方法概述

1，文章将多源概念引入到了无监督域适应行人重识别当中。
2，文章提出了一种修订域特定批归一化模块(RDSBN), 可以同时减少域特定信息和增加人物特征的可区别性。
3，我们基于多域信息融合（MDIF）开发了图神经网络（GCN），以此来拉近特征空间中的多域。

文章目录

方法概述
内容概要
- 工作概述
- 成果概述
方法详解
- 方法框架
- 具体实现
实验结果
总体评价
引用格式
参考文献

内容概要

论文名称	简称	会议/期刊	出版年份	baseline	backbone	数据集
Unsupervised Multi-Source Domain Adaptation for Person Re-Identification	UMSDA	CVPR	2021	【MMT】Ge, Y., Chen, D., Li, H.: Mutual mean-teaching: Pseudo label refinery for unsu- pervised domain adaptation on person re-identification. In: ICLR (2020)	ResNet50 [11]	Market1501[49], DukeMTMC- reID[31],CUHK03[18] and MSMT[38]

在线链接：https://openaccess.thecvf.com/content/CVPR2021/html/Bai_Unsupervised_Multi-Source_Domain_Adaptation_for_Person_Re-Identification_CVPR_2021_paper.html
源码链接：

工作概述

1， we introduce the multi-source concept into UDA person re-ID field, where multiple source datasets are used during train- ing.
2， we try to address this problem from two perspectives, i.e. domain-specific view and domain-fusion view.
3，Two con- structive modules are proposed, and they are compatible with each other. First, a rectification domain-specific batch normalization (RDSBN) module is explored to simultane- ously reduce domain-specific characteristics and increase the distinctiveness ofperson features. Second, a graph con- volutional network (GCN) based multi-domain information fusion (MDIF) module is developed, which minimizes do- main distances by fusing features ofdifferent domains.

成果概述

The proposed method outperforms state-of-the-art UDA person re-ID methods by a large margin, and even achieves com- parable performance to the supervised approaches with- out any post-processing techniques.

方法详解

方法框架

具体实现

1，有K个带标注的源域，和一个无标注的目标域。

2，提出无监督多源域适应框架，如图1（a）所示。框架可以分为三个部分：伪标签生成器，骨干网络和头部。

3，文章在标签生成器中使用DBSCAN算法进行聚类，输入目标域特征，输出带有伪标签的聚类。backbone用于提取域无关的特征，Head中的MDIF是用于减少域间差距的。
4，批归一化BN在backbone之后取代了全连接层，其计算如公式1所示。在训练的过程中 mu 和 sigmod 通过公式2和公式3更新。

5，文章在图1（b）中提出了修正程序，RDSBN的计算如公式4所示。与公式1不同的地方在于增加了 a 参数，其表示通道权重，其计算如公式 5所示。

6，文章提出了domain-agent-node作为全局域表达，其形式化为公式6所示，在训练过程中的更新如公式7所示。

7，在图结构中，边的链接矩阵的计算如公式8所示，接着链接矩阵按照同时9进行标准化。 GCN 层可以被重写为非线性的转化（公式10）

8，在训练过程中，首先进行源域的预训练，损失函数采用了交叉熵损失和三元组损失的同等结合（公式12），在fine-tuning阶段的总体损失如公式13所示。

实验结果

总体评价

1，文章的创新出发点是想到用多个域来作为源域进行预训练。因为样本量的增加，性能可想而知的提高了。但是作者考虑多个源域之间的差异性问题，在backbone后设置BN层，提取域相关的信息，随后又加入了MDIF 对多域信息进行融合。
2，文章公式比较多，设计考虑的细节比较细致，整篇读下来，信息量有点大了。
3，虽然公式多，细节细，但每个地方仔细看起来也是些常规处理方式。
4，图画得不错，值得学习。把简单的内容高端化，明确化。

