M2TR: Multi-modal Multi-scale Transformers for DeepfakeDetection

Introduction

This is the official pytorch implementation of Multi-modal Multi-scale for Deepfake detection, which is accepted by ICMR 2022.

Model Zoo

The baseline models on three versions of FF-DF dataset and Celeb-DF are provided.

Dataset	Accuracy	model
FF++ (Raw)	99.50	FF-RAW
FF++ (C23)	97.93	FF-C23
FF++ (C40)	92.89	FF-C40
CelebDF	99.76	CelebDF

Training and Evaluation

python run.py --cfg m2tr.yaml

License

This project is released under the MIT license.

Citations

@article{wang2021m2tr,
  inproceedings={M2TR: Multi-modal Multi-scale Transformers for Deepfake Detection},
  author={Wang, Junke and Wu, Zuxuan and Chen, Jingjing and Jiang, Yu-Gang},
  booktitle={ICMR},
  year={2022}
}