Scholarlink

<p>Fire poses a persistent threat to human life and critical infrastructure, highlighting the need for accurate and real-time fire and smoke detection algorithms. This paper proposes a lightweight and high-performance fire and smoke detection model, termed YOLOv11-FSNet. By introducing a C3k2-RVB-EMA module that integrates depthwise separable convolution and dynamic attention into the backbone and neck, the proposed model enhances feature representation while reducing computational cost. The neck further adopts a Feature Focused Diffusion Pyramid Network (FSNet) to improve multi-scale feature fusion in complex scenes. In addition, a lightweight Mixed Local Channel Attention (MLCA) mechanism is incorporated to enhance sensitivity to ambiguous targets such as smoke. Experimental results demonstrate that, compared with the baseline model, YOLOv11-FSNet improves mAP@0.5, precision, and recall by 1.6%, 2.5%, and 1.7%, respectively. Although the model complexity is slightly increased, the inference speed is improved by 25.9 FPS, indicating the potential of the proposed method for real-time fire and smoke detection in high-performance computing scenarios.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on Fire Smoke Detection Algorithm Based on YOLOv11-FSNet

  • 期刊论文

  • 发表日期：

    2026年5月15日

查看全文
  • Xin Chen,
  • Bole Zhang,
  • Yaolin Zhu,
  • Yan Fu,
  • Kai Han,
  • Bing Liu

摘要

Fire poses a persistent threat to human life and critical infrastructure, highlighting the need for accurate and real-time fire and smoke detection algorithms. This paper proposes a lightweight and high-performance fire and smoke detection model, termed YOLOv11-FSNet. By introducing a C3k2-RVB-EMA module that integrates depthwise separable convolution and dynamic attention into the backbone and neck, the proposed model enhances feature representation while reducing computational cost. The neck further adopts a Feature Focused Diffusion Pyramid Network (FSNet) to improve multi-scale feature fusion in complex scenes. In addition, a lightweight Mixed Local Channel Attention (MLCA) mechanism is incorporated to enhance sensitivity to ambiguous targets such as smoke. Experimental results demonstrate that, compared with the baseline model, YOLOv11-FSNet improves mAP@0.5, precision, and recall by 1.6%, 2.5%, and 1.7%, respectively. Although the model complexity is slightly increased, the inference speed is improved by 25.9 FPS, indicating the potential of the proposed method for real-time fire and smoke detection in high-performance computing scenarios.