小样本学习与智能前沿（下方↓公众号）后台回复“UMSDA"，即可获得论文电子资源。

引用格式

@inproceedings{DBLP:conf/cvpr/BaiWW0D21,
author = {Zechen Bai and
Zhigang Wang and
Jian Wang and
Di Hu and
Errui Ding},
title = {Unsupervised Multi-Source Domain Adaptation for Person Re-Identification},
booktitle = {{CVPR}},
pages = {12914–12923},
publisher = {Computer Vision Foundation / {IEEE}},
year = {2021}
}

参考文献

[1] Konstantinos Bousmalis, George Trigeorgis, Nathan Silber- man, Dilip Krishnan, and Dumitru Erhan. Domain separa- tion networks. In Advances in Neural Information Process- ing Systems 29: Annual Conference on Neural Information Processing Systems 2016, December 5-10, 2016, Barcelona, Spain, pages 343–351, 2016. 1
[2] Woong-Gi Chang, Tackgeun You, Seonguk Seo, Suha Kwak, and Bohyung Han. Domain-specific batch normalization for unsupervised domain adaptation. In IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2019, Long Beach, CA, USA, June 16-20, 2019, pages 7354–7362, 2019. 1, 2, 4, 6, 7
[3] Weijian Deng, Liang Zheng, Qixiang Ye, Guoliang Kang, Yi Yang, and Jianbin Jiao. Image-image domain adaptation with preserved self-similarity and domain-dissimilarity for person re-identification. In CVPR, 2018. 2
[4] Weijian Deng, Liang Zheng, Qixiang Ye, Yi Yang, and Jian- bin Jiao. Similarity-preserving image-image domain adap- tation for person re-identification. CoRR, abs/1811.10551, 2018. 2
[5] Martin Ester, Hans-Peter Kriegel, J¨org Sander, and Xiaowei Xu. A density-based algorithm for discovering clusters in large spatial databases with noise. In Proceedings ofthe Sec- ond International Conference on Knowledge Discovery and Data Mining (KDD-96), Portland, Oregon, USA, pages 226– 231, 1996. 3
[6] Hehe Fan, Liang Zheng, Chenggang Yan, and Yi Yang. Unsupervised person re-identification: Clustering and fine- tuning. ACM Trans. Multim. Comput. Commun. Appl., 14(4):83:1–83:18, 2018. 2
[7] Yang Fu, Yunchao Wei, Guanshuo Wang, Yuqian Zhou, Honghui Shi, and Thomas S. Huang. Self-similarity group- ing: A simple unsupervised cross domain adaptation ap- proach for person re-identification. In The IEEE Inter- national Conference on Computer Vision (ICCV), October 2019. 2, 7
[8] Yaroslav Ganin and Victor Lempitsky. Unsupervised domain adaptation by backpropagation. In International conference on machine learning, pages 1180–1189. PMLR, 2015. 7
[9] Yaroslav Ganin and Victor S. Lempitsky. Unsupervised do- main adaptation by backpropagation. In Proceedings of the 32nd International Conference on Machine Learning, ICML 2015, Lille, France, 6-11 July 2015, volume 37, pages 1180– 1189, 2015. 2
[10] Yixiao Ge, Dapeng Chen, and Hongsheng Li. Mutual mean- teaching: Pseudo label refinery for unsupervised domain adaptation on person re-identification. In International Con- ference on Learning Representations, 2020. 1, 2, 5, 7, 8
[11] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2016, Las Vegas, NV, USA, June 27-30, 2016, pages 770–778, 2016. 5
[12] Alexander Hermans*, Lucas Beyer*, and Bastian Leibe. In Defense of the Triplet Loss for Person Re-Identification. arXiv preprint arXiv:1703.07737, 2017. 5
[13] Xun Huang and Serge Belongie. Arbitrary style transfer in real-time with adaptive instance normalization. In ICCV, 2017. 4
[14] Bo Jiang, Xixi Wang, and Bin Luo. PH-GCN: person re- identification with part-based hierarchical graph convolu- tional network. CoRR, 2019. 2
[15] Xin Jin, Cuiling Lan, Wenjun Zeng, Zhibo Chen, and Li Zhang. Style normalization and restitution for generalizable person re-identification. CVPR, 2020. 7
[16] Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, 2015. 5
[17] Thomas N. Kipf and Max Welling. Semi-supervised classi- fication with graph convolutional networks. In International Conference on Learning Representations (ICLR), 2017. 2, 3
