基于PDEW-YOLOv8n的反光衣穿戴检测
DOI:
作者:
作者单位:

四川轻化工大学计算机科学与工程学院

作者简介:

通讯作者:

中图分类号:

基金项目:

四川省科技研发重点项目(2023YFS0371); 四川省科技创新(苗子工程)培育项目(2022049); 企业信息化与物联网测控技术四川省高校重点实验室基金项目(2022WYY03);


Research on wearing detection of reflective clothing based on PDEW-YOLOv8n
Author:
Affiliation:

1.School of Computer Science &2.Engineering,Sichuan University of Science &3.Engineering

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    为解决传统反光衣穿戴检测算法在施工、军队和国防科技等方面存在的检测精度低、错检、漏检等问题,提出了一种适用于反光衣穿戴检测的PDEW-YOLOv8n算法。首先,添加了小目标检测层与检测头,提高检测精度;接着,为了使模型适应反光衣的不规则形变,在骨干网络中将普通卷积替换为可变形卷积;然后,为了使模型更加关注反光衣信息,提高网络的特征表达能力,在颈部网络中引入高效多尺度注意模块(Efficient Multi-Scale Attention Module,EMA);最后,为了加速网络收敛,将损失函数CIOU替换为WIOU。结果表明:相较于YOLOv8n,PDEW-YOLOv8n在模型复杂度基本不变的同时,准确率、召回率、mAP@0.5、mAP@0.5:0.95分别提升了0.6%、5.7%、4.2%、5.2%。可见基本满足反光衣穿戴实时检测的低复杂度兼高精度等要求。

    Abstract:

    To address the issues of low detection accuracy, false positives, and missed detections in traditional algorithms for reflective vest wearing detection in construction, military, and defense technology fields, a PDEW-YOLOv8n algorithm suitable for such applications is proposed. Firstly, a small target detection layer and detection head were added to improve detection accuracy; subsequently, to adapt to irregular deformations of reflective vests, standard convolutions in the backbone network were replaced with deformable convolutions. Additionally, to enhance the model’s focus on reflective vest information, an Efficient Multi-Scale Attention Module (EMA) was introduced in the neck network to improve the network"s feature representation ability. Finally, to accelerate network convergence, the CIOU loss function was replaced with WIOU. Experimental results indicate that compared to YOLOv8n, PDEW-YOLOv8n maintains a similar model complexity while improving accuracy, recall rate, mAP@0.5, and mAP@0.5:0.95 by 0.6%, 5.7%, 4.2%, and 5.2% respectively, effectively meeting the requirements for real-time detection of reflective vest wear with low complexity and high accuracy.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-06-26
  • 最后修改日期:2024-06-28
  • 录用日期:2024-07-08
  • 在线发布日期:
  • 出版日期: