ISSN 0253-2778

CN 34-1054/N

Open AccessOpen Access JUSTC Info.& intellengence 18 January 2023

Unsupervised person re-identification based on removal of camera bias and dynamic updating of the memory bank

Cite this:
https://doi.org/10.52396/JUSTC-2022-0015
More Information
  • Author Bio:

    Jun Zhang received the B.E. degree from the University of Science and Technology of China (USTC) in 2019. He is currently pursuing a master’s degree at the College of Information Engineering in USTC. His main research interests include deep learning and computer vision

    Xinmei Tian is an Associate Professor in the CAS Key Laboratory of Technology in Geo-spatial Information Processing and Application Systems, University of Science and Technology of China (USTC). She received her B.E. degree and Ph.D. degree from USTC in 2005 and 2010, respectively. She received the Excellent Doctoral Dissertation of Chinese Academy of Sciences award in 2012 and the Nomination of National Excellent Doctoral Dissertation award in 2013. Her current research interests include multimedia information retrieval and machine learning

  • Corresponding author: E-mail: xinmei@ustc.edu.cn
  • Received Date: 21 January 2022
  • Accepted Date: 28 July 2022
  • Available Online: 18 January 2023
  • In recent years, unsupervised person reidentification technology has made great strides. The technology retrieves images of interested persons under different cameras from massive repositories of unlabeled images. However, in the current research, there are some existing problems, such as the influence of pedestrians appearing across cameras and pseudo-label noise. To solve these problems, we conduct research in two ways: removing the camera bias and dynamically updating the memory model. In removing the camera bias, based on a learnable channel attention module, the features that are only related to cameras can be extracted from the feature map, thereby removing the camera bias in the global features and obtaining the features that can represent the pedestrians. In regards to dynamically updating the memory model, since the instance features do not necessarily belong to the identity represented by the pseudo-label, we adopt a method to update the memory dynamically according to the distance between the instance features and the category features so that the category features tend to be true. We combine the removal of the camera bias and the dynamic updating of the memory model to better solve problems in this field. Extensive experimentation demonstrates the superiority of our method over the state-of-the-art approaches on fully unsupervised Re-ID tasks.
    This paper improves unsupervised person reidentification by removing the camera bias and dynamically updating the memory model.
    In recent years, unsupervised person reidentification technology has made great strides. The technology retrieves images of interested persons under different cameras from massive repositories of unlabeled images. However, in the current research, there are some existing problems, such as the influence of pedestrians appearing across cameras and pseudo-label noise. To solve these problems, we conduct research in two ways: removing the camera bias and dynamically updating the memory model. In removing the camera bias, based on a learnable channel attention module, the features that are only related to cameras can be extracted from the feature map, thereby removing the camera bias in the global features and obtaining the features that can represent the pedestrians. In regards to dynamically updating the memory model, since the instance features do not necessarily belong to the identity represented by the pseudo-label, we adopt a method to update the memory dynamically according to the distance between the instance features and the category features so that the category features tend to be true. We combine the removal of the camera bias and the dynamic updating of the memory model to better solve problems in this field. Extensive experimentation demonstrates the superiority of our method over the state-of-the-art approaches on fully unsupervised Re-ID tasks.
    • We design a channel attention module to separate camera-related features from global ones and obtain intra-class shared features.
    • We propose a mechanism to dynamically update the memory dictionary according to the distance between the instance and cluster features.
    • Extensive experiments on mainstream datasets Market1501 and DukeMTMC-Re-ID prove that our method outperforms state-of-the-art methods.

