Lightweight object detection for UAV Images based on multidimensional attention fusion

Unmanned Aerial Vehicle (UAV) image object detection holds broad application prospects in environmental protection, remote sensing, intelligent transportation, and many other fields. However, drone-captured aerial images present challenges such as small targets, low resolution, and limited device power consumption, which pose significant demands on the precision of object detection and the lightweight nature of algorithms. This paper proposes an improved lightweight object detection method tailored for drone-captured aerial images. To better extract the features of small objects, a C2fCF module is introduced into the backbone component, and a C2fMDAF module is introduced into the neck component. Experimental results show that the lightweight object detection method proposed in this paper can significantly improve the accuracy of object detection while almost not affecting the detection efficiency.