【ICCV 2019】Self-similarity Grouping: A Simple Unsupervised Cross DA Approach for Person Re-id（SSG）

下方↓公众号后台回复“SSG”，即可获得论文电子资源。

背景知识

Person re-identification (re-ID)

Person re-identification (re-ID) aims at matching images of a person in one camera with the images of this person from other different cameras. Because

unsupervised domain adaptation(UDA)

Problem #1

However, the traditional UDA approaches [4, 5, 27] always have an assumption that the source and target domain share the same set of classes, which does not hold for the person re-ID problem.

Problem #2

The main reason is that most previous works focus on increasing the training samples or comparing the similarity or dissim- ilarity between the source dataset and the target dataset but ignoring the similar natural characteristics existing in the training samples from the target domain. //其主要原因是以往的工作多集中于增加训练样本，或者比较源数据集和目标数据集的相似度或不相似度，而忽略了目标域训练样本中存在的相似的自然特征。

person re-ID problem is actually an open set problem

Problem #3

person re-ID problem is actually an open set prob- lem. In other words, we cannot know in advance how many identities are included in a given unlabeled target dataset. Thus, the superior characteristics from traditional one shot setting cannot be directly applied to the re-ID case. //人再识别问题实际上是一个开集问题。换句话说，我们无法预先知道给定的未标记目标数据集中包含了多少身份。因此，传统的one-shot设置的优越特性不能直接应用到re-ID的情况下。

内容概要

本文工作

we explore how to harness the similar natural characteris- tics existing in the samples from the target domain for learning to conduct person re-ID in an unsupervised man- ner. //我们探讨了如何利用目标域样本中存在的相似的自然特征来学习在无监督的人中进行人的再识别。

• we propose a Self-similarity Group- ing (SSG) approach, which exploits the potential simi- larity (from the global body to local parts) of unlabeled samples to build multiple clusters from different views automatically. //我们提出了一种自相似性分组(SSG)方法，该方法利用未标记样本的潜在相似性(从全局主体到局部部分)，从不同的视图自动构建多个聚类。
• We repeatedly and alternatively conduct such a grouping and training process until the model is stable. //我们反复交替地进行这样的分组和训练过程，直到模型稳定为止。
• Upon our SSG, we further introduce a clustering-guided semi- supervised approach named SSG++ to conduct the one- shot domain adaption in an open set setting (i.e. the num- ber of independent identities from the target domain is unknown).

实验效果

• our SSG outperforms the state-of-the-arts by more than 4.6% (DukeMTMC→Market1501) and 4.4% (Market1501→DukeMTMC) in mAP, respectively.
• our SSG++ can further promote the mAP upon SSG by 10.7% and 6.9%, respectively.

相关工作

• Unsupervised domain adaptation.
• Unsupervised re-ID
• Semi-supervised re-ID

数据集

• Market1501 [45]
• DukeMTMC-ReID [31, 46]
• MSMT17 [40]

方法提要

In order to address Problem #2 and discover the similarities among person images in target dataset, we pro- posed unsupervised Self-similarity Grouping (SSG) to mine the potential similarities from global to local manner.

• we extract fea- tures of all persons in target dataset and group them by three different cues, whole bodies (A), upper parts (B) and lower parts © independently
  
• By assigning a pseudo label to each group, we can pair every person with different pseudo labels.

Upon our SSG, we further present a semi-supervised so- lution based clustering-guided annotation to approach the performance of the fully-supervised counterpart and effi- ciently achieve the adaption from the source domain to the target one. //在此基础上，我们进一步提出了一种基于聚类引导的半监督注释，以接近完全监督对等体的性能，并有效地实现了从源域到目标域的自适应。

• To tackle Problem #3, we innovatively provide a clustering-guided semi-supervised solution.
• The proposed semi-supervised training strategy is based on the clustering- guided annotations, which samples a single image from each clustering.

