SRAF-Net: A Scene-Relevant Anchor-Free Object Detection Network in Remote Sensing Images

Abstract

Object detection is a fundamental and important task in the analysis of <i>remote sensing images</i> (RSIs), and existing deep learning-based object detection models in this literature strongly rely on predefined anchor boxes and encounter redesigned difficulties related to anchors. In addition, they often ignore the scene-contextual information that objects are usually closely related to their surrounding scene. To deal with these problems, we propose an anchor-free network, referred to as <i>scene-relevant anchor-free network</i> (SRAF-Net), for object detection in RSIs. The SRAF-Net first captures the scene-contextual features of objects by using a designed <i>scene-enhanced feature pyramid network</i> (SE-FPN) and then performs more accurate detection by implementing a <i>scene auxiliary detection head</i> (SADH), which can predict the existence of the objects with the help of the scene-contextual features extracted from the SE-FPN. To deal with insufficient scene diversity in the training stage, a simple yet effective data augmentation module, termed <i>balanced mixup data augment</i> (BMDA), is introduced by linearly expanding the training dataset to improve the generalization of SRAF-Net. Comprehensive experiments on three publicly available challenging remote sensing datasets demonstrate the effectiveness of the proposed method. The codes will be made publicly available at <uri>https://github.com/Complicateddd/SRAF-Net</uri>.