Reproduce “SOVGaussian: Sparse-View 3D Gaussian Splatting for Open-Vocabulary Scene Understanding” 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
Peng Ling, Tiao Tan, Jiaqi Lin, Wenming Yang
SOVGaussian uses object masks to make monocular depth supervision locally meaningful, then prunes isolated 3D Gaussian primitives inconsistent with nearby elements sharing language features. On 3DOVS, using the same three training images and novel-view tests, it achieves 65.9% overall mean intersection-over-union, versus 42.0% for LangSplat. This comparison is limited to the tested three-view protocol and static scenes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.