摘要
遮挡的人物再识别ReID是一项人物检索任务,旨在将遮挡的人物图像与完整的图像进行匹配。为了解决遮挡的再识别问题,基于部分的方法已经被证明是有效的,因为它们提供了细粒度的信息,并且很适合表示部分可见的人体。提出一种基于HRNet的遮挡行人重识别算法,采用ODConv增加对输入数据的适应性,并添加了注意力机制以增强特征表示。
Occluded person re-identification ReID is a person retrieval task that aims to match occluded person images with complete images.To address the challenge of occluded re-identification,part-based methods have been proven beneficial as they provide fine-grained information and are suitable for representing partially visible human body parts.In this paper,we propose an occluded person re-identification algorithm based on HRNet.We utilize ODConv to enhance adaptability to input data and incorporate attention mechanisms to enhance feature representation.
作者
彭晓聪
周卫
段卓
Peng Xiaocong;Zhou Wei;Duan Zhuo(School of Electronic Information,Guangxi Minzu University,Nanning 530006,China)
出处
《现代计算机》
2024年第24期67-72,共6页
Modern Computer
关键词
遮挡
行人重识别
注意力机制
HRNet
occluded
person re-identification
attention mechanisms
HRNet
