Reproduce “CVTSplat: Multi-view Reconstruction Using an Adaptive Centroidal Voronoi Tessellation” 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

Briac Toussaint, Diego Thomas, Jean-Sébastien Franco, Edmond Boyer, Christian Theobalt

CVTSplat reconstructs objects from multiple views using a surface-adaptive Voronoi representation. It biases cell sites toward signed-distance-field-dependent targets, producing smaller cells near the surface and larger cells farther away. On DTU, it reports a mean Chamfer distance of 0.48 mm across 15 scenes and 2.5-minute training on an RTX 5090; it is not best on every benchmark, and runtimes are not hardware-normalized.

Summary by alphaXiv.

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