Semantically Selective Augmentation for Deep Compact Person Re-Identification

Víctor Ponce-López; Tilo Burghardt; Sion Hannunna; Dima Damen; Alessandro Masullo; Majid Mirmehdi

doi:10.1007/978-3-030-11012-3_41

Semantically Selective Augmentation for Deep Compact Person Re-Identification

Víctor Ponce-López^*, Tilo Burghardt, Sion Hannunna, Dima Damen, Alessandro Masullo, Majid Mirmehdi

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

2 Citations (Scopus)

108 Downloads (Pure)

Abstract

We present a deep person re-identification approach that combines semantically selective, deep data augmentation with clustering-based network compression to generate high performance, light and fast inference networks. In particular, we propose to augment limited training data via sampling from a deep convolutional generative adversarial network (DCGAN), whose discriminator is constrained by a semantic classifier to explicitly control the domain specificity of the generation process. Thereby, we encode information in the classifier network which can be utilized to steer adversarial synthesis, and which fuels our CondenseNet ID-network training. We provide a quantitative and qualitative analysis of the approach and its variants on a number of datasets, obtaining results that outperform the state-of-the-art on the LIMA dataset for long-term monitoring in indoor living spaces.

Original language	English
Title of host publication	Computer Vision – ECCV 2018 Workshops, Proceedings
Editors	Stefan Roth, Laura Leal-Taixé
Publisher	Springer Verlag
Pages	551-561
Number of pages	11
Volume	11130
ISBN (Electronic)	978-3-030-11012-3
ISBN (Print)	9783030110116
DOIs	https://doi.org/10.1007/978-3-030-11012-3_41
Publication status	Published - 29 Jan 2019
Event	15th European Conference on Computer Vision, ECCV 2018 - Munich, Germany Duration: 8 Sept 2018 → 14 Sept 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11130 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	15th European Conference on Computer Vision, ECCV 2018
Country/Territory	Germany
City	Munich
Period	8/09/18 → 14/09/18

Research Groups and Themes

Digital Health
SPHERE

Keywords

Adversarial synthesis
Deep compression
Face filtering
Person re-identification
Selective augmentation

Access to Document

10.1007/978-3-030-11012-3_41

Full-text PDF (accepted author manuscript)
This is the accepted author manuscript (AAM). The final published version (version of record) is available online via Springer Link at https://doi.org/10.1007/978-3-030-11012-3_41 . Please refer to any applicable terms of use of the publisher.
Accepted author manuscript, 1.64 MBLicence: Other

Handle.net

Persistent link

Cite this

Ponce-López, V., Burghardt, T., Hannunna, S., Damen, D., Masullo, A., & Mirmehdi, M. (2019). Semantically Selective Augmentation for Deep Compact Person Re-Identification. In S. Roth, & L. Leal-Taixé (Eds.), Computer Vision – ECCV 2018 Workshops, Proceedings (Vol. 11130, pp. 551-561). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11130 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-11012-3_41

Ponce-López, Víctor ; Burghardt, Tilo ; Hannunna, Sion et al. / Semantically Selective Augmentation for Deep Compact Person Re-Identification. Computer Vision – ECCV 2018 Workshops, Proceedings. editor / Stefan Roth ; Laura Leal-Taixé. Vol. 11130 Springer Verlag, 2019. pp. 551-561 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{8fffbb1fb5144898817a358a1fb2d0d1,

title = "Semantically Selective Augmentation for Deep Compact Person Re-Identification",

abstract = "We present a deep person re-identification approach that combines semantically selective, deep data augmentation with clustering-based network compression to generate high performance, light and fast inference networks. In particular, we propose to augment limited training data via sampling from a deep convolutional generative adversarial network (DCGAN), whose discriminator is constrained by a semantic classifier to explicitly control the domain specificity of the generation process. Thereby, we encode information in the classifier network which can be utilized to steer adversarial synthesis, and which fuels our CondenseNet ID-network training. We provide a quantitative and qualitative analysis of the approach and its variants on a number of datasets, obtaining results that outperform the state-of-the-art on the LIMA dataset for long-term monitoring in indoor living spaces. ",

keywords = "Adversarial synthesis, Deep compression, Face filtering, Person re-identification, Selective augmentation",

author = "V{\'i}ctor Ponce-L{\'o}pez and Tilo Burghardt and Sion Hannunna and Dima Damen and Alessandro Masullo and Majid Mirmehdi",

year = "2019",

month = jan,

day = "29",

doi = "10.1007/978-3-030-11012-3_41",

language = "English",

isbn = "9783030110116",

volume = "11130",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "551--561",

editor = "Stefan Roth and Laura Leal-Taix{\'e}",

booktitle = "Computer Vision – ECCV 2018 Workshops, Proceedings",

address = "Germany",

note = "15th European Conference on Computer Vision, ECCV 2018 ; Conference date: 08-09-2018 Through 14-09-2018",

