With the development of artificial intelligence, railway enterprises hope to add intelligent detection functions to their detection equipment, realizing the intelligence of railway detection to solve problems such as low efficiency in traditional manual detection methods. However, due to limitations in the performance of mobile devices, it is difficult to deploy network models with large amounts of computation and weight parameters on mobile devices. Based on the YOLOv5 network model, this article proposes the YOLOv5-SC model for railway catenary dropper location detection, which has less computation and smaller network model size while still maintaining good detection accuracy. Through experimental comparison, the lightweighted network model has a small decrease in detection accuracy, and YOLOv5-SC network model size and computation are half that of the original YOLOv5 network model, with an improved detection speed by14.5 % .