How about generate texture by encode-decode network

[zhangqijun]

Due to self-occlusion texture is always bad。Can we train a model direct generate texture？Some Invisible part should be better。 like this paper UV-GAN: Adversarial Facial UV Map Completion for Pose-invariant Face Recognition

[Wjiajie]

Have you tried UV-GAN? How is the effect?

[hanyanghong86]

Hi，have you found the reference paper of the method generate uv position map?