Self-Supervised Learning for Invariant Representations from Multi-Spectral and SAR Images

TL;DR

This paper introduces RSDnet, a self-supervised learning method using distillation networks for remote sensing images, demonstrating that single spectral bands and multi-modal data can produce highly effective invariant feature representations.

Contribution

It applies BYOL-based self-supervised learning to remote sensing data, exploring single-band and multi-modal training, and shows improved performance over supervised ImageNet models.

Findings

Single spectral bands achieve high F1 scores and mIoU.

RS SSL model outperforms supervised ImageNet models.

Multi-modal data enhances invariant feature learning.

Abstract

Self-Supervised learning (SSL) has become the new state-of-art in several domain classification and segmentation tasks. Of these, one popular category in SSL is distillation networks such as BYOL. This work proposes RSDnet, which applies the distillation network (BYOL) in the remote sensing (RS) domain where data is non-trivially different from natural RGB images. Since Multi-spectral (MS) and synthetic aperture radar (SAR) sensors provide varied spectral and spatial resolution information, we utilised them as an implicit augmentation to learn invariant feature embeddings. In order to learn RS based invariant features with SSL, we trained RSDnet in two ways, i.e., single channel feature learning and three channel feature learning. This work explores the usefulness of single channel feature learning from random MS and SAR bands compared to the common notion of using three or more bands. In our linear evaluation, these single channel features reached a 0.92 F1 score on the EuroSAT classification task and 59.6 mIoU on the DFC segmentation task for certain single bands. We also compared our results with ImageNet weights and showed that the RS based SSL model outperforms the supervised ImageNet based model. We further explored the usefulness of multi-modal data compared to single modality data, and it is shown that utilising MS and SAR data learn better invariant representations than utilising only MS data.