Yen-Siang Wu (James)

I'm currently a research assistant in the Vision Learning Lab at National Taiwan University, working with Prof. Yu-Chiang Frank Wang and Prof. Wei-Chiu Ma at Cornell University.

I received my Bachelor’s degree in computer science from National Taiwan University in 2024. Following that, I started working as a research assistant. My research interests include deep generative modeling, 3D computer vision, and other areas that enable machines to understand the world on their own.

Email  /  LinkedIn  /  CV   

Publications

MotionMatcher is a feature-level fine-tuning framework that achieves state-of-the-art performance in motion customization. It can customize pre-traind T2V diffusion models using the motion of the given reference video. Once customized, the diffusion model is able to transfer this motion to a variety of scenes.

VideoMage enables general-purpose concept customization for text-to-video diffusion models. Using the proposed sampling method, VideoMage can support subject customization, motion customization, and their combination without interference.