| 摘 要: 为了提高多人人体姿态检测的准确率,本研究采用YOLOv5s模型用于多人人体姿态检测并对模型进行改进。首先,引入坐标注意力(Coordinate Attention)模块改进骨干网络,将注意力资源分配给关键区域,降低复杂环境中的背景干扰,增强模型对多人目标的精准定位能力。其次,使用双向特征金字塔网络改进YOLOv5s的特征融合网络,增强网络的信息表达能力。实验结果表明:在多人人体姿态MS COCO2017验证集上,经改进的YOLOv5s算法的检测平均精度高达61.9%,相比原始YOLOv5s网络,平均精度提升了1.5%。由此可见,改进后的网络能更加精准、有效地检测多人人体姿态。 |
| 关键词: 多人人体姿态检测;YOLOv5s;双向特征金字塔网络;检测精度 |
|
中图分类号: TP183
文献标识码: A
|
| 基金项目: 福建工程学院科研启动基金(GY-Z21064) |
|
| Multi-person Pose Estimation Based on Improved YOLOv5s-pose |
|
JIANG Jinhua, ZHUANG Liping, CHEN Jin, YAO Hongze, CAI Zhiming
|
(Fujian University of Technology, Fuzhou 350118, China)
1422881869@qq.com; 1504879132@qq.com; 464525151@qq.com; 894130700@qq.com; caizm@fjut.edu.cn
|
| Abstract: In order to improve the accuracy of multi-person pose detection, this paper proposes to use and improve the YOLOv5s model for multi-person pose detection. Firstly, the Coordinate Attention module is introduced to improve the backbone network by allocating attention resources to key areas, reducing background interference in complex environments, and enhancing the model 's precise localization ability for multi-person targets. Secondly, a bidirectional feature pyramid network is used to improve the feature fusion network of YOLOv5s and enhance the network's information expression ability. The experimental results show that on the MS COCO2017 validation set for multi-person poses, the improved YOLOv5s algorithm achieves an average detection precision of 61. 9% and the average accuracy increases by 1.5% , compared to the original YOLOv5s network. It can be seen that the improved network can more accurately and effectively detect multiple human body postures. |
| Keywords: multi-person pose detection; YOLOv5s; bidirectional feature pyramid network; detection precision |
