Leveraging Domain Adaptation for Low-Resource Geospatial Machine Learning

TL;DR

This paper explores how domain adaptation techniques can address the challenge of limited labeled data in geospatial machine learning, especially across diverse regions and conditions, by testing on multiple benchmarks.

Contribution

It investigates the application of modern domain adaptation methods to geospatial benchmarks, identifying unique challenges and proposing tailored solutions.

Findings

Domain adaptation improves model performance on diverse geospatial datasets

Identified specific challenges in applying domain adaptation to remote sensing data

Proposed solutions enhance generalization across different regions and conditions

Abstract

Machine learning in remote sensing has matured alongside a proliferation in availability and resolution of geospatial imagery, but its utility is bottlenecked by the need for labeled data. What's more, many labeled geospatial datasets are specific to certain regions, instruments, or extreme weather events. We investigate the application of modern domain-adaptation to multiple proposed geospatial benchmarks, uncovering unique challenges and proposing solutions to them.