Intelligent Fire Recognition Algorithm Based on Yolov7 Algorithm

Fire detection is important to ensure the safety of property and life of enterprises and people. Firstly, the similarity between fire and light features can lead to the phenomenon of false detection. Secondly, small target class recognition like small fire is not well recognized in most models. Finally, the recognition and inference speed of fire needs to be faster. To address the above problems, this paper proposes a fire recognition method based on Yolo framework. Firstly, an attention module SE is added to the backbone layer of the model to solve the problem that the fire is easily confused with the light. Then the CotNet Transformer module is added to the original network module to optimize the ResNet structure, strengthen the feature extraction ability of the model, and improve the recognition accuracy of the small fire. Finally, the loss function SIOU is used to accelerate the model’s fire identification and convergence speed, and the accuracy and reasoning speed of the model’s fire identification are improved. Experimental results show that this method can effectively improve the accuracy of fire detection.