Dual Attention Matching Network for Context-Aware Feature Sequence based Person Re-Identification

CVPR, pp. 5363-5372, 2018.

Cited by: 148|Bibtex|Views104|DOI:https://doi.org/10.1109/cvpr.2018.00562
EI
Other Links: dblp.uni-trier.de|academic.microsoft.com|arxiv.org
Weibo:
We proposed an end-to-end trainable framework, namely Dual ATtention Matching network, to learn context-aware feature sequences and to perform dually attentive comparison for person ReID

Abstract:

Typical person re-identification (ReID) methods usually describe each pedestrian with a single feature vector and match them in a task-specific metric space. However, the methods based on a single feature vector are not sufficient enough to overcome visual ambiguity, which frequently occurs in real scenario. In this paper, we propose a no...More

Code:

Data:

0
Introduction
  • Person Re-Identification (ReID) aims at associating the same pedestrian across multiple cameras [13, 63], which has attracted rapidly increased attention in the computer vision community due to its importance for many potential applications, such as video surveillance analysis and content-based image/video retrieval.
  • A typical person ReID pipeline usually describes each pedestrian image or video footage with a single feature vector firstly and match them in a task-specific metric space, where the feature vectors of same pedestrian are expected to have smaller dis- X (a) (b) (c).
  • As shown in Fig. 1 (a), different individuals are very similar to each other in appearance, except for some local patterns on skirts.
  • The single feature vector based methods usually pay more attention on the overall appearance rather than the local discriminative parts and fail to yield accurate matching results.
  • As shown in Fig. 1 (d), there are some interference frames in each video sequence, which will heavily contaminate the whole feature vector and lead to mismatching
Highlights
  • Person Re-Identification (ReID) aims at associating the same pedestrian across multiple cameras [13, 63], which has attracted rapidly increased attention in the computer vision community due to its importance for many potential applications, such as video surveillance analysis and content-based image/video retrieval
  • To tackle the challenges mentioned above, we propose a novel end-to-end trainable framework, named Dual ATtention Matching network (DuATM), to jointly learn contextaware feature sequences and perform attentive sequences comparison
  • An obstacle in training DuATM with triplet loss is lack of positive pairs compared with negative ones
  • We find that DuATM achieves the best performance with the settings of (λ1, λ2, p) as (0.3, 0.9, 0.4), (0.1, 0.5, 0.6), and (0.5, 0.5, 0.4), on Market-1501, DukeMTMC-reID, and MARS, respectively
  • We proposed an end-to-end trainable framework, namely Dual ATtention Matching network (DuATM), to learn context-aware feature sequences and to perform dually attentive comparison for person ReID
  • DuATM is trained via a triplet loss assisted with a de-correlation loss and a cross-entropy loss
Methods
  • RNN [69], and QAN [26], in which the salient local patterns are extracted by the attention strategy and aggregated into a single comparable feature vector.
  • DuATM can adaptively infer the semantic correspondence structure between local patterns and automatically remove local corruptions within sequence, the method achieves better performance than body-part based (e.g., SpindleNet [59], DRLPL [52]) and densely-matching based (e.g., SCSP [3]) methods on dataset Market-1501
Results
  • The authors employ the standard metrics as in most person ReID literatures, namely the cumulative matching cure (CMC) and the mean Average Precision.
  • To compute these scores, the authors reimplement the evaluation code provided by [61] in Python.
  • In the loss function of DuATM, there are two parameters λ1 and λ2.
  • To evaluate the influence of these parameters, the authors conduct experiments mAP(%) mAP(%) mAP(%)
Conclusion
  • The authors proposed an end-to-end trainable framework, namely Dual ATtention Matching network (DuATM), to learn context-aware feature sequences and to perform dually attentive comparison for person ReID.
  • The core component of DuATM is a dual attention block, which simultaneously performs feature refinement and feature-pair alignment.
  • DuATM is trained via a triplet loss assisted with a de-correlation loss and a cross-entropy loss.
  • Experiments conducted on large-scale image and video data sets have confirmed the significant advantages of the proposal
Summary
  • Introduction:

