FlowFace: Semantic Flow-Guided Shape-Aware Face Swapping

In this work, we propose a semantic flow-guided two-stage framework for shape-aware face swapping, namely FlowFace. Unlike most previous methods that focus on transferring the source inner facial features but neglect contours, our FlowFace can transfer both of them to target face, thus leading more realistic swapping. Concretely, consists reshaping network and swapping network. The addresses shape outline differences between faces. It first estimates flow (i.e. differences) then explicitly warps with estimated flow. After reshaping, generates exhibit identity face. We employ pre-trained masked autoencoder (MAE) extract from contrast use embedding preserve information, extracted by encoder better capture appearances information. Then, develop cross-attention fusion module adaptively fuse attributes, preservation. Extensive quantitative qualitative experiments in-the-wild faces demonstrate outperforms state-of-the-art significantly.