  • loading
  • [1]
    Huang Y, Peng P, Jin Y, et al. Domain adaptive attention learning for unsupervised person re-identification. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34: 11069–11076. doi: 10.1609/aaai.v34i07.6762
    [2]
    Ge Y, Zhu F, Chen D, et al. Self-paced contrastive learning with hybrid memory for domain adaptive object re-ID. In: NIPS’20: 34th International Conference on Neural Information Processing Systems. BC, Canada: Curran Associates Inc, 2020: 11309–11321.
    [3]
    Chen H, Lagadec B, Brémond F. ICE: Inter-instance contrastive encoding for unsupervised person re-identification. In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, Canada: IEEE, 2021: 14940–14949.
    [4]
    Wang M, Lai B, Huang J, et al. Camera-aware proxies for unsupervised person re-identification. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35 (4): 2764–2772. doi: 10.1609/aaai.v35i4.16381
    [5]
    Ge Y, Chen D, Li H. Mutual mean-teaching: Pseudo label refinery for unsupervised domain adaptation on person re-identification. 2020. https://doi.org/10.48550/arXiv.2001.01526. Accessed December 3, 2021
    [6]
    Zheng K, Lan C, Zeng W, et al. Exploiting sample uncertainty for domain adaptive person re-identification. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35 (4): 3538–3546. doi: 10.1609/aaai.v35i4.16468
    [7]
    Lin Y, Xie L, Wu Y, et al. Unsupervised person re-identification via softened similarity learning. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, USA: IEEE, 2020: 3387–3396.
    [8]
    Tian J, Tang Q, Li R, et al. A camera identity-guided distribution consistency method for unsupervised multi-target domain person re-identification. ACM Transactions on Intelligent Systems and Technology, 2021, 12 (4): 1–18. doi: 10.1145/3454130
    [9]
    Xuan S, Zhang S. Intra-inter camera similarity for unsupervised person re-identification. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, USA: IEEE, 2021: 11921–11930.
    [10]
    Zheng L, Shen L, Tian L, et al. Scalable person re-identification: A benchmark. In: 2015 IEEE International Conference on Computer Vision (ICCV). Santiago, Chile: IEEE, 2015: 1116–1124.
    [11]
    Ristani E, Solera F, Zou R, et al. Performance measures and a data set for multi-target, multi-camera tracking. In: Hua G, Jégou H, editors. Computer Vision–ECCV 2016 Workshops. Cham: Springer International Publishing, 2016: 17–35.
    [12]
    Song L, Wang C, Zhang L, et al. Unsupervised domain adaptive re-identification: Theory and practice. Pattern Recognition, 2020, 102: 107173. doi: 10.1016/j.patcog.2019.107173
    [13]
    Fu Y, Wei Y, Wang G, et al. Self-similarity grouping: A simple unsupervised cross domain adaptation approach for person re-identification. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, South Korea: IEEE, 2019: 6111–6120.
    [14]
    Lin Y, Dong X, Zheng L, et al. A bottom-up clustering approach to unsupervised person re-identification. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33 (1): 8738–8745. doi: 10.1609/aaai.v33i01.33018738
    [15]
    Zeng K, Ning M, Wang Y, et al. Hierarchical clustering with hard-batch triplet loss for person re-identification. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, USA: IEEE, 2020: 13654–13662.
    [16]
    Zhang X, Cao J, Shen C, et al. Self-training with progressive augmentation for unsupervised cross-domain person re-identification. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, South Korea: IEEE, 2019: 8221–8230.
    [17]
    Yu H X, Zheng W S, Wu A, et al. Unsupervised person re-identification by soft multilabel learning. In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, USA: IEEE, 2019: 2143–2152.
    [18]
    Wang D, Zhang S. Unsupervised person re-identification via multi-label classification. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, USA: IEEE, 2020: 10978–10987.
    [19]
    Wei L, Zhang S, Gao W, et al. Person transfer GAN to bridge domain gap for person re-identification. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018: 79–88.
    [20]
    Zhao L, Li X, Zhuang Y, et al. Deeply-learned part-aligned representations for person re-identification. In: 2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy: IEEE, 2017: 3239–3248.
    [21]
    Zou Y, Yang X, Yu Z, et al. Joint disentangling and adaptation for cross-domain person re-identification. In: Vedaldi A, Bischof H, Brox T, editors. Computer Vision-ECCV 2020. Cham: Springer International Publishing, 2020: 87–104.
    [22]
    Choi Y, Choi M, Kim M, et al. StarGAN: Unified generative adversarial networks for multi-domain image-to-image translation. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018: 8789–8797.
    [23]
    Yang F, Zhong Z, Luo Z, et al. Joint noise-tolerant learning and meta camera shift adaptation for unsupervised person re-identification. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, USA: IEEE, 2021: 4853–4862.
    [24]
    van den Oord A, Li Y, Vinyals O. Representation learning with contrastive predictive coding. 2018. https://doi.org/10.48550/arXiv.1807.03748. Accessed December 12, 2021.
    [25]
    Zheng Y, Tang S, Teng G, et al. Online pseudo label generation by hierarchical cluster dynamics for adaptive person re-identification. In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, Canada: IEEE, 2021: 8351–8361.
    [26]
    Wu Y, Huang T, Yao H, et al. Multi-centroid representation network for domain adaptive person re-ID. Proceedings of the AAAI Conference on Artificial Intelligence, 2022, 36 (5): 2750–2758. doi: 10.1609/aaai.v36i3.20178
    [27]
    Wu Y, Wu X, Li X, et al. MGH: metadata guided hypergraph modeling for unsupervised person re-identification. In: MM '21: Proceedings of the 29th ACM International Conference on Multimedia. New York: Association for Computing Machinery, 2021: 1571–1580.
  • 加载中