    Person Re-Identification (ReID) aims at associating the same pedestrian across multiple cameras [13, 63], which has attracted rapidly increased attention in the computer vision community due to its importance for many potential applications, such as video surveillance analysis and content-based image/video retrieval.
  • A typical person ReID pipeline usually describes each pedestrian image or video footage with a single feature vector firstly and match them in a task-specific metric space, where the feature vectors of same pedestrian are expected to have smaller dis- X (a) (b) (c).
  • As shown in Fig. 1 (a), different individuals are very similar to each other in appearance, except for some local patterns on skirts.
  • The single feature vector based methods usually pay more attention on the overall appearance rather than the local discriminative parts and fail to yield accurate matching results.
  • As shown in Fig. 1 (d), there are some interference frames in each video sequence, which will heavily contaminate the whole feature vector and lead to mismatching
  • Methods:

    RNN [69], and QAN [26], in which the salient local patterns are extracted by the attention strategy and aggregated into a single comparable feature vector.
  • DuATM can adaptively infer the semantic correspondence structure between local patterns and automatically remove local corruptions within sequence, the method achieves better performance than body-part based (e.g., SpindleNet [59], DRLPL [52]) and densely-matching based (e.g., SCSP [3]) methods on dataset Market-1501
  • Results:

    The authors employ the standard metrics as in most person ReID literatures, namely the cumulative matching cure (CMC) and the mean Average Precision.
  • To compute these scores, the authors reimplement the evaluation code provided by [61] in Python.
  • In the loss function of DuATM, there are two parameters λ1 and λ2.
  • To evaluate the influence of these parameters, the authors conduct experiments mAP(%) mAP(%) mAP(%)
  • Conclusion:

    The authors proposed an end-to-end trainable framework, namely Dual ATtention Matching network (DuATM), to learn context-aware feature sequences and to perform dually attentive comparison for person ReID.
  • The core component of DuATM is a dual attention block, which simultaneously performs feature refinement and feature-pair alignment.
  • DuATM is trained via a triplet loss assisted with a de-correlation loss and a cross-entropy loss.
  • Experiments conducted on large-scale image and video data sets have confirmed the significant advantages of the proposal
Tables
  • Table1: Comparison to the baseline model. ∗ We adjust the parameters of loss functions to more appropriate values as obtained in the parameter analysis experiments. ∗∗ The data augmentation is also adopted during the evaluation stage
  • Table2: Ablation study of DuATM on Market1501
  • Table3: Comparison to other attention methods
  • Table4: Comparison to other feature sequence/set based methods
  • Table5: Comparison to state-of-the-art on Market-1501
  • Table6: Comparison to state-of-the-art on DukeMTMC-reID
  • Table7: Comparison to state-of-the-art on MARS. DuATM∗: trained with a larger image size 256 × 128 as suggested in [<a class="ref-link" id="c14" href="#r14">14</a>]
Download tables as Excel
Related work
  • Person ReID systems usually consist of two major components: a) feature extraction and b) metric learning. Previous works on person ReID focus on either constructing informative features, or finding a discriminative distance metric. According to the used representation forms in matching stage, we roughly divide the existing methods into two groups: feature vector based methods, e.g., [10, 4, 19, 37, 41, 17, 12, 45, 35]; and feature set or feature sequence based methods, e.g., [69, 60, 59, 57, 36, 18, 1, 38].

