Reproduce “Enhancing Glass Surface Reconstruction via Depth Prior for Robot Navigation” 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
Jiamin Zheng, Jingwen Yu, Guangcheng Chen, Hong Zhang
The authors recover metric depth for transparent glass by aligning a pretrained monocular depth prior to RGB-D measurements with patch-wise sampling and local scale-and-shift estimation, rather than a global fit. On 316 hard GlassRecon images, DA3 local alignment reduces AbsRel from 0.323 to 0.172 and raises δ<1.25 from 0.651 to 0.883, but can fail when priors or sensor readings misrepresent glass.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.