YOLOv5s-Transformer: Improved YOLOv5 Network for Real-Time Detection of Cigarette Smoking Based on Image processing

Abstract

Cigarette smoking is a significant cause of fires, resulting in increasingly devastating effects. However, the current detection methods fail to meet the real-time smoking detection requirements. The main contribution in this paper is to address the problems of low detection accuracy and inaccurate positioning of smoking detection. First, for the real-time detection speed, this paper proposes a lightweight smoking detection model based on YOLOv5s. Second, for better smoking detection accuracy, this paper uses image processing technology and combines C3TR block into the network architecture. Finally the proposed model, named YOLOv5s-Transformer, is deployed into edge computing platform NVIDIA Jeston Nano. The mean average precision of the proposed model is 18.3% higher than the baseline YOLOv5s, thereby achieving an improved balance between detection speed and accuracy.