SimCLR for pre-training VGG19 weights

Fork of PyTorch SimCLR adapted for pre-training VGG19 weights for usage in Medical Out-of-Distribution Analysis Challenge (MOOD).

See original code and documentation in https://github.com/sthalles/SimCLR

See a description of our solution to the MOOD Challenge in https://github.com/ninatu/mood_challenge

Usage

1. Install the anomaly detection framework

pip install git+https://github.com/ninatu/mood_challenge.git

2. Perform "Data Preparation" step. See [email protected]:ninatu/mood_challenge.git
3. Put correct paths in config/exp_1_mood.yaml and run

python run.py configs/exp_1_mood_tmp.yaml