doodleai / scripts /README.md
alanoee's picture
Upload folder using huggingface_hub
346f830 verified
# Scripts
Data preparation and model training scripts.
## Files
- `prepare_data.py` - Download and preprocess the Quick Draw! dataset
- `train_model.py` - Train the CNN model
- `check_classes_mapping.py` - Verify the number of output classes in a trained model
## Usage
```bash
# 1. Download and preprocess data (~2GB download, configurable sample count)
python scripts/prepare_data.py
# 2. Train the model
python scripts/train_model.py
```
## Configuration
Paths and training parameters are defined in `config.py` at the project root.
## Output
- `prepare_data.py` writes processed splits to `dataset/processed/` and class mappings to `model/classes.json`
- `train_model.py` saves the best model to `model/best_model.keras` and training plots to `outputs/`