Reproduce “AnchorComplete: Recovering Complete Scene Meshes from Partial Point Clouds” 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

Weirong Chen, Ismail El-Yamany, Ganlin Zhang, Chuanxia Zheng, Andrea Vedaldi, Daniel Cremers

AnchorComplete completes partial indoor point clouds into meshes by using observed geometry to guide a generative model. Global anchor tokens provide scene context, while local spatial evidence preserves observed geometry; coarse-to-fine refinement extends the layout into unseen regions. On four-view Replica scenes, it reports an F-score of 0.684, versus Seen2Scene’s 0.647, with precision evaluated only inside input camera frustums.

Summary by alphaXiv.

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