WCSE

A Flame Recognition Method Based on Data Enhancement and Yolo-v5s

Abstract— To identify fires occurring in complex environments faster and locate the flames more accurately, the synthesized data enhancement methods are proposed, then a self-built flame dataset is constructed, and a flame recognition model based on Yolo-v5s is developed. The results show that after data enhancement, the false alarm rate decreases by 18.38% and the F1 score improves by 19.79%, which achieves a significant improvement in the quality of flame recognition.

Index Terms— Flame Recognition; Data Enhancement; Yolo-v5s; Flame Dataset; Data Augmentation

Chao Huang
Civil Aviation University of China, CHINA
Shangyuan Li
Civil Aviation University of China, CHINA
Zhangcheng Yin
Civil Aviation University of China, CHINA
Runxia Guo
Civil Aviation University of China, CHINA
Guihang Liu
Civil Aviation University of China, CHINA
Jiusheng Chen
Civil Aviation University of China, CHINA

[Download]

Cite: Chao Huang, Shangyuan Li, Zhangcheng Yin, Runxia Guo, Guihang Liu, Jiusheng Chen, "A Flame Recognition Method Based on Data Enhancement and Yolo-v5s," 2024 The 14th International Workshop on Computer Science and Engineering (WCSE 2024), pp. 42-49, Phuket Island, Thailand, June 19-21, 2024.