1 paper · 1 filter
Zhuoqian Feng, Weixing Chen, Ziliang Chen +2
Feed-forward models regress dense 3D point trajectories directly from monocular video, yet the residual after global alignment is substantial and lacks a structural explanation. Me…