Research on fire detection based on YOLOv5

Abstract

Fire is a frequent and hugely harmful disaster accident in people's daily life, which will seriously threaten people's life and property safety. Therefore, fire detection is very important in safety inspection, and people can effectively reduce losses based on early warning. Traditional fire detection methods have obvious limitations, such as temperature detectors, smoke detectors, thermal cameras, etc. These devices are often expensive, vulnerable, inaccurate, and prone to excessive false positives. In recent years, human beings have entered a highly information age. Fire detection technology based on computer vision has begun to enter people's field of vision, especially the rapid development of deep learning provides a new direction for fire detection. In recent years, human beings have entered a highly information age. Fire detection and monitoring platforms based on UAV technology, 5G technology, ultra-broadband transmission technology, and high-speed Internet of Things technology are more and more widely used. Although this solves the problem of the data transmission rate, the detection performance of the terminal needs to be improved. Fire detection technology based on computer vision has begun to enter people's field of vision, especially the rapid development of deep learning provides a new direction for fire detection. This paper first analyzes the advantages and disadvantages of four commonly used deep learning-based target detection algorithms, then uses the latest YOLOv5 target detection algorithm for experiments, and finally uses PyQT5 to make a fire detection system demo. The experimental results show that fire detection based on YOLOv5 can achieve better detection accuracy.