行人重识别工作汇总

person re-identification

2018CVPR

paper	keyword	link	note
1. Pose Transferrable Person Re-Identification		paper
2. Deep Spatial Feature Reconstruction for Partial Person Re-identification: Freestyle Approach	部分行人再识别:不需要行人对齐的方法,借鉴字典学习中重建误差来计算不同的空域特征图的相似度(CASIA)	paper/github	note(论文2)/my-note
3. Adversarially Occluded Samples For Improving Generalization of Person Re-identification Models	为了提高模型的泛化性能，提出特对抗式遮挡样本来扩充数据集(CASIA)	paper/github	note(论文3)
4. Mask-guided Contrastive Attention Model for Person Re-Identification	引进了二值化的行人分割轮廓/区域级别的三元组损失函数，分别约束不同区域(CASIA)	paper/github	note(论文9)
5. Harmonious Attention Network for Person Re-Identification	HA-CNN	paper
6. Transferable Joint Attribute-Identity Deep Learning for Unsupervised Person Re-Identification		paper
7. Camera Style Adaptation for Person Re-identfication	cycleGAN完成镜头间图片转换/增加数据防过拟合，但是也会产生噪音（SydneyUniversity）	paper/github	my-note
8. Exploit the Unknown Gradually: One-Shot Video-Based Person Re-Identification by Stepwise Learning	用未标注样本解决单标注样本下的视频行人重识别问题(SydneyUniversity)	paper/github	note/my-comment
9. Person Transfer GAN to Bridge Domain Gap for Person Re-Identification	数据集和图像风格迁移算法/无监督GAN（PTGAN+MSMT17）(spotlight)(PKU)	paper/github	my-note
10. Unsupervised Cross-dataset Person Re-identification by Transfer Learning of Spatial-Temporal Patterns		paper/github
11. Image-Image Domain Adaptation with Preserved Self-Similarity and Domain-Dissimilarity for Person Re-identification	(SPGAN)	paper
12. A Pose-Sensitive Embedding for Person Re-Identification with Expanded Cross Neighborhood Re-Ranking	re-ranking策略/结合了人的精细和粗略姿态信息来学习判别嵌入（姿态估计点和rgb并成通道/加入前后侧粗精度估计）	paper/github	my-note
14. Disentangled Person Image Generation	一种基于自监督的解耦表达人体前景、背景、姿态特征的生成式模型	paper/project	note
15. Deep Mutual Learning		paper
16. Efficient and Deep Person Re-Identification using Multi-Level Similarity		paper
17. Dual Attention Matching Network for Context-Aware Feature Sequence based Person Re-Identification		paper
18. Exploiting Transitivity for Learning Person Re-identification Models on a Budget		paper
19. Good Appearance Features for Multi-Target Multi-Camera Tracking		paper/github
20. Human Semantic Parsing for Person Re-identification	parsing工作/多数据集训练细节的讨论	paper	my-note
21. Multi-Level Factorisation Net for Person Re-Identification		paper
22. Multi-shot Pedestrian Re-identification via Sequential Decision Making	视频连续帧有很多冗余。出发点：效果好的时候，看一张图片就好了，出现模糊和遮挡的现象时，需要动态的调节看多少图像。使用增强学习框架	paper	my-note
23. Person Re-identification with Cascaded Pairwise Convolutions		paper
24. Resource Aware Person Re-identification across Multiple Resolutions		paper
25. Group Consistent Similarity Learning via Deep CRFs for Person Re-Identification	判断图像群是否相似 而不像传统工作只判断两张图是否相似/使用pairwise loss/(sensetime oral)	paper/github	my-note(1)
26. Deep Group-shuffling Random Walk for Person Re-identification	(sensetime)	paper/github
27. Eliminating Background-bias for Robust Person Re-identification	parsing方法/CRF用来传递正、侧、背面相似的置信度。中间传递不准确的话也会有影响（所以CRF的方法使用可能不是很有用）(sensetime)	paper	my-note(2)
28. Video Person Re-identification with Competitive Snippet-similarity Aggregation and Co-attentive Snippet Embedding	(sensetime)	paper
29. End-to-End Deep Kronecker-Product Matching for Person Re-identification	(sensetime)	paper/github
30. Attention-aware Compositional Network for Person Re-Identification	Attention-Aware Compositional Network (AACN), Pose-guided Part Attention (PPA), Attention-aware Feature Composition (AFC)/(sensetime)	paper
31. Diversity Regularized Spatiotemporal Attention for Video-based Person Re-identification	(CUHK-SenseTime)	paper
32. Easy Identification from Better Constraints: Multi-Shot Person Re-Identification from Reference Constraints		paper

person search

paper （日期，会议，机构）	keyword	link	note
Joint Detection and Identification Feature Learning for Person Search	person search	paper/github	my-note
Person Re-identification in the Wild	person search	paper/github
person search with natural language description	person search/language description		note

其他工作(要按时间整理)

paper	keyword	link	note
Pose-Normalized Image Generation for Person Re-identification	姿态归一化GAN	paper/github	my-note
Multi-pseudo Regularized Label for Generated Samples in Person Re-Identification(18.01/Tans)	多伪正则化标签(MpRL)、GAN	paper	my-note
Multi-Channel Pyramid Person Matching Network for Person Re-Identification	空洞卷积多尺度
SphereReID: Deep Hypersphere Manifold Embedding for Person Re-Identification	Sphere Loss,balanced sampling,训练技巧	paper	my-note

扩展研究

paper	keyword	link	note
聚类方法结合卷积网络，实现无监督端到端图像分类			note

参考link:

https://github.com/handong1587/handong1587.github.io/blob/master/_posts/deep_learning/2015-10-09-re-id.md

https://zhuanlan.zhihu.com/p/34716633

https://zhuanlan.zhihu.com/personReid