Optimizing UAV Detection Performance with YOLOv5 Series Algorithms

---

Abstract

This paper provides a comprehensive evaluation of the YOLOv5 series, including YOLOv5n, YOLOv5s, YOLOv5m, and YOLOv5l, with a specific focus on their effectiveness in drone detection applications. By analyzing key performance metrics such as precision, recall, mAP50, mAP90, and F1-score, the study identifies YOLOv5s as the standout performer, achieving the highest precision (98.75%) and mAP50 (97.8%). These results highlight its exceptional object detection capabilities. The study also examines inference speed and computational requirements, revealing the trade-offs between accuracy and computational efficiency when selecting YOLOv5 models for drone detection tasks. Despite these tradeoffs, YOLOv5s proves to be a promising candidate, offering a well-balanced combination of precision, recall, and computational efficiency. This balance makes YOLOv5s particularly well-suited for practical deployment in real-world drone detection systems, where both accuracy and operational efficiency are critical. Overall, the findings of this study provide valuable insights for researchers and practitioners aiming to optimize drone detection capabilities using YOLOv5 models.

---

Keywords

UAV Detection, Drone Detection, YOLOv5, Deep Learning, Object Detection, Autonomous Drone, Computer Vison