Abstract
Most existing methods of unsupervised person re-identification (re-ID) still suffer from two aspects of challenges: inter-domain inconsistency and pseudo-label inaccuracy. To alleviate the two problems, we propose an reinforced domain adaptation (RDA) re-ID method by innovatively employing adversarial learning and spatial-channel attention. Specifically, to handle the inter-domain inconsistency problem, we specially design an adversarial learning module to reduce the feature discrepancy between target domain image and translated source domain image, and take Wasserstein distance as the discriminative function because that it can provide an effective gradient for model optimization regardless of the distribution difference between the source domain and the target domain. To handle the pseudo-label inaccuracy problem, we design an attention module to highlight the person region of the image so as to improve accuracy of person clustering and matching. An improved re-ID model can therefore be obtained by jointly training the translated source-domain images with ground-truth identities and target-domain images with pseudo identities. In addition, in order to maintain the semantic consistency of source domain images before and after style translation, we design a closed-loop training mechanism to refine the style translation based on the feedback from person re-ID result, finally making the style translation and person re-ID collaboratively converge to their best state. In the experiments, our proposed framework is shown to outperform state-of-the-art methods on multiple tasks of unsupervised person re-ID.
Similar content being viewed by others
References
Deng W, Zheng L, Ye Q, Kang G, Yang Y, Jiao J (2018) Image-image domain adaptation with preserved self-similarity and domain-dissimilarity for person re-identification. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 994–1003
Wei L, Zhang S, Gao W, Tian Q (2018) Person transfer gan to bridge domain gap for person re-identification. In: Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), pp 79–88
Zhong Z, Zheng L, Zheng Z, Li S, Yang Y (2019) Camstyle: A novel data augmentation method for person re-identification. IEEE Trans Image Process 28(3):1176–1190. https://doi.org/10.1109/TIP.2018.2874313
Li Y-J, Lin C-S, Lin Y-B, Wang Y-C F (2019) Cross-dataset person re-identification via unsupervised pose disentanglement and adaptation. In: Proceedings of the IEEE/CVF international conference on computer vision (ICCV), pp 7918–7928
Zhong Z, Zheng L, Li S, Yang Y (2018) Generalizing a person retrieval model hetero-and homogeneously. In: Proceedings of the European conference on computer vision (ECCV), pp 172–188
Zhu J-Y, Park T, Isola P, Efros A (2017) Unpaired image-to-image translation using cycle-consistent adversarial networks. In: Proceedings of the IEEE international conference on computer vision (ICCV), pp 2242–2251
Arjovsky M, Chintala S, Bottou L (2017) Wasserstein gan
Fan H, Zheng L, Yang Y (2017) Unsupervised person re-identification: Clustering and fine-tuning. ACM Trans Multimed Comput Commun Appl 14 https://doi.org/10.1145/3243316
Zhang X, Cao J, Shen C, You M (2019) Self-training with progressive augmentation for unsupervised cross-domain person re-identification. In: Proceedings of the IEEE/CVF international conference on computer vision (ICCV), pp 8221–8230
Fu Y, Wei Y, Wang G, Zhou Y, Shi H, Uiuc U, Huang T (2019) Self-similarity grouping: A simple unsupervised cross domain adaptation approach for person re-identification. In: Proceedings of the IEEE/CVF international conference on computer vision (ICCV), pp 6111–6120
Chen S, Fan Z, Yin J (2020) Pseudo label based on multiple clustering for unsupervised cross-domain person re-identification. IEEE Signal Process Lett 27:1460–1464. https://doi.org/10.1109/LSP.2020.3016528https://doi.org/10.1109/LSP.2020.3016528
Lin Y, Dong X, Zheng L, Yang Y (2019) A bottom-up clustering approach to unsupervised person re-identification. Proc AAAI Conf Artif Intell 33:8738–8745. https://doi.org/10.1609/aaai.v33i01.33018738https://doi.org/10.1609/aaai.v33i01.33018738
Yang F, Li K, Zhong Z, Luo Z, Sun X, Cheng H, Guo X, Huang F, Ji R, Li S (2020) Asymmetric co-teaching for unsupervised cross-domain person re-identification. Proc AAAI Conf Artif Intell 34:12597–12604. https://doi.org/10.1609/aaai.v34i07.6950
Duan L, Xu D, Chang S-F (2012) Exploiting web images for event recognition in consumer videos: A multiple source domain adaptation approach. In: 2012 IEEE Conference on Computer Vision and Pattern Recognition, pp 1338–1345
Rozantsev A, Salzmann M, Fua P (2019) Beyond sharing weights for deep domain adaptation. IEEE Trans Pattern Anal Mach Intell 41(4):801–814. https://doi.org/10.1109/TPAMI.2018.2814042
Ghifary M, Kleijn W B, Zhang M (2014) Domain adaptive neural networks for object recognition. In: Pacific Rim international conference on artificial intelligence. Springer, pp 898–904
Tzeng E, Hoffman J, Zhang N, Saenko K, Darrell T (201412) Deep domain confusion: Maximizing for domain invariance. arXiv:1412.3474
Ganin Y, Ustinova E, Ajakan H, Germain P, Larochelle H, Laviolette F, Marchand M, Lempitsky V (2016) Domain-adversarial training of neural networks. J Mach Learn Res 17(1):2096–2030
Geng B, Tao D (2011) Daml: Domain adaptation metric learning. IEEE Trans Image Process 20:2980–2989. https://doi.org/10.1109/TIP.2011.2134107https://doi.org/10.1109/TIP.2011.2134107
