Drone Small Target Detection Based on YOLOV8-RAWM

With the widespread use of drones, detecting small targets in aerial images captured by drones poses significant challenges. Issues such as small size and low resolution make these targets difficult to detect. In response to this, this paper proposes an improved object detection algorithm based on YOLOv8. Firstly, a lightweight convolution and Manhattan self-attention mechanism were introduced into the backbone network, along with a feature fusion module in the neck, and the loss function was optimized to enhance the model's performance and robustness in small target detection. Experiments show that the model achieved a 6.2% improvement in [email protected] and a 5% increase in recall on the VisDrone2019 dataset, demonstrating its effectiveness in detecting small targets in drone applications.