    In feature vector based methods, an image or video is represented by a feature vector and the metric learning is performed based on feature vectors. For example, in [2, 20, 29, 47, 56, 51, 54, 64, 70, 21], hand-crafted local features are integrated into a feature vector, and distance metric is learned by simultaneously maximizing inter-class margins and minimizing intra-class variations. Meanwhile, many recent works directly learn comparable feature embedding from the raw input data via a neural network. For example, in [33, 26], high-quality local patterns are explored from images or videos firstly and then aggregated into informative feature vectors; in [28, 39, 49], local features of recurrent appearance data are extracted and integrated using temporal-pooling strategy; in [14, 5], to enhance the generalization capability of the learned embeddings, the pairwise similarity criterion is extended to triplet or quadruplet. Although these methods mentioned above are able to learn task-specific compact embeddings, these methods still suffer from the mismatching problem, especially when some vital visual details fail to be captured.
Funding
  • The ROSE Lab is supported by the National Research Foundation and the Infocomm Media Development Authority, Singapore
  • Li are supported by Beijing Municipal Science and Technology Commission Project under Grant No Z181100001918005
  • Li is also partially supported by the Open Project Fund from Key Laboratory of Machine Perception (MOE), Peking University
Study subjects and analysis
large-scale data sets: 3
This can be regarded as a simplified version of the interpolation method [11] to augment training dataset. To evaluate our proposal, we conduct extensive experiments on three large-scale data sets, including Market-1501 [62], DukeMTMC-reID [65], and MARS [61]. 4.1

data sets: 3
Experimental results are presented in Table 1. As observed from Table 1 that, the results of DuATM+l(0) consistently outperform that of AvePool+l(0) on all three data sets. This confirms the effectiveness of using dual attention block in DuATM: using context-aware feature sequences with dual attentive matching mechanism is more effective than the average-pooling based single feature vector method

data sets: 3
Rank-5 Rank-10. Sequence Length (b) on three data sets by changing one parameter while fixing the other two. Experimental results are shown in Fig. 6