Redko I, Habrard A, Sebban M (2017) Theoretical analysis of domain adaptation with optimal transport. In: Joint european conference on machine learning and knowledge discovery in databases. Springer, pp 737–753
Cuturi M (2013) Sinkhorn distances: Lightspeed computation of optimal transportation distances. Adv Neural Information Process Syst 26
Tang Y (2020) Cgan-tm: A novel domain-to-domain transferring method for person re-identification. IEEE Trans Image Process PP. https://doi.org/10.1109/TIP.2020.2985545
Astha V, Venkata S, Wang Z, Satoh S, Shah R (2021) Unsupervised domain adaptation for person re-identification via individual-preserving and environmental-switching cyclic generation. IEEE Trans Multimed PP:1–1. https://doi.org/10.1109/TMM.2021.3126404https://doi.org/10.1109/TMM.2021.3126404
Ge Y, Chen D, Li H (2020) Mutual mean-teaching: Pseudo label refinery for unsupervised domain adaptation on person re-identification. arXiv:2001.01526
Tay C-P, Roy S, Yap K-H (2019) Aanet: Attribute attention network for person re-identifications. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp 7127–7136
Gao S, Wang J, Lu H, Zimo L (2020) Pose-guided visible part matching for occluded person reid. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), pp 11741–11749
Yang J, Zhang C, Tang Y, Li Z (2022) Pafm: pose-drive attention fusion mechanism for occluded person re-identification. Neural Comput Appl:1–12. https://doi.org/10.1007/s00521-022-06903-4
Wang F, Jiang M, Qian C, Yang S, Li C, Zhang H, Wang X, Tang X (2017) Residual attention network for image classification. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 6450–6458
Taigman Y, Polyak A, Wolf L (2016) Unsupervised cross-domain image generation
Villani C (2009) Optimal transport: old and new, vol 338. Springer
Gulrajani I, Ahmed F, Arjovsky M, Dumoulin V, Courville A (2017) Improved training of wasserstein gans
Ester M, Kriegel H-P, Sander J, Xu X et al (1996) A density-based algorithm for discovering clusters in large spatial databases with noise.. In: kdd, vol 96, pp 226–231
Hu J, Shen L, Sun G, Albanie S (2017) Squeeze-and-excitation networks. IEEE Trans Pattern Anal Mach Intell PP. https://doi.org/10.1109/TPAMI.2019.2913372
Zheng L, Shen L, Tian L, Wang S, Wang J, Tian Q (2015) Scalable person re-identification: A benchmark. In: Proceedings of the IEEE international conference on computer vision (ICCV), pp 1116–1124
Ristani E, Solera F, Zou R, Cucchiara R, Tomasi C (2016) Performance measures and a data set for multi-target, multi-camera tracking. In: European conference on computer vision, vol 9914
He K, Zhang X, Ren S, Sun J (2016) Deep residual learning for image recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 770–778
Zhang H, Wu C, Zhang Z, Zhu Y, Lin H, Zhang Z, Sun Y, He T, Mueller J, Manmatha R et al (2020) Resnest: Split-attention networks. arXiv:2004.08955
Deng J, Dong W, Socher R, Li L-J, Li K, Li F-F (2009) Imagenet: a large-scale hierarchical image database. In: 2009 IEEE conference on computer vision and pattern recognition, pp 248–255
Radford A, Metz L, Chintala S (2016) Unsupervised representation learning with deep convolutional generative adversarial networks
Zhong Z, Zheng L, Kang G, Li S, Yang Y (2017) Random erasing data augmentation. Proc AAAI Conf Artif Intell 34. https://doi.org/10.1609/aaai.v34i07.7000
Zhong Z, Zheng L, Luo Z, Li S, Yang Y (2020) Learning to adapt invariance in memory for person re-identification. IEEE Trans Pattern Anal Mach Intell PP:1–1. https://doi.org/10.1109/TPAMI.2020.2976933https://doi.org/10.1109/TPAMI.2020.2976933
Zou Y, Yang X, Yu Z, Kumar BVK, Kautz J (2020) Joint disentangling and adaptation for cross-domain person re-identification. In: European conference on computer vision. Springer, pp 87– 104
Song L, Wang C, Zhang L, Du B, Zhang Q, Huang C, Wang X (2020) Unsupervised domain adaptive re-identification: Theory and practice. Pattern Recogn 102:107173. https://doi.org/10.1016/j.patcog.2019.107173https://doi.org/10.1016/j.patcog.2019.107173, https://www.sciencedirect.com/science/article/pii/S003132031930473X
Yu H-X, Zheng W-S, Wu A, Guo X, Gong S, Lai J-H (2019) Unsupervised person re-identification by soft multilabel learning. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp 2143–2152
Yuan Y, Chen W, Chen T, Yang Y, Ren Z, Wang Z, Hua G (2020) Calibrated domain-invariant learning for highly generalizable large scale re-identification. In: Proceedings of the IEEE/CVF winter conference on applications of computer vision, pp 3578–3587
Yang Q, Yu H-X, Wu A, Zheng W-S (2019) Patch-based discriminative feature learning for unsupervised person re-identification. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp 3628–3637
Acknowledgments
This work is supported by the National Natural Science Foundation of China (Nos. 61866004, 61966004, 61962007), the Guangxi Natural Science Foundation (Nos. 2018GXNSFDA281009, 2019GXNSFDA245018, 2018GXNSFDA294001), Research Fund of Guangxi Key Lab of Multi-source Information Mining & Security (No.20-A-03-01), Guangxi “Bagui Scholar” Teams for Innovation and Research Project, and Innovation Project of Guangxi Graduate Education(JXXYYJSCXXM-2021-007).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Wei, P., Zhang, C., Tang, Y. et al. Reinforced domain adaptation with attention and adversarial learning for unsupervised person Re-ID. Appl Intell 53, 4109–4123 (2023). https://doi.org/10.1007/s10489-022-03640-y
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10489-022-03640-y