Reproduce “3D Point Tracking with State Space Models” 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
Masahiro Ogawa, Qi An, Atsushi Yamashita
The authors present a pose-free 3D tracker that combines optical flow with monocular metric depth, then learns corrections along tracked pixels’ rays. Its temporal refiner processes whole clips, not online. On TAPVid-3D minival, its best DA3-l system scores mean absolute metric-AJ 0.256 versus 0.183 for DELTA plus DA3-l, but DELTA leads on Aria Digital Twin and the normalized leaderboard.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.