This is an unofficial implementations for Latent Distribution Adjusting for Face Anti-Spoofing. To refer the paper please click here.
Put your dataset in the folder data/, and corresponding [dataset.py] in datasets/(for instance, define
MyDataset in datasets/dataset), and import your customized dataset in [trainer.py].
Use Bash train.sh to train LDA model. The model consists many hyperparameters, which is stored in hyp/LDA.json.
During training model's parameter will be automatically stored in results/ on every epoch, denoted with validation loss.