/idrnd-anti-spoofing-challenge

ID R&D Anti-spoofing Challenge 1st Place Solution

Primary LanguagePythonMIT LicenseMIT

ID R&D Anti-spoofing Challenge 1st place Solution (link)

Train model: python train.py --config 'configuration file'

Predict: python test.py --path-images-csv 'annotation' --path-test-dir 'path to images' --path-submission-csv 'submission'

Final solution is based on ensemble of two models: se_resnext50 and densnet121.

The following configs are used to train this models: se_resnext50_focal.yaml and densnet121_focal.yaml

Pretrained models: se_resnext50_focal.pth and densnet121_focal.pth