Self-supervised Domain-agnostic Domain Adaptation for Satellite Images

TL;DR

This paper introduces SS(DA)2, a self-supervised, domain-agnostic method for satellite image domain adaptation that enhances model generalization across diverse data sources without explicit domain labels.

Contribution

The paper proposes a novel self-supervised domain adaptation approach that does not require domain definitions, using contrastive generative adversarial training for image translation and data augmentation.

Findings

Effective on public benchmarks

Improves model generalization across domains

No need for explicit domain labels

Abstract

Domain shift caused by, e.g., different geographical regions or acquisition conditions is a common issue in machine learning for global scale satellite image processing. A promising method to address this problem is domain adaptation, where the training and the testing datasets are split into two or multiple domains according to their distributions, and an adaptation method is applied to improve the generalizability of the model on the testing dataset. However, defining the domain to which each satellite image belongs is not trivial, especially under large-scale multi-temporal and multi-sensory scenarios, where a single image mosaic could be generated from multiple data sources. In this paper, we propose an self-supervised domain-agnostic domain adaptation (SS(DA)2) method to perform domain adaptation without such a domain definition. To achieve this, we first design a contrastive generative adversarial loss to train a generative network to perform image-to-image translation between any two satellite image patches. Then, we improve the generalizability of the downstream models by augmenting the training data with different testing spectral characteristics. The experimental results on public benchmarks verify the effectiveness of SS(DA)2.