Reproduce “ReLift: Dense 3D Tracking as Lifted Trajectory Repair” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Assaf Singer, Maroun Khriesh, Noam Rotstein, Yoni Kasten, Or Litany

ReLift repairs and completes dense 3D point trajectories by lifting 2D tracks with estimated depth and camera motion, then correcting positions and filling gaps. On WorldTrack, ReLift with CoWTracker and 4RC geometry achieves mean APD3D of 89.01% on all points and 86.95% on dynamic points. Its performance remains sensitive to input geometry, and training does not simulate camera-pose errors.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.