Papers › semantic image synthesis of anime characters based on conditional generative...
semantic image synthesis of anime characters based on conditional generative adversarial networks
Xuhui Zhu; Feng Jiang; Jing Wen; Yi Wang; Qiang Gao
The goal of semantic image synthesis is to generate realistic images from semantic label maps. However, current approaches for generating anime characters from semantic label maps still encounter some issues, particularly the inability to directly generate a specific anime character from the semantic label map, as well as blurred colors and chaotic textures in the generated images. To address these issues, we propose a Conditional Generative Adversarial Network for Semantic Image Synthesis of Anime Characters. Specifically, in the generator, we propose character identity tensor to control the generation of specified anime characters, and introduce conditional noise to enable the generated images to have natural colors. Additionally, we redesign the discriminator as a network based on semantic segmentation and edge detection, which effectively supervises the texture details, guiding the generator to generate images with higher-quality textures. Experimental results show the superiority of our proposed method in generating specific and realistic anime characters compared to existing methods.
Code
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
Code Syntology ran Syntology
Not run by Syntology. Nothing on this page verifies that the listed code works.
Tasks
1 archive task tag without a task page not shown.
Results from the paper archive 2025-07-28
No leaderboard rows for this paper in the archive.
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections