KAUST DepartmentComputer Science Program
Visual Computing Center (VCC)
Computer, Electrical and Mathematical Science and Engineering (CEMSE) Division
Permanent link to this recordhttp://hdl.handle.net/10754/677923
MetadataShow full item record
AbstractWhile GANs can produce photo-realistic images in ideal conditions for certain domains, the generation of full-body human images remains difficult due to the diversity of identities, hairstyles, clothing, and the variance in pose. In-stead of modeling this complex domain with a single GAN, we propose a novel method to combine multiple pretrained GANs, where one GAN generates a global canvas (e.g., human body) and a set of specialized GANs, or insets, focus on different parts (e.g., faces, shoes) that can be seamlessly inserted onto the global canvas. We model the problem as jointly exploring the respective latent spaces such that the generated images can be combined, by inserting the parts from the specialized generators onto the global canvas, without introducing seams. We demonstrate the setup by combining a full body GAN with a dedicated high-quality face GAN to produce plausible-looking humans. We evalu-ate our results with quantitative metrics and user studies.
CitationFruhstuck, A., Singh, K. K., Shechtman, E., Mitra, N. J., Wonka, P., & Lu, J. (2022). InsetGAN for Full-Body Image Generation. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). https://doi.org/10.1109/cvpr52688.2022.00757
Conference/Event name2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)