A Comparative Study of YOLO Architectures for Poles Detection on Agricultural Land UAVs Images

Abstract

Using unmanned aerial vehicles (UAVs) in modern agriculture faces with the problem of overcoming obstacles. One of the most widespread type of these obstacles on the agricultural land is powerlines and aerial communication lines, which are crucial for agriculture itself as well as for other economic activities, and thus could not be removed from the agricultural lands. The initial subtask in overcoming poles is detection of such objects. Recent neural network detection architectures such as YOLO has shown promising results in general object detection task, however, the results of comparative studies of YOLO architectures in a specific task of poles detection are not presented in scientific literature. In this work, we present results of comparative study of a set of YOLO architectures performance on a custom dataset of powerlines and aerial communication lines poles on the agricultural land obtained using UAV. The dataset consists of 3508 images with 1691 wooden poles and 1750 concrete poles. We consider five recent YOLO architectures from v8 to v12. Comparative analysis of the considering architectures has shown that YOLOv11 achieved the best performance in average according to recall (0.765), precision (0.798), mAP@50 (0.809) and mAP@50-90 (0.484) metrics. This results along with the least required computational resources (6.5 GFLOPS) makes YOLOv11 the most appropriate architecture for poles detection on the agricultural land.