Reproduce “ComplementMe: Weakly-Supervised Component Suggestions for 3D Modeling” 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

Minhyuk Sung, Hao Su, Vladimir G. Kim, Siddhartha Chaudhuri, Leonidas Guibas

ComplementMe learns to propose complementary parts from partial assemblies without consistent semantic part labels, using existing componentized models. It maps each partial assembly to a mixture of Gaussians over learned component embeddings, then predicts translation-only placements. On six-category functional retrieval, mean top-five average precision is 0.81, versus 0.79 for fixed MVCNN and 0.55 for CK10; gains are modest and uneven.

Summary by alphaXiv.

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