Reproduce “DIGS: Distribution-Informed Gaussian Splatting for Training-Free Open-Vocabulary 3D Segmentation” 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
Zhifan Zeng, Wenbo Xiao, Arcot Sowmya, Changming Sun
DIGS maps persistent 3D objects from multiple views without scene-specific semantic training. It matches class-agnostic SAM masks to rendered regions, retains label distributions on Gaussians, and chooses labels after rendering; language-based naming follows mapping. On LERF-OVS, under the OpenGaussian 3D-native protocol, DIGS with LLM and CLIP reaches 71.95% mean mIoU across four scenes, versus reproduced SFS (62.16%) and LaGa (60.55%).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.