huahua123313/

AVH-Align

Official PyTorch Implementation of the Paper:

Ștefan Smeu, Dragoș-Alexandru Boldisor, Dan Oneață and Elisabeta OneațăCircumventing shortcuts in audio-visual deepfake detection datasets with unsupervised learningCVPR, 2025

To set up your data, follow these steps: Download the datasets:

AV-Deepfake1M(AV1M) Dataset: Follow instructions from AV-Deepfake1M FakeAVCeleb Dataset: Follow instructions from FakeAVCeleb GitHub repo AVLips Dataset: Follow… See the full description on the dataset page: https://huggingface.co/datasets/huahua123313/AVH-Align.