Ryota Natsume, Tatsuya Yatagawa, Shigeo Morishima

RSGAN: Face Swapping and Editing Via Region Separation in Latent Spaces

SIGGRAPH 2018, Posters


This poster proposes a new deep generative model that we refer to as region separative generative adversarial network (RSGAN), which can achieve face swapping between arbitrary image pairs and can robustly perform the swapping compared to previous methods using 3DMM.