DeeperForensics-1.0: A Large-Scale Dataset for Real-World Face Forgery Detection

This repository provides the dataset and code for the following paper:

DeeperForensics-1.0: A Large-Scale Dataset for Real-World Face Forgery Detection
Liming Jiang, Ren Li, Wayne Wu, Chen Qian and Chen Change Loy
To appear in CVPR, 2020.
Project Page | Paper | YouTube Demo

Abstract: In this paper, we present our on-going effort of constructing a large-scale benchmark, DeeperForensics-1.0, for face forgery detection. Our benchmark represents the largest face forgery detection dataset by far, with 60, 000 videos constituted by a total of 17.6 million frames, 10 times larger than existing datasets of the same kind. Extensive real-world perturbations are applied to obtain a more challenging benchmark of larger scale and higher diversity. All source videos in DeeperForensics-1.0 are carefully collected, and fake videos are generated by a newly proposed end-to-end face swapping framework. The quality of generated videos outperforms those in existing datasets, validated by user studies. The benchmark features a hidden test set, which contains manipulated videos achieving high deceptive scores in human evaluations. We further contribute a comprehensive study that evaluates five representative detection baselines and make a thorough analysis of different settings. We believe this dataset will contribute to real-world face forgery detection research.

Updates

[05/2020] The perturbation codes of DeeperForensics-1.0 are released.
[05/2020] The dataset of DeeperForensics-1.0 is released.
[02/2020] The paper of DeeperForensics-1.0 is accepted by CVPR 2020.

Dataset

DeeperForensics-1.0 dataset has been made publicly available for non-commercial research purposes. Please visit the dataset download and document page for more details.

We plan to host a challenge, DeeperForensics Challenge, based on DeeperForensics-1.0 dataset. Thus, the hidden test set will not be released. Looking forward to your participation!

Code

The code to implement the diverse perturbations in our dataset has been released. Please see the perturbation implementation for more details.

Summary

Data Collection

We invite 100 paid actors from 26 countries to record the source videos. Our high-quality collected data vary in identities, poses, expressions, emotions, lighting conditions, and 3DMM blendshapes.

Face Manipulation

We also propose a new learning-based many-to-many face swapping method, DeepFake Variational Auto-Encoder (DF-VAE). DF-VAE improves scalability, style matching, and temporal continuity to ensure face swapping quality.

Several face manipulation results:

Many-to-many (three-to-three) face swapping by a single model:

Real-World Perturbation

We apply 7 types (transmission errors, compression, etc.) of distortions at 5 intensity levels. Some videos are subjected to a mixture of more than one distortion. These perturbations make DeeperForensics-1.0 better simulate real-world scenarios.

Benchmark

We benchmark five representative forgery detection methods using the DeeperForensics-1.0 dataset. Please refer to our paper for more information.

Citation

If you find this work useful for your research, please cite our paper:

@inproceedings{jiang2020deeperforensics10,
  title={DeeperForensics-1.0: A Large-Scale Dataset for Real-World Face Forgery Detection},
  author={Jiang, Liming and Li, Ren and Wu, Wayne and Qian, Chen and Loy, Chen Change},
  booktitle={CVPR},
  year={2020}
}

Acknowledgments

We gratefully acknowledge the exceptional help from Hao Zhu and Keqiang Sun for their contribution on source data collection and coordination.

Contact

If you have any questions, please contact us by sending an email to deeperforensics@gmail.com.

Terms of Use

The use of DeeperForensics-1.0 is bounded by the Terms of Use: DeeperForensics-1.0 Dataset.
The code is released under the MIT license.

EYcab / DeeperForensics-1.0