[18] W. Li, R. Zhao, T. Xiao, and X. Wang. Deepreid: Deep filter pairing neural network for person re-identification. In 2014 IEEE Conference on Computer Vision and Pattern Recogni- tion, pages 152–159, 2014. 5
[19] Yutian Lin, Yu Wu, Chenggang Yan, Mingliang Xu, and Yi Yang. Unsupervised person re-identification via cross- camera similarity exploration. IEEE Transactions on Image Processing, 29:5481–5490, 2020. 2
[20] Yutian Lin, Lingxi Xie, Yu Wu, Chenggang Yan, and Qi Tian. Unsupervised person re-identification via softened similarity learning. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 3390–3399, 2020. 2
[21] Mingsheng Long, Yue Cao, Jianmin Wang, and Michael Jor- dan. Learning transferable features with deep adaptation net- works. In International conference on machine learning, pages 97–105. PMLR, 2015. 2
[22] Mingsheng Long, Han Zhu, Jianmin Wang, and Michael I Jordan. Unsupervised domain adaptation with residual trans- fer networks. In Advances in neural information processing systems, pages 136–144, 2016. 2
[23] Mingsheng Long, Han Zhu, Jianmin Wang, and Michael I Jordan. Deep transfer learning with joint adaptation net- works. In International conference on machine learning, pages 2208–2217. PMLR, 2017. 2
[24] Hao Luo, Youzhi Gu, Xingyu Liao, Shenqi Lai, and Wei Jiang. Bag of tricks and a strong baseline for deep person re-identification. In Proceedings ofthe IEEE Conference on Computer Vision and Pattern Recognition Workshops, pages 0–0, 2019. 7, 8
[25] Ping Luo, Jiamin Ren, Zhanglin Peng, Ruimao Zhang, and Jingyu Li. Differentiable learning-to-normalize via switch- able normalization. International Conference on Learning Representation (ICLR), 2019. 3, 6
[26] Chen Ma, Liheng Ma, Yingxue Zhang, Jianing Sun, Xue Liu, and Mark Coates. Memory augmented graph neural networks for sequential recommendation. In The Thirty- Fourth AAAI Conference on Artificial Intelligence, pages 5045–5052, 2020. 3
[27] Hyeonseob Nam and Hyo-Eun Kim. Batch-instance normal- ization for adaptively style-invariant neural networks. In Ad-vances in Neural Information Processing Systems, 2018. 3, 6
[28] Xingang Pan, Ping Luo, Jianping Shi, and Xiaoou Tang. Two at once: Enhancing learning and generalization capacities via ibn-net. In ECCV, 2018. 2, 6
[29] Xingchao Peng, Qinxun Bai, Xide Xia, Zijun Huang, Kate Saenko, and Bo Wang. Moment matching for multi-source domain adaptation. In 2019 IEEE/CVF International Con- ference on Computer Vision, ICCV 2019, Seoul, Korea (South), October 27 - November 2, 2019, pages 1406–1415, 2019. 2
[30] Xingchao Peng, Qinxun Bai, Xide Xia, Zijun Huang, Kate Saenko, and Bo Wang. Moment matching for multi-source domain adaptation. In Proceedings ofthe IEEE International Conference on Computer Vision, pages 1406–1415, 2019. 7
[31] Ergys Ristani, Francesco Solera, Roger Zou, Rita Cucchiara, and Carlo Tomasi. Performance measures and a data set for multi-target, multi-camera tracking. In European Confer- ence on Computer Vision workshop on Benchmarking Multi- Target Tracking, 2016. 5
[32] Yantao Shen, Hongsheng Li, Shuai Yi, Dapeng Chen, and Xiaogang Wang. Person re-identification with deep similarity-guided graph neural network. In Computer Vision - ECCV 2018 - 15th European Conference, Munich, Ger- many, September 8-14, 2018, Proceedings, Part XV, volume 11219, pages 508–526, 2018. 2, 3
[33] Liangchen Song, Cheng Wang, Lefei Zhang, Bo Du, Qian Zhang, Chang Huang, and Xinggang Wang. Unsupervised domain adaptive re-identification: Theory and practice. Pat- tern Recognition, 102:107173, 2020. 1, 2, 3
[34] Yifan Sun, Liang Zheng, Yi Yang, Qi Tian, and Shengjin Wang. Beyond part models: Person retrieval with refined part pooling (and A strong convolutional baseline). In Com- puter Vision - ECCV2018 - 15th European Conference, Mu- nich, Germany, September 8-14, 2018, Proceedings, Part IV, volume 11208, pages 501–518, 2018. 1, 7, 8
