Reproduce “Prepare Lighter and Faster for Open-Vocabulary Queries: A Query-Wise 3D Segmenter for 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

Nhat-Quynh Le-Pham, Khoi Nguyen, Binh-Son Hua

The authors present a method for selecting queried objects in reconstructed 3D Gaussian scenes without first learning scene-wide semantic features. The system localizes a text-relevant image region, tracks it across selected views, and combines visibility-conditioned evidence into a 3D mask. On LERF-OVS, it reaches 63.02 mean intersection-over-union versus LUDVIG’s 50.58, with time-to-first-query of 1.5 minutes versus 10 minutes.

Summary by alphaXiv.

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