Fire Detection System Based On YOLOv5

Fire is a disaster with great harm in real life, and traditional fire detection methods have many shortcomings. In recent years, the development of computer vision and deep learning has provided new solutions for fire detection. This paper proposes a fire monitoring method based on the YOLOv5 target detection algorithm, which uses monitoring equipment to achieve real-time fire detection. The research content includes the construction of self-built data sets, YOLOv5 model training and optimization, detection system design and performance verification. By embedding the CBAM module in the C3 layer of the YOLOv5s backbone network, the channel and spatial processing of the feature map is performed, the feature refinement is achieved, and the construction and visualization of the detection system are completed. The experimental results show that the detection accuracy of flame targets is 0.745, and the detection accuracy of smoke targets is 0.582; on the test set, the average detection accuracy of flames and smoke is 0.823 and 0.673 respectively, and the overall average detection rate is 0.748. The fire detection system based on YOLOv5 implemented in this work has a significant effect on the current fire prevention and control, and it meets the actual application scenarios to the greatest extent.