@@ -18,6 +18,6 @@ We introduce WaveConViT, a novel spatio-temporal architecture for deepfake detec
Additionally, we introduce and evaluate a temporal sampling strategy based on frame skipping.
## Datasets
We train WaveConViT end-to-end on [FaceForensics++](https://arxiv.org/pdf/1901.08971) in a binary image sequence classification task.
We train WaveConViT end-to-end on [FaceForensics++](https://arxiv.org/abs/1901.08971) in a binary image sequence classification task.
Cross-manipulation testing is done on [Celeb-DF](https://arxiv.org/abs/1909.12962) and [DeeperForensics-1.0](https://arxiv.org/abs/2001.03024)(standard set).
