Abstract: Insulator defects seriously affect the security of transmission lines,and effective identification of insulator loss in aerial images is an important link in the UAV line inspection.This paper proposes a lightweight network model for insulator loss detection,which replaces the backbone network part of YOLOv4 with a lightweight network MobileNetV3. Based on the segmentation performance and calculation speed, the performance of the YOLOv4 model and the model which replaces its trunk network with a lightweight network in the insulator loss detection are comprehensively analyzed and compared,and the test results show: the selected YoloV4-Mobilenet V3 lightweight network insulator loss detection model can accurately locate single and multiple target insulators in the image. The improved YOLOv4-MobileNetV3 detection model is78% smaller than the original model in volume,and the FPS is increased by 4.85,and the corresponding mAP is decreased by0.6%. The insulator loss detection method proposed in this study can meet the needs of the UAV power line inspection.