DSS-YOLO: an improved lightweight real-time fire detection model based on YOLOv8 DOI Creative Commons
Hongjie Wang,

Xiaoyang Fu,

Zixuan Yu

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 15, 2025

Fire disasters pose significant risks to human life, economic development, and social stability. The early stages of a fire, often characterized by small flames, diffuse smoke, obstructed objects, can lead challenges such as missed detections poor real-time performance. To address these issues, we propose DSS-YOLO model based on an improved YOLOv8n architecture, designed enhance the recognition accuracy obscured objects targets while reducing computational overhead. Specifically, replace all C2f modules in Backbone with DynamicConv reduce computation without sacrificing feature extraction capabilities. We also introduce SEAM attention mechanism improve detection targets, SPPELAN module at end across different scales. is evaluated using public dataset mytest-hrswj, which contains diverse fire scenarios, including indoors, forests, buildings. Compared original YOLOv8n, DSS-YOLOv8 proposed this paper improves mAP 0.6% Recall 1.6%, size FLOPs 3.4% 12.3% respectively. results study provide effective technical support for intelligent monitoring systems, significantly cost model. It enhances capabilities complex facilitating hazards helping minimize damage caused fires.

Language: Английский

DSS-YOLO: an improved lightweight real-time fire detection model based on YOLOv8 DOI Creative Commons
Hongjie Wang,

Xiaoyang Fu,

Zixuan Yu

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 15, 2025

Fire disasters pose significant risks to human life, economic development, and social stability. The early stages of a fire, often characterized by small flames, diffuse smoke, obstructed objects, can lead challenges such as missed detections poor real-time performance. To address these issues, we propose DSS-YOLO model based on an improved YOLOv8n architecture, designed enhance the recognition accuracy obscured objects targets while reducing computational overhead. Specifically, replace all C2f modules in Backbone with DynamicConv reduce computation without sacrificing feature extraction capabilities. We also introduce SEAM attention mechanism improve detection targets, SPPELAN module at end across different scales. is evaluated using public dataset mytest-hrswj, which contains diverse fire scenarios, including indoors, forests, buildings. Compared original YOLOv8n, DSS-YOLOv8 proposed this paper improves mAP 0.6% Recall 1.6%, size FLOPs 3.4% 12.3% respectively. results study provide effective technical support for intelligent monitoring systems, significantly cost model. It enhances capabilities complex facilitating hazards helping minimize damage caused fires.

Language: Английский

Citations

0