A Flame Detection Algorithm Based on Improved YOLOv7

Abstract

Flame recognition is of great significance in fire prevention. However, current algorithms for flame detection have some problems, such as missing detection and false detection, and the detection accuracy cannot satisfy the requirements for fire prevention. In order to further the above problems, we propose a flame detection algorithm based on an improved YOLOv7 network. In our algorithm, we replace a convolution of the MP-1 module with a SimAM structure, which is a parameter-free attention mechanism. In this way, the missing detection problem can be improved. Furthermore, we use a ConvNeXt-based CNeB module to replace a convolution of the ELAN-W module for increasing detection accuracy and the false detection problem in complex environments. Finally, we evaluate the performance of our algorithm through a large number of test cases, and the data set used in our experiments was constructed by combining several publicly available data sets for various application scenarios. The experimental results indicate that compared with the original YOLOv7 algorithm, our proposed algorithm can achieve a 7% increase in the aspect of mAP_0.5 and a 4.1% increase in the aspect of F1 score.

Keywords

• flame detection
• deep learning
• YOLO series
• convolutional neural networks
• target detection