A Lightweight Object Detection Method in Aerial Images Based on Dense Feature Fusion Path Aggregation Network

In recent years, significant progress has been obtained in object detection using Convolutional Neural Networks (CNNs). However, owing to the particularity of Remote Sensing Images (RSIs), common methods are not well suited for RSIs. Aiming at difficulties RSIs, this paper proposes an method based on Dense Feature Fusion Path Aggregation Network (DFF-PANet). Firstly, better improving performance small and medium-sized instances, we propose Reuse Module (FRM), which can integrate semantic location information contained feature maps; module reuse maps backbone enhance capability instances. After that, design DFF-PANet, help extracted from be fused more efficiently, thus cope with problem external interference factors. We performed experiments Dataset Object deTection Aerial images (DOTA) dataset HRSC2016 dataset; accuracy reached 71.5% mAP, exceeds most detectors one-stage two-stages present. Meanwhile, size our model is only 9.2 M, satisfies requirement being lightweight. The experimental results demonstrate that but also maintains high efficiency