Mahalanobis Distance Learning for Person Re-Identification

Peter Roth, Martin Hirzer, Martin Köstinger, Csaba Beleznai, Horst Bischof

Research output: Chapter in Book/Report/Conference proceedingChapterResearch

Abstract

Recently, Mahalanobis metric learning has gained a considerable interest for single-shot person re-identification. The main idea is to build on an existing image representation and to learn a metric that reflects the visual camera-to-camera transitions, allowing for a more powerful classification. The goal of this chapter is twofold. We first review the main ideas of Mahalanobis metric learning in general and then give a detailed study on different approaches for the task of single-shot person re-identification, also comparing to the state-of-the-art. In particular, for our experiments we used Linear Discriminant Metric Learning (LDML), Information Theoretic Metric Learning (ITML), Large Margin Nearest Neighbor (LMNN), Large Margin Nearest Neighbor with Rejection (LMNN-R), Efficient Impostor-based Metric Learning (EIML), and KISSME. For our evaluations we used four different publicly available datasets (i.e., VIPeR, ETHZ, PRID 2011, and CAVIAR4REID). Additionally, we generated the new, more realistic PRID 450S dataset, where we also provide detailed segmentations. For the latter one, we also evaluated the influence of using well segmented foreground and background regions. Finally, the corresponding results are presented and discussed.
Original languageEnglish
Title of host publicationPerson Re-Identification
Place of PublicationLondon
PublisherSpringer
Pages247-267
Edition1
ISBN (Print)978-1-4471-6295-7
Publication statusPublished - 2014

Publication series

NameAdvances in Computer Vision and Pattern Recognition
PublisherSpringer

Fingerprint

Distance education
Cameras
Experiments

Fields of Expertise

  • Information, Communication & Computing

Cite this

Roth, P., Hirzer, M., Köstinger, M., Beleznai, C., & Bischof, H. (2014). Mahalanobis Distance Learning for Person Re-Identification. In Person Re-Identification (1 ed., pp. 247-267). (Advances in Computer Vision and Pattern Recognition). London: Springer.

Mahalanobis Distance Learning for Person Re-Identification. / Roth, Peter; Hirzer, Martin; Köstinger, Martin; Beleznai, Csaba; Bischof, Horst.

Person Re-Identification. 1. ed. London : Springer, 2014. p. 247-267 (Advances in Computer Vision and Pattern Recognition).

Research output: Chapter in Book/Report/Conference proceedingChapterResearch

Roth, P, Hirzer, M, Köstinger, M, Beleznai, C & Bischof, H 2014, Mahalanobis Distance Learning for Person Re-Identification. in Person Re-Identification. 1 edn, Advances in Computer Vision and Pattern Recognition, Springer, London, pp. 247-267.
Roth P, Hirzer M, Köstinger M, Beleznai C, Bischof H. Mahalanobis Distance Learning for Person Re-Identification. In Person Re-Identification. 1 ed. London: Springer. 2014. p. 247-267. (Advances in Computer Vision and Pattern Recognition).
Roth, Peter ; Hirzer, Martin ; Köstinger, Martin ; Beleznai, Csaba ; Bischof, Horst. / Mahalanobis Distance Learning for Person Re-Identification. Person Re-Identification. 1. ed. London : Springer, 2014. pp. 247-267 (Advances in Computer Vision and Pattern Recognition).
@inbook{5cee1981c9a442a080138caf461b9654,
title = "Mahalanobis Distance Learning for Person Re-Identification",
abstract = "Recently, Mahalanobis metric learning has gained a considerable interest for single-shot person re-identification. The main idea is to build on an existing image representation and to learn a metric that reflects the visual camera-to-camera transitions, allowing for a more powerful classification. The goal of this chapter is twofold. We first review the main ideas of Mahalanobis metric learning in general and then give a detailed study on different approaches for the task of single-shot person re-identification, also comparing to the state-of-the-art. In particular, for our experiments we used Linear Discriminant Metric Learning (LDML), Information Theoretic Metric Learning (ITML), Large Margin Nearest Neighbor (LMNN), Large Margin Nearest Neighbor with Rejection (LMNN-R), Efficient Impostor-based Metric Learning (EIML), and KISSME. For our evaluations we used four different publicly available datasets (i.e., VIPeR, ETHZ, PRID 2011, and CAVIAR4REID). Additionally, we generated the new, more realistic PRID 450S dataset, where we also provide detailed segmentations. For the latter one, we also evaluated the influence of using well segmented foreground and background regions. Finally, the corresponding results are presented and discussed.",
author = "Peter Roth and Martin Hirzer and Martin K{\"o}stinger and Csaba Beleznai and Horst Bischof",
year = "2014",
language = "English",
isbn = "978-1-4471-6295-7",
series = "Advances in Computer Vision and Pattern Recognition",
publisher = "Springer",
pages = "247--267",
booktitle = "Person Re-Identification",
edition = "1",

}

TY - CHAP

T1 - Mahalanobis Distance Learning for Person Re-Identification

AU - Roth, Peter

AU - Hirzer, Martin

AU - Köstinger, Martin

AU - Beleznai, Csaba

AU - Bischof, Horst

PY - 2014

Y1 - 2014

N2 - Recently, Mahalanobis metric learning has gained a considerable interest for single-shot person re-identification. The main idea is to build on an existing image representation and to learn a metric that reflects the visual camera-to-camera transitions, allowing for a more powerful classification. The goal of this chapter is twofold. We first review the main ideas of Mahalanobis metric learning in general and then give a detailed study on different approaches for the task of single-shot person re-identification, also comparing to the state-of-the-art. In particular, for our experiments we used Linear Discriminant Metric Learning (LDML), Information Theoretic Metric Learning (ITML), Large Margin Nearest Neighbor (LMNN), Large Margin Nearest Neighbor with Rejection (LMNN-R), Efficient Impostor-based Metric Learning (EIML), and KISSME. For our evaluations we used four different publicly available datasets (i.e., VIPeR, ETHZ, PRID 2011, and CAVIAR4REID). Additionally, we generated the new, more realistic PRID 450S dataset, where we also provide detailed segmentations. For the latter one, we also evaluated the influence of using well segmented foreground and background regions. Finally, the corresponding results are presented and discussed.

AB - Recently, Mahalanobis metric learning has gained a considerable interest for single-shot person re-identification. The main idea is to build on an existing image representation and to learn a metric that reflects the visual camera-to-camera transitions, allowing for a more powerful classification. The goal of this chapter is twofold. We first review the main ideas of Mahalanobis metric learning in general and then give a detailed study on different approaches for the task of single-shot person re-identification, also comparing to the state-of-the-art. In particular, for our experiments we used Linear Discriminant Metric Learning (LDML), Information Theoretic Metric Learning (ITML), Large Margin Nearest Neighbor (LMNN), Large Margin Nearest Neighbor with Rejection (LMNN-R), Efficient Impostor-based Metric Learning (EIML), and KISSME. For our evaluations we used four different publicly available datasets (i.e., VIPeR, ETHZ, PRID 2011, and CAVIAR4REID). Additionally, we generated the new, more realistic PRID 450S dataset, where we also provide detailed segmentations. For the latter one, we also evaluated the influence of using well segmented foreground and background regions. Finally, the corresponding results are presented and discussed.

M3 - Chapter

SN - 978-1-4471-6295-7

T3 - Advances in Computer Vision and Pattern Recognition

SP - 247

EP - 267

BT - Person Re-Identification

PB - Springer

CY - London

ER -