This is a project that explores the performance increase from using a backbone trained using a self-supervised method (SimSiam) before training in a more traditional supervised way.
You should use a virtual environment. Thus do the following steps for the first time setup:
python -m venv venv
source venv/bin/activate # windows: venv\Scripts\activate
pip install -r requirements.txt
Cat breeds image dataset: https://www.kaggle.com/datasets/shawngano/gano-cat-breed-image-collection
Animals10: https://www.kaggle.com/datasets/alessiocorrado99/animals10
- Update the parameters in
train_ssl.py
to your liking. - Run
python train_ssl.py
- Update the parameters in
train_supervised.py
to your liking. - Run
python train_supervised.py