SGMNet: Scene Graph Matching Network for Few-Shot Remote Sensing Scene Classification

TL;DR

This paper introduces SGMNet, a novel scene graph matching network that leverages object co-occurrence and spatial correlations in remote sensing images to improve few-shot scene classification accuracy.

Contribution

The paper proposes a scene graph matching-based meta-learning framework that explicitly models object co-occurrence and spatial relationships for better few-shot remote sensing scene classification.

Findings

SGMNet outperforms previous state-of-the-art methods on multiple datasets.

Explicit modeling of object relationships enhances classification performance.

The approach effectively addresses data scarcity in remote sensing scene classification.

Abstract

Few-Shot Remote Sensing Scene Classification (FSRSSC) is an important task, which aims to recognize novel scene classes with few examples. Recently, several studies attempt to address the FSRSSC problem by following few-shot natural image classification methods. These existing methods have made promising progress and achieved superior performance. However, they all overlook two unique characteristics of remote sensing images: (i) object co-occurrence that multiple objects tend to appear together in a scene image and (ii) object spatial correlation that these co-occurrence objects are distributed in the scene image following some spatial structure patterns. Such unique characteristics are very beneficial for FSRSSC, which can effectively alleviate the scarcity issue of labeled remote sensing images since they can provide more refined descriptions for each scene class. To fully exploit these characteristics, we propose a novel scene graph matching-based meta-learning framework for FSRSSC, called SGMNet. In this framework, a scene graph construction module is carefully designed to represent each test remote sensing image or each scene class as a scene graph, where the nodes reflect these co-occurrence objects meanwhile the edges capture the spatial correlations between these co-occurrence objects. Then, a scene graph matching module is further developed to evaluate the similarity score between each test remote sensing image and each scene class. Finally, based on the similarity scores, we perform the scene class prediction via a nearest neighbor classifier. We conduct extensive experiments on UCMerced LandUse, WHU19, AID, and NWPU-RESISC45 datasets. The experimental results show that our method obtains superior performance over the previous state-of-the-art methods.