github.com/YZY-stack/DF40 @main
Chat with this repoWhat it actually does AI analysis from the code graph — generated when you open this
README
DF40: Toward Next-Generation Deepfake Detection (Project Page; Paper; Download DF40; Checkpoints)
🎉🎉🎉 Our DF40 has been accepted by NeurIPS 2024 D&B track!
Welcome to our work DF40, for next-generation deepfake detection.
In this work, we propose: (1) a diverse deepfake dataset with 40 distinct generations methods; and (2) a comprehensive benchmark for training, evaluation, and analysis.
"Expanding Your Evaluation with 40 distinct High-Quality Fake Data from the FF++ and CDF domains!!"
DF40 Dataset Highlight: The key features of our proposed DF40 dataset are as follows
✅ Forgery Diversity: DF40 comprises 40 distinct deepfake techniques (both representive and SOTA methods are included), facilitating the detection of nowadays' SOTA deepfakes and AIGCs. We provide 10 face-swapping methods, 13 face-reenactment methods, 12 entire face synthesis methods, and 5 face editing.
✅ Forgery Realism: DF40 includes realistic deepfake data created by highly popular generation software and methods, e.g., HeyGen, MidJourney, DeepFaceLab, to simulate real-world deepfakes. We even include the just-released DiT, SiT, PixArt-$\alpha$, etc.
✅ Forgery Scale: DF40 offers million-level deepfake data scale for both images and videos.
✅ Data Alignment: DF40 provides alignment between fake methods and data domains. Most methods (31) are generated under the FF++ and CDF domains. Using our fake data, you can further expand your evaluation (training on FF++ and testing on CDF).
The figure below provides a brief introduction to our DF40 dataset.
The following table displays the statistical description and illustrates the details of our DF40 dataset. Please check our paper for details.
💥 DF40 Dataset
| Type | ID-Number | Generation Method | Original Data Source | Visual Examples |
|---|---|---|---|---|
| Face-swapping (FS) | 1 | FSGAN | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 2 | FaceSwap | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 3 | SimSwap | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 4 | InSwapper | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 5 | BlendFace | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 6 | UniFace | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 7 | MobileSwap | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 8 | e4s | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 9 | FaceDancer | FF++ and Celeb-DF |  |
| Face-swapping (FS) | 10 | DeepFaceLab | UADFV |  |
| Face-reenactment (FR) | 11 | FOMM | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 12 | FS_vid2vid | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 13 | Wav2Lip | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 14 | MRAA | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 15 | OneShot | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 16 | PIRender | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 17 | TPSM | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 18 | LIA | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 19 | DaGAN | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 20 | SadTalker | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 21 | MCNet | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 22 | HyperReenact | FF++ and Celeb-DF |  |
| Face-reenactment (FR) | 23 | HeyGen | FVHQ |  |
| Entire Face Synthesis (EFS) | 24 | VQGAN | Finetuning on FF++ and Celeb-DF |  |
| Entire Face Synthesis (EFS) | 25 | StyleGAN2 | Finetuning on FF++ and Celeb-DF |  |
| Entire Face Synthesis (EFS) | 26 | StyleGAN3 | Finetuning on FF++ and Celeb-DF |  |
| Entire Face Synthesis (EFS) | 27 | StyleGAN-XL | Finetuning on [FF++](https://github.com/ondy |