Real-time Detection of Smoking Behavior Based on Improved YOLOv5s Model

Real time detection of smoking behavior is very important for the safety of non-smoking places. For real-time detection of smoking behavior in different environments, a smoking behavior detection model based on the improved YOLOv5s algorithm is proposed. The improved algorithm is used to analyze and detect smoking behavior frame by frame on the images obtained by the camera. Firstly, the Transformer is introduced into the deep learning network to enhance the network's ability to detect multi-scale objects; Secondly, the CBAM attention mechanism is introduced to make the model more focused on the content and location information of the cigarette target area. At the same time, a Coordinate Attention module is added to enrich the feature map information extracted by the network and enhance the ability to express feature map information; Finally, improve the multi-scale detection head by adding smaller detection layers suitable for cigarettes, and enhance the model's ability to detect cigarettes. The experimental results show that the improved algorithm has improved accuracy, recall, and average accuracy compared to the original model. The actual detection performance in multi-target and small target scenarios is significantly improved compared to the original model. At the same time, the detection speed meets the real-time requirements. The improved model can be better applied to the real-time detection task of smoking behavior.