Human Motion Transfer: Techniques, Models and Applications
Toturial, International Conference on Neural Computing for Advanced Applications，NCAA, Guilin, China
This tutorial mainly introduces the theory, methods and algorithms of human motion transfer with different task settings, and their applications in virtual reality, teleconferencing and digital human. It includes the introduction of one-shot, few-shot, and video-specific motion transfer tasks. The method level mainly includes feature warping, appearance style injection using generative adversarial models and stable diffusion models, person-specific models using neural scene representation and neural radiance field and Gaussian splatting. Finally, a brief introduction is made to the current research topics and some achievements of the research group.