Reproduce “RelateGS: Reliable Semantic and Relational Evidence for 3D Scene Understanding with Gaussian Splatting” 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

Seongjun Choi, Sangyoun Lee

RelateGS improves language-based object localization in reconstructed indoor scenes by filtering weak multi-view semantic evidence and using a persistent object graph to select existing camera views for missing or low-confidence relations. It retains GaussianGraph’s downstream matcher. On Sr3D+, [email protected] rises from 7.4% to 15.3%, though gains are not uniform across metrics and splits.

Summary by alphaXiv.

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