Reference
  • 2017 TIP[24] 2017 ICCV[25] This paper
    Google ScholarFindings
  • 2017 CVPR[69] 2017 CVPR[26] This paper
    Google ScholarFindings
  • 2016 CVPR 2017 CVPR[59] 2017 ICCV[60] 2017 Arxiv[52] This paper
    Google ScholarFindings
  • 2015 ICCV[62] 2016 CVPR[55] 2017 CVPR[66] 2017 CVPR[2]
    Google ScholarFindings
  • 2016 ECCV[39] 2016 ECCV[38] 2017 TPAMI[8] 2017 CVPR[68] 2017 CVPR[22] 2017 ICCV[65] 2017 CVPR[17] 2017 ICCV[37] 2017 Arxiv[14] 2017 IJCAI[19] 2017 Arxiv[58] 2017 Arxiv[67] 2017 Arxiv[6]
    Google ScholarFindings
  • 2015 ICCV[62] 2015 CVPR[21]
    Google ScholarFindings
  • 2017 ICCV[65] 2017 CVPR[46] 2017 Arxiv[23] 2017 ICCV[37] 2017 ICCVW 2017 Arxiv[67]
    Google ScholarFindings
  • 2017 ICCV[27] 2015 ICCV[62] 2017 ICCV[53] 2017 CVPR[66]
    Google ScholarFindings
  • 2016 ECCV[61] 2017 CVPR[17] 2017 Arxiv[14]
    Google ScholarFindings
  • [1] E. Ahmed, M. Jones, and T. K. Marks. An improved deep learning architecture for person re-identification. In CVPR, pages 3908–3916, 2015. 1, 2
    Google ScholarLocate open access versionFindings
  • [2] S. Bai, X. Bai, and Q. Tian. Scalable person re-identification on supervised smoothed manifold. In CVPR, 2017. 2, 8
    Google ScholarLocate open access versionFindings
  • [3] D. Chen, Z. Yuan, B. Chen, and N. Zheng. Similarity learning with spatial constraints for person re-identification. In CVPR, pages 1268–1277, 2016. 2, 7
    Google ScholarLocate open access versionFindings
  • [4] J. Chen, Y. Wang, J. Qin, L. Liu, and L. Shao. Fast person re-identification via cross-camera semantic binary transformation. In CVPR, 2017. 2
    Google ScholarLocate open access versionFindings
  • [5] W. Chen, X. Chen, J. Zhang, and K. Huang. Beyond triplet loss: a deep quadruplet network for person re-identification. In CVPR, 2017. 2
    Google ScholarLocate open access versionFindings
  • [6] Y. Chen, N. Wang, and Z. Zhang. Darkrank: Accelerating deep metric learning via cross sample similarities transfer. arXiv, 2017. 8
    Google ScholarLocate open access versionFindings
  • [7] Y. Chen, X. Zhu, and S. Gong. Person re-identification by deep learning multi-scale representations. In ICCVW, 2017. 8
    Google ScholarLocate open access versionFindings
  • [8] Y.-C. Chen, X. Zhu, W.-S. Zheng, and J.-H. Lai. Person reidentification by camera correlation aware feature augmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 208
    Google ScholarLocate open access versionFindings
  • [9] M. Cogswell, F. Ahmed, R. Girshick, L. Zitnick, and D. Batra. Reducing overfitting in deep networks by decorrelating representations. In ICLR, 2016. 5
    Google ScholarLocate open access versionFindings
  • [10] L. L. A. G. H. Y. G. N. Z. De Cheng, Xiaojun Chang. Discriminative dictionary learning with ranking metric embedded for person re-identification. In IJCAI, pages 964–970, 2017. 2
    Google ScholarLocate open access versionFindings
  • [11] T. DeVries and G. W. Taylor. Dataset augmentation in feature space. ICLRW, 2017. 5
    Google ScholarLocate open access versionFindings
  • [12] M. Geng, Y. Wang, T. Xiang, and Y. Tian. Deep transfer learning for person re-identification. arXiv, 2016. 2
    Google ScholarFindings
  • [13] S. Gong, M. Cristani, S. Yan, and C. C. Loy. Person reidentification. Springer, 2014. 1
    Google ScholarFindings
  • [14] A. Hermans, L. Beyer, and B. Leibe. In defense of the triplet loss for person re-identification. arXiv, 2017. 1, 2, 6, 7, 8
    Google ScholarLocate open access versionFindings
  • [15] G. Huang, Z. Liu, L. van der Maaten, and K. Q. Weinberger. Densely connected convolutional networks. In CVPR, 2017. 3, 5
    Google ScholarLocate open access versionFindings
  • [16] S. Ioffe and C. Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In ICML, pages 448–456, 2015. 4
