Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications

Jonathan Eichild Schmidt; Oscar Edvard Mäkinen; Simon Gørtz Flou Nielsen; Anders Skaarup Johansen; Kamal Nasrollahi; Thomas B. Moeslund

doi:10.1117/12.2622634

Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications

Jonathan Eichild Schmidt, Oscar Edvard Mäkinen, Simon Gørtz Flou Nielsen, Anders Skaarup Johansen, Kamal Nasrollahi, Thomas B. Moeslund

Publikation: Bidrag til bog/antologi/rapport/konference proceeding › Konferenceartikel i proceeding › Forskning › peer review

Abstract

As Re-Identification (Re-ID) is becoming more and more common in today’s world, the need for more optimizedalgorithms also becomes more wanted. This is due to the importance of high accuracy as the consequences ofan incorrect match can mean security issues, if used to gain access or result in incorrect findings in science dueto wrong data. This paper explores enhancing the performance of Siamese Neural Networks by exploring theperformance of loss functions to better suit the user’s Re-IDing needs. These loss functions are Triplet loss,Triplet Hard loss and Quadruplet loss. Results show that the Triplet hard loss function performs better thanthe two others. The functions were tested on a human dataset as well as on animal datasets

Originalsprog	Engelsk
Titel	Fourteenth International Conference on Machine Vision, ICMV 2021
Redaktører	Wolfgang Osten, Dmitry Nikolaev, Jianhong Zhou
Antal sider	7
Forlag	SPIE - International Society for Optical Engineering
Publikationsdato	5 mar. 2022
Artikelnummer	120840W
ISBN (Trykt)	9781510650442
ISBN (Elektronisk)	9781510650459
DOI	https://doi.org/10.1117/12.2622634
Status	Udgivet - 5 mar. 2022
Begivenhed	Fourteenth International Conference on Machine Vision (ICMV 2021), 2021 - Rome, Italien Varighed: 8 nov. 2021 → 12 nov. 2021 http://icmv.org/

Konference

Konference	Fourteenth International Conference on Machine Vision (ICMV 2021), 2021
Land/Område	Italien
By	Rome
Periode	08/11/2021 → 12/11/2021
Internetadresse	http://icmv.org/

Navn	Proceedings of SPIE, the International Society for Optical Engineering
Vol/bind	12084
ISSN	0277-786X

Adgang til dokumentet

10.1117/12.2622634

AUB Link

Søg efter materialet i Aalborg Universitetsbiblioteks søgemaskine

Andre filer og links

Link to publication in Scopus

Citationsformater

Schmidt, J. E., Mäkinen, O. E., Flou Nielsen, S. G., Johansen, A. S., Nasrollahi, K., & Moeslund, T. B. (2022). Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications. I W. Osten, D. Nikolaev, & J. Zhou (red.), Fourteenth International Conference on Machine Vision, ICMV 2021 Artikel 120840W SPIE - International Society for Optical Engineering. https://doi.org/10.1117/12.2622634

Schmidt, Jonathan Eichild ; Mäkinen, Oscar Edvard ; Flou Nielsen, Simon Gørtz et al. / Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications. Fourteenth International Conference on Machine Vision, ICMV 2021. red. / Wolfgang Osten ; Dmitry Nikolaev ; Jianhong Zhou. SPIE - International Society for Optical Engineering, 2022. (Proceedings of SPIE, the International Society for Optical Engineering, Bind 12084).

@inproceedings{5c707a2501d94cac9b80f1a31b9be171,

title = "Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications",

abstract = "As Re-Identification (Re-ID) is becoming more and more common in today{\textquoteright}s world, the need for more optimizedalgorithms also becomes more wanted. This is due to the importance of high accuracy as the consequences ofan incorrect match can mean security issues, if used to gain access or result in incorrect findings in science dueto wrong data. This paper explores enhancing the performance of Siamese Neural Networks by exploring theperformance of loss functions to better suit the user{\textquoteright}s Re-IDing needs. These loss functions are Triplet loss,Triplet Hard loss and Quadruplet loss. Results show that the Triplet hard loss function performs better thanthe two others. The functions were tested on a human dataset as well as on animal datasets",

keywords = "Re-Identification, Siamese Neural Networks, Few Shot Learning, Triplet Loss, Metric Learning",

author = "Schmidt, {Jonathan Eichild} and M{\"a}kinen, {Oscar Edvard} and {Flou Nielsen}, {Simon G{\o}rtz} and Johansen, {Anders Skaarup} and Kamal Nasrollahi and Moeslund, {Thomas B.}",

year = "2022",

month = mar,

day = "5",

doi = "10.1117/12.2622634",

language = "English",

isbn = "9781510650442",

series = "Proceedings of SPIE, the International Society for Optical Engineering",

