Reproduce “SVNet: Where SO(3) Equivariance Meets Binarization on Point Cloud Representation” 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
Zhuo Su, Max Welling, Matti Pietikäinen, Li Liu
SVNet combines rotation-invariant scalar features with rotation-equivariant vector features, using binary operations for much of the network. On ModelNet40, binary SVNet-DGCNN reaches 83.8% accuracy when trained with random z-axis rotations and tested on arbitrary 3D rotations, compared with 33.8% for DGCNN. Parameter storage falls from 57.7M to 3.4M bits; these are benchmark and accounting results, not measured inference speed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.