Reproduce “Three LiDARs, Twenty-Five Datasets: Cross-Source Generalization in Tree-Species Recognition” 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

Jae Joong Lee, Bedrich Benes

The authors develop a multi-resolution LiDAR classifier and test whether it transfers across data sources. They withhold source–species pairings while keeping each scored species represented in training. Across eight species and 5,566 trees, unweighted species-mean accuracy falls from 82% to 36%, while tree-weighted accuracy falls from 95% to 60%. This evaluates supported cross-source cases, not species absent from training.

Summary by alphaXiv.

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