,4AA
+0059_02.png,3BA
+0059_03.png,4BB
+0060_01.png,4AA
+0063_01.png,3AB
+0063_02.png,4AA
+0063_04.png,3BB
+0063_05.png,3AB
+0063_07.png,4AA
+0064_01.png,4AB
+0065_01.png,4AB
+0065_02.png,3AB
+0065_04.png,4AA
+0066_01.png,4AA
+0066_02.png,4AB
+0066_04.png,3AA
+0067_02.png,4AB
+0068_01.png,4AB
+0068_02.png,4BA
+0068_03.png,4AB
+0068_04.png,4AA
+0068_05.png,4BB
+0069_01.png,3AB
+0069_02.png,4AA
+0069_03.png,4AB
+0071_01.png,4AA
+0071_02.png,3AA
+0072_01.png,4BA
+0072_04.png,3AB
+0074_01.png,4AA
+0074_02.png,3AA
+0074_03.png,4AA
+0074_04.png,4AB
+0075_01.png,4AA
+0075_03.png,4BA
+0076_01.png,4AA
+0076_02.png,3BA
+0077_01.png,4AA
+0077_02.png,3AB
+0078_01.png,3AB
+0079_01.png,4AA
+0079_02.png,3BB
+0079_04.png,3BA
+0080_01.png,3BB
+0080_02.png,4BA
+0080_03.png,4AB
+0081_01.png,3BB
+0081_02.png,4AA
+0082_01.png,5AA
+0083_01.png,4BA
+0083_02.png,4BB
+0083_03.png,3BA
+0083_04.png,3CB
+0084_01.png,4AA
+0085_01.png,4BA
+0085_02.png,4AA
+0085_03.png,4AA
+0085_04.png,4AA
+0086_01.png,3BB
+0086_02.png,4AA
+0087_01.png,4AA
+0087_02.png,4AC
+0088_01.png,4AC
+0088_02.png,3BB
+0088_03.png,4BA
+0089_01.png,4AB
+0090_01.png,4AB
+0090_02.png,4AA
+0090_03.png,4AB
+0091_01.png,5AA
+0091_03.png,4BC
+0092_02.png,4AB
+0092_03.png,5AA
+0092_04.png,4AA
+0092_05.png,3AB
+0092_07.png,4AB
+0092_08.png,4BA
+0092_09.png,3BA
+0092_10.png,4BA
+0093_01.png,4AA
+0093_02.png,4AA
+0094_01.png,4AA
+0094_02.png,4AA
+0094_03.png,4BA
+0095_01.png,5BA
+0095_02.png,4BA
+0097_01.png,3AB
+0098_01.png,4AB
+0099_01.png,4AA
+0100_01.png,4AA
+0100_02.png,4AB
+0100_03.png,4BB
+0101_01.png,4AA
+0101_03.png,4AA
+0101_04.png,4AA
+0102_01.png,5BB
+0102_02.png,3BA
+0103_01.png,5AB
+0104_01.png,4AB
+0104_02.png,4AA
+0104_03.png,3BA
+0105_02.png,4AA
+0105_04.png,5AB
+0105_05.png,3BB
+0105_07.png,4AA
+0106_01.png,4BA
+0106_02.png,3AA
+0107_01.png,3AA
+0108_01.png,5BB
+0108_02.png,3BB
+0108_03.png,3AB
+0108_05.png,4AA
+0109_01.png,4AB
+0109_02.png,4AA
+0110_01.png,4AA
+0110_02.png,4AB
+0110_03.png,4AA
+0110_04.png,4AA
+0111_01.png,4AA
+0112_01.png,5AA
+0112_02.png,3BB
+0112_03.png,4BB
+0112_04.png,4BA
+0114_01.png,4AB
+0114_02.png,3AA
+0114_03.png,3BA
+0114_04.png,4AA
+0115_01.png,4AC
+0116_01.png,4AB
+0116_02.png,4BB
+0116_04.png,4AA
+0118_01.png,5AA
+0119_01.png,4AB
+0120_01.png,3BB
+0121_01.png,4AA
+0122_01.png,5AA
+0123_01.png,4AB
+0123_02.png,4AB
+0124_01.png,4AA
+0125_01.png,3AB
+0126_01.png,4AA
+0126_03.png,4AB
+0126_04.png,4BB
+0126_07.png,4AA
+0126_08.png,3AA
+0126_09.png,3AB
+0127_01.png,4AA
+0127_02.png,4BA
+0128_01.png,4AB
+0130_02.png,4AB
+0130_05.png,4AB
+0130_06.png,4AC
+0130_07.png,3BA
+0131_01.png,3BB
+0131_02.png,4AA
+0131_03.png,4BC
+0132_01.png,4AA
+0132_03.png,4BB
+0133_01.png,4AB
+0134_01.png,4AA
+0135_01.png,4AA
+0135_03.png,4AB
+0135_04.png,5AA
+0136_01.png,4AA
+0136_02.png,4BB
+0136_03.png,5AA
+0137_01.png,4AA
+0137_02.png,5BA
+0138_01.png,4AA
+0139_01.png,4AB
+0139_02.png,3BB
+0139_03.png,4AC
+0140_01.png,4AB
+0141_01.png,3BB
+0142_01.png,3BB
+0142_02.png,3BB
+0142_03.png,4AB
+0142_04.png,4AA
+0142_05.png,4BB
+0143_01.png,3AA
+0143_02.png,4AA
+0144_01.png,4AA
+0145_01.png,4BA
+0145_02.png,5AA
+0146_01.png,4BB
+0146_02.png,4AB
+0146_03.png,4AB
+0146_04.png,4AA
+0147_01.png,4AA
+0147_02.png,4AB
+0147_03.png,3BB
+0147_04.png,4AB
+0148_01.png,4AA
+0150_01.png,4AB
+0150_02.png,4AB
+0150_04.png,4AB
+0151_01.png,4AA
+0151_02.png,3BB
+0151_04.png,4BA
+0152_01.png,4AA
+0152_02.png,4BA
+0153_01.png,4AA
+0153_02.png,4AA
+0154_01.png,4AA
+0155_01.png,4AA
+0156_01.png,4AA
+0156_02.png,4AA
+0156_04.png,4BA
+0157_01.png,5BA
+0157_02.png,4AA
+0158_01.png,4AA
+0161_01.png,4BB
+0162_01.png,4AA
+0163_03.png,3BB
+0164_01.png,4AB
+0164_02.png,4BB
+0164_03.png,3BB
+0166_01.png,4AA
+0168_01.png,4AA
+0168_02.png,3AB
+0168_03.png,3BB
+0168_04.png,3BB
+0168_05.png,4AA
+0168_06.png,4AA
+0169_01.png,5BB
+0169_02.png,3CB
+0169_03.png,4AA
+0169_04.png,4AB
+0169_06.png,3BB
+0170_01.png,4AA
+0171_01.png,4AB
+0172_01.png,5AA
+0172_02.png,5BB
+0172_03.png,4AB
+0172_04.png,4AB
+0172_05.png,5AB
+0172_06.png,4AA
+0173_03.png,3AB
+0173_04.png,4AB
+0173_05.png,4AB
+0173_06.png,4AB
+0174_02.png,3BA
+0174_03.png,4BA
+0175_01.png,4BA
+0175_03.png,3BB
+0175_05.png,4BB
+0176_01.png,4AA
+0176_02.png,5BA
+0176_04.png,4AB
+0177_01.png,4BB
+0177_02.png,4BC
+0177_03.png,4AA
+0178_01.png,4BB
+0179_01.png,5AA
+0179_02.png,5AB
+0180_01.png,4AB
+0180_02.png,4AA
+0181_01.png,4AA
+0181_02.png,4AB
+0181_03.png,4AB
+0182_01.png,5AA
+0182_02.png,4BB
+0182_04.png,4AA
+0183_01.png,3AA
+0185_01.png,4AB
+0186_01.png,4AB
+0186_02.png,4AA
+0187_02.png,4AB
+0189_02.png,3BB
+0190_01.png,3AB
+0190_03.png,4AA
+0190_04.png,3AA
+0190_05.png,4AB
