Face swapping is the task of generating images with source face identity and the attributes from the target image. Current approaches face difficulties in providing photo-realistic results or conserving the original face shape.
A recent study proposes a novel end-to-end learning framework, which can preserve the face shape and generate high fidelity face-swapping results. A 3D shape-aware identity extractor generates an identity vector with exact shape information to enforce precise face shape transfer. A semantic facial fusion module is proposed to achieve a better combination in feature-level and image-level. That helps to solve occlusion and lighting problems.
Extensive experiments show that the suggested method can generate higher fidelity results than previous approaches. It can be applied in the film industry, computer games, or face forgery detection.
In this work, we propose a high fidelity face swapping method, called HifiFace, which can well preserve the face shape of the source face and generate photo-realistic results. Unlike other existing face swapping works that only use face recognition model to keep the identity similarity, we propose 3D shape-aware identity to control the face shape with the geometric supervision from 3DMM and 3D face reconstruction method. Meanwhile, we introduce the Semantic Facial Fusion module to optimize the combination of encoder and decoder features and make adaptive blending, which makes the results more photo-realistic. Extensive experiments on faces in the wild demonstrate that our method can preserve better identity, especially on the face shape, and can generate more photo-realistic results than previous state-of-the-art methods.
Research paper: Wang, Y., “HifiFace: 3D Shape and Semantic Prior Guided High Fidelity Face Swapping”, 2021 . Link to the article: https://arxiv.org/abs/2106.09965