Detection of Insulator Burst Position of Lightweight YOLOv5
Shiyi Huang, Xiaojie Dong, Yifan Wang, Longhuan Yang
Abstract
Insulator is a key insulating device. With the development of artificial intelligence technology, traditional manual inspection has been gradually replaced by image inspection technology. Therefore, this paper proposes an improved algorithm based on YOLOv5 for detecting the burst position of lightweight insulators. Firstly, redundant convolutional layers are deleted and channels are clipped to make the model lightweight and occupy less memory. Secondly, adaptive attention modules are added between adjacent residual modules of the network to make the key feature targets gain more weight and enhance the learning feature capability of the network. Experimental results show that compared with the original model, the Map of the improved model is reduced by 0.52%, but the size of model memory is reduced by 69.4%, and the time of each round of training is increased by 20.8%.