DeFCN-nano: An End-to-End Real-Time Object Detection for Insulator Defects

Abstract

It is important to achieve real-time and accurate detection of the insulator defects via the unmanned aerial vehicles (UAVs) so as to improve the inspection efficiency of the large-scale power grids. This paper studies the You Only Look Once version 8 (YOLOv8) and DeFCN object detection algorithms based on the deep learning techniques, then an end-to-end real-time insulator defect detection method is proposed based on the DeFCN. The DeFCN model is lightweighted referring to the design of YOLOv8-n, which can balance the accuracy and real-time performance in the model structure. The proposed DeFCN-nano is validated on an open-source dataset and the experimental results demonstrate that the mean Average Precision (mAP) of the insulator detection is 97.51%, the mAP for detecting defects is 99.26% and the overall mAP is 98.39%. Compared with the baseline models, the proposed model has higher detection speed with a real-time detection speed of 58 frames per second.