Fire Prevention and Control Monitoring System based on Improved YOLOv8

In this paper, a lightweight CBAM_YOLOv8 algorithm is proposed to address the problems of low accuracy, poor detection of small targets, large model volume and computational complexity of existing flame detection methods. By introducing the CBAM attention mechanism in the network backbone and feature extraction module, the detection accuracy is improved and the model volume and computation are reduced. Experimental results show that the improved algorithm achieves a mean accuracy (mAP) of 95.1%, which is 1.6 percentage points higher than that of the original YOLOv8, with a detection speed of 51FPS, and a model size of 6.5M, which is 16.1M less than the original model, and is suitable for real-time deployment in mobile devices.