Enhanced YOLOv8-Based Lightweight Algorithm for Flame Detection

Shichen Duan, Jun Zhou
Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, PMLR 278:144-154, 2025.

Abstract

This paper presents an enhanced real-time lightweight fire flame detection algorithm based on improved YOLOv8n. To address the challenges of flame detection in complex scenarios, the algorithm integrates the RVB Block and EMA attention mechanism into the YOLOv8n backbone, enhancing its ability to capture fire features accurately. Additionally, a lightweight Slim-Neck structure is introduced to reduce computational complexity and parameters, facilitating embedded deployment. The proposed WiseIoU loss function further accelerates convergence and optimizes bounding box loss. Experiments demonstrate that the improved algorithm achieves a precision rate of 97.7%, a mAP@50 of 98% and a recall rate of 94.4%, with a 16% reduction in parameters and a 1.7 reduction in GFLOPs. The algorithm’s lightweight nature and high accuracy provide strong technical support for early fire warning and control.

Cite this Paper

BibTeX
@InProceedings{pmlr-v278-duan25a, title = {Enhanced YOLOv8-Based Lightweight Algorithm for Flame Detection}, author = {Duan, Shichen and Zhou, Jun}, booktitle = {Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing}, pages = {144--154}, year = {2025}, editor = {Zeng, Nianyin and Pachori, Ram Bilas and Wang, Dongshu}, volume = {278}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v278/main/assets/duan25a/duan25a.pdf}, url = {https://proceedings.mlr.press/v278/duan25a.html}, abstract = {This paper presents an enhanced real-time lightweight fire flame detection algorithm based on improved YOLOv8n. To address the challenges of flame detection in complex scenarios, the algorithm integrates the RVB Block and EMA attention mechanism into the YOLOv8n backbone, enhancing its ability to capture fire features accurately. Additionally, a lightweight Slim-Neck structure is introduced to reduce computational complexity and parameters, facilitating embedded deployment. The proposed WiseIoU loss function further accelerates convergence and optimizes bounding box loss. Experiments demonstrate that the improved algorithm achieves a precision rate of 97.7%, a mAP@50 of 98% and a recall rate of 94.4%, with a 16% reduction in parameters and a 1.7 reduction in GFLOPs. The algorithm’s lightweight nature and high accuracy provide strong technical support for early fire warning and control.} }
Endnote
%0 Conference Paper %T Enhanced YOLOv8-Based Lightweight Algorithm for Flame Detection %A Shichen Duan %A Jun Zhou %B Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing %C Proceedings of Machine Learning Research %D 2025 %E Nianyin Zeng %E Ram Bilas Pachori %E Dongshu Wang %F pmlr-v278-duan25a %I PMLR %P 144--154 %U https://proceedings.mlr.press/v278/duan25a.html %V 278 %X This paper presents an enhanced real-time lightweight fire flame detection algorithm based on improved YOLOv8n. To address the challenges of flame detection in complex scenarios, the algorithm integrates the RVB Block and EMA attention mechanism into the YOLOv8n backbone, enhancing its ability to capture fire features accurately. Additionally, a lightweight Slim-Neck structure is introduced to reduce computational complexity and parameters, facilitating embedded deployment. The proposed WiseIoU loss function further accelerates convergence and optimizes bounding box loss. Experiments demonstrate that the improved algorithm achieves a precision rate of 97.7%, a mAP@50 of 98% and a recall rate of 94.4%, with a 16% reduction in parameters and a 1.7 reduction in GFLOPs. The algorithm’s lightweight nature and high accuracy provide strong technical support for early fire warning and control.
APA
Duan, S. & Zhou, J.. (2025). Enhanced YOLOv8-Based Lightweight Algorithm for Flame Detection. Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, in Proceedings of Machine Learning Research 278:144-154 Available from https://proceedings.mlr.press/v278/duan25a.html.

Related Material