Evaluating the Label Efficiency of Contrastive Self-Supervised Learning for Multi-Resolution Satellite Imagery

TL;DR

This paper investigates the effectiveness of contrastive self-supervised learning methods in improving land use classification accuracy on multi-resolution satellite imagery with limited labeled data, demonstrating superior performance over traditional pretraining.

Contribution

It introduces a benchmark for contrastive self-supervised methods in remote sensing and shows their advantages over out-of-domain pretraining for multi-resolution land use classification.

Findings

Contrastive self-supervised methods perform well with limited labels.

They outperform out-of-domain pretraining approaches.

Pretraining on large-scale datasets improves transferability.

Abstract

The application of deep neural networks to remote sensing imagery is often constrained by the lack of ground-truth annotations. Adressing this issue requires models that generalize efficiently from limited amounts of labeled data, allowing us to tackle a wider range of Earth observation tasks. Another challenge in this domain is developing algorithms that operate at variable spatial resolutions, e.g., for the problem of classifying land use at different scales. Recently, self-supervised learning has been applied in the remote sensing domain to exploit readily-available unlabeled data, and was shown to reduce or even close the gap with supervised learning. In this paper, we study self-supervised visual representation learning through the lens of label efficiency, for the task of land use classification on multi-resolution/multi-scale satellite images. We benchmark two contrastive self-supervised methods adapted from Momentum Contrast (MoCo) and provide evidence that these methods can be perform effectively given little downstream supervision, where randomly initialized networks fail to generalize. Moreover, they outperform out-of-domain pretraining alternatives. We use the large-scale fMoW dataset to pretrain and evaluate the networks, and validate our observations with transfer to the RESISC45 dataset.