+0190_06.png,4AB
+0190_07.png,4AA
+0190_08.png,4AB
+0192_01.png,4BA
+0193_01.png,4AA
+0194_01.png,4AA
+0194_02.png,4AA
+0194_03.png,4AA
+0195_01.png,4AA
+0195_02.png,4BB
+0195_04.png,3AA
+0195_05.png,3BB
+0196_01.png,5AA
+0197_01.png,4AA
+0197_02.png,4BA
+198_01.png,5AA
+200_04.png,4AA
+200_05.png,3AA
+201_01.png,4AB
+201_02.png,3AB
+201_04.png,3AB
+201_05.png,4AA
+201_06.png,4AB
+202_01.png,3AA
+202_07.png,3AA
+202_08.png,4AA
+203_02.png,3AA
+203_04.png,3AB
+204_01.png,4AA
+204_02.png,4AB
+204_03.png,4AA
+204_06.png,4AA
+206_01.png,4BA
+206_02.png,4BB
+208_01.png,3AA
+208_03.png,4AA
+208_04.png,5AA
+208_05.png,3AA
+208_06.png,4AA
+209_02.png,3CB
+210_01.png,3AB
+210_02.png,4AA
+210_03.png,4AA
+210_04.png,4BB
+210_05.png,4AB
+210_06.png,4AB
+211_02.png,4BA
+211_03.png,4BA
+211_04.png,4AA
+211_05.png,4AB
+211_06.png,3AA
+212_04.png,3AB
+212_05.png,4AA
+212_06.png,4CA
+212_07.png,4AB
+213_01.png,4BA
+214_03.png,4AA
+214_04.png,3AA
+215_02.png,5AA
+215_04.png,4AB
+215_05.png,3AB
+216_01.png,4AA
+216_02.png,4AB
+216_04.png,4AA
+217_01.png,5AA
+217_02.png,4AA
+218_01.png,4AB
+218_02.png,4AB
+218_03.png,4AB
+218_04.png,4AA
+219_01.png,4AB
+220_01.png,4AA
+220_02.png,4AB
+221_08.png,4AA
+221_09.png,4AB
+232_03.png,4AB
+234_02.png,3AB
+235_01.png,3AA
+235_02.png,3AA
+236_01.png,3AA
+237_01.png,3AA
+237_02.png,5AB
+237_03.png,4AB
+237_05.png,4AB
+237_06.png,4AA
+240_01.png,4AB
+240_02.png,4AB
+240_03.png,4AB
+241_01.png,3AA
+241_03.png,4AA
+241_05.png,5AB
+242_02.png,3AB
+244_01.png,4AA
+244_02.png,5BB
+246_01.png,3AA
+247_01.png,3AA
+247_04.png,3AB
+248_01.png,3BB
+248_02.png,5AA
+248_03.png,5AB
+250_02.png,4BA
+252_02.png,3AA
+256_01.png,3AA
+256_02.png,4AB
+256_03.png,5AA
+256_04.png,4AA
+257_01.png,4BA
+257_05.png,4AB
+258_01.png,4AB
+258_02.png,4AB
+258_04.png,3AA
+260_01.png,3AB
+260_02.png,3BB
+260_03.png,4BB
+261_03.png,5BA
+261_04.png,4BA
+262_01.png,4AA
+265_01.png,4AA
+266_01.png,3AA
+266_03.png,4AC
+268_02.png,4BB
+268_04.png,4AA
+269_01.png,5BC
+269_02.png,3BA
+271_01.png,3AA
+271_02.png,5AA
+273_02.png,4AA
+273_03.png,4AA
+274_01.png,4AB
+275_01.png,5AA
+275_02.png,3BA
+275_03.png,3AB
+275_04.png,4AB
+275_05.png,3AB
+275_06.png,3BB
+276_01.png,5AB
+276_02.png,4AB
+276_03.png,3AA
+279_02.png,3BA
+280_01.png,4AB
+283_02.png,4AA
+283_04.png,3AB
+283_05.png,4AB
+284_01.png,4AB
+284_02.png,4AB
+285_01.png,5AB
+285_02.png,4AA
+286_01.png,4AB
+286_02.png,3AB
+286_03.png,4BA
+286_04.png,3BA
+287_01.png,5AA
+287_03.png,4BB
+288_02.png,4BB
+289_01.png,4AA
+290_01.png,4AA
+291_01.png,4AA
+291_02.png,3BA
+292_01.png,4AC
+292_04.png,4AB
+293_01.png,4AA
+293_02.png,4AA
+293_03.png,4BA
+293_04.png,4AB
+293_06.png,4BA
+294_01.png,4AA
+294_02.png,4AA
+294_04.png,4AA
+294_05.png,5AA
+294_06.png,3AA
+294_07.png,4BA
+294_08.png,4AA
+295_01.png,3AB
+296_01.png,4AA
+296_02.png,4AB
+297_01.png,4AA
+297_02.png,4AB
+297_05.png,3AB
+298_01.png,4AA
+298_02.png,4AA
+299_01.png,3AB
+300_01.png,5AA
+300_02.png,4AB
+300_03.png,4AA
+300_04.png,3AB
+300_05.png,4AB
+300_06.png,3BA
+301_01.png,4AA
+301_02.png,4AA
+301_03.png,5AB
+301_04.png,4AA
+301_06.png,4AA
+302_01.png,4AA
+302_03.png,3BA
+302_04.png,4AA
+303_01.png,4AA
+303_02.png,4AA
+303_03.png,4BB
+303_04.png,5AB
+304_02.png,3AB
+305_01.png,4AA
+305_02.png,3AB
+306_01.png,4AB
+306_03.png,4AA
+308_01.png,4AA
+309_01.png,5AA
+309_02.png,4AA
+309_03.png,4AB
+310_01.png,4BB
+310_04.png,3BA
+311_01.png,4AB
+311_03.png,3BA
+313_01.png,4AA
+313_03.png,3AB
+313_04.png,4AA
+314_01.png,4AA
+315_01.png,5AA
+316_01.png,4AA
+318_01.png,4AB
+318_02.png,4AB
+318_03.png,3AA
+319_04.png,4AA
+320_01.png,3AB
+321_01.png,4AB
+321_02.png,5AA
+321_03.png,4AA
+321_04.png,4AB
+323_01.png,4AA
+323_02.png,5AA
+323_03.png,4BB
+323_04.png,4AB
+324_01.png,5AA
+324_03.png,5BB
+324_04.png,3AB
+324_05.png,4AA
+324_06.png,4AA
+325_01.png,4AA
+326_01.png,4AA
+326_02.png,3AB
+327_01.png,5AA
+327_02.png,5BA
+327_03.png,4AA
+327_04.png,4AA
+327_05.png,4BA
+327_06.png,4AA
+328_02.png,4AB
+328_03.png,5AB
+328_04.png,5AB
+329_01.png,4AA
+329_02.png,5AA
+329_04.png,4BB
+330_01.png,4AA
+330_02.png,3AA
+330_03.png,4AA
+331_01.png,4AA
+331_02.png,4AA
+331_03.png,3AA
+331_04.png,4AA
+331_05.png,4AA
+333_01.png,3BA
+333_03.png,4BB
+333_04.png,3AB
+333_05.png,4AA
+334_01.png,4AA
+334_02.png,4AB
+335_01.png,3AB
+336_01.png,4AA
+336_02.png,4AA
+337_01.png,4AB
+337_02.png,4AA
+338_01.png,4AA
+338_02.png,5BB
+340_01.png,3AB
+340_03.png,4AB
+340_05.png,3AA
+341_01.png,4AA
+341_05.png,3BA
+342_01.png,4AA
+342_03.png,3AA
+342_06.png,4AB
+342_08.png,4AB
+343_01.png,4AA
+343_02.png,4AA
+343_03.png,5AB
+344_01.png,4AA
+344_03.png,4AA
+344_04.png,3AA
+344_05.png,4BB
+345_02.png,3AA
