# Models Directory

This directory will contain the ONNX models after conversion.

## Structure

After running the conversion scripts, this directory will contain:

```
models/
├── classifier_model_compressed/
│   ├── model.onnx              # ONNX model for embryo detection
│   ├── config.json              # Model configuration
│   └── preprocessor_config.json # Image preprocessing config
│
├── poor_good_compressed/
│   ├── model.onnx              # ONNX model for quality assessment
│   ├── config.json              # Model configuration
│   └── preprocessor_config.json # Image preprocessing config
│
├── grader_model_compressed/
│   ├── model.onnx              # ONNX model for Gardner grading
│   ├── config.json              # Model configuration
│   └── preprocessor_config.json # Image preprocessing config
│
└── yolo-cropper/
    └── best.onnx                # ONNX model for embryo detection
```

## How to Generate Models

Run the conversion scripts from the parent directory:

```bash
# Convert SigLIP models
python convert_to_onnx.py

# Convert YOLO model
python convert_yolo_to_onnx.py
```

## Model Sizes (Approximate)

- Classifier: ~95 MB
- Poor/Good: ~95 MB
- Grader: ~95 MB
- YOLO: ~6 MB

Total: ~290 MB

## Note

These models are not included in the repository due to their size.
You must convert them from the original PyTorch models.