Reproduce “Improving Zero-shot 3D Semantic Mapping in Offroad with Geometric Priors” 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
Eerik Alamikkotervo, Risto Ojala
The authors refine SAM 3 point labels by reweighting image-model scores with lidar-derived elevation and sphericity patterns learned from confident predictions. For points within 25 m and camera field of view, mIoU, the mean per-class intersection-over-union, rises from 28.1 to 34.4 on RELLIS-3D and 47.9 to 49.6 on WildScenes; accuracy rises too. Results cover evaluated points, not navigation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.