Reproduce “ASCoR: Asymmetric Semantic-Spatial Learning for Remote Sensing Domain Generalization” 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
Zitong Wang, Shuchang Lyu, Meng Li, Qiang Wang, Guangliang Cheng, Zhenwei Shi
ASCoR combines SAM3’s class-name-guided predictions with a dense spatial-detail branch and a learned router for spatially varying corrections, trained on labeled source data without target labels. Across six directed ISPRS transfers, it achieves 68.61% average mIoU (mean class-wise intersection-over-union over valid target pixels) versus EarthAdapter’s 65.92%, ranking first in four directions; results cover specified benchmarks, not arbitrary domains.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.