Reproduce “MetricTrail: Revisiting the Power of Metric Depth and Local Optimization for Long-Sequence Streaming 3D Reconstruction” 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

Jingnan Gao, Jiahui Huang, Sven Elflein, Jianhao Zheng, Bingxin Ke, Xiaokang Yang, Yichao Yan, Laura Leal-Taixé and 3 more

MetricTrail uses single-image metric depth to initialize sparse patch depths and retain it as a soft constraint while a bounded local optimizer refines poses and depths; frozen DINOv3 features support matching. On RGB-only, uncalibrated Oxford Spires video, it reports average full-sequence ATE of 1.77 m versus 5.15 m for LingBot-MAP. ATE follows one Sim(3) alignment per sequence, not absolute scale accuracy.

Summary by alphaXiv.

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