Fire Detection Based on Improved YOLOv8 Network Mode

In the traditional field of fire detection, a significant amount of specialized equipment and highly trained professionals are typically required. However, the operational efficiency of these methods often tends to be low, and they are susceptible to false alarms, making it challenging to effectively meet the urgent demands of routine fire safety inspections. Additionally, current fire detection technologies based on neural network image recognition also face challenges in terms of realtime performance and accuracy. To enhance the performance of fire detection, this study proposes a fire detection algorithm based on improved YOLOv8, termed FD-YOLOv8 (Fire detection- YOLOv8). On a self-built fire dataset, Fire_dataset, the mean Average Precision (mAP) at intersection over union (IoU) thresholds of 0.5 and 0.5 to 0.95 are 53.6% and 30.7%, respectively.