DBA-RMCL: Refined Metric Contrastive Learning With Dual-Branch Attention for Few-Shot Remote Sensing Image Scene Classification

Remote sensing scene classification plays a vital role in identifying Earth’s surface features. However, traditional methods necessitate vast quantities of labeled training data, which significantly limits their applicability when labeled samples are scarce. To overcome this limitation, remote sensing few-shot classification (RSFSC) methods have arisen as a promising approach to maintaining classification accuracy with minimal data. In this article, we present a refined metric contrastive learning framework with dual-branch attention (DBA-RMCL), a novel framework that tackles two fundamental challenges in RSFSC: intraclass diversity and interclass similarity. Our framework employs a dual-branch architecture consisting of way and shot branches to reduce intraclass variance while enhancing feature consistency and interclass separability. The architecture incorporates feature fusion between branches to optimize feature learning and classification performance. We further enhance feature representations through our refined Siamese contrastive learning representation (RSCLR) module, which implements self-supervised contrastive learning. The framework also includes a combined attention-based prototypical head (CAP-Head) module with a compositional attention computation unit to model complex relationships between query sets and prototypes. Experimental results on six publicly available datasets demonstrate that DBA-RMCL achieves state-of-the-art performance compared to existing FSC methods. For 5-way 1-shot tasks, our framework attains accuracies of 77.15% $\pm ~0.34$ % on NWPU-RESISC45 and 87.33% $\pm ~0.22$ % on WHU-RS19. Additionally, we established the first benchmark for few-shot scene classification, MEET-FSL, using the MEET dataset. These outcomes not only underscore the efficacy of our dual-branch architecture but also represent a significant advancement in the field of few-shot learning for remote sensing classification.