Grasping the Future: 3D Hand-Object Motion Prediction from Web Videos

Researchers from UW, UNC and JHU released and open-sourced 'Grasping the Future,' a 3D world model that predicts joint hand-object motion from just half a second of video, forecasting about 4.5 seconds of future movement.

2026-11-17 ~ 2026-11-17 · 3 related posts