簡易檢索 / 詳目顯示

研究生: 蔡詠祺
Tsai, Yung-Chi
論文名稱: 一個增強版之跨域行人重辨識架構
An Enhanced Cross-Domain Person Re-Identification Framework
指導教授: 戴顯權
Tai, Shen-Chuan
學位類別: 碩士
Master
系所名稱: 電機資訊學院 - 電機工程學系
Department of Electrical Engineering
論文出版年: 2021
畢業學年度: 109
語文別: 英文
論文頁數: 83
中文關鍵詞: 深度學習行人重辨識資料隔閡非監督式學習跨域偽標籤
外文關鍵詞: deep learning, person-reidentification, domain gap, unsupervised learning, cross-domain, pseudo labels
相關次數: 點閱:149下載:5
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報
  • 行人重辨識為近年來的重要議題,其可以用來快速查找出感興趣的行人,因此在公共安全上以及智慧檢索等應用可以看到該研究的蹤影。行人重辨識建立在跨相機的情境下,去匹配出同樣行人身分的影像。然而一個資料集中,不同相機之間的影像會因為成像的變化性以及內容的不一致性導致傳統方法萃取的特徵不夠全面。因此許多研究利用深度學習抓取出相機不變性的行人特徵,並且達到不錯的表現。但是考慮到實際應用場景,將一個預訓練過的行人重辨識模型測試在另外一個行人資料集時,卻發現模型的表現大幅降低,這並不符合期待。而這個問題來自於資料隔閡,近年來跨域的行人重辨識研究針對該議題改進,並透過非監督式學習來去除人工標記資料所需要付出的心力以進一步接近實際應用。然而非監督式學習的通病在於未知領域的探索會影響模型在訓練過程中的正確性,因此這篇論文透過增強模型架構、設計更全面的優化方式、改善特徵分布的正確性,讓萃取完的特徵能更具有代表性。為了評估方法的有效性,進一步將本論文提出的方法和近期最好的研究做了比較。當使用DukeMTMC-reID為源域、Market1501為目標域時,本論文的方法在mAP上超越該研究3.7%。當使用Market1501為源域、DukeMTMC-reID為目標域時,本論文的方法在mAP上超越該研究1.5%。

    Person re-identification (Re-ID) has been an important issue in recent years. Re-ID can be used to quickly find a person of interest. Thus, the traces of Re-ID can be founded in public safety and smart retrieval. Re-ID is based on a cross-camera scenario to match images of the same identity. However, in a dataset, the images among different cameras cause the features extracted by conventional methods to be insufficient due to the variance of imaging and the inconsistency of content. Thus, many studies utilize deep learning to capture features that are camera-invariant. But considering the actual application scenarios, when testing a pre-trained Re-ID model on another pedestrian dataset, the performance of the model is greatly reduced, which does not meet the expectations. This problem comes from the domain gap. In recent years, many cross-domain Re-ID studies have improved on this topic and used unsupervised learning to remove the effort required for manual labeling to further achieve practical applications. However, the common problem of unsupervised learning is that the exploration of unknown domains affects the correctness of the model during the training process. Therefore, the proposed methods enhance the model architecture, design more comprehensive optimization methods, and elevate the effectiveness of feature distribution, so the features can be more representative. To evaluate the effectiveness of the proposed method, it is compared with the best recent research. When leveraging DukeMTMC-reID as the source domain and Market1501 as the target domain, the proposed method surpasses the study by 3.7% on mAP. When using Market1501 as the source domain and DukeMTMC-reID as the target domain, the proposed method surpasses the study by 1.5% on mAP.

    摘 要 i Abstract ii Acknowledgments iv Contents v List of Tables viii List of Figures x Chapter 1 Introduction 1 1.1 Overview of Person Re-identification 1 1.2 Overview of Learning Methodology 5 1.2.1 Conventional Methods 5 1.2.2 Deep Learning 6 1.3 Issues of Re-ID 7 Chapter 2 Related Works 11 2.1 Supervised Re-ID 11 2.2 Unsupervised Cross-domain Re-ID 12 2.2.1 Pseudo Label Enhancement 12 2.2.2 Domain Discrepancy Mitigation 14 2.2.3 Image Transformation 14 2.3 Network Architecture 15 2.3.1 ResNet50 15 2.3.2 ResNet50-IBN 17 Chapter 3 Proposed Methods 20 3.1 Training Framework 20 3.1.1 Purposes and Advantages 20 3.1.2 Realization 22 3.2 Pseudo Label Enhancement 24 3.2.1 Clustering Method 24 3.2.2 Distance Improvement 27 3.2.3 Backbone Refinement 29 3.3 Loss Functions 35 3.3.1 Classification Loss 35 3.3.2 Batch Triplet Loss 36 3.3.3 Batch Center Loss 39 3.3.4 Overall Loss 41 3.4 Algorithm Flow 42 3.4.1 Flowcharts 42 3.4.2 Pseudo Codes 44 Chapter 4 Experimental Results 46 4.1 Datasets and Evaluation Metric 46 4.2 Implementation Details 49 4.3 Ablation Study 51 4.3.1 Effectiveness of Each Component 51 4.3.2 Placement of Advanced Attention Mechanism 54 4.3.3 Parameter Study 57 4.3.4 Recognition Capability for Images of Different Cases 60 4.3.5 Self-collected Pedestrian Dataset 63 4.4 Comparison with State-of-the-arts 68 4.4.1 Performance of Framework 68 4.4.2 Number of Parameters of Backbone 72 Chapter 5 Conclusion and Future Work 75 5.1 Conclusion 75 5.2 Future Work 76 References 77

    [1] Plantinga, A. (1961). "Things and Persons." The Review of Metaphysics: 493-519.
