StyleMapGAN: Exploiting Spatial Dimensions of Latent in GAN for Real-time Image Editing

The official demo of StyleMapGAN (CVPR21) We recommend watching the video at 1080p. The paper and code are available at the links below. Paper (arXiv): Paper (PDF): Code (GitHub): Authors: Hyunsu Kim¹² Yunjey Choi¹ Junho Kim¹ Sungjoo Yoo² Youngjung Uh³ ¹Naver AI Lab ²Seoul National University ³Yonsei University Abstract: Generative adversarial networks (GANs) synthesize realistic images from random latent vectors. Although manipulating the latent vectors controls the synthesized outputs, editing real images with GANs suffers from i) time-consuming optimization for projecting real images to the latent vectors, ii) or inaccurate embedding through an encoder. We propose StyleMapGAN: the intermediate latent space has spatial dimensions, and a spatially variant modulation replaces AdaIN. It makes the embedding through an encoder more accurate than existing optimization-based met