+346_01.png,4AB
+346_02.png,4AA
+346_03.png,3BA
+346_04.png,3AA
+347_02.png,4BA
+347_03.png,4AB
+348_01.png,4AA
+349_01.png,4AA
+349_02.png,3BA
+350_01.png,4AA
+350_02.png,3AA
+350_03.png,3BA
+350_05.png,4AB
+350_06.png,4AA
+351_01.png,4AA
+351_02.png,4BA
+351_03.png,4AB
+352_03.png,4AA
+352_05.png,3BA
+353_01.png,4AA
+353_02.png,4AA
+353_03.png,4AA
+353_04.png,5AA
+354_01.png,4AA
+354_02.png,4AB
+354_03.png,4AA
+355_01.png,5AA
+355_02.png,4AA
+355_03.png,4AA
+356_01.png,5AA
+357_01.png,4AA
+357_03.png,4AA
+358_01.png,4AB
+358_02.png,4AA
+359_01.png,4AA
+359_02.png,3AA
+360_01.png,4AA
+361_01.png,5AA
+362_01.png,4AA
+362_02.png,4AA
+362_03.png,4AB
+362_04.png,4AA
+363_01.png,4AA
+363_02.png,4AA
+363_03.png,3AA
+363_05.png,4AA
+363_06.png,4BA
+364_01.png,4AA
+366_01.png,4AA
+366_02.png,4AA
+366_03.png,3BB
+366_04.png,4BB
+368_02.png,4AA
+368_03.png,4AA
+368_04.png,3BA
+368_05.png,3AA
+369_02.png,4AA
+369_04.png,4AA
+369_05.png,4BA
+370_01.png,5AA
+371_01.png,4AA
+371_02.png,4AA
+371_03.png,4AA
+372_01.png,4AA
+372_03.png,4AA
+372_04.png,5AA
+372_05.png,4AA
+372_06.png,4AA
+373_01.png,4AA
+374_01.png,4BA
+375_01.png,3AB
+376_01.png,4AA
+377_01.png,4AB
+377_02.png,4AA
+377_03.png,4AB
+378_01.png,4BC
+379_02.png,4AC
+380_01.png,3AA
+381_01.png,5AA
+381_02.png,3AB
+381_03.png,5AA
+382_01.png,5AA
+383_01.png,4AA
+383_03.png,4AA
+383_04.png,3AB
+384_01.png,4AA
+384_02.png,4AB
+384_03.png,3AB
+384_04.png,3AB
+384_05.png,4AB
+386_01.png,4BA
+387_01.png,4AA
+388_01.png,4AA
+388_02.png,4AA
+388_03.png,4AA
+389_01.png,4AB
+389_02.png,4BA
+390_01.png,4AB
+390_02.png,4AA
+390_03.png,4BA
+390_04.png,3AA
+391_01.png,4AA
+391_02.png,4AB
+392_02.png,4AA
+392_04.png,3AA
+392_05.png,4AA
+393_01.png,4AA
+393_05.png,3AA
+393_06.png,4AA
+393_07.png,4AA
+393_09.png,4AA
+393_10.png,4AA
+393_11.png,4BB
+394_01.png,4AA
+395_01.png,4AA
+395_02.png,3AB
+396_01.png,4AA
+397_01.png,3AA
+398_01.png,4BA
+399_04.png,4BA
+400_02.png,4AA
+400_03.png,4AA
+401_03.png,4AA
+401_04.png,3AB
+401_05.png,4AA
+402_02.png,4AA
+403_01.png,4AC
+405_01.png,5AA
+405_02.png,4AA
+405_03.png,3AB
+405_04.png,4AC
+405_05.png,3AB
+406_01.png,4AA
+406_04.png,3AB
+406_05.png,3AB
+406_06.png,3AA
+407_01.png,3AA
+407_02.png,4AB
+408_01.png,4AA
+409_01.png,4AB
+409_02.png,4AA
+409_03.png,4BB
+409_04.png,4AB
+409_05.png,3AA
+409_06.png,4AA
+410_01.png,4AA
+411_01.png,4AA
+411_02.png,4CB
+411_03.png,4AC
+411_04.png,3AB
+412_01.png,4AA
+412_03.png,3AA
+412_04.png,3AA
+413_01.png,4AA
+414_01.png,5AA
+414_02.png,4AB
+414_03.png,3BA
+414_04.png,4AA
+414_05.png,4AA
+414_06.png,4AA
+414_08.png,4AB
+414_09.png,4AA
+414_10.png,4AA
+415_01.png,4AA
+416_01.png,5AA
+417_01.png,5AA
+418_01.png,4AA
+419_01.png,4AA
+419_02.png,3BA
+420_01.png,4AA
+420_02.png,4AA
+420_03.png,4AA
+420_04.png,3AA
+420_05.png,3AA
+420_06.png,4AA
+420_07.png,4AA
+420_08.png,4AA
+421_01.png,4AA
+421_03.png,4AB
+421_04.png,4AA
+421_05.png,4AB
+421_06.png,4AA
+421_07.png,4AA
+423_01.png,4AA
+424_01.png,4AA
+425_01.png,5AA
+426_01.png,3AC
+427_01.png,4AA
+427_02.png,3AB
+429_01.png,5AA
+430_05.png,4AA
+430_06.png,4AA
+430_07.png,5AA
+430_08.png,4AA
+430_09.png,5AA
+431_01.png,3BA
+431_02.png,3AB
+431_05.png,4AA
+432_01.png,4AA
+432_02.png,4AA
+432_03.png,3BA
+433_01.png,4BA
+433_03.png,4AA
+433_04.png,3AA
+434_03.png,3AB
+434_04.png,3AA
+434_08.png,4AA
+435_01.png,4AA
+436_01.png,4AA
+436_02.png,4AA
+436_03.png,3AB
+436_05.png,4BA
+436_07.png,4AA
+436_08.png,4BB
+437_01.png,3AB
+437_02.png,3BA
+437_04.png,5AA
+438_02.png,4AA
+438_03.png,3AB
+439_01.png,5AA
+439_02.png,4AA
+439_06.png,4AB
+440_01.png,4BA
+440_02.png,4AA
+441_02.png,3AA
+442_01.png,4AC
+442_02.png,4AB
+442_04.png,4AB
+443_01.png,4AA
+443_02.png,5AB
+443_03.png,4AA
+443_04.png,4AA
+443_05.png,4BC
+444_02.png,4AB
+444_03.png,4AA
+444_04.png,4AA
+445_01.png,4BA
+445_02.png,4AA
+445_03.png,4AA
+446_02.png,4AC
+447_01.png,3CA
+447_02.png,4AB
+449_01.png,4CB
+453_01.png,3AA
+454_01.png,3AA
+456_01.png,3AA
+456_03.png,3AA
+456_05.png,4AB
+456_06.png,4AA
+456_07.png,4AA
+457_01.png,3AA
+457_02.png,4AA
+457_03.png,4AA
+458_01.png,3AB
+458_02.png,4AA
+458_03.png,4BB
+458_04.png,4BA
+459_03.png,3AA
+460_01.png,3AA
+463_01.png,3AA
+463_02.png,4BB
+463_03.png,4AB
+463_04.png,4AA
+463_05.png,4AA
+464_01.png,3AA
+465_01.png,4AA
+465_02.png,4AA
+466_01.png,4AA
+466_02.png,3BB
+467_01.png,4AA
+467_03.png,4AA
+467_04.png,4BA
+468_01.png,4AA
+468_02.png,4AB
+468_03.png,3BA
+469_01.png,3AA
+469_02.png,3BA
+469_03.png,3AB
+470_01.png,4AA
+471_01.png,4AA
+472_01.png,4AB
+473_01.png,3BA
+474_02.png,3AB
+474_04.png,4AB
+474_05.png,3BA
+475_01.png,4AA
+475_02.png,4AB
+475_03.png,5AA
+475_05.png,4AB
+475_06.png,3AA
