deepfake-models

List some popular deepfake models e.g. DeepFake, FaceSwap, IPGAN, FaceShifter, Nirkin et al, FSGAN, etc.

Deepfake is a tool that utilizes deep learning to recognize and swap faces in pictures and videos. [code] | [forum]

FaceSwap is an app that have originally created as an exercise for students in "Mathematics in Multimedia". [code] | [homepage]

It uses face alignment, Gauss-Newton optimization, and image blending to swap the face of a person seen by the camera with a face of a person in a provided image.

Towards Open-Set Identity Preserving Face Synthesis. CVPR 2018 [paper] | [homepage]

Jianmin Bao, Dong Chen, Fang Wen, Houqiang Li, and Gang Hua.

propose an Open-Set Identity Preserving Generative Adversarial Network framework for disentangling the identity and attributes of faces, synthesizing faces from the recombined identity and attributes.

FaceShifter: Towards High Fidelity And Occlusion Aware Face Swapping. CVPR 2020 [paper] | [homepage]

Lingzhi Li, Jianmin Bao, Hao Yang, Dong Chen, Fang Wen.

Faceshifter is a novel two-stage framework for high fidelity and occlusion aware face-swapping. It's able to generate high fidelity identity preserving face swap results and, in comparison to previous methods, deal with facial occlusions using a second synthesis stage consisting of a Heuristic Error Acknowledging Refinement Network (HEAR-Net).
- in the first stage, generate the swapped face in high-fidelity by exploiting and integrating the target attributes thoroughly and adaptively.
- in the second stage, propose a novel Heuristic Error Acknowledging Refinement Network (HEAR-Net) to address the challenging facial occlusions.

On face segmentation, face swapping, and face perception.. F&G 2018 [paper] | [code] [homepage]

Yuval Nirkin, Iacopo Masi, Anh Tran Tuan, Tal Hassner, and Gerard Medioni.
- Instead of tailoring systems for face segmentation, as others previously proposed, this work shows that a standard fully convolutional network (FCN) can achieve remarkably fast and accurate segmentation, provided that it is trained on a rich enough example set.
- use special image segmentation to enable robust face-swapping under unprecedented conditions.
- fit 3D face shapes
- measure the effect of intra- and inter-subject face swapping on recognition. Generally speaking, intra-subject swapped faces remain as recognizable as their sources, while better face-swapping produces less recognizable inter-subject results.

FSGAN: Subject Agnostic Face Swapping and Reenactment. ICCV 2019 [paper] | [code] | [homepage-Nirkin] | [homepage-Hassner]

Yuval Nirkin, Yosi Keller, Tal Hassner.

Unlike previous work, FSGAN is subject agnostic and can be applied to pairs of faces without requiring training on those faces.

chaerlo / deepfake-models