    Google ScholarLocate open access versionFindings
  • [17] D. Li, X. Chen, Z. Zhang, and K. Huang. Learning deep context-aware features over body and latent parts for person re-identification. In CVPR, 2017. 2, 8
    Google ScholarLocate open access versionFindings
  • [18] W. Li, R. Zhao, T. Xiao, and X. Wang. Deepreid: Deep filter pairing neural network for person re-identification. In CVPR, pages 152–159, 2014. 1, 2
    Google ScholarLocate open access versionFindings
  • [19] W. Li, X. Zhu, and S. Gong. Person re-identification by deep joint learning of multi-loss classification. In IJCAI, 2017. 2, 7, 8
    Google ScholarLocate open access versionFindings
  • [20] Z. Li, S. Chang, F. Liang, T. S. Huang, L. Cao, and J. R. Smith. Learning locally-adaptive decision functions for person verification. In CVPR, pages 3610–3617, 2013. 1, 2
    Google ScholarLocate open access versionFindings
  • [21] S. Liao, Y. Hu, X. Zhu, and S. Z. Li. Person re-identification by local maximal occurrence representation and metric learning. In CVPR, pages 2197–2206, 2015. 1, 2, 8
    Google ScholarLocate open access versionFindings
  • [22] J. Lin, L. Ren, J. Lu, J. Feng, and J. Zhou. Consistent-aware deep learning for person re-identification in a camera network. In CVPR, pages 5771–5780, 2017. 8
    Google ScholarLocate open access versionFindings
  • [23] Y. Lin, L. Zheng, Z. Zheng, Y. Wu, and Y. Yang. Improving person re-identification by attribute and identity learning. arXiv, 2017. 8
    Google ScholarLocate open access versionFindings
  • [24] H. Liu, J. Feng, M. Qi, J. Jiang, and S. Yan. End-to-end comparative attention networks for person re-identification. IEEE Transactions on Image Processing, 2017. 3, 7
    Google ScholarLocate open access versionFindings
  • [25] X. Liu, H. Zhao, M. Tian, L. Sheng, J. Shao, S. Yi, J. Yan, and X. Wang. Hydraplus-net: Attentive deep features for pedestrian analysis. In ICCV, 2017. 7
    Google ScholarLocate open access versionFindings
  • [26] Y. Liu, Y. Junjie, and W. Ouyang. Quality aware network for set to set recognition. In CVPR, 2017. 1, 2, 7
    Google ScholarLocate open access versionFindings
  • [27] Z. Liu, D. Wang, and H. Lu. Stepwise metric promotion for unsupervised video person re-identification. In ICCV, 2017. 8
    Google ScholarLocate open access versionFindings
  • [28] N. McLaughlin, J. Martinez del Rincon, and P. Miller. Recurrent convolutional network for video-based person reidentification. In CVPR, pages 1325–1334, 2016. 1, 2, 6
    Google ScholarLocate open access versionFindings
  • [29] A. Mignon and F. Jurie. Pcca: A new approach for distance learning from sparse pairwise constraints. In CVPR, pages 2666–2672, 2012. 2
    Google ScholarLocate open access versionFindings
  • [30] E. Ristani, F. Solera, R. Zou, R. Cucchiara, and C. Tomasi. Performance measures and a data set for multi-target, multicamera tracking. In ECCVW, 2016. 5
    Google ScholarLocate open access versionFindings
  • [31] F. Schroff, D. Kalenichenko, and J. Philbin. Facenet: A unified embedding for face recognition and clustering. In CVPR, pages 815–823, 2015. 6
    Google ScholarLocate open access versionFindings
  • [32] Y. Shen, W. Lin, J. Yan, M. Xu, J. Wu, and J. Wang. Person re-identification with correspondence structure learning. In CVPR, pages 3200–3208, 2015. 1, 2
    Google ScholarLocate open access versionFindings
  • [33] K. G. Y. Y. S. C. Shuangjie Xu, Yu Cheng and P. Zhou. Jointly attentive spatial-temporal pooling networks for video-based person re-identification. In ICCV, 2017. 2
    Google ScholarLocate open access versionFindings
  • [34] J. Si, H. Zhang, and C.-G. Li. Regularization in metric learning for person re-identification. In ICIP, pages 2309–2313, 2015. 1