+475_07.png,4AA
+476_01.png,4AA
+477_01.png,4AA
+478_01.png,4AA
+478_02.png,4AB
+478_03.png,4AA
+478_04.png,3AB
+479_01.png,4AA
+479_02.png,4AA
+479_03.png,3AA
+480_01.png,4BA
+481_01.png,4AA
+481_02.png,5AA
+481_03.png,4AB
+483_01.png,4AA
+483_03.png,4AA
+483_04.png,4AB
+484_02.png,4AA
+485_01.png,4AB
+485_02.png,4BA
+486_01.png,5AA
+487_01.png,5AA
+488_01.png,4AA
+488_02.png,3AB
+489_01.png,4AA
+489_02.png,4AB
+489_03.png,3AA
+490_01.png,4AA
+490_02.png,4AA
+490_03.png,4AA
+491_01.png,4AA
+492_01.png,4AA
+492_02.png,4AC
+492_03.png,3AA
+493_01.png,4AA
+494_01.png,4AA
+494_02.png,4BA
+495_01.png,4AA
+495_02.png,4AB
+495_04.png,4BB
+496_05.png,4AA
+496_07.png,3AA
+497_01.png,4AA
+497_02.png,4AA
+497_03.png,5AB
+497_04.png,4BB
+498_01.png,4AA
+498_02.png,5AA
+498_03.png,5BA
+499_01.png,4AB
+499_02.png,3AA
+500_01.png,4BA
+501_01.png,4AB
+501_02.png,4AA
+501_03.png,4AA
+501_05.png,4AA
+501_06.png,4AA
+501_07.png,4AA
+502_01.png,4AA
+502_02.png,5AA
+504_02.png,4BA
+506_01.png,4AA
+506_02.png,4AA
+507_01.png,4BA
+507_02.png,4AB
+507_03.png,4BA
+508_03.png,4AB
+508_04.png,3BB
+508_05.png,4AC
+509_01.png,4BB
+510_01.png,3BB
+510_02.png,4AA
+512_02.png,4AA
+513_01.png,4AA
+513_02.png,3AA
+513_03.png,4AA
+514_01.png,3AB
+515_01.png,3AA
+515_02.png,3BB
+515_03.png,4AB
+516_02.png,4AA
+516_03.png,5AB
+516_04.png,3BB
+517_01.png,5AA
+517_02.png,4AA
+518_01.png,3AA
+520_01.png,4AA
+520_03.png,4AA
+520_04.png,4AA
+521_02.png,3AA
+523_01.png,4BA
+523_03.png,3BA
+523_04.png,4AA
+524_01.png,3AB
+524_04.png,3BA
+525_01.png,4AA
+525_03.png,4AA
+526_01.png,4AA
+527_01.png,4AA
+528_01.png,4AA
+528_02.png,4AB
+528_04.png,3AA
+528_05.png,5BA
+528_06.png,4AA
+529_01.png,4AA
+529_03.png,4AA
+529_04.png,4AA
+530_01.png,4BA
+530_02.png,4AB
+531_01.png,4BB
+532_01.png,4AA
+532_02.png,3BB
+533_01.png,4AA
+533_02.png,4BA
+533_03.png,4AB
+534_01.png,3AA
+535_01.png,3AC
+535_02.png,4AA
+535_03.png,4AB
+535_04.png,4AB
+536_01.png,3AA
+538_01.png,4AA
+539_01.png,4AA
+541_02.png,3BB
+542_01.png,5AA
+542_02.png,4AA
+543_01.png,4AA
+543_02.png,4AB
+543_03.png,4BA
+544_01.png,4AB
+544_02.png,3AC
+546_01.png,3AA
+546_02.png,4AA
+547_01.png,4AB
+547_02.png,4BA
+547_03.png,4AB
+547_04.png,3AA
+547_06.png,3AA
+548_01.png,5AA
+548_02.png,3CA
+550_01.png,4AA
+550_03.png,4AA
+552_01.png,4AA
+553_01.png,4AB
+553_02.png,3AA
+553_03.png,3AA
+553_06.png,3AB
+554_01.png,4AA
+554_02.png,5AA
+554_03.png,4AA
+554_04.png,4BC
+556_02.png,4AA
+558_01.png,4AA
+558_03.png,4CA
+558_04.png,4BB
+559_01.png,5BA
+559_02.png,4AA
+559_03.png,5AA
+561_01.png,4AA
+562_01.png,4AA
+562_02.png,3AB
+563_02.png,4AA
+564_01.png,4AA
+564_02.png,3BA
+566_01.png,4AA
+567_01.png,4AB
+568_01.png,4AA
+570_01.png,4AA
+571_01.png,4AA
+571_03.png,4AB
+572_01.png,4AA
+572_02.png,4AC
+572_03.png,3CB
+575_01.png,4AA
+576_01.png,5AA
+578_01.png,4AA
+578_02.png,4BA
+580_01.png,4AA
+581_01.png,4AA
+581_02.png,4BB
+582_01.png,4BA
+583_01.png,3BA
+583_02.png,4AA
+584_01.png,4AA
+585_01.png,4AB
+586_02.png,3AB
+588_01.png,4AA
+588_02.png,4AA
+588_03.png,4AA
+588_05.png,4AB
+588_06.png,4BC
+588_07.png,3AA
+588_09.png,3BA
+588_11.png,4AA
+589_02.png,4BA
+589_03.png,5AA
+589_04.png,3AB
+590_01.png,4BA
+590_03.png,4AB
+590_04.png,4AA
+591_01.png,4AA
+591_04.png,4AA
+593_01.png,3AA
+593_03.png,3AB
+594_01.png,4AB
+596_01.png,4AB
+597_01.png,4BB
+597_02.png,3BA
+598_01.png,4AA
+598_02.png,4BB
+599_02.png,3AA
+599_03.png,4BA
+599_04.png,4AA
+600_02.png,3AB
+600_03.png,3BC
+601_03.png,4AA
+601_04.png,3AA
+602_01.png,4AB
+602_02.png,4BA
+602_04.png,4BC
+603_01.png,4CB
+603_02.png,4AA
+603_04.png,4AA
+603_05.png,4AB
+603_06.png,4AA
+603_07.png,4AB
+603_08.png,4AA
+603_09.png,4AB
+604_01.png,5AA
+604_02.png,4AA
+604_03.png,5AB
+605_01.png,4AB
+605_02.png,4AB
+605_03.png,4AA
+605_04.png,4AA
+606_02.png,4AA
+606_03.png,4AA
+607_01.png,3BA
+607_02.png,3AA
+607_04.png,4AA
+607_05.png,4BB
+607_07.png,4BA
+607_08.png,5BA
+607_09.png,4AC
+607_10.png,3AA
+608_04.png,4CB
+608_05.png,4AB
+609_01.png,4BA
+609_02.png,4AB
+610_01.png,5AA
+610_02.png,4BA
+610_03.png,3BA
+610_05.png,3AC
+611_02.png,4BB
+612_01.png,5AA
+613_01.png,3BB
+615_01.png,4AB
+615_04.png,5AA
+616_01.png,4AA
+616_03.png,4BA
+617_01.png,4AA
+617_03.png,4AA
+619_01.png,3AC
+619_02.png,4AC
+620_01.png,3AA
+621_01.png,3AA
+621_02.png,4AB
+622_01.png,4AB
+623_01.png,4AC
+623_02.png,4AA
+624_01.png,5AB
+624_02.png,5AB
+624_03.png,5BA
+624_04.png,5AC
+625_01.png,4AB
+625_02.png,4BB
+625_03.png,4AA
+626_03.png,3AA
+626_04.png,3BA
+626_06.png,4AA
+626_07.png,4AA
+626_08.png,3AA
+627_02.png,3AA
+627_03.png,4AA
+628_01.png,3AB
+628_03.png,3BA
+628_05.png,4AA
+628_06.png,4AA
+628_08.png,4AA
+629_01.png,3AA
