Attention-based Insulator Defect Detection

With the rapid development of the electric system, the demand for auto-inspection of the electric transmission system, e.g., transmission lines, and insulators, is increasing. To achieve automatic insulator inspection, we propose a two-stage insulator defect inspection framework which consists of the attention-based insulator detection network and the defect inspection network. The attention-based insulator detection network is responsible for the positioning of the insulator and the defect inspection model determines whether the insulator is corrupted. The two-stage detect-then-inspect design avoids the influence of background and enables high-precision inspection. Experimental results on our collected dataset show that our model outperforms previous object detection method.