Translate the Facial Regions You Like Using Region-Wise Normalization
This work addresses the problem of precise control in facial region translation for applications like expression changes and makeup, representing an incremental advancement in GAN-based face translation.
The paper tackles region-level face translation by proposing a region-wise normalization framework, achieving large improvements over state-of-the-art methods like StarGAN, SEAN, and FUNIT on datasets such as Morph, RaFD, and CelebAMask-HQ.
Though GAN (Generative Adversarial Networks) based technique has greatly advanced the performance of image synthesis and face translation, only few works available in literature provide region based style encoding and translation. We propose in this paper a region-wise normalization framework, for region level face translation. While per-region style is encoded using available approach, we build a so called RIN (region-wise normalization) block to individually inject the styles into per-region feature maps and then fuse them for following convolution and upsampling. Both shape and texture of different regions can thus be translated to various target styles. A region matching loss has also been proposed to significantly reduce the inference between regions during the translation process. Extensive experiments on three publicly available datasets, i.e. Morph, RaFD and CelebAMask-HQ, suggest that our approach demonstrate a large improvement over state-of-the-art methods like StarGAN, SEAN and FUNIT. Our approach has further advantages in precise control of the regions to be translated. As a result, region level expression changes and step by step make up can be achieved. The video demo is available at https://youtu.be/ceRqsbzXAfk.