Gated Path Aggregation Feature Pyramid Network for Object Detection in Remote Sensing Images

Object detection in remote sensing images is a challenge because remote sensing targets have characteristics such as small geometries, an unfixed direction and multiple poses. Recent studies have shown that the accuracy of object detection can be improved using feature fusion. However, direct fusion methods regard each layer as being of equal importance and rarely consider the hierarchical structure of multiple convolutional layers, leading to redundancy and rejected information being rarely applied during the fusion process. To address these issues, we propose a gated path aggregate (GPA) network that integrates path enhancement and information filtering into an end-to-end integrated network. Specifically, we first quantitatively analyze the performance of different gating functions to select the most suitable gating function. Then, we explore the embedding of soft switchable atrous convolution (SSAC) in the topmost feature layer. Finally, we validate our proposed model by combining it with experiments using the public NWPU VHR-10 dataset. The experimental results show that our proposed GPAFPN structure has significant improvement compared to the FPN structure. Compared with the mainstream networks, it achieved state-of-the-art performance.