Research on Multi-target Detection of Lightweight Substation Based on YOLOv5

Mobile terminal devices such as unmanned aerial vehicles and patrol robots are gradually popularized and applied in domestic substations. The storage space of this kind of mobile terminal device is small, which limits the effective deployment of the depth model. Therefore, this paper proposes a research on multi-target detection of lightweight substation based on YOLOv5, pruning and compressing the BN layer of YOLOv5s model. Compared with the unpruned model, the compressed model reduces the size of the model by 70%, accelerates the computing time by 17.1%, and reduces the average accuracy by only 3%. It shows that the study is beneficial to the deployment of the model.