publisher = "SPIE - International Society for Optical Engineering",

editor = "Wolfgang Osten and Dmitry Nikolaev and Jianhong Zhou",

booktitle = "Fourteenth International Conference on Machine Vision, ICMV 2021",

address = "United States",

note = "Fourteenth International Conference on Machine Vision (ICMV 2021), 2021, ICMV 2021 ; Conference date: 08-11-2021 Through 12-11-2021",

url = "http://icmv.org/",

}

Schmidt, JE, Mäkinen, OE, Flou Nielsen, SG, Johansen, AS , Nasrollahi, K & Moeslund, TB 2022, Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications. i W Osten, D Nikolaev & J Zhou (red), Fourteenth International Conference on Machine Vision, ICMV 2021., 120840W, SPIE - International Society for Optical Engineering, Proceedings of SPIE, the International Society for Optical Engineering, bind 12084, Fourteenth International Conference on Machine Vision (ICMV 2021), 2021, Rome, Italien, 08/11/2021. https://doi.org/10.1117/12.2622634

Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications. / Schmidt, Jonathan Eichild; Mäkinen, Oscar Edvard; Flou Nielsen, Simon Gørtz et al.
Fourteenth International Conference on Machine Vision, ICMV 2021. red. / Wolfgang Osten; Dmitry Nikolaev; Jianhong Zhou. SPIE - International Society for Optical Engineering, 2022. 120840W (Proceedings of SPIE, the International Society for Optical Engineering, Bind 12084).

Publikation: Bidrag til bog/antologi/rapport/konference proceeding › Konferenceartikel i proceeding › Forskning › peer review

TY - GEN

T1 - Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications

AU - Schmidt, Jonathan Eichild

AU - Mäkinen, Oscar Edvard

AU - Flou Nielsen, Simon Gørtz

AU - Johansen, Anders Skaarup

AU - Nasrollahi, Kamal

AU - Moeslund, Thomas B.

PY - 2022/3/5

Y1 - 2022/3/5

N2 - As Re-Identification (Re-ID) is becoming more and more common in today’s world, the need for more optimizedalgorithms also becomes more wanted. This is due to the importance of high accuracy as the consequences ofan incorrect match can mean security issues, if used to gain access or result in incorrect findings in science dueto wrong data. This paper explores enhancing the performance of Siamese Neural Networks by exploring theperformance of loss functions to better suit the user’s Re-IDing needs. These loss functions are Triplet loss,Triplet Hard loss and Quadruplet loss. Results show that the Triplet hard loss function performs better thanthe two others. The functions were tested on a human dataset as well as on animal datasets

AB - As Re-Identification (Re-ID) is becoming more and more common in today’s world, the need for more optimizedalgorithms also becomes more wanted. This is due to the importance of high accuracy as the consequences ofan incorrect match can mean security issues, if used to gain access or result in incorrect findings in science dueto wrong data. This paper explores enhancing the performance of Siamese Neural Networks by exploring theperformance of loss functions to better suit the user’s Re-IDing needs. These loss functions are Triplet loss,Triplet Hard loss and Quadruplet loss. Results show that the Triplet hard loss function performs better thanthe two others. The functions were tested on a human dataset as well as on animal datasets

KW - Re-Identification

KW - Siamese Neural Networks

KW - Few Shot Learning

KW - Triplet Loss

KW - Metric Learning

UR - http://www.scopus.com/inward/record.url?scp=85134331033&partnerID=8YFLogxK

U2 - 10.1117/12.2622634

DO - 10.1117/12.2622634

M3 - Article in proceeding

SN - 9781510650442

T3 - Proceedings of SPIE, the International Society for Optical Engineering

BT - Fourteenth International Conference on Machine Vision, ICMV 2021

A2 - Osten, Wolfgang

A2 - Nikolaev, Dmitry

A2 - Zhou, Jianhong

PB - SPIE - International Society for Optical Engineering

T2 - Fourteenth International Conference on Machine Vision (ICMV 2021), 2021

Y2 - 8 November 2021 through 12 November 2021

ER -

Schmidt JE, Mäkinen OE, Flou Nielsen SG, Johansen AS , Nasrollahi K , Moeslund TB. Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications. I Osten W, Nikolaev D, Zhou J, red., Fourteenth International Conference on Machine Vision, ICMV 2021. SPIE - International Society for Optical Engineering. 2022. 120840W. (Proceedings of SPIE, the International Society for Optical Engineering, Bind 12084). doi: 10.1117/12.2622634

Exploring loss functions for optimising the accuracy of Siamese Neural Networks in Re-Identification applications

Abstract

Konference

Adgang til dokumentet

AUB Link

Andre filer og links

Fingeraftryk

Citationsformater