[35] Petar Veliˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Li`o, and Yoshua Bengio. Graph Attention Networks. International Conference on Learning Representations, 2018. 3
[36] DongkaiWang and Shiliang Zhang. Unsupervised person re- identification via multi-label classification. In CVPR, 2020. 7
[37] GuanshuoWang, Yufeng Yuan, Xiong Chen, Jiwei Li, and Xi Zhou. Learning discriminative features with multiple granu- larities for person re-identification. In 2018 ACM Multime- dia Conference on Multimedia Conference, MM2018, Seoul, Republic of Korea, October 22-26, 2018, pages 274–282, 2018. 1
[38] L. Wei, S. Zhang, W. Gao, and Q. Tian. Person transfer gan to bridge domain gap for person re-identification. In 2018 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 79–88, 2018. 2, 5
[39] Bing Xu, Naiyan Wang, Tianqi Chen, and Mu Li. Empirical evaluation of rectified activations in convolutional network. arXiv preprint arXiv:1505.00853, 2015. 5
[40] Sijie Yan, Yuanjun Xiong, and Dahua Lin. Spatial tempo- ral graph convolutional networks for skeleton-based action
recognition. recognition. In Proceedings ofthe Thirty-Second AAAI Con- ference on Artificial Intelligence, pages 7444–7452, 2018. 3
[41] Fengxiang Yang, Ke Li, Zhun Zhong, Zhiming Luo, Xing Sun, Hao Cheng, Xiaowei Guo, Feiyue Huang, Rongrong Ji, and Shaozi Li. Asymmetric co-teaching for unsuper- vised cross-domain person re-identification. In AAAI, pages 12597–12604, 2020. 1, 2, 3
[42] Luyu Yang, Yogesh Balaji, Ser-Nam Lim, and Abhinav Shri- vastava. Curriculum manager for source selection in multi- source domain adaptation. CoRR, abs/2007.01261, 2020. 2
[43] Werner Zellinger, Thomas Grubinger, Edwin Lughofer, Thomas Natschl¨ager, and Susanne Saminger-Platz. Central moment discrepancy (cmd) for domain-invariant representa- tion learning. arXiv preprint arXiv:1702.08811, 2017. 2
[44] Yunpeng Zhai, Shijian Lu, Qixiang Ye, Xuebo Shan, Jie Chen, Rongrong Ji, and Yonghong Tian. Ad-cluster: Aug- mented discriminative clustering for domain adaptive person re-identification. In 2020 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, CVPR 2020, Seattle, WA, USA, June 13-19, 2020, pages 9018–9027, 2020. 2, 7
[45] Weichen Zhang, Wanli Ouyang, Wen Li, and Dong Xu. Collaborative and adversarial network for unsupervised do- main adaptation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3801– 3809, 2018. 1
[46] Xinyu Zhang, Jiewei Cao, Chunhua Shen, and Mingyu You. Self-training with progressive augmentation for unsuper- vised cross-domain person re-identification. In Proceedings of the IEEE International Conference on Computer Vision, pages 8222–8231, 2019. 2, 7
[47] Zhizheng Zhang, Cuiling Lan, Wenjun Zeng, Xin Jin, and Zhibo Chen. Relation-aware global attention for person re- identification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2020. 1
[48] Fang Zhao, Shengcai Liao, Guo-Sen Xie, Jian Zhao, Kai- hao Zhang, and Ling Shao. Unsupervised domain adap- tation with noise resistible mutual-training for person re- identification. 2020. 7
[49] Liang Zheng, Liyue Shen, Lu Tian, Shengjin Wang, Jing- dong Wang, and Qi Tian. Scalable person re-identification: A benchmark. In Proceedings of the 2015 IEEE Interna- tional Conference on Computer Vision (ICCV), ICCV ’15, page 1116–1124, 2015. 5
[50] Jia-Xing Zhong, Nannan Li, Weijie Kong, Shan Liu, Thomas H. Li, and Ge Li. Graph convolutional label noise cleaner: Train a plug-and-play action classifier for anomaly detection. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 3
[51] Zhun Zhong, Liang Zheng, Zhiming Luo, Shaozi Li, and Yi Yang. Learning to adapt invariance in memory for person re-identification. TPAMI, 2020. 7
[52] Yang Zou, Xiaodong Yang, Zhiding Yu, B. V. K. Vijaya Kumar, and Jan Kautz. Joint disentangling and adap- tation for cross-domain person re-identification. CoRR, abs/2007.10315, 2020. 2, 7

