Utilising domain transformations in multi-camera re-identification scenarios beyond data augmentation

Glen Brown; Jesus Martinez del Rincon; Paul Miller

Utilising domain transformations in multi-camera re-identification scenarios beyond data augmentation

School of Electronics, Electrical Engineering and Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

47 Downloads (Pure)

Abstract

GANs and CycleGANs, such as CamStyle, have proved to be very effective when augmenting multi-camera datasets in re-identification scenarios. They achieve this by transforming the training images into the domain of each camera. However, this learned domain adaptation is not exploited at the re-identification stage. In this work we propose an extension to CamStyle where the domain transformation is not only used in training for data augmentation, but also further integrated intesting for improving the re-identification performance when different cameras in the scenario are indistinct domains.

Original language	English
Title of host publication	22nd Irish Machine Vision and Image Processing Conference, IMVIP 2020: Proceedings
Editors	Sean Mullery
Publisher	Irish Pattern Recognition & Classification Society
Pages	133-136
Number of pages	4
ISBN (Electronic)	9780993420757
Publication status	Published - 24 Sept 2020
Event	22nd Irish Machine Vision and Image Processing Conference 2020 - Sligo, Ireland Duration: 31 Aug 2020 → 02 Sept 2020 https://imvipconference.github.io/

Conference

Conference	22nd Irish Machine Vision and Image Processing Conference 2020
Abbreviated title	IMVIP 2020
Country/Territory	Ireland
City	Sligo
Period	31/08/2020 → 02/09/2020
Internet address	https://imvipconference.github.io/

Access to Document

Utilising domain transformations in multi-camera re-identification scenarios beyond data augmentation
Copyright 2020 The Authors. This is an open access article published under a Creative Commons Attribution-NonCommercial-NoDerivatives License (https://creativecommons.org/licenses/by-nc-nd/3.0/), which permits distribution and reproduction for non-commercial purposes, provided the author and source are cited.
Final published version, 1.61 MBLicence: CC BY-NC-ND

Face-based biometric systems with deep learning in non-homogenous settings
Brown, G. (Author), Martinez del Rincon, J. (Supervisor) & Miller, P. (Supervisor), Jul 2023
Student thesis: Doctoral Thesis › Doctor of Philosophy

File

Cite this

@inproceedings{e3b3fc22475743c8aae607566ae55c93,

title = "Utilising domain transformations in multi-camera re-identification scenarios beyond data augmentation",

abstract = "GANs and CycleGANs, such as CamStyle, have proved to be very effective when augmenting multi-camera datasets in re-identification scenarios. They achieve this by transforming the training images into the domain of each camera. However, this learned domain adaptation is not exploited at the re-identification stage. In this work we propose an extension to CamStyle where the domain transformation is not only used in training for data augmentation, but also further integrated intesting for improving the re-identification performance when different cameras in the scenario are indistinct domains. ",

author = "Glen Brown and \{Martinez del Rincon\}, Jesus and Paul Miller",

year = "2020",

month = sep,

day = "24",

language = "English",

pages = "133--136",

editor = "Sean Mullery",

booktitle = "22nd Irish Machine Vision and Image Processing Conference, IMVIP 2020: Proceedings",

publisher = "Irish Pattern Recognition \& Classification Society",

note = "22nd Irish Machine Vision and Image Processing Conference 2020, IMVIP 2020 ; Conference date: 31-08-2020 Through 02-09-2020",

url = "https://imvipconference.github.io/",

}

Brown, G, Martinez del Rincon, J & Miller, P 2020, Utilising domain transformations in multi-camera re-identification scenarios beyond data augmentation. in S Mullery (ed.), 22nd Irish Machine Vision and Image Processing Conference, IMVIP 2020: Proceedings. Irish Pattern Recognition & Classification Society, pp. 133-136, 22nd Irish Machine Vision and Image Processing Conference 2020, Sligo, Ireland, 31/08/2020.

Utilising domain transformations in multi-camera re-identification scenarios beyond data augmentation. / Brown, Glen; Martinez del Rincon, Jesus ; Miller, Paul.
22nd Irish Machine Vision and Image Processing Conference, IMVIP 2020: Proceedings. ed. / Sean Mullery. Irish Pattern Recognition & Classification Society, 2020. p. 133-136.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Utilising domain transformations in multi-camera re-identification scenarios beyond data augmentation

AU - Brown, Glen

AU - Martinez del Rincon, Jesus

AU - Miller, Paul

PY - 2020/9/24

Y1 - 2020/9/24

N2 - GANs and CycleGANs, such as CamStyle, have proved to be very effective when augmenting multi-camera datasets in re-identification scenarios. They achieve this by transforming the training images into the domain of each camera. However, this learned domain adaptation is not exploited at the re-identification stage. In this work we propose an extension to CamStyle where the domain transformation is not only used in training for data augmentation, but also further integrated intesting for improving the re-identification performance when different cameras in the scenario are indistinct domains.

AB - GANs and CycleGANs, such as CamStyle, have proved to be very effective when augmenting multi-camera datasets in re-identification scenarios. They achieve this by transforming the training images into the domain of each camera. However, this learned domain adaptation is not exploited at the re-identification stage. In this work we propose an extension to CamStyle where the domain transformation is not only used in training for data augmentation, but also further integrated intesting for improving the re-identification performance when different cameras in the scenario are indistinct domains.

M3 - Conference contribution

SP - 133

EP - 136

BT - 22nd Irish Machine Vision and Image Processing Conference, IMVIP 2020: Proceedings

A2 - Mullery, Sean

PB - Irish Pattern Recognition & Classification Society

T2 - 22nd Irish Machine Vision and Image Processing Conference 2020

Y2 - 31 August 2020 through 2 September 2020

ER -

Utilising domain transformations in multi-camera re-identification scenarios beyond data augmentation

Abstract

Conference

Access to Document

Fingerprint

Student theses

Face-based biometric systems with deep learning in non-homogenous settings

Cite this