Improved Strategy of Lightweight Fire Early Detection Algorithm Based on GBM-YOLO

DOI

10.2991/978-94-6463-780-9_40How to use a DOI?

Keywords

fire detection; YOLOv8; Lightweight improvement; Small object detection

Abstract

Fire causes serious damage to buildings and infrastructure, threatening social security. Early fire feature detection can effectively reduce the harm caused by fire. However, traditional fire detection methods have problems such as difficult early detection, high false alarm rate, high missed detection rate and high maintenance cost. In order to solve the above problems, this paper proposes a lightweight fire detection model GBM-YOLO based on YOLOv8n. By designing a lightweight feature extraction network G-HGNetV2 to replace the backbone network of YOLOv8, the model complexity is reduced and the feature extraction ability is improved. A lightweight cross-scale adaptive weighted fusion module is designed in the neck network, and the DySample dynamic upsampling module is introduced to effectively fuse shallow and deep features to retain more feature details. The lightweight C2f-MPF module is designed to improve the feature extraction ability of the model for small fire targets and multi-scale targets. The results show that the average accuracy (mAP@50) of the improved model on the self-built fire fireworks data set reaches 86.1%, and the number of parameters (Params), computational complexity (GFLOPs) and model Size (Size) are reduced by 53.3%, 27.1% and 49.1%, respectively. The effect of the improved model in fireworks detection is improved. And the model is more lightweight and easier to deploy to edge devices to provide technical support for early warning of fire.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Shuogui Zeng
AU  - Ye Chen
AU  - Jun Li
AU  - Tao Liu
AU  - Fuwei He
AU  - Zhenxi Liu
PY  - 2025
DA  - 2025/07/03
TI  - Improved Strategy of Lightweight Fire Early Detection Algorithm Based on GBM-YOLO
BT  - Proceedings of the 2025 International Conference on Engineering Management and Safety Engineering (EMSE 2025)
PB  - Atlantis Press
SP  - 440
EP  - 451
SN  - 2352-5401
UR  - https://doi.org/10.2991/978-94-6463-780-9_40
DO  - 10.2991/978-94-6463-780-9_40
ID  - Zeng2025
ER  -