Reproduce “OpenEarthMap: A Benchmark Dataset for Global High-Resolution Land Cover Mapping” 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
Junshi Xia, Naoto Yokoya, Bruno Adriano, Clifford Broni-Bediako
The authors introduce OpenEarthMap, a labeled benchmark of 5,000 images at 0.25–0.5 m resolution, spanning 97 regions in 44 countries across six continents. It provides eight land-cover classes and 2.2 million connected segments. On six shared classes in a Chesapeake Bay test, a U-Net trained on it achieves 56.91% mean intersection over union versus 46.04% for LoveDA, supporting transfer in this area.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.