Diffusion-Deepfake-Detection-Datasets_2023

Since there is no conventially selected DFD benchmarks hitherto, this repository is a thorough survey of useful diffusion deepfake detection datasets to foster further researches.

🔎 Diffusion Deepfake Detection Datasets

1. DiffusionForensics (ICCV'23)
   • Paper: https://arxiv.org/abs/2303.09295
   • Github: https://github.com/ZhendongWang6/DIRE
   • Generation Models: Diffusion Models, GAN
   • Type: General
   • Potential Issues: Real,Fake image size and image type mismatched
2. ArtiFact (ICIP'23)
   • Paper: https://arxiv.org/abs/2302.11970
   • Github: https://github.com/awsaf49/artifact
   • Generation Models: Diffusion Models, GAN
   • Type: General
3. DeepFakeFace (arXiv'23)
   • Paper: https://arxiv.org/abs/2309.02218
   • Github: https://github.com/OpenRL-Lab/DeepFakeFace
   • Generation Models: Stable Diffusion
   • Type: Face
4. GenImage (arXiv'23)
   • Paper: https://arxiv.org/abs/2306.08571
   • Github: https://github.com/GenImage-Dataset/GenImage
   • Generation Models: Stable Diffusion
   • Type: General
5. CIFAKE (arXiv'23)
   • Paper: https://arxiv.org/pdf/2303.14126.pdf
   • Github: -
   • Kaggle: https://www.kaggle.com/datasets/birdy654/cifake-real-and-ai-generated-synthetic-images
   • Generation Models: SDM(LDM)
   • Type: General (Follow CIFAR-10)
6. DMimageDetection (ICASSP'23)
   • Paper: https://arxiv.org/abs/2211.00680
   • Github: https://github.com/grip-unina/DMimageDetection
   • Generation Models: Diffusion Models, GAN
   • Type: General
   • Potential Issues: Real parts should be downloaded from IMAGENET,COCO,UCID on your own.(Possible mismatch subset based on your own selection)
7. Towards the Detection of Diffusion Model Deepfakes (arXiv'22)
   • Paper: https://arxiv.org/pdf/2210.14571.pdf
   • Github: https://github.com/grip-unina/DMimageDetection
   • Generation Models: Diffusion Models, GAN
   • Type: General
   • Potential Issues: You have to recreate the train and validation set on your own

🌌 Other synthetic Data by Diffusion Models

1. (SFHQ)Synthetic Faces High Quality
   • Paper: -
   • Github: https://github.com/SelfishGene/SFHQ-dataset
   • Kaggle:
     • Part2: https://www.kaggle.com/datasets/selfishgene/synthetic-faces-high-quality-sfhq-part-2
     • Part3: https://www.kaggle.com/datasets/selfishgene/synthetic-faces-high-quality-sfhq-part-4
   • Generation Models:
     • Part2: Stable Diffusion v1.4
     • Part4: Stable Diffusion v2.1
   • Type: Face

Acknowledgement

Sincerely appreciate for all the contributions from original authors. If you find there are any needed updates, please feel free to fork and pull request.