+629_03.png,3AB
+629_04.png,4AA
+629_07.png,3AB
+629_08.png,4AB
+629_09.png,4BC
+630_01.png,4AA
+630_03.png,3AA
+630_04.png,4AA
+631_01.png,3BA
+631_02.png,4AA
+631_03.png,4AA
+631_04.png,4AA
+631_05.png,4AA
+631_06.png,4AB
+631_07.png,4AB
+631_08.png,4BB
+631_09.png,4AB
+632_01.png,3AA
+632_02.png,3AA
+632_04.png,4AA
+633_01.png,3AA
+634_02.png,4AA
+634_04.png,3AB
+635_01.png,4AA
+635_02.png,4AB
+635_03.png,4BB
+636_01.png,4AA
+636_02.png,3AA
+636_04.png,4AA
+636_05.png,3AA
+636_06.png,3AA
+636_09.png,4AB
+637_01.png,4AA
+637_02.png,4AA
+637_03.png,4AA
+637_04.png,3CB
+638_01.png,5AB
+639_02.png,3BA
+639_03.png,4AA
+639_04.png,3AA
+639_05.png,3AA
+639_07.png,4AA
+640_01.png,3AA
+640_02.png,4AB
+640_04.png,5AA
+642_01.png,3BA
+642_02.png,4AA
+642_03.png,3AA
+642_04.png,4AA
+643_01.png,4AA
+643_03.png,4AA
+644_01.png,3AA
+644_02.png,4AA
+644_03.png,4AA
+645_01.png,4AA
+646_01.png,4AB
+646_02.png,5AA
+646_03.png,4BA
+646_04.png,3AB
+646_05.png,3BB
+647_01.png,4AA
+647_02.png,5AA
+648_01.png,4BA
+648_02.png,4AA
+648_03.png,4BA
+648_04.png,4AA
+648_05.png,4AA
+648_06.png,3AA
+648_07.png,4AA
+648_08.png,5AA
+649_01.png,3AA
+649_02.png,3AA
+649_03.png,3AA
+649_04.png,3AA
+649_06.png,3BA
+650_01.png,3BC
+651_01.png,5BA
+652_01.png,4AA
+652_02.png,4AA
+652_03.png,4AA
+653_01.png,4AA
+654_01.png,4BC
+655_01.png,4AA
+655_02.png,4BA
+655_03.png,4AB
+656_01.png,4BA
+656_02.png,3AB
+657_01.png,3AA
+657_02.png,3AA
+657_05.png,4AB
+657_06.png,4BA
+658_02.png,5AA
+658_03.png,4BA
+658_04.png,3AB
+658_05.png,3AA
+658_06.png,4BA
+658_07.png,4AA
+658_08.png,4AA
+660_01.png,4AB
+661_01.png,4AA
+662_02.png,3BC
+663_01.png,4AA
+665_01.png,4AA
+665_02.png,4AB
+666_01.png,4BB
+666_02.png,4AB
+666_03.png,4BC
+666_04.png,3AB
+666_05.png,4AA
+666_06.png,4AA
+667_01.png,3BB
+668_01.png,4AA
+668_02.png,4AA
+669_01.png,3BC
+670_01.png,4AB
+672_01.png,3AB
+674_01.png,4AA
+674_02.png,3AB
+676_01.png,4AB
+678_02.png,4BB
+678_03.png,4AB
+679_01.png,4AB
+680_01.png,3AA
+680_02.png,4AB
+681_01.png,4AA
+684_01.png,4AA
+685_01.png,4BB
+686_01.png,4AA
+686_03.png,3AA
+688_01.png,4AA
+690_01.png,4AA
+690_02.png,3AB
+691_01.png,4AA
+691_02.png,3AA
+692_01.png,3AA
+694_01.png,4AA
+695_01.png,3AB
+695_02.png,3AB
+696_01.png,4AA
+696_02.png,4AB
+697_01.png,4AA
+698_01.png,3AA
+698_02.png,3AB
+699_01.png,4BB
+701_01.png,5AA
+701_03.png,3AB
+703_01.png,4AA
+704_01.png,4AA
+705_01.png,3AB
+706_01.png,3BA
+706_02.png,4BA
+707_01.png,4BA
+707_02.png,4AB
+708_01.png,4AA
+709_01.png,4AA
+709_02.png,4AB
+710_01.png,4AB
+710_03.png,4BA
+710_04.png,4AA
+710_05.png,4BA
+710_06.png,4AA
+711_01.png,4AA
+712_01.png,4AA
+712_02.png,4AA
+713_01.png,4AA
+713_02.png,4BB
+714_01.png,4AA
+714_02.png,3BA
+715_01.png,4AA
+715_02.png,3BA
+715_03.png,4AA
+716_01.png,4AA
+716_02.png,3AA
+716_03.png,3AA
+719_01.png,4AA
+719_02.png,4AA
+721_01.png,5AA
+721_02.png,4AA
+721_03.png,4AA
+723_01.png,4AA
+724_01.png,5AA
+724_02.png,4AA
+724_03.png,4AA
+724_04.png,3AA
+725_01.png,3AA
+726_01.png,5AB
+726_02.png,5AA
+726_03.png,3AA
+727_01.png,4BC
+728_01.png,4AA
+728_03.png,4AB
+728_04.png,4AB
+729_01.png,3AA
+730_01.png,4BB
+730_02.png,3BA
+731_01.png,4BA
+731_02.png,4AB
+732_01.png,5AA
+732_02.png,4AB
+732_03.png,4AB
+732_04.png,4AB
+734_01.png,4AB
+735_01.png,4AA
+736_01.png,3AA
+738_01.png,4AA
+739_01.png,4AA
+739_02.png,4BB
+740_01.png,4AA
+741_01.png,4AA
+741_02.png,4BA
+742_01.png,4AA
+742_02.png,4AB
+743_01.png,4AA
+743_02.png,3BA
+744_01.png,3AB
+745_01.png,3AB
+747_01.png,3BA
+748_01.png,4BB
+748_02.png,4BA
+748_03.png,4AA
+749_01.png,5AA
+749_02.png,4AB
+750_01.png,4AA
+750_02.png,4AA
+751_01.png,5AA
+751_02.png,4AB
+751_03.png,5AA
+752_01.png,4AA
+752_02.png,4AA
+753_01.png,4AA
+753_02.png,4AB
+753_03.png,3AB
+753_04.png,4AA
+753_05.png,3AB
+754_01.png,4AA
+754_02.png,4AB
+756_01.png,4AA
+758_01.png,5AA
+758_02.png,3BB
+758_03.png,4AA
+759_02.png,4AA
+759_04.png,3AA
+760_01.png,3BB
+761_01.png,4AA
+761_03.png,4AB
+763_01.png,4AB
+763_02.png,4BB
+763_03.png,3BA
+765_01.png,3AA
+766_01.png,3CA
+767_01.png,5AA
+767_02.png,4AA
+767_03.png,4AA
+767_04.png,4AA
+769_01.png,4AA
+771_01.png,4AB
+772_02.png,3AA
+772_03.png,3AA
+773_01.png,4AA
+773_02.png,3BB
+773_03.png,3AB
+773_04.png,3BA
+774_01.png,3BC
+775_01.png,4AA
+775_02.png,4AA
+775_03.png,4AA
+775_04.png,4AA
+776_01.png,4AA
+776_02.png,4BA
+777_02.png,4AA
+778_01.png,5AA
+778_02.png,4AA
+779_01.png,4AB
+780_01.png,4AA
+781_01.png,4AA
+781_02.png,4AA
+781_03.png,3AA
+781_04.png,3BA
+782_01.png,4AB
+782_02.png,4AB
+783_01.png,4AA
+785_01.png,4AA
+785_02.png,3BA
+785_03.png,3AA
+786_01.png,4BA
+789_02.png,3BA
+790_01.png,4AA
+790_02.png,3AA
+792_01.png,3BA
