Siyu Tang

Multi-stereo reconstruction outperforms feed-forward models by achieving 67% coverage of ground-truth surgical surfaces with just three viewpoints, highlighting the critical role of viewpoint diversity in surgical perception.

ADAPT enables humanoid robots to seamlessly execute diverse motion skills in real-time based on text commands, revolutionizing interactive control.

ReViV reconstructs 4D viewer and view dynamics from a single monocular video, achieving unprecedented accuracy and speed without heavy task-specific priors.