Reproduce “METRIC-GS: Selective Metric-Scale Estimation for 3D 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
Sooyeun Yang, Jongseong Brad Choi
METRIC-GS estimates metric scale for indoor 3D Gaussian Splatting reconstructions from semantic object and architectural size cues, using visibility-filtered votes and abstaining when cues disagree. On 47 ScanNet++ scenes, it reduces mean absolute relative scale error from 21.1% to 16.6% versus a quantile heuristic. Routing refusals to MASt3R gives 9.4% mean error with 60% fewer invocations, within benchmark scope.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.