+793_01.png,4AB
+794_01.png,3AA
+795_01.png,4AB
+796_01.png,3AA
+796_02.png,3AA
+797_01.png,4AA
+798_01.png,5AA
+798_02.png,4AA
+799_01.png,4AA
+800_01.png,5AA
+801_01.png,3AA
+801_02.png,4AB
+802_01.png,4AA
+802_02.png,4AA
+803_01.png,4AA
+803_02.png,3AB
+804_01.png,4AA
+805_01.png,4AA
+805_02.png,3AA
+808_01.png,4AA
+808_02.png,4AA
+808_03.png,4AA
+808_04.png,4AB
+808_05.png,4AA
+809_01.png,4AA
+810_01.png,4AA
+810_02.png,4AB
+810_03.png,3AA
+811_01.png,4AA
+811_04.png,3AA
+812_01.png,4AA
+812_02.png,3AA
+813_01.png,3AB
+814_01.png,3BA
+815_01.png,4AA
+815_02.png,4AA
+816_01.png,3BA
+816_02.png,4BA
+817_01.png,4BA
+818_01.png,4AA
+820_01.png,4BA
+821_01.png,4AA
+821_02.png,4AB
+822_01.png,3AA
+823_01.png,4AB
+823_02.png,3AA
+823_03.png,4AA
+824_01.png,3AA
+824_04.png,5AA
+824_05.png,4AA
+825_01.png,3AA
+826_01.png,4AA
+826_02.png,4AA
+827_01.png,5AA
+828_01.png,4AA
+828_02.png,3AB
+829_01.png,4AA
+830_02.png,4AA
+831_01.png,4BA
+834_01.png,5AA
+834_02.png,4AA
+834_03.png,4AA
+835_01.png,4AA
+835_02.png,4AB
+835_03.png,4AA
+835_04.png,4AA
+835_06.png,4AA
+835_07.png,4AA
+835_08.png,5BA
+835_09.png,3AB
+835_10.png,4AA
+836_02.png,3AA
+836_03.png,4AA
+836_04.png,5AB
+836_05.png,5AA
+838_01.png,5AA
+838_02.png,4AA
+839_01.png,4AA
+839_04.png,4BB
+840_01.png,4AA
+841_01.png,4BA
+841_02.png,4AA
+841_03.png,4BB
+841_04.png,3BA
+842_01.png,4AA
+842_02.png,4AA
+842_03.png,3BB
+843_01.png,4AA
+843_02.png,3BA
+843_03.png,4AB
+843_04.png,4AA
+844_01.png,4AA
+844_02.png,4AA
+845_01.png,4AA
+845_02.png,4AB
+846_02.png,4AA
+846_04.png,4AA
+847_01.png,4AB
+847_02.png,3BB
+848_01.png,4AA
+848_02.png,4AA
+848_03.png,4AA
+849_01.png,5AA
+849_02.png,3AA
+851_01.png,3BB
+851_02.png,4AB
+853_01.png,4AA
+853_02.png,3AA
+853_03.png,3AA
+854_01.png,4AA
+854_02.png,3AB
+855_01.png,4AA
+855_02.png,3AA
+856_01.png,5AA
+856_02.png,4AA
+856_03.png,3AB
+856_04.png,4AA
+857_01.png,4AA
+857_02.png,3AA
+858_01.png,4AA
+858_02.png,4AB
+858_03.png,4AB
+858_04.png,4AA
+858_05.png,4AA
+858_06.png,4AA
+858_07.png,5BA
+858_08.png,4AB
+858_09.png,4AA
+858_10.png,4AA
+859_02.png,4AB
+860_01.png,3AB
+860_02.png,4AB
+861_01.png,3AA
+861_02.png,3AA
+861_03.png,5AA
+861_04.png,3AA
+861_05.png,4AA
+861_06.png,4AA
+861_07.png,4CA
+861_09.png,4AA

v1/finetune.py

@@ -0,0 +1,170 @@
+import os
+import warnings
+warnings.filterwarnings("ignore")
+
+import gc
+import numpy as np
+import pandas as pd
+from collections import Counter
+from sklearn.metrics import accuracy_score
+from imblearn.over_sampling import RandomOverSampler
+import evaluate
+from datasets import Dataset, Image, ClassLabel
+from transformers import (
+    TrainingArguments,
+    Trainer
+)
+from transformers import AutoImageProcessor
+from transformers import SiglipForImageClassification
+from torch.utils.data import DataLoader
+from torchvision.transforms import (
+    CenterCrop,
+    Compose,
+    Normalize,
+    RandomRotation,
+    RandomResizedCrop,
+    RandomHorizontalFlip,
+    RandomAdjustSharpness,
+    Resize,
+    ToTensor
+)
+from PIL import Image as PILImage
+from PIL import ImageFile
+ImageFile.LOAD_TRUNCATED_IMAGES = True
+import torch
+
+# Paths as provided
+csv_path = r"D:\AI\Interns\Anish\Gardner Grading\3. model creation\AI approaches\siglip2\filtered_data.csv"
+image_dir = r"D:\AI\Interns\Anish\Gardner Grading\3. model creation\AI approaches\siglip2\Images"
+model_path = r"D:\AI\Interns\Anish\Gardner Grading\3. model creation\AI approaches\siglip2\model"
+
+# Load CSV
+df = pd.read_csv(csv_path)
+print("DataFrame shape:", df.shape)
+print(df.head())
+print("Unique grades:", df['grade'].unique())
+
+# Oversample to balance classes
+y = df[['grade']]
+df_no_label = df.drop(['grade'], axis=1)
+ros = RandomOverSampler(random_state=83)
+df_resampled, y_resampled = ros.fit_resample(df_no_label, y)
+df_resampled['grade'] = y_resampled
+df = df_resampled  # use the oversampled DataFrame
+del y, y_resampled, df_no_label
+gc.collect()
+
+# Define labels list from unique grades
+labels_list = sorted(df['grade'].unique().tolist())
+label2id = {label: i for i, label in enumerate(labels_list)}
+id2label = {i: label for i, label in enumerate(labels_list)}
+ClassLabels = ClassLabel(num_classes=len(labels_list), names=labels_list)
+print("Mapping of IDs to Labels:", id2label)
+print("Mapping of Labels to IDs:", label2id)
+
+# Create Hugging Face Dataset from oversampled df, rename 'Image' to 'image_path'
+df = df.rename(columns={'Image': 'image_path'})
+dataset = Dataset.from_pandas(df)
+
+# Map labels to integers
+def map_label2id(examples):
+    examples['label'] = [ClassLabels.str2int(grade) for grade in examples['grade']]
