A Target Detection Algorithm for UAV Aerial Photography Based on Improved Yolov5

Unmanned Aerial Vehicle (UAV) has been widely used in various fields such as agriculture, environmental monitoring and urban planning. Aiming to improve the accuracy and robustness of target detection, this paper proposes an improved method based on Yolov5 model. by introducing an attention mechanism and fusing it with the C3 module in Yolov5s backbone network to form the C3_SE module. The experiment results show that on the official dataset VisDrone2019, the new improved model has better performance compared to the baseline model Yolov5, which can provide more accurate results for different applications.