行人重识别(ReID)开源代码

转自：https://blog.csdn.net/Layumi1993/article/details/55798568?locationNum=2&fps=1

关于行人重识别综述，推荐一下liang zheng 2016年的综述: Past, Present and Future 写了从传统方法到深度学习，从图片到视频的行人重识别的方法。

推荐链接：http://www.liangzheng.org/

• 作者：郑哲东
• 链接：https://www.zhihu.com/question/46943328/answer/139931140
  来源：知乎
  著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。
  
  

  以下为一些公开的行人重识别代码链接：（我会持续更新）
  

  • LOMO: Person Re-identification by Local Maximal Occurrence Representation and Metric Learning
  • Improved CNN: [python implementation (tensorFlow) ]: Ning-Ding/Implementation-CVPR2015-CNN-for-ReID [c++ dlib version ]: ptran516/idla-person-reid
  • 2stream-CNN: [Matconvnet]: layumi/2016_person_re-ID [Caffe]: D-X-Y/caffe-reid
  • CNN+Pose: erichuang0771/PoseBox-Reid
  • DomainDropout: Cysu/dgd_person_reid
  • PersonSearch: ShuangLI59/person_search
  • Baseline: zhunzhong07/IDE-baseline-Market-1501
  • k-reciprocal re-ranking: zhunzhong07/person-re-ranking
  • Triplet: VisualComputingInstitute/triplet-reid

行人重识别 数据集链接：

• Re-ID Dataset Summary: https://robustsystems.coe.neu.edu/sites/robustsystems.coe.neu.edu/files/systems/projectpages/reiddataset.html

图像数据集（Image-based）：

• CUHK01, 02, 03: Rui Zhao's Homepage
• Market-1501, PRW: Liang Zheng's Homepage
• DukeMTMC_reID: layumi/DukeMTMC-reID_evaluation

Attribute相关数据集：

• RAP : An introduction to the RAP dataset
• Attribute for Market-1501and DukeMTMC_reID: vana-yutian (组里妹子标的，大家可以支持一下～)

视频相关数据集：

• Mars：http://liangzheng.org/Project/project_mars.html
• PRID2011：ICG - Downloads

NLP相关数据集：

• 自然语言搜图像：http://xiaotong.me/static/projects/person-search-language/dataset.html
• 自然语言搜索行人所在视频：http://www.mi.t.u-tokyo.ac.jp/projects/person_search

在这里列出这些paper和新paper的arXiv链接，和大家共同学习。（和自然语言结合；svd做更好更短的embedding；Reranking对map提升很大，pratical的work；GAN做数据增强，防止过拟合；Attribute 传统multitask，IDlevel提供数据集；Triplet loss扶我起来，我还能干；PoseBox检测part然后fusion）

• Caption：[1702.05729] Person Search with Natural Language Description
• SVDNet：[1703.05693] SVDNet for Pedestrian Retrieval
• Reranking：[1701.08398] Re-ranking Person Re-identification with k-reciprocal Encoding
• GAN：[1701.07717] Unlabeled Samples Generated by GAN Improve the Person Re-identification Baseline in vitro
• Attribute：[1703.07220] Improving Person Re-identification by Attribute and Identity Learning
• Triplet Loss：[1703.07737] In Defense of the Triplet Loss for Person Re-Identification
• PoseBox：[1701.07732] Pose Invariant Embedding for Deep Person Re-identification
• NLPVideo：[1704.07945] Spatio-temporal Person Retrieval via Natural Language Queries

欢迎关注 知乎 行人重识别专栏 

更多相关文章：

1. 行人重识别综述：从哈利波特地图说起

2. 用GAN做行人重识别

3. 行人对齐和行人重识别一起做

行人重识别PPT：总结state-of-the-art 附上代码链接。

https://www.slideshare.net/ZhedongZheng1/deep-reid