}

Ponce-López, V, Burghardt, T, Hannunna, S, Damen, D, Masullo, A & Mirmehdi, M 2019, Semantically Selective Augmentation for Deep Compact Person Re-Identification. in S Roth & L Leal-Taixé (eds), Computer Vision – ECCV 2018 Workshops, Proceedings. vol. 11130, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11130 LNCS, Springer Verlag, pp. 551-561, 15th European Conference on Computer Vision, ECCV 2018, Munich, Germany, 8/09/18. https://doi.org/10.1007/978-3-030-11012-3_41

Semantically Selective Augmentation for Deep Compact Person Re-Identification. / Ponce-López, Víctor; Burghardt, Tilo; Hannunna, Sion et al.
Computer Vision – ECCV 2018 Workshops, Proceedings. ed. / Stefan Roth; Laura Leal-Taixé. Vol. 11130 Springer Verlag, 2019. p. 551-561 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11130 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - Semantically Selective Augmentation for Deep Compact Person Re-Identification

AU - Ponce-López, Víctor

AU - Burghardt, Tilo

AU - Hannunna, Sion

AU - Damen, Dima

AU - Masullo, Alessandro

AU - Mirmehdi, Majid

PY - 2019/1/29

Y1 - 2019/1/29

N2 - We present a deep person re-identification approach that combines semantically selective, deep data augmentation with clustering-based network compression to generate high performance, light and fast inference networks. In particular, we propose to augment limited training data via sampling from a deep convolutional generative adversarial network (DCGAN), whose discriminator is constrained by a semantic classifier to explicitly control the domain specificity of the generation process. Thereby, we encode information in the classifier network which can be utilized to steer adversarial synthesis, and which fuels our CondenseNet ID-network training. We provide a quantitative and qualitative analysis of the approach and its variants on a number of datasets, obtaining results that outperform the state-of-the-art on the LIMA dataset for long-term monitoring in indoor living spaces.

AB - We present a deep person re-identification approach that combines semantically selective, deep data augmentation with clustering-based network compression to generate high performance, light and fast inference networks. In particular, we propose to augment limited training data via sampling from a deep convolutional generative adversarial network (DCGAN), whose discriminator is constrained by a semantic classifier to explicitly control the domain specificity of the generation process. Thereby, we encode information in the classifier network which can be utilized to steer adversarial synthesis, and which fuels our CondenseNet ID-network training. We provide a quantitative and qualitative analysis of the approach and its variants on a number of datasets, obtaining results that outperform the state-of-the-art on the LIMA dataset for long-term monitoring in indoor living spaces.

KW - Adversarial synthesis

KW - Deep compression

KW - Face filtering

KW - Person re-identification

KW - Selective augmentation

UR - http://www.scopus.com/inward/record.url?scp=85061827958&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-11012-3_41

DO - 10.1007/978-3-030-11012-3_41

M3 - Conference Contribution (Conference Proceeding)

SN - 9783030110116

VL - 11130

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 551

EP - 561

BT - Computer Vision – ECCV 2018 Workshops, Proceedings

A2 - Roth, Stefan

A2 - Leal-Taixé, Laura

PB - Springer Verlag

T2 - 15th European Conference on Computer Vision, ECCV 2018

Y2 - 8 September 2018 through 14 September 2018

ER -

Ponce-López V, Burghardt T, Hannunna S, Damen D, Masullo A, Mirmehdi M. Semantically Selective Augmentation for Deep Compact Person Re-Identification. In Roth S, Leal-Taixé L, editors, Computer Vision – ECCV 2018 Workshops, Proceedings. Vol. 11130. Springer Verlag. 2019. p. 551-561. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-11012-3_41

Semantically Selective Augmentation for Deep Compact Person Re-Identification

Abstract

Publication series

Conference

Research Groups and Themes

Keywords

Access to Document

Handle.net

Other files and links

Fingerprint

SPHERE (EPSRC IRC)

Cite this

Semantically Selective Augmentation for Deep Compact Person Re-Identification

Abstract

Publication series

Conference

Research Groups and Themes

Keywords

Access to Document

Handle.net

Other files and links

Fingerprint

Projects

SPHERE (EPSRC IRC)

Cite this