Reproduce “VLSplat: Vision-Language Guided Object-Centric 3D Gaussian Splatting via Scene Graph” 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
Yuntae Jeon, Younho Jeon, Sujin Jin, Sungho Jo, Seunghee Park
VLSplat refines indoor 3D Gaussian splat labels by treating coarse object groups as candidates and using scene-graph context and language-guided checks to test uncertain and nearby splats. It changes semantic features rather than Gaussian geometry. On ScanNet++ rendered Thing segmentation, it reaches 69.81 mIoU versus 64.42 for a matched baseline, with refinement targeting six indoor categories rather than all labels.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.