【CVPR 2021】Unsupervised Multi-Source Domain Adaptation for Person Re-Identification (UMSDA)相关推荐

【CVPR 2021】树状决策知识蒸馏：Tree-like Decision Distillation
[CVPR 2021]树状决策知识蒸馏:Tree-like Decision Distillation 论文地址: 主要问题: 主要思路: 具体实现: 基本符号: Tree-like Decision ...
【CVPR 2021】Knowledge Review：知识蒸馏新解法
[CVPR 2021]Knowledge Review:知识蒸馏新解法论文地址: 主要问题: 主要思路: 符号假设: 具体实现: 实验结果: 关注我的公众号: 联系作者: 论文地址: https:/ ...
【CVPR 2021】基于Wasserstein Distance对比表示蒸馏方法：Wasserstein Contrastive Representation Distillation
[CVPR 2021]基于Wasserstein Distance对比表示蒸馏方法:Wasserstein Contrastive Representation Distillation 论文地址: ...
【CVPR 2021】自我知识蒸馏：Self-distillation with Batch Knowledge Ensembling Improves ImageNet Classification
[CVPR 2021]自我知识蒸馏:Self-distillation with Batch Knowledge Ensembling Improves ImageNet Classification ...
【CVPR 2021】基于解耦特征的目标检测知识蒸馏：Distilling Object Detectors via Decoupled Features
[CVPR 2021]基于解耦特征的目标检测知识蒸馏:Distilling Object Detectors via Decoupled Features 论文地址: 主要问题: 主要思路: 具体实现 ...
【CVPR 2021】剪枝篇（一）：Network Pruning via Performance Maximization
[CVPR 2021]剪枝篇(一):Network Pruning via Performance Maximization 论文地址: 主要问题: 主要思路: 具体实现: 基本符号: 子网络生成: ...
【CVPR 2021】剪枝篇（五）：基于关键通路的神经网络可解释剪枝
[CVPR 2021]剪枝篇(五):Neural Response Interpretation through the Lens of Critical Pathways 论文地址: 主要问题: 主 ...
【CVPR 2021】通用的实例级蒸馏：General Instance Distillation for Object Detection
[CVPR 2021]通用的实例级蒸馏:General Instance Distillation for Object Detection 论文地址: 主要问题: 主要思路: 主要贡献: 具体实现: ...
【CVPR 2021】剪枝篇（二）：Convolutional Neural Network Pruning with Structural Redundancy Reduction
[CVPR 2021]剪枝篇(二):Convolutional Neural Network Pruning with Structural Redundancy Reduction 论文地址: 主要 ...
【CVPR 2021】Joint Generative and Contrastive Learning for Unsupervised Person Re-identification
方法概述 1,提出了一种用于无监督行人重识别的联合生成对比学习框架,生成和对比模块互相提高对方的性能. 2, 在生成模块中,我们引入了3D网格生成器. 3, 在对比模块,我们提出了一种视角无关的损失, ...

【CVPR 2021】Unsupervised Multi-Source Domain Adaptation for Person Re-Identification (UMSDA)