Abstract
Few-shot remote sensing scene classification (FSRSSC) is a task of categorizing remote sensing images (RSIs) with insufficient labeled samples. This task becomes particularly challenging due to low intra-class similarity and high inter-class similarity in RSIs. To overcome these challenges, we introduce a novel Masked Second-Order Pooling (MSoP) module to exploit the masked second-order features, enhancing feature representation and classification performance for FSRSSC. The MSoP module comprises two key components, a Learnable Dropblock (LDB) and a Compressed Second-Order Pooling (CSoP). The LDB selectively masks discriminative regions in feature maps, effectively alleviating the low intra-class similarity problem. On the other hand, the CSoP enhances second-order statistics by computing and aggregating channel-wise similarity, thereby reducing the impact of inter-class similarity. We construct an MSoP-Net based on the proposed MSoP module. Experiment results demonstrate its superior performance on two widely-used datasets, i.e., NWPU-RESISC45 and UCM.
Funding Information
  • Natural Science Foundation of Sichuan Province (2022NSFSC0957)
  • NSFC (62301455)

This publication has 15 references indexed in Scilit: