DFIL: Deepfake Incremental Learning by Exploiting Domain-invariant Forgery Clues

This is official code for DFIL.
Paper

Overview of DFIL

Detail of training

Firstly, you should download all datasets including(FF++,DFDC-p,CDF,DFD).

Secondly, you can use file 'train_CNN_SupCon_and_CE.py' to train your first detection model with FF++ dataset.

Thirdly, you can use 'get_feature.py' , 'get_image_info.py' and 'create_memory.py' to construct your memory set.

Finally, you could randomly pick up 25 train video in your new dataset and add them into your memery set to train new model by file 'TaskN_KD.py'.

Trained Model

First Task Model

Citation

@inproceedings{pan2023dfil,
  title={DFIL: Deepfake Incremental Learning by Exploiting Domain-invariant Forgery Clues},
  author={Pan, Kun and Yin, Yifang and Wei, Yao and Lin, Feng and Ba, Zhongjie and Liu, Zhenguang and Wang, Zhibo and Cavallaro, Lorenzo and Ren, Kui},
  booktitle={Proceedings of the 31st ACM International Conference on Multimedia},
  pages={8035--8046},
  year={2023}
}