    [2] Li, X., W.-S. Zheng, X. Wang, T. Xiang and S. Gong (2015). Multi-Scale Learning for Low-Resolution Person Re-Identification. Proceedings of the IEEE International Conference on Computer Vision.
    [3] Ye, M., J. Shen, G. Lin, T. Xiang, L. Shao and S. C. Hoi (2021). "Deep Learning for Person Re-Identification: A Survey and Outlook." IEEE Transactions on Pattern Analysis and Machine Intelligence.
    [4] Gheissari, N., T. B. Sebastian and R. Hartley (2006). Person Reidentification Using Spatiotemporal Appearance. 2006 IEEE computer society conference on computer vision and pattern recognition (CVPR'06), IEEE.
    [5] Huang, Y., Q. Wu, J. Xu and Y. Zhong (2019). Sbsgan: Suppression of Inter-Domain Background Shift for Person Re-Identification. Proceedings of the IEEE/CVF International Conference on Computer Vision.
    [6] Song, C., Y. Huang, W. Ouyang and L. Wang (2018). Mask-Guided Contrastive Attention Model for Person Re-Identification. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.
    [7] Tian, M., S. Yi, H. Li, S. Li, X. Zhang, J. Shi, J. Yan and X. Wang (2018). Eliminating Background-Bias for Robust Person Re-Identification. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.
    [8] Liu, J., B. Ni, Y. Yan, P. Zhou, S. Cheng and J. Hu (2018). Pose Transferrable Person Re-Identification. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.
    [9] Zheng, L., Y. Huang, H. Lu and Y. Yang (2019). "Pose-Invariant Embedding for Deep Person Re-Identification." IEEE Transactions on Image Processing 28(9): 4500-4509.
    [10] Karanam, S., Y. Li and R. J. Radke (2015). Person Re-Identification with Discriminatively Trained Viewpoint Invariant Dictionaries. Proceedings of the IEEE international conference on computer vision.
    [11] Huang, Y., Z.-J. Zha, X. Fu and W. Zhang (2019). Illumination-Invariant Person Re-Identification. Proceedings of the 27th ACM International Conference on Multimedia.
    [12] Zheng, L., L. Shen, L. Tian, S. Wang, J. Wang and Q. Tian (2015). Scalable Person Re-Identification: A Benchmark. Proceedings of the IEEE international conference on computer vision.
    [13] Ma, B., Y. Su and F. Jurie (2012). Local Descriptors Encoded by Fisher Vectors for Person Re-Identification. European conference on computer vision, Springer.
    [14] Li, W., R. Zhao and X. Wang (2012). Human Reidentification with Transferred Metric Learning. Asian conference on computer vision, Springer.
    [15] Hirzer, M., P. M. Roth, M. Köstinger and H. Bischof (2012). Relaxed Pairwise Learned Metric for Person Re-Identification. European conference on computer vision, Springer.
    [16] Gray, D. and H. Tao (2008). Viewpoint Invariant Pedestrian Recognition with an Ensemble of Localized Features. European conference on computer vision, Springer.
    [17] Zheng, W.-S., S. Gong and T. Xiang (2011). Person Re-Identification by Probabilistic Relative Distance Comparison. CVPR 2011, IEEE.
    [18] Tay, C.-P., S. Roy and K.-H. Yap (2019). Aanet: Attribute Attention Network for Person Re-Identifications. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.
    [19] Martinel, N., G. Luca Foresti and C. Micheloni (2019). Aggregating Deep Pyramidal Representations for Person Re-Identification. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops.
    [20] Zhu, Z., X. Jiang, F. Zheng, X. Guo, F. Huang, X. Sun and W. Zheng (2020). Aware Loss with Angular Regularization for Person Re-Identification. Proceedings of the AAAI Conference on Artificial Intelligence.