Catalog

    Figure  1.  Illustration of our pipeline. We first apply the channel attention module after the backbone network to explicitly separate the camera-related information from the feature maps. Then, we utilize a new mechanism of dynamically updating the memory dictionary according to the distance between the instance feature and the cluster feature.

    Figure  2.  The CA module.

    Figure  3.  Visualization of the attention map in the CA module.

    Figure  4.  T-SNE visualization of 7 classes in the Market1501 test set between our method with d-up removed (left) and our method (right). We utilize a new mechanism of dynamically updating the memory dictionary according to the distance between the instance and cluster feature.

    Figure  5.  The effect of batch size.

    [1]
    Huang Y, Peng P, Jin Y, et al. Domain adaptive attention learning for unsupervised person re-identification. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34: 11069–11076. doi: 10.1609/aaai.v34i07.6762
    [2]
    Ge Y, Zhu F, Chen D, et al. Self-paced contrastive learning with hybrid memory for domain adaptive object re-ID. In: NIPS’20: 34th International Conference on Neural Information Processing Systems. BC, Canada: Curran Associates Inc, 2020: 11309–11321.
    [3]
    Chen H, Lagadec B, Brémond F. ICE: Inter-instance contrastive encoding for unsupervised person re-identification. In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, Canada: IEEE, 2021: 14940–14949.
    [4]
    Wang M, Lai B, Huang J, et al. Camera-aware proxies for unsupervised person re-identification. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35 (4): 2764–2772. doi: 10.1609/aaai.v35i4.16381
    [5]
    Ge Y, Chen D, Li H. Mutual mean-teaching: Pseudo label refinery for unsupervised domain adaptation on person re-identification. 2020. https://doi.org/10.48550/arXiv.2001.01526. Accessed December 3, 2021
    [6]
    Zheng K, Lan C, Zeng W, et al. Exploiting sample uncertainty for domain adaptive person re-identification. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35 (4): 3538–3546. doi: 10.1609/aaai.v35i4.16468
    [7]
    Lin Y, Xie L, Wu Y, et al. Unsupervised person re-identification via softened similarity learning. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, USA: IEEE, 2020: 3387–3396.
    [8]
    Tian J, Tang Q, Li R, et al. A camera identity-guided distribution consistency method for unsupervised multi-target domain person re-identification. ACM Transactions on Intelligent Systems and Technology, 2021, 12 (4): 1–18. doi: 10.1145/3454130
    [9]
    Xuan S, Zhang S. Intra-inter camera similarity for unsupervised person re-identification. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, USA: IEEE, 2021: 11921–11930.
    [10]
    Zheng L, Shen L, Tian L, et al. Scalable person re-identification: A benchmark. In: 2015 IEEE International Conference on Computer Vision (ICCV). Santiago, Chile: IEEE, 2015: 1116–1124.
    [11]
    Ristani E, Solera F, Zou R, et al. Performance measures and a data set for multi-target, multi-camera tracking. In: Hua G, Jégou H, editors. Computer Vision–ECCV 2016 Workshops. Cham: Springer International Publishing, 2016: 17–35.
    [12]
    Song L, Wang C, Zhang L, et al. Unsupervised domain adaptive re-identification: Theory and practice. Pattern Recognition, 2020, 102: 107173. doi: 10.1016/j.patcog.2019.107173
    [13]
