Flame smoke detection method based on the improved YOLOv5s

To solve the problems of untimely traditional flame smoke detection， difficult pyrotechnic detection and low detection accuracy of small targets， an improved flame smoke detection method of YOLOv5s was proposed. Firstly， the SE attention mechanism was introduced into the backbone layer of the YOLOv5s model， which can adaptively adjust the feature weight of each channel， enhance the useful features and suppresses the useless features， improve the network’s ability to extract the features of flame and smoke. Secondly， BiFPN module was introduced as a feature fusion module in the Neck layer of the YOLOv5s model， and the bidirectional connection was introduced through BiFPN module， which can make full use of the feature information of different levels and improve the richness of features by combining the bottom‑up and top‑down feature fusion methods. Finally， the improved YOLOv5s model was applied to the actual flame smoke dataset， and the experimental results showed that the accuracy， recall rate and mAP value of the improved YOLOv5s model were increased by 1.8%， 2.6% and 1.5%， respectively， which can meet the accuracy requirements of flame smoke detection.