Reproduce “UniMetric: Universal Metric Scale Recovery for Visual Geometry Foundation 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
Lei Yang, Yuner Zhang, Ziyang Liu, Ziying Song, Kai WU, Yiyang Gao, Jiangtao Su, Yiming Li
UniMetric estimates a shared metric scale for video reconstructions by combining uncertain object-height intervals with measured object extents. It applies the factor to depth, point clouds, and camera motion. Across nine models and twelve subsets, it reduces depth error by 54.2%, Chamfer distance by 83.8%, and trajectory error by 65.6%, but cannot correct local distortions or drift.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.