    Google ScholarLocate open access versionFindings
  • [35] J. Si, H. Zhang, C.-G. Li, and J. Guo. Spatial pyramid-based statistical features for person re-identification: A comprehensive evaluation. IEEE Transactions on Systems, Man, and Cybernetics: Systems, to appear, 2017. 2
    Google ScholarLocate open access versionFindings
  • [36] A. Subramaniam, M. Chatterjee, and A. Mittal. Deep neural networks with inexact matching for person re-identification. In NIPS, pages 2667–2675, 2016. 2
    Google ScholarLocate open access versionFindings
  • [37] Y. Sun, L. Zheng, W. Deng, and S. Wang. Svdnet for pedestrian retrieval. In ICCV, 2017. 1, 2, 7, 8
    Google ScholarLocate open access versionFindings
  • [38] R. R. Varior, M. Haloi, and G. Wang. Gated siamese convolutional neural network architecture for human reidentification. In ECCV, pages 791–808, 2016. 2, 8
    Google ScholarLocate open access versionFindings
  • [39] R. R. Varior, B. Shuai, J. Lu, D. Xu, and G. Wang. A siamese long short-term memory architecture for human reidentification. In ECCV, pages 135–153, 2016. 2, 8
    Google ScholarLocate open access versionFindings
  • [40] O. Vinyals, S. Bengio, and M. Kudlur. Order matters: Sequence to sequence for sets. In ICLR, 2016. 3
    Google ScholarLocate open access versionFindings
  • [41] F. Wang, W. Zuo, L. Lin, D. Zhang, and L. Zhang. Joint learning of single-image and cross-image representations for person re-identification. In CVPR, pages 1288–1296, 2016. 2
    Google ScholarLocate open access versionFindings
  • [42] S. Wang and J. Jiang. A compare-aggregate model for matching text sequences. In ICLR, 2017. 3
    Google ScholarLocate open access versionFindings
  • [43] T. Wang, S. Gong, X. Zhu, and S. Wang. Person reidentification by video ranking. In ECCV, pages 688–703, 2014. 1, 2
    Google ScholarLocate open access versionFindings
  • [44] T. Wang, S. Gong, X. Zhu, and S. Wang. Person reidentification by discriminative selection in video ranking. IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(12):2501–2514, 2016. 1
    Google ScholarLocate open access versionFindings
  • [45] T. Xiao, H. Li, W. Ouyang, and X. Wang. Learning deep feature representations with domain guided dropout for person re-identification. In CVPR, pages 1249–1258, 2016. 2
    Google ScholarLocate open access versionFindings
  • [46] T. Xiao, S. Li, B. Wang, L. Lin, and X. Wang. Joint detection and identification feature learning for person search. In CVPR, 2017. 8
    Google ScholarLocate open access versionFindings
  • [47] F. Xiong, M. Gou, O. Camps, and M. Sznaier. Person reidentification using kernel-based metric learning methods. In ECCV, pages 1–16, 2014. 2
    Google ScholarLocate open access versionFindings
  • [48] K. Xu, J. Ba, R. Kiros, K. Cho, A. Courville, R. Salakhudinov, R. Zemel, and Y. Bengio. Show, attend and tell: Neural image caption generation with visual attention. In ICML, pages 2048–2057, 2015. 3
    Google ScholarLocate open access versionFindings
  • [49] Y. Yan, B. Ni, Z. Song, C. Ma, Y. Yan, and X. Yang. Person re-identification via recurrent feature aggregation. In ECCV, pages 701–716, 2016. 2
    Google ScholarLocate open access versionFindings
  • [50] J. Yang, P. Ren, D. Zhang, D. Chen, F. Wen, H. Li, and G. Hua. Neural aggregation network for video face recognition. In CVPR, 2017. 3
    Google ScholarLocate open access versionFindings
  • [51] Y. Yang, S. Liao, Z. Lei, and S. Z. Li. Large scale similarity learning using similar pairs for person verification. In AAAI, pages 3655–3661, 2016. 2
    Google ScholarLocate open access versionFindings
  • [52] H. Yao, S. Zhang, Y. Zhang, J. Li, and Q. Tian. Deep representation learning with part loss for person re-identification. arXiv, 2017. 7