    [21] Zhang, Z., C. Lan, W. Zeng and Z. Chen (2019). Densely Semantically Aligned Person Re-Identification. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.
    [22] Wang, H., Y. Fan, Z. Wang, L. Jiao and B. Schiele (2018). "Parameter-Free Spatial Attention Network for Person Re-Identification." arXiv preprint arXiv:1811.12150.
    [23] Jin, X., C. Lan, W. Zeng, Z. Chen and L. Zhang (2020). Style Normalization and Restitution for Generalizable Person Re-Identification. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.
    [24] Zheng, Z., L. Zheng and Y. Yang (2017). Unlabeled Samples Generated by Gan Improve the Person Re-Identification Baseline in Vitro. Proceedings of the IEEE International Conference on Computer Vision.
    [25] Quiñonero-Candela, J., M. Sugiyama, N. D. Lawrence and A. Schwaighofer (2009). Dataset Shift in Machine Learning, Mit Press.
    [26] Fan, H., L. Zheng, C. Yan and Y. Yang (2018). "Unsupervised Person Re-Identification: Clustering and Fine-Tuning." ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM) 14(4): 1-18.
    [27] Zhang, X., J. Cao, C. Shen and M. You (2019). Self-Training with Progressive Augmentation for Unsupervised Cross-Domain Person Re-Identification. Proceedings of the IEEE/CVF International Conference on Computer Vision.
    [28] Zhong, Z., L. Zheng, D. Cao and S. Li (2017). Re-Ranking Person Re-Identification with K-Reciprocal Encoding. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.
    [29] Hermans, A., L. Beyer and B. Leibe (2017). "In Defense of the Triplet Loss for Person Re-Identification." arXiv preprint arXiv:1703.07737.
    [30] Chen, G., Y. Lu, J. Lu and J. Zhou (2020). Deep Credible Metric Learning for Unsupervised Domain Adaptation Person Re-Identification. Proc. Eur. Conf. Comput. Vis, Springer.
    [31] Pereira, T. d. C. and T. E. de Campos (2021). "Learn by Guessing: Multi-Step Pseudo-Label Refinement for Person Re-Identification." arXiv preprint arXiv:2101.01215.
    [32] Fu, Y., Y. Wei, G. Wang, Y. Zhou, H. Shi and T. S. Huang (2019). Self-Similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-Identification. Proceedings of the IEEE/CVF International Conference on Computer Vision.
    [33] Zhao, F., S. Liao, G.-S. Xie, J. Zhao, K. Zhang and L. Shao (2020). Unsupervised Domain Adaptation with Noise Resistible Mutual-Training for Person Re-Identification. European Conference on Computer Vision, Springer.
    [34] Ge, Y., D. Chen and H. Li (2020). "Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-Identification." arXiv preprint arXiv:2001.01526.
    [35] Zhai, Y., Q. Ye, S. Lu, M. Jia, R. Ji and Y. Tian (2020). "Multiple Expert Brainstorming for Domain Adaptive Person Re-Identification." arXiv preprint arXiv:2007.01546.
    [36] Zheng, K., W. Liu, L. He, T. Mei, J. Luo and Z.-J. Zha (2021). "Group-Aware Label Transfer for Domain Adaptive Person Re-Identification." arXiv preprint arXiv:2103.12366.
    [37] Yang, F., Z. Zhong, Z. Luo, Y. Cai, Y. Lin, S. Li and N. Sebe (2021). "Joint Noise-Tolerant Learning and Meta Camera Shift Adaptation for Unsupervised Person Re-Identification." arXiv preprint arXiv:2103.04618.
    [38] Zhong, Z., L. Zheng, Z. Luo, S. Li and Y. Yang (2019). Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-Identification. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.
    [39] Qi, L., L. Wang, J. Huo, L. Zhou, Y. Shi and Y. Gao (2019). A Novel Unsupervised Camera-Aware Domain Adaptation Framework for Person Re-Identification. Proceedings of the IEEE/CVF International Conference on Computer Vision.
    [40] Luo, C., C. Song and Z. Zhang (2020). Generalizing Person Re-Identification by Camera-Aware Invariance Learning and Cross-Domain Mixup. European Conference on Computer Vision, Springer.
    [41] Deng, W., L. Zheng, Q. Ye, G. Kang, Y. Yang and J. Jiao (2018). Image-Image Domain Adaptation with Preserved Self-Similarity and Domain-Dissimilarity for Person Re-Identification. Proceedings of the IEEE conference on computer vision and pattern recognition.
    [42] Wei, L., S. Zhang, W. Gao and Q. Tian (2018). Person Transfer Gan to Bridge Domain Gap for Person Re-Identification. Proceedings of the IEEE conference on computer vision and pattern recognition.
