Substation Personnel Smoking Detection Based On GhostNetV2-YOLOv5

In substations, violations such as smoking may lead to equipment damage and even threaten the safety of personnel. In this paper, a GhostNetV2-YOLOv5 based algorithm for detecting smoking by substation personnel is proposed. The algorithm uses GhostNetV2 as the backbone network to improve the accuracy of the model while achieving real-time detection of smoking behavior. The experimental results show that the method can improve the detection speed of the model without reducing the detection accuracy, and compared with the original YOLOv5 algorithm, the total mAP is improved by 2.58% and the prediction speed is improved by 1.61 times, which is superior to other smoking detection networks and has practical application value.