方法框架

实验结果

方法详解

。。。。

参考文献

[1] Slawomir Bak and Peter Carr. One-shot metric learning for person re-identification. In IEEE CVPR, pages 1571–1580, 2017. 3
[2] Loris Bazzani, Marco Cristani, and Vittorio Murino. Symmetry-driven accumulation of local features for human characterization and re-identification. Computer Vision and Image Understanding, 2013. 3
[3] Konstantinos Bousmalis, Nathan Silberman, David Dohan, Dumitru Erhan, and Dilip Krishnan. Unsupervised pixel- level domain adaptation with generative adversarial net- works. In IEEE CVPR, 2017. 2
[4] Yuhua Chen, Wen Li, Christos Sakaridis, Dengxin Dai, and Luc Van Gool. Domain adaptive faster r-cnn for object de- tection in the wild. In IEEE CVPR, 2018. 1
[5] Yuhua Chen, Wen Li, and Luc Van Gool. Road: Reality ori- ented adaptation for semantic segmentation of urban scenes. In IEEE CVPR, 2018. 1
[6] Brian Chu, Vashisht Madhavan, Oscar Beijbom, Judy Hoff- man, and Trevor Darrell. Best practices for fine-tuning visual classifiers to new domains. In ECCV, pages 435–442, 2016. 2
[7] Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In IEEE CVPR, pages 248–255, 2009. 4
[8] Weijian Deng, Liang Zheng, Guoliang Kang, Yi Yang, Qix- iang Ye, and Jianbin Jiao. Image-image domain adaptation with preserved self-similarity and domain-dissimilarity for person reidentification. In IEEE CVPR, 2018. 1, 2, 3, 4, 6, 7, 8
[9] Martin Ester, Hans-Peter Kriegel, J¨org Sander, Xiaowei Xu, et al. A density-based algorithm for discovering clusters in large spatial databases with noise. In Kdd, 1996. 4
[10] Hehe Fan, Liang Zheng, Chenggang Yan, and Yi Yang. Unsupervised person re-identification: Clustering and fine- tuning. ACMTransactions on Multimedia Computing, Com- munications, and Applications (TOMM), 2018. 1, 8
[11] Li Fei-Fei, Rob Fergus, and Pietro Perona. One-shot learning of object categories. IEEE TPAMI, 2006. 3
[12] Pedro F Felzenszwalb, Ross B Girshick, David McAllester, and Deva Ramanan. Object detection with discriminatively trained part-based models. IEEE TPAMI, 2010. 6
[13] Dario Figueira, Loris Bazzani, Ha Quang Minh, Marco Cristani, Alexandre Bernardino, and Vittorio Murino. Semi-supervised multi-feature learning for person re- identification. In AVSS, pages 111–116, 2013. 3
[14] Yang Fu, Xiaoyang Wang, Yunchao Wei, and Thomas Huang. Sta: Spatial-temporal attention for large-scale video- based person re-identification. In AAAI, 2019. 4
[15] Yang Fu, Yunchao Wei, Yuqian Zhou, Honghui Shi, Gao Huang, Xinchao Wang, Zhiqiang Yao, and Thomas Huang. Horizontal pyramid matching for person re-identification. In AAAI, 2019. 4
[16] Yaroslav Ganin and Victor Lempitsky. Unsupervised domain adaptation by backpropagation. In IEEE ICML, 2015. 2
[17] Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pas- cal Germain, Hugo Larochelle, Franc¸ois Laviolette, Mario
Marchand,Marchand, and Victor Lempitsky. Domain-adversarial train- ing of neural networks. The Journal of Machine Learning Research, 2016. 2
[18] Douglas Gray and Hai Tao. Viewpoint invariant pedestrian recognition with an ensemble of localized features. In ECCV, 2008. 3
[19] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In IEEE CVPR, pages 770–778, 2016. 4
[20] Alexander Hermans, Lucas Beyer, and Bastian Leibe. In de- fense of the triplet loss for person re-identification. arXiv preprint arXiv:1703.07737, 2017. 1, 4, 5
[21] Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. Image-to-image translation with conditional adver- sarial networks. arXiv preprint, 2017. 2
[22] Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014. 6
[23] Yu-Jhe Li, Fu-En Yang, Yen-Cheng Liu, Yu-Ying Yeh, Xi- aofei Du, and Yu-Chiang Frank Wang. Adaptation and re- identification network: An unsupervised deep transfer learn- ing approach to person re-identification. In IEEE CVPRW, 2018. 7, 8
[24] Shengcai Liao, Yang Hu, Xiangyu Zhu, and Stan Z Li. Per- son re-identification by local maximal occurrence represen- tation and metric learning. In IEEE CVPR, pages 2197– 2206, 2015. 3, 8
[25] Ming-Yu Liu and Oncel Tuzel. Coupled generative adversar- ial networks. In NIPS, pages 469–477, 2016. 2
[26] Xiao Liu, Mingli Song, Dacheng Tao, Xingchen Zhou, Chun Chen, and Jiajun Bu. Semi-supervised coupled dictionary learning for person re-identification. In IEEE CVPR, pages 3550–3557, 2014. 3
