行人重识别github开源库——HJL-re-id

更新

2020年11月15日：添加了新的论文复现

• Pyramid [CVPR2019] Pyramidal Person Re-IDentification via Multi-Loss Dynamic Training

---

目录

一、简介

二、实现的Re-ID模型

三、MDRS

四、遮挡的行人重识别

---

项目地址：GitHub - nickhuang1996/HJL-re-id: A awesome project for person re-id. There are re-implements of PCB, MGN, MDRS(ours), PGFA, Pyramidal and HOReID.

---

一、简介

这是由博主自己完成的行人重识别代码库，包含了博主自己研究的MDRS模型。

该项目包含对日志记录、损失监测和可视化Rank图像的充分支持。

项目中的各个模型都是Pytorch实现的， 包含了PCB，MGN，PGFA， Pyramid和HOReID等顶会论文模型。

二、实现的Re-ID模型

• PCB [ECCV2018] Beyond Part Models: Person Retrieval with Refined Part Pooling (and A Strong Convolutional Baseline)
• MGN [ACM Multimedia 2018] Learning Discriminative Features with Multiple Granularities for Person Re-Identification
• PGFA [ICCV2019] Pose-Guided Feature Alignment for Occluded Person Re-Identification
• Pyramid [CVPR2019] Pyramidal Person Re-IDentification via Multi-Loss Dynamic Training
• HOReID [CVPR2020] High-Order Information Matters: Learning Relation and Topology for Occluded Person Re-Identification

三、MDRS

• 名称: Joint multi-scale discrimination and region segmentation for person re-ID
• 期刊: Pattern Recognition Letters
• Volume 138, October 2020, Pages 540-547
• JCR: Q2
• web of science

框架

四、遮挡的行人重识别

博主提供了 PGFAReIDTrainer.py，PyramidalReIDTrainer.py和 HOReIDTrainer.py 来实现模型PGFA和模型HOReID的训练和测试