    Google ScholarLocate open access versionFindings
  • [53] M. Ye, A. J. Ma, L. Zheng, J. Li, and P. C. Yuen. Dynamic label graph matching for unsupervised video re-identification. In ICCV, 2017. 8
    Google ScholarLocate open access versionFindings
  • [54] J. You, A. Wu, X. Li, and W.-S. Zheng. Top-push videobased person re-identification. In CVPR, pages 1345–1353, 2016. 1, 2
    Google ScholarLocate open access versionFindings
  • [55] L. Zhang, T. Xiang, and S. Gong. Learning a discriminative null space for person re-identification. In CVPR, pages 1239–1248, 2016. 8
    Google ScholarLocate open access versionFindings
  • [56] Y. Zhang, B. Li, H. Lu, A. Irie, and X. Ruan. Sample-specific svm learning for person re-identification. In CVPR, pages 1278–1287, 2016. 2
    Google ScholarLocate open access versionFindings
  • [57] Y. Zhang, X. Li, L. Zhao, and Z. Zhang. Semantics-aware deep correspondence structure learning for robust person reidentification. In IJCAI, pages 3545–3551, 2016. 2
    Google ScholarLocate open access versionFindings
  • [58] Y. Zhang, T. Xiang, T. M. Hospedales, and H. Lu. Deep mutual learning. arXiv, 2017. 8
    Google ScholarLocate open access versionFindings
  • [59] H. Zhao, M. Tian, S. Sun, J. Shao, J. Yan, S. Yi, X. Wang, and X. Tang. Spindle net: Person re-identification with human body region guided feature decomposition and fusion. CVPR, 2017. 1, 2, 7
    Google ScholarLocate open access versionFindings
  • [60] L. Zhao, X. Li, J. Wang, and Y. Zhuang. Deeply-learned part-aligned representations for person re-identification. In ICCV, 2017. 1, 2, 7
    Google ScholarLocate open access versionFindings
  • [61] L. Zheng, Z. Bie, Y. Sun, J. Wang, C. Su, S. Wang, and Q. Tian. Mars: A video benchmark for large-scale person re-identification. In ECCV, 2016. 5, 8
    Google ScholarLocate open access versionFindings
  • [62] L. Zheng, L. Shen, L. Tian, S. Wang, J. Wang, and Q. Tian. Scalable person re-identification: A benchmark. In ICCV, 2015. 5, 8
    Google ScholarLocate open access versionFindings
  • [63] L. Zheng, Y. Yang, and A. G. Hauptmann. Person reidentification: Past, present and future. arXiv, 2016. 1
    Google ScholarLocate open access versionFindings
  • [64] W.-S. Zheng, S. Gong, and T. Xiang. Reidentification by relative distance comparison. IEEE Transactions on Pattern Analysis and Machine Intelligence, 35(3):653–668, 2013. 1, 2
    Google ScholarLocate open access versionFindings
  • [65] Z. Zheng, L. Zheng, and Y. Yang. Unlabeled samples generated by gan improve the person re-identification baseline in vitro. In ICCV, 2017. 5, 8
    Google ScholarLocate open access versionFindings
  • [66] Z. Zhong, L. Zheng, D. Cao, and S. Li. Re-ranking person re-identification with k-reciprocal encoding. In CVPR, 2017. 8
    Google ScholarLocate open access versionFindings
  • [67] Z. Zhong, L. Zheng, G. Kang, S. Li, and Y. Yang. Random erasing data augmentation. arXiv, 2017. 7, 8
    Google ScholarLocate open access versionFindings
  • [68] S. Zhou, J. Wang, J. Wang, Y. Gong, and N. Zheng. Point to set similarity based deep feature learning for person reidentification. In CVPR, 2017. 8
    Google ScholarLocate open access versionFindings
  • [69] Z. Zhou, Y. Huang, W. Wang, L. Wang, and T. Tan. See the forest for the trees: Joint spatial and temporal recurrent neural networks for video-based person re-identification. In CVPR, 2017. 1, 2, 3, 7
    Google ScholarLocate open access versionFindings
  • [70] X. Zhu, X.-Y. Jing, F. Wu, and H. Feng. Video-based person re-identification by simultaneously learning intra-video and inter-video distance metrics. In IJCAI, pages 3552–3559, 2016. 2
    Google ScholarLocate open access versionFindings
Full Text
Your rating :
0

 

Tags
Comments