Region-Aware Face Swapping
This work addresses face swapping for applications like entertainment and security, but it is incremental as it builds on existing methods like StyleGAN2.
The paper tackles the problem of generating identity-consistent high-resolution swapped faces by proposing a Region-Aware Face Swapping network, achieving a 96.70 ID retrieval score that outperforms the state-of-the-art MegaFS by 5.87.
This paper presents a novel Region-Aware Face Swapping (RAFSwap) network to achieve identity-consistent harmonious high-resolution face generation in a local-global manner: \textbf{1)} Local Facial Region-Aware (FRA) branch augments local identity-relevant features by introducing the Transformer to effectively model misaligned cross-scale semantic interaction. \textbf{2)} Global Source Feature-Adaptive (SFA) branch further complements global identity-relevant cues for generating identity-consistent swapped faces. Besides, we propose a \textit{Face Mask Predictor} (FMP) module incorporated with StyleGAN2 to predict identity-relevant soft facial masks in an unsupervised manner that is more practical for generating harmonious high-resolution faces. Abundant experiments qualitatively and quantitatively demonstrate the superiority of our method for generating more identity-consistent high-resolution swapped faces over SOTA methods, \eg, obtaining 96.70 ID retrieval that outperforms SOTA MegaFS by 5.87$\uparrow$.