    [43] Zhai, Y., S. Lu, Q. Ye, X. Shan, J. Chen, R. Ji and Y. Tian (2020). Ad-Cluster: Augmented Discriminative Clustering for Domain Adaptive Person Re-Identification. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.
    [44] Liu, J., W. Li, H. Pei, Y. Wang, F. Qu, Y. Qu and Y. Chen (2019). "Identity Preserving Generative Adversarial Network for Cross-Domain Person Re-Identification." IEEE Access 7: 114021-114032.
    [45] Zou, Y., X. Yang, Z. Yu, B. Kumar and J. Kautz (2020). "Joint Disentangling and Adaptation for Cross-Domain Person Re-Identification." arXiv preprint arXiv:2007.10315.
    [46] He, K., X. Zhang, S. Ren and J. Sun (2016). Deep Residual Learning for Image Recognition. Proceedings of the IEEE conference on computer vision and pattern recognition.
    [47] Pan, X., P. Luo, J. Shi and X. Tang (2018). Two at Once: Enhancing Learning and Generalization Capacities Via Ibn-Net. Proceedings of the European Conference on Computer Vision (ECCV).
    [48] Ulyanov, D., A. Vedaldi and V. Lempitsky (2016). "Instance Normalization: The Missing Ingredient for Fast Stylization." arXiv preprint arXiv:1607.08022.
    [49] Xuan, S. and S. Zhang (2021). "Intra-Inter Camera Similarity for Unsupervised Person Re-Identification." arXiv preprint arXiv:2103.11658.
    [50] Ge, Y., F. Zhu, D. Chen, R. Zhao and H. Li (2020). "Self-Paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-Id." arXiv preprint arXiv:2006.02713.
    [51] Ester, M., H.-P. Kriegel, J. Sander and X. Xu (1996). A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise. Kdd.
    [52] Hu, J., L. Shen and G. Sun (2018). Squeeze-and-Excitation Networks. Proceedings of the IEEE conference on computer vision and pattern recognition.
    [53] Li, W., R. Zhao, T. Xiao and X. Wang (2014). Deepreid: Deep Filter Pairing Neural Network for Person Re-Identification. Proceedings of the IEEE conference on computer vision and pattern recognition.
    [54] Luo, H., Y. Gu, X. Liao, S. Lai and W. Jiang (2019). Bag of Tricks and a Strong Baseline for Deep Person Re-Identification. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops.
    [55] Kingma, D. P. and J. Ba (2014). "Adam: A Method for Stochastic Optimization." arXiv preprint arXiv:1412.6980.
    [56] Li, Y.-J., C.-S. Lin, Y.-B. Lin and Y.-C. F. Wang (2019). Cross-Dataset Person Re-Identification Via Unsupervised Pose Disentanglement and Adaptation. Proceedings of the IEEE/CVF International Conference on Computer Vision.
    [57] Chen, H., Y. Wang, B. Lagadec, A. Dantcheva and F. Bremond (2020). "Joint Generative and Contrastive Learning for Unsupervised Person Re-Identification." arXiv preprint arXiv:2012.09071.
    [58] Jiang, K., T. Zhang, Y. Zhang, F. Wu and Y. Rui (2020). "Self-Supervised Agent Learning for Unsupervised Cross-Domain Person Re-Identification." IEEE Transactions on Image Processing 29: 8549-8560.
    [59] Wang, G., J.-H. Lai, W. Liang and G. Wang (2020). Smoothing Adversarial Domain Attack and P-Memory Reconsolidation for Cross-Domain Person Re-Identification. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.
    [60] Zhang, H., H. Cao, X. Yang, C. Deng and D. Tao (2021). "Self-Training with Progressive Representation Enhancement for Unsupervised Cross-Domain Person Re-Identification." IEEE Transactions on Image Processing.
    [61] Chen, Y., X. Zhu and S. Gong (2019). Instance-Guided Context Rendering for Cross-Domain Person Re-Identification. Proceedings of the IEEE/CVF International Conference on Computer Vision.
    [62] Chen, S., Z. Fan and J. Yin (2020). "Pseudo Label Based on Multiple Clustering for Unsupervised Cross-Domain Person Re-Identification." IEEE Signal Processing Letters 27: 1460-1464.
    [63] Huang, H., X. Chen and K. Huang (2020). Proxy Task Learning for Cross-Domain Person Re-Identification. 2020 IEEE International Conference on Multimedia and Expo (ICME), IEEE.
    [64] Yang, F., K. Li, Z. Zhong, Z. Luo, X. Sun, H. Cheng, X. Guo, F. Huang, R. Ji and S. Li (2020). Asymmetric Co-Teaching for Unsupervised Cross-Domain Person Re-Identification. Proceedings of the AAAI Conference on Artificial Intelligence.

    下載圖示 校內:2023-09-01公開
    校外:2023-09-01公開
    QR CODE