Comparative Analysis of YOLO8 and YOLO5 Variants in Detecting Cracks on Solar Panels

Solar panels are highly efficient electricity generators, durable over long periods. However, the existence of surface defects on these panels can lead to a reduction in energy production, leading to an overall reduction in efficiency. In this context, the need to effectively detect defects in solar panels is very important in order to improve the overall production efficiency of Photovoltaic (PV) systems. This paper presents a comprehensive comparison between the YOLOv5 and YOLOv8 object detection models in various configurations, including small, large, nano, medium and xtralarge. The main objective is to evaluate the effectiveness of these models in detecting defects, specifically cracks. The assessment encompasses various aspects such as precision, recall, and overall computational efficiency. The results show that the YOLOv8 outperforms the YOLOv5 in terms of precision, recall, mean Average Precision (mAP) mAP_50 and mAP_95. It achieves 87, 90, 86 and 93%, respectively. However, the YOLOv5 achieves 83, 65, 54 and 72% for the corresponding measurements. Continuous improvement in detection methods is essential to minimize the impact of defects on solar panels, thus contributing to the overall reliability and durability of the solar energy as a clean and renewable power source.