Reproduce “Point-LN: A Lightweight Framework for Efficient Point Cloud Classification Using Non-Parametric Positional Encoding” 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
Marzieh Mohammadi, Amir Salarpour, Pedram MohajerAnsari
Point-LN combines fixed geometric operations and positional encodings with learned linear layers and a classifier for 3D point-cloud classification. It samples points, groups local neighborhoods, and aggregates features across four stages. On ScanObjectNN, it reports 91.7% accuracy with 0.8 million parameters on PB-T50-RS, compared with Point-PN’s 87.1% at the same count; this establishes a benchmark comparison, not faster inference or causal attribution.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.