An Unmanned Aerial Vehicle Image Object Detection Algorithm Based on Improved YOLOv8

In the object detection task of unmanned aerial vehicle (UAV) images, the object distribution is dense and the size is small. It is challenging to extract feature information from the UAV images, resulting in low detection accuracy. To solve these problems, we proposed an improved UAV image object detection algorithm based on YOLOv8. Firstly, the P2 detection layer is introduced to the model, so as to enhance the model's perception of small objects and improve the detection performance. Then, the improved BiFPN is used to fuse the shallow and deep feature information through rich feature information branches. Finally, the RCS-OSA module is introduced to the model to enhance the feature learning ability and inference speed of the model through reparameterized convolution and one-shot aggregation. The experimental results on the VisDrone2019 dataset show that the improved YOLOv8 algorithm that we proposed has a 6.4% increase in mAP50, and mAP50-95 is also improved by 4%. The overall performance is pretty excellent which verifies the effectiveness of the improved methods proposed in this paper in UAV image object detection.