multi-attention deepfake detection https://arxiv.org/abs/2103.02406 I will continue to update full evaluating code ,models ,training code and document.