- train.py: it trains different networks specified by config.py (where the real network models are in models.py)
- predict.py: Given a list of trained model, it loades all the models and make predictions
- SETTINGS.py specifis the data directories
- CNN_models: the specific CNN models
- img_process: realtime loading and img augmentation (including rotation, shift, shear, flip, PCA_color noise, croping and more)