PAC-GAN: An Effective Pose Augmentation Scheme for Unsupervised Cross-View Person Re-identification

Chengyuan Zhang, Lei Zhu, Shichao Zhang, Weiren Yu

Research output: Contribution to journalArticle

Abstract

Person re-identification (person Re-Id) aims to retrieve the pedestrian images of the same person that captured by disjoint and non-overlapping cameras. Lots of researchers recently focused on this hot issue and proposed deep learning based methods to enhance the recognition rate in a supervised or unsupervised manner. However,there are two limitations that cannot be ignored: firstly, compared with other image retrieval benchmarks, the size of existing person Re-Id datasets is far from meeting the requirement, which cannot provide sufficient pedestrian samples for the training of deep model; secondly, the samples in existing datasets do not have sufficient human motions or postures coverage to provide more priori knowledges for learning. In this paper, we introduce a novel unsupervised pose augmentation cross-view person Re-Id scheme called PAC-GAN to overcome these limitations. We firstly present the formal definition of cross-view pose augmentation and then propose the framework of PAC-GAN that is a novel conditional generative adversarial network (CGAN) based approach to improve the performance of unsupervised corss-view person Re-Id. Specifically, the pose generation model in PAC-GAN called CPG-Net is to generate enough quantity of pose-rich samples from original image and skeleton samples. The pose augmentation dataset is produced by combining the synthesized pose-rich samples with the original samples, which is fed into the corss-view person Re-Id model named Cross-GAN. Besides, we use weight-sharing strategy in the CPG-Net to improve the quality of new generated samples. To the best of our knowledge, we are the first to enhance the unsupervised cross-view person Re-Id by pose augmentation, and the results of extensive experiments show that the proposed scheme can combat the state-of-the-arts with recognition rate.
Original languageEnglish
JournalNeurocomputing
Early online date27 Dec 2019
DOIs
Publication statusE-pub ahead of print - 27 Dec 2019

Fingerprint

Image retrieval
Cameras
Learning
Benchmarking
Experiments
Posture
Skeleton
Research Personnel
Weights and Measures
Datasets
Deep learning
Pedestrians
Recognition (Psychology)

Bibliographical note

© 2019, Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International http://creativecommons.org/licenses/by-nc-nd/4.0/

Funding: National Natural Science Foundation of China (61702560, 61836016, 61672177), project (2018JJ3691, 2016JC2011) of Science and Technology Plan of Hunan Province, and the Research and Innovation Project of Central South University Graduate Students(2018zzts177, 2018zzts588).

Keywords

  • Cross-view Person Re-Id
  • Pose Augmentation
  • Generative Adversarial Networks
  • Unsupervised Learning

Cite this

@article{c9e840eb674f49a8876d342d706040a1,
title = "PAC-GAN: An Effective Pose Augmentation Scheme for Unsupervised Cross-View Person Re-identification",
abstract = "Person re-identification (person Re-Id) aims to retrieve the pedestrian images of the same person that captured by disjoint and non-overlapping cameras. Lots of researchers recently focused on this hot issue and proposed deep learning based methods to enhance the recognition rate in a supervised or unsupervised manner. However,there are two limitations that cannot be ignored: firstly, compared with other image retrieval benchmarks, the size of existing person Re-Id datasets is far from meeting the requirement, which cannot provide sufficient pedestrian samples for the training of deep model; secondly, the samples in existing datasets do not have sufficient human motions or postures coverage to provide more priori knowledges for learning. In this paper, we introduce a novel unsupervised pose augmentation cross-view person Re-Id scheme called PAC-GAN to overcome these limitations. We firstly present the formal definition of cross-view pose augmentation and then propose the framework of PAC-GAN that is a novel conditional generative adversarial network (CGAN) based approach to improve the performance of unsupervised corss-view person Re-Id. Specifically, the pose generation model in PAC-GAN called CPG-Net is to generate enough quantity of pose-rich samples from original image and skeleton samples. The pose augmentation dataset is produced by combining the synthesized pose-rich samples with the original samples, which is fed into the corss-view person Re-Id model named Cross-GAN. Besides, we use weight-sharing strategy in the CPG-Net to improve the quality of new generated samples. To the best of our knowledge, we are the first to enhance the unsupervised cross-view person Re-Id by pose augmentation, and the results of extensive experiments show that the proposed scheme can combat the state-of-the-arts with recognition rate.",
keywords = "Cross-view Person Re-Id, Pose Augmentation, Generative Adversarial Networks, Unsupervised Learning",
author = "Chengyuan Zhang and Lei Zhu and Shichao Zhang and Weiren Yu",
note = "{\circledC} 2019, Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International http://creativecommons.org/licenses/by-nc-nd/4.0/ Funding: National Natural Science Foundation of China (61702560, 61836016, 61672177), project (2018JJ3691, 2016JC2011) of Science and Technology Plan of Hunan Province, and the Research and Innovation Project of Central South University Graduate Students(2018zzts177, 2018zzts588).",
year = "2019",
month = "12",
day = "27",
doi = "10.1016/j.neucom.2019.12.094",
language = "English",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",

}

PAC-GAN: An Effective Pose Augmentation Scheme for Unsupervised Cross-View Person Re-identification. / Zhang, Chengyuan; Zhu, Lei; Zhang, Shichao; Yu, Weiren.