+    return examples
+
+dataset = dataset.map(map_label2id, batched=True)
+dataset = dataset.cast_column('label', ClassLabels)
+
+# Load images
+def load_image(examples):
+    examples['image'] = [PILImage.open(os.path.join(image_dir, path)).convert("RGB") for path in examples['image_path']]
+    return examples
+
+dataset = dataset.map(load_image, batched=True)
+
+# Remove unnecessary columns
+dataset = dataset.remove_columns(['grade', 'image_path'])
+
+full_data = dataset
+
+# Load processor
+processor = AutoImageProcessor.from_pretrained(model_path)
+
+# Extract preprocessing parameters
+image_mean, image_std = processor.image_mean, processor.image_std
+size = processor.size["height"]
+
+# Define training transforms with augmentations
+_train_transforms = Compose([
+    Resize((size, size)),
+    RandomRotation(90),
+    RandomAdjustSharpness(2),
+    ToTensor(),
+    Normalize(mean=image_mean, std=image_std)
+])
+
+def train_transforms(examples):
+    examples['pixel_values'] = [_train_transforms(image) for image in examples['image']]
+    return examples
+
+# Apply transforms
+train_data = full_data.with_transform(train_transforms)
+
+# Data collator
+def collate_fn(examples):
+    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+    labels = torch.tensor([example['label'] for example in examples])
+    return {"pixel_values": pixel_values, "labels": labels}
+
+# Load model
+model = SiglipForImageClassification.from_pretrained(
+    model_path,
+    num_labels=len(labels_list),
+    id2label=id2label,
+    label2id=label2id,
+    ignore_mismatched_sizes=True
+)
+
+print(model.num_parameters(only_trainable=True) / 1e6)
+
+# Metrics (even if no eval)
+accuracy_metric = evaluate.load("accuracy")
+def compute_metrics(eval_pred):
+    predictions = eval_pred.predictions
+    label_ids = eval_pred.label_ids
+    predicted_labels = predictions.argmax(axis=1)
+    acc_score = accuracy_metric.compute(predictions=predicted_labels, references=label_ids)['accuracy']
+    return {"accuracy": acc_score}
+
+# Training arguments for train-only
+args = TrainingArguments(
+    output_dir="./siglip2_finetuned",
+    logging_dir='./logs',
+    eval_strategy="no",
+    do_eval=False,
+    learning_rate=2e-5,
+    per_device_train_batch_size=32,
+    num_train_epochs=3,
+    weight_decay=0.02,
+    warmup_steps=50,
+    remove_unused_columns=False,
+    save_strategy='epoch',
+    load_best_model_at_end=False,
+    save_total_limit=2,
+    report_to="none",
+    fp16=True  # If GPU supports
+)
+
+# Initialize Trainer
+trainer = Trainer(
+    model=model,
+    args=args,
+    train_dataset=train_data,
+    data_collator=collate_fn,
+    tokenizer=processor,
+)
+
+# Train
+trainer.train()
+
+# Save the model at the end
+trainer.save_model()

v1/infer.py

@@ -0,0 +1,78 @@
+import tkinter as tk
+from tkinter import filedialog, messagebox
+from PIL import Image, ImageTk
+import torch
+from transformers import AutoImageProcessor, SiglipForImageClassification
+
+# Model and processor paths (adjust if needed; assumes final model saved in output_dir)
+model_path = "./siglip2_finetuned"  # Or "./siglip2_finetuned/checkpoint-1284" if using a specific checkpoint
+
+# Load processor and model
+processor = AutoImageProcessor.from_pretrained(model_path)
+model = SiglipForImageClassification.from_pretrained(model_path)
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model.to(device)
+model.eval()
+
+# Get label mappings from model config
+id2label = model.config.id2label
+
+# Tkinter GUI
+class ImageClassifierApp:
+    def __init__(self, root):
+        self.root = root
+        self.root.title("SigLIP2 Gardner Grading Classifier")
+        self.root.geometry("600x600")
+
+        # Label for instructions
+        self.instruction_label = tk.Label(root, text="Select an image to classify")
+        self.instruction_label.pack(pady=10)
+
+        # Button to load image
+        self.load_button = tk.Button(root, text="Load Image", command=self.load_image)
+        self.load_button.pack(pady=10)
+
+        # Canvas to display image
+        self.image_canvas = tk.Canvas(root, width=400, height=400, bg="white")
+        self.image_canvas.pack(pady=10)
+
+        # Label to display prediction
+        self.prediction_label = tk.Label(root, text="", font=("Arial", 14))
+        self.prediction_label.pack(pady=10)
+
+    def load_image(self):
+        file_path = filedialog.askopenfilename(filetypes=[("Image files", "*.png *.jpg *.jpeg")])
+        if file_path:
+            try:
+                # Open and convert image to RGB
+                img = Image.open(file_path).convert("RGB")
