Reproduce “PointTransformerX: Portable and Efficient 3D Point Cloud Processing without Sparse Algorithms” 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
Laurenz Reichardt, Nikolas Ebert, Oliver Wasenmüller
PointTransformerX (PTX) replaces sparse spatial operators with a PyTorch-native point-cloud transformer that encodes 3D position in attention and scales attention windows at inference. On ScanNet semantic segmentation, it reports 0.765 mIoU, compared with PointTransformer V3’s 0.775, using fewer parameters, less peak memory, and less H100 inference time. Portability is demonstrated across devices, but sorting and pooling remain bottlenecks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.