OpenForensics: Large-Scale Challenging Dataset For Multi-Face Forgery Detection And Segmentation In-The-Wild

Robust Deepfake On Unrestricted Media: Generation And Detection

“…Natsume et al [29] encoded latent vectors for face and hair as two separate variational AEs (VAEs) and then conditionally swapped or edited the identity of the target. Le et al [22] proposed a framework for generating fake identities by combining facial attribute editing and face-swapping. These methods are subject agnostic and can be applied to any pair of faces without retraining.…”

“…large pose variations. Although the improved blending with recently developed methods [4,22] helps to reduce artifacts, it does not completely solve the problem.…”

“…Zhou et al [53] trained an attention framework to detect face forgeries in multi-person scenes. Le et al [22] used transfer learning to build end-to-end models for multiple face forgery detection and segmentation for in-the-wild images.…”

“…Indeed, although strong deep learning models have attained very high accuracy [48,58] in laboratory environments with a simple background and a single clear front-view face [61,80], they generally do not work well in diverse contexts that do not share a close distribution with the training dataset. Le et al [22] showed that even end-to-end deepfake detectors are easily degraded on unseen image scenarios, i.e., ones distant from the training distribution.…”

“…In addition, it insightfully discusses potential ways of improving the robustness of deepfake detection methods. Face2Face [5] Deepfakes [2] FaceSwapGAN [3] ZAO [14] Reface [13] Zhu et al [18] FaceApp [12] ProGAN [19] StyleGAN [20] DeepFaceLab [21] Le et al [22] Korshunova et al [23] StarGAN [24] NeuralTextures [6] StyleGANv2 [25] ReenactGAN [26] GANimation [27] StarGANv2 [28] RsGAN [29] FSGAN [7] InterFaceGAN [30] X2Face [31] Zhang et al [32] StyleALAE [33] FSNet [34] Egor et al [11] FaceShifter [4] Zhixin et al [8] ICface [35] Kim et al [36] FaR-GAN [37] The remainder of this chapter is organized as follows. Section 2 introduces deepfake generation methods.…”