+                img_resized = img.resize((400, 400))  # For display
+                self.photo_img = ImageTk.PhotoImage(img_resized)
+                self.image_canvas.create_image(200, 200, image=self.photo_img)
+
+                # Preprocess with explicit settings
+                inputs = processor(
+                    images=img,
+                    return_tensors="pt",
+                    do_resize=True,
+                    size={"height": 224, "width": 224},  # Adjust based on model's expected size (common for SigLIP)
+                    do_normalize=True
+                ).to(device)
+
+                # Inference
+                with torch.no_grad():
+                    outputs = model(**inputs)
+                    logits = outputs.logits
+                    predicted_id = logits.argmax(-1).item()
+                    predicted_label = id2label[predicted_id]
+
+                # Display prediction
+                self.prediction_label.config(text=f"Predicted Grade: {predicted_label}")
+
+            except Exception as e:
+                messagebox.showerror("Error", f"Failed to process image: {str(e)}")
+
+if __name__ == "__main__":
+    root = tk.Tk()
+    app = ImageClassifierApp(root)
+    root.mainloop()

v1/inference.py

@@ -0,0 +1,70 @@
+import tkinter as tk
+from tkinter import filedialog, messagebox
+from PIL import Image, ImageTk
+import torch
+from transformers import AutoImageProcessor, SiglipForImageClassification
+
+# Model and processor paths (adjust if needed; assumes final model saved in output_dir)
+model_path = "./siglip2_finetuned"  # Or "./siglip2_finetuned/checkpoint-1284" if using a specific checkpoint
+
+# Load processor and model
+processor = AutoImageProcessor.from_pretrained(model_path)
+model = SiglipForImageClassification.from_pretrained(model_path)
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model.to(device)
+model.eval()
+
+# Get label mappings from model config
+id2label = model.config.id2label
+
+# Tkinter GUI
+class ImageClassifierApp:
+    def __init__(self, root):
+        self.root = root
+        self.root.title("SigLIP2 Gardner Grading Classifier")
+        self.root.geometry("600x600")
+
+        # Label for instructions
+        self.instruction_label = tk.Label(root, text="Select an image to classify")
+        self.instruction_label.pack(pady=10)
+
+        # Button to load image
+        self.load_button = tk.Button(root, text="Load Image", command=self.load_image)
+        self.load_button.pack(pady=10)
+
+        # Canvas to display image
+        self.image_canvas = tk.Canvas(root, width=400, height=400, bg="white")
+        self.image_canvas.pack(pady=10)
+
+        # Label to display prediction
+        self.prediction_label = tk.Label(root, text="", font=("Arial", 14))
+        self.prediction_label.pack(pady=10)
+
+    def load_image(self):
+        file_path = filedialog.askopenfilename(filetypes=[("Image files", "*.png *.jpg *.jpeg")])
+        if file_path:
+            try:
+                # Open and display image
+                img = Image.open(file_path)
+                img_resized = img.resize((400, 400))
+                self.photo_img = ImageTk.PhotoImage(img_resized)
+                self.image_canvas.create_image(200, 200, image=self.photo_img)
+
+                # Preprocess and infer
+                inputs = processor(images=img, return_tensors="pt").to(device)
+                with torch.no_grad():
+                    outputs = model(**inputs)
+                    logits = outputs.logits
+                    predicted_id = logits.argmax(-1).item()
+                    predicted_label = id2label[predicted_id]
+
+                # Display prediction
+                self.prediction_label.config(text=f"Predicted Grade: {predicted_label}")
+
+            except Exception as e:
+                messagebox.showerror("Error", f"Failed to process image: {str(e)}")
+
+if __name__ == "__main__":
+    root = tk.Tk()
+    app = ImageClassifierApp(root)
+    root.mainloop()

v1/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25dc1e164ee1e85fd09314fae638557bb38b03a4f07adeb4a1492be808e8c9f1
+size 371629520

v1/preprocessor_config.json

@@ -0,0 +1,24 @@
+{
+  "do_convert_rgb": null,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "SiglipImageProcessor",
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "processor_class": "SiglipProcessor",
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 224,
+    "width": 224
+  }
+}

v1/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d3142bd300351d5b2a53e1c77655b4c6c0503cb026feec26d0ac94dd37597d9
+size 5713