In: Neurocomputing, 27.12.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - PAC-GAN: An Effective Pose Augmentation Scheme for Unsupervised Cross-View Person Re-identification

AU - Zhang, Chengyuan

AU - Zhu, Lei

AU - Zhang, Shichao

AU - Yu, Weiren

N1 - © 2019, Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International http://creativecommons.org/licenses/by-nc-nd/4.0/ Funding: National Natural Science Foundation of China (61702560, 61836016, 61672177), project (2018JJ3691, 2016JC2011) of Science and Technology Plan of Hunan Province, and the Research and Innovation Project of Central South University Graduate Students(2018zzts177, 2018zzts588).

PY - 2019/12/27

Y1 - 2019/12/27

N2 - Person re-identification (person Re-Id) aims to retrieve the pedestrian images of the same person that captured by disjoint and non-overlapping cameras. Lots of researchers recently focused on this hot issue and proposed deep learning based methods to enhance the recognition rate in a supervised or unsupervised manner. However,there are two limitations that cannot be ignored: firstly, compared with other image retrieval benchmarks, the size of existing person Re-Id datasets is far from meeting the requirement, which cannot provide sufficient pedestrian samples for the training of deep model; secondly, the samples in existing datasets do not have sufficient human motions or postures coverage to provide more priori knowledges for learning. In this paper, we introduce a novel unsupervised pose augmentation cross-view person Re-Id scheme called PAC-GAN to overcome these limitations. We firstly present the formal definition of cross-view pose augmentation and then propose the framework of PAC-GAN that is a novel conditional generative adversarial network (CGAN) based approach to improve the performance of unsupervised corss-view person Re-Id. Specifically, the pose generation model in PAC-GAN called CPG-Net is to generate enough quantity of pose-rich samples from original image and skeleton samples. The pose augmentation dataset is produced by combining the synthesized pose-rich samples with the original samples, which is fed into the corss-view person Re-Id model named Cross-GAN. Besides, we use weight-sharing strategy in the CPG-Net to improve the quality of new generated samples. To the best of our knowledge, we are the first to enhance the unsupervised cross-view person Re-Id by pose augmentation, and the results of extensive experiments show that the proposed scheme can combat the state-of-the-arts with recognition rate.

AB - Person re-identification (person Re-Id) aims to retrieve the pedestrian images of the same person that captured by disjoint and non-overlapping cameras. Lots of researchers recently focused on this hot issue and proposed deep learning based methods to enhance the recognition rate in a supervised or unsupervised manner. However,there are two limitations that cannot be ignored: firstly, compared with other image retrieval benchmarks, the size of existing person Re-Id datasets is far from meeting the requirement, which cannot provide sufficient pedestrian samples for the training of deep model; secondly, the samples in existing datasets do not have sufficient human motions or postures coverage to provide more priori knowledges for learning. In this paper, we introduce a novel unsupervised pose augmentation cross-view person Re-Id scheme called PAC-GAN to overcome these limitations. We firstly present the formal definition of cross-view pose augmentation and then propose the framework of PAC-GAN that is a novel conditional generative adversarial network (CGAN) based approach to improve the performance of unsupervised corss-view person Re-Id. Specifically, the pose generation model in PAC-GAN called CPG-Net is to generate enough quantity of pose-rich samples from original image and skeleton samples. The pose augmentation dataset is produced by combining the synthesized pose-rich samples with the original samples, which is fed into the corss-view person Re-Id model named Cross-GAN. Besides, we use weight-sharing strategy in the CPG-Net to improve the quality of new generated samples. To the best of our knowledge, we are the first to enhance the unsupervised cross-view person Re-Id by pose augmentation, and the results of extensive experiments show that the proposed scheme can combat the state-of-the-arts with recognition rate.

KW - Cross-view Person Re-Id

KW - Pose Augmentation

KW - Generative Adversarial Networks

KW - Unsupervised Learning

UR - https://linkinghub.elsevier.com/retrieve/pii/S0925231219318090

U2 - 10.1016/j.neucom.2019.12.094

DO - 10.1016/j.neucom.2019.12.094

M3 - Article

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

ER -