Application of YOLOv5 in Device Detection of Hydropower Station

yEfficient detection and classification of devices are very important for effective maintenance in hydropower stations. However, the conventional manual approach suffers from low accuracy, efficiency, and reliability. Nowadays, object detection based on deep learning has achieved great development. In particular, the YOLO series models demonstrate significant advantages in terms of high accuracy and speed and robustness in complex image backgrounds, which have been widely used in numerous contexts for multi-object recognition tasks. Thus, in this paper, a YOLOv5-based algorithm is proposed to realize real-time multi-device recognition in hydropower stations. We labelled about 600 device photos of 27 categories collected from the site, based on which comparative experiments were carried out to evaluate the performance of YOLOv5 models of different configurations. The experimental results show that the mAP of YOLOv5m outperforms the others, the mAP of YOLOv5m can reach 95.3% and its precision can reach 94.1%. The study tests and reveals the effectiveness of YOLOv5 for device detection in hydropower stations. As such, the study on the one hand provides a useful asset management tool for the maintenance team, for another provides valuable empirical data for other studies that apply deep learning models in similar industrial situations.