    Fu Y, Wei Y, Wang G, et al. Self-similarity grouping: A simple unsupervised cross domain adaptation approach for person re-identification. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, South Korea: IEEE, 2019: 6111–6120.
    [14]
    Lin Y, Dong X, Zheng L, et al. A bottom-up clustering approach to unsupervised person re-identification. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33 (1): 8738–8745. doi: 10.1609/aaai.v33i01.33018738
    [15]
    Zeng K, Ning M, Wang Y, et al. Hierarchical clustering with hard-batch triplet loss for person re-identification. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, USA: IEEE, 2020: 13654–13662.
    [16]
    Zhang X, Cao J, Shen C, et al. Self-training with progressive augmentation for unsupervised cross-domain person re-identification. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, South Korea: IEEE, 2019: 8221–8230.
    [17]
    Yu H X, Zheng W S, Wu A, et al. Unsupervised person re-identification by soft multilabel learning. In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, USA: IEEE, 2019: 2143–2152.
    [18]
    Wang D, Zhang S. Unsupervised person re-identification via multi-label classification. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, USA: IEEE, 2020: 10978–10987.
    [19]
    Wei L, Zhang S, Gao W, et al. Person transfer GAN to bridge domain gap for person re-identification. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018: 79–88.
    [20]
    Zhao L, Li X, Zhuang Y, et al. Deeply-learned part-aligned representations for person re-identification. In: 2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy: IEEE, 2017: 3239–3248.
    [21]
    Zou Y, Yang X, Yu Z, et al. Joint disentangling and adaptation for cross-domain person re-identification. In: Vedaldi A, Bischof H, Brox T, editors. Computer Vision-ECCV 2020. Cham: Springer International Publishing, 2020: 87–104.
    [22]
    Choi Y, Choi M, Kim M, et al. StarGAN: Unified generative adversarial networks for multi-domain image-to-image translation. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018: 8789–8797.
    [23]
    Yang F, Zhong Z, Luo Z, et al. Joint noise-tolerant learning and meta camera shift adaptation for unsupervised person re-identification. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, USA: IEEE, 2021: 4853–4862.
    [24]
    van den Oord A, Li Y, Vinyals O. Representation learning with contrastive predictive coding. 2018. https://doi.org/10.48550/arXiv.1807.03748. Accessed December 12, 2021.
    [25]
    Zheng Y, Tang S, Teng G, et al. Online pseudo label generation by hierarchical cluster dynamics for adaptive person re-identification. In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, Canada: IEEE, 2021: 8351–8361.
    [26]
    Wu Y, Huang T, Yao H, et al. Multi-centroid representation network for domain adaptive person re-ID. Proceedings of the AAAI Conference on Artificial Intelligence, 2022, 36 (5): 2750–2758. doi: 10.1609/aaai.v36i3.20178
    [27]
    Wu Y, Wu X, Li X, et al. MGH: metadata guided hypergraph modeling for unsupervised person re-identification. In: MM '21: Proceedings of the 29th ACM International Conference on Multimedia. New York: Association for Computing Machinery, 2021: 1571–1580.

    Article Metrics

    Article views (257) PDF downloads(1283)
    Proportional views

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return