[27] Saeid Motiian, Marco Piccirilli, Donald A Adjeroh, and Gi- anfranco Doretto. Unified deep supervised domain adapta- tion and generalization. In IEEE ICCV, volume 2, page 3, 2017. 1, 2
[28] Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Al- ban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in pytorch. In NIPS-W, 2017. 6
[29] Peixi Peng, Tao Xiang, Yaowei Wang, Massimiliano Pon- til, Shaogang Gong, Tiejun Huang, and Yonghong Tian. Unsupervised cross-dataset transfer learning for person re- identification. In IEEE CVPR, pages 1306–1315, 2016. 3, 8
[30] Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. In NIPS, 2015. 6
[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 ECCV, 2016. 2, 6
[32] Florian Schroff, Dmitry Kalenichenko, and James Philbin. Facenet: A unified embedding for face recognition and clus- tering. In IEEE CVPR, 2015. 5
[33] Liangchen Song, Cheng Wang, Lefei Zhang, Bo Du, Qian Zhang, Chang Huang, and Xinggang Wang. Unsuper-vised domain adaptive re-identification: Theory and practice. arXiv preprint arXiv:1807.11334, 2018. 7, 8
[34] Baochen Sun, Jiashi Feng, and Kate Saenko. Return of frus- tratingly easy domain adaptation. In AAAI, 2016. 2
[35] Baochen Sun and Kate Saenko. Deep coral: Correlation alignment for deep domain adaptation. In ECCV, pages 443– 450, 2016. 2
[36] Yifan Sun, Liang Zheng, Yi Yang, Qi Tian, and Shengjin Wang. Beyond part models: Person retrieval with refined part pooling. arXiv preprint arXiv:1711.09349, 2017. 1
[37] Eric Tzeng, Coline Devin, Judy Hoffman, Chelsea Finn, Pieter Abbeel, Sergey Levine, Kate Saenko, and Trevor Darrell. Adapting deep visuomotor representations with weak pairwise constraints. arXiv preprint arXiv:1511.07111, 2015. 2
[38] GuanshuoWang, Yufeng Yuan, Xiong Chen, Jiwei Li, and Xi Zhou. Learning discriminative features with multiple gran- ularities for person re-identification. In ACM Multimedia, 2019. 1, 4
[39] Jingya Wang, Xiatian Zhu, Shaogang Gong, and Wei Li. Transferable joint attribute-identity deep learning for unsu- pervised person re-identification. In IEEE CVPR, 2018. 3, 4, 8
[40] Longhui Wei, Shiliang Zhang, Wen Gao, and Qi Tian. Person transfer gan to bridge domain gap for person re- identification. In IEEE CVPR, 2018. 2, 3, 4, 6, 8
[41] Yu Wu, Yutian Lin, Xuanyi Dong, Yan Yan, Wanli Ouyang, and Yi Yang. Exploit the unknown gradually: One-shot video-based person re-identification by stepwise learning. In IEEE CVPR, 2018. 3, 5
[42] Jason Yosinski, Jeff Clune, Yoshua Bengio, and Hod Lipson. How transferable are features in deep neural networks? In Advances in neural information processing systems, pages 3320–3328, 2014. 2
[43] Hong-Xing Yu, Ancong Wu, and Wei-Shi Zheng. Cross- view asymmetric metric learning for unsupervised person re- identification. In IEEE ICCV, 2017. 8
[44] Hong-Xing Yu, Wei-Shi Zheng, Ancong Wu, Xiaowei Guo, Shaogang Gong, and Jian-Huang Lai. Unsupervised person re-identification by soft multilabel learning. In IEEE CVPR, 2019. 7, 8
[45] Liang Zheng, Liyue Shen, Lu Tian, Shengjin Wang, Jing- dong Wang, and Qi Tian. Scalable person re-identification: A benchmark. In IEEE ICCV, pages 1116–1124, 2015. 2, 6, 8
[46] Zhedong Zheng, Liang Zheng, and Yi Yang. Unlabeled sam- ples generated by gan improve the person re-identification baseline in vitro. arXiv preprint arXiv:1701.07717, 2017. 2, 6
[47] Zhun Zhong, Liang Zheng, Donglin Cao, and Shaozi Li. Re- ranking person re-identification with k-reciprocal encoding. In IEEE CVPR, pages 3652–3661, 2017. 4, 5, 6
[48] Zhun Zhong, Liang Zheng, Guoliang Kang, Shaozi Li, and Yi Yang. Random erasing data augmentation. arXiv preprint arXiv:1708.04896, 2017. 6
[49] Zhun Zhong, Liang Zheng, Shaozi Li, and Yi Yang. Gener- alizing a person retrieval model hetero-and homogeneously. In ECCV, pages 172–188, 2018. 2, 3, 4, 8
[50] Zhun Zhong, Liang Zheng, Zhiming Luo, Shaozi Li, and Yi Yang. Invariance matters: Exemplar memory for domain adaptive person re-identification. In IEEE CVPR, 2019. 8
[51] Zhun Zhong, Liang Zheng, Zhedong Zheng, Shaozi Li, and Yi Yang. Camera style adaptation for person re- identification. In IEEE CVPR, pages 5157–5166, 2018. 4, 6