WildDeepfake is a dataset for real-world Deepfake detection. The dataset contains 7,314 facial action sequences extracted from 707 Deepfake videos. The WildDeepfake dataset is different from previous datasets in some ways: (1) The videos are all sourced from the Internet, and the content is rich and diverse. The visual effects are more realistic and more in line with real-life scenes. (2) Face-swapped videos are synthesized using a variety of methods, making them more difficult to detect.

Citation

@inproceedings{zi2020wilddeepfake, title={Wilddeepfake: A challenging real-world dataset for deepfake detection}, author={Zi, Bojia and Chang, Minghao and Chen, Jingjing and Ma, Xingjun and Jiang, Yu-Gang} booktitle={Proceedings of the 28th ACM International Conference on Multimedia}, pages={2382–2390}, year={2020} }