A survey of facial deepfake detection in the context of generative AI

The resources are organized according to the structure of "research category method paper code dataset evaluation indicators", covering facial generation and forgery methods such as explicit 3D models, neural rendering, autoencoders, generative adversarial networks, diffusion models, as well as main technical routes such as image spatial domain detection, frequency domain detection, spatial frequency fusion detection, video spatiotemporal detection, physiological feature detection, and multimodal detection. Meanwhile, this resource has been compiled FaceForensics++、Celeb-DF、DFDC、ForgeryNet、DF40、Deepfake-Eval-2024、Omni-Fake Collect representative datasets and evaluation resources, and organize commonly used evaluation indicators such as ACC, AUROC, EER, F1, mIoU, and related experimental report conditions.

The resources provide access to representative papers, open code, and datasets for researchers to conduct literature searches, compare methods, select datasets, and reproduce experiments. We will continue to supplement and update the newly published results in the field of facial deepfake generation and detection in the future.