Smoke detection often has low detection accuracy,high missed detection rate and false de-tection rate,in order to solve these problems,an improved YOLOv5s smoke detection model was pro-posed.Firstly,the C3＿PSA structure was added to the backbone of YOLOv5s to improve the feature ex-traction ability of the model in harsh environments and reduce the missed detection rate.Secondly,BiFu-sion structure was used to replace the neck structure of YOLOv5s model,so as to enhance the model's ability to detect details and location information,improve the model detection accuracy,and reduce the model false detection rate.Finally,the loss function is improved to further improve the accuracy of model detection.The experimental results show that the accuracy of the improved model is improved by 5.6%,the mAP is improved by 3.5%,and the FPS is 369 frames/second,which indicates that the improved model can accurately detect the smoke characteristics in the complex background environment,and at the same time meet the requirements of high real-time deployment of the model on the end side.
NETL
NSTL
万方数据
