Leveraging Domain Adaptation for Low-Resource Geospatial Machine Learning

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.

arXiv commentTackling Climate Change with Machine Learning Workshop at ICML 2021

References in corpus (10)

• UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
• ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation
• Training Generative Adversarial Networks with Limited Data
• Barlow Twins: Self-Supervised Learning via Redundancy Reduction
• VisDA: The Visual Domain Adaptation Challenge
• SpaceNet: A Remote Sensing Dataset and Challenge Series
• Capsules for Object Segmentation
• Playing for Data: Ground Truth from Computer Games
• Building Damage Detection in Satellite Imagery Using Convolutional Neural Networks
• Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation