Metric learning based on outlier sample elimination for person re-identification

Rima Ruktiari Ismail, Seiichiro Kamata

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Person re-identification is a task that aims to recognize a person of interest between two non-overlapping cameras. However, because of different camera angles and quality of the camera, person re-identification is becoming challenging task until now. Many existing methods have been proposed to solve these kinds of problems. Metric learning is already becoming a hot research topic especially for the effectiveness in matching person images problem. For metric learning process, most of them utilize all sample pairs without considering the ratio between distance of each pair. However, not all pairs are useful for the training process. We consider that there are some outliers that can not give good effect in the learning process. For example, a distance too far or too close can influence other samples and mislead the training process, resulting in longer training process time, less accuracy, and other bad effects. In this paper, we propose a new method based on eliminating the outliers, which is called Outlier Sample Elimination. Our method divides negative pairs into three groups using some thresholds to find the proper sample for learning process. During the learning process, we only use the best sample pairs to take into account in our loss function. We eliminate other samples that are considered as outliers.We try to evaluate our proposed method using the challenging VIPeR dataset. Our experiment shows that our method achieves a competitive performance.

Original languageEnglish
Title of host publication2018 International Conference on Image and Video Processing, and Artificial Intelligence
EditorsRuidan Su
PublisherSPIE
ISBN (Electronic)9781510623101
DOIs
Publication statusPublished - 2018 Jan 1
Event2018 International Conference on Image and Video Processing, and Artificial Intelligence, IVPAI 2018 - Shanghai, China
Duration: 2018 Aug 152018 Aug 17

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume10836
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

Conference2018 International Conference on Image and Video Processing, and Artificial Intelligence, IVPAI 2018
CountryChina
CityShanghai
Period18/8/1518/8/17

Fingerprint

learning
Outlier
Elimination
elimination
Person
Cameras
Learning Process
Metric
education
Camera
cameras
Loss Function
Divides
Learning
Eliminate
Experiments
Angle
thresholds
Evaluate
Experiment

Keywords

  • Metric Learning
  • Outlier
  • Person Re-Identification

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Computer Science Applications
  • Applied Mathematics
  • Electrical and Electronic Engineering

Cite this

Ismail, R. R., & Kamata, S. (2018). Metric learning based on outlier sample elimination for person re-identification. In R. Su (Ed.), 2018 International Conference on Image and Video Processing, and Artificial Intelligence [1083603] (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 10836). SPIE. https://doi.org/10.1117/12.2326993

Metric learning based on outlier sample elimination for person re-identification. / Ismail, Rima Ruktiari; Kamata, Seiichiro.

2018 International Conference on Image and Video Processing, and Artificial Intelligence. ed. / Ruidan Su. SPIE, 2018. 1083603 (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 10836).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Ismail, RR & Kamata, S 2018, Metric learning based on outlier sample elimination for person re-identification. in R Su (ed.), 2018 International Conference on Image and Video Processing, and Artificial Intelligence., 1083603, Proceedings of SPIE - The International Society for Optical Engineering, vol. 10836, SPIE, 2018 International Conference on Image and Video Processing, and Artificial Intelligence, IVPAI 2018, Shanghai, China, 18/8/15. https://doi.org/10.1117/12.2326993
Ismail RR, Kamata S. Metric learning based on outlier sample elimination for person re-identification. In Su R, editor, 2018 International Conference on Image and Video Processing, and Artificial Intelligence. SPIE. 2018. 1083603. (Proceedings of SPIE - The International Society for Optical Engineering). https://doi.org/10.1117/12.2326993
Ismail, Rima Ruktiari ; Kamata, Seiichiro. / Metric learning based on outlier sample elimination for person re-identification. 2018 International Conference on Image and Video Processing, and Artificial Intelligence. editor / Ruidan Su. SPIE, 2018. (Proceedings of SPIE - The International Society for Optical Engineering).
@inproceedings{ddd3b66af9b545578e0d2c6ae8490ff0,
title = "Metric learning based on outlier sample elimination for person re-identification",
abstract = "Person re-identification is a task that aims to recognize a person of interest between two non-overlapping cameras. However, because of different camera angles and quality of the camera, person re-identification is becoming challenging task until now. Many existing methods have been proposed to solve these kinds of problems. Metric learning is already becoming a hot research topic especially for the effectiveness in matching person images problem. For metric learning process, most of them utilize all sample pairs without considering the ratio between distance of each pair. However, not all pairs are useful for the training process. We consider that there are some outliers that can not give good effect in the learning process. For example, a distance too far or too close can influence other samples and mislead the training process, resulting in longer training process time, less accuracy, and other bad effects. In this paper, we propose a new method based on eliminating the outliers, which is called Outlier Sample Elimination. Our method divides negative pairs into three groups using some thresholds to find the proper sample for learning process. During the learning process, we only use the best sample pairs to take into account in our loss function. We eliminate other samples that are considered as outliers.We try to evaluate our proposed method using the challenging VIPeR dataset. Our experiment shows that our method achieves a competitive performance.",
keywords = "Metric Learning, Outlier, Person Re-Identification",
author = "Ismail, {Rima Ruktiari} and Seiichiro Kamata",
year = "2018",
month = "1",
day = "1",
doi = "10.1117/12.2326993",
language = "English",
series = "Proceedings of SPIE - The International Society for Optical Engineering",
publisher = "SPIE",
editor = "Ruidan Su",
booktitle = "2018 International Conference on Image and Video Processing, and Artificial Intelligence",

}

TY - GEN

T1 - Metric learning based on outlier sample elimination for person re-identification

AU - Ismail, Rima Ruktiari

AU - Kamata, Seiichiro

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Person re-identification is a task that aims to recognize a person of interest between two non-overlapping cameras. However, because of different camera angles and quality of the camera, person re-identification is becoming challenging task until now. Many existing methods have been proposed to solve these kinds of problems. Metric learning is already becoming a hot research topic especially for the effectiveness in matching person images problem. For metric learning process, most of them utilize all sample pairs without considering the ratio between distance of each pair. However, not all pairs are useful for the training process. We consider that there are some outliers that can not give good effect in the learning process. For example, a distance too far or too close can influence other samples and mislead the training process, resulting in longer training process time, less accuracy, and other bad effects. In this paper, we propose a new method based on eliminating the outliers, which is called Outlier Sample Elimination. Our method divides negative pairs into three groups using some thresholds to find the proper sample for learning process. During the learning process, we only use the best sample pairs to take into account in our loss function. We eliminate other samples that are considered as outliers.We try to evaluate our proposed method using the challenging VIPeR dataset. Our experiment shows that our method achieves a competitive performance.

AB - Person re-identification is a task that aims to recognize a person of interest between two non-overlapping cameras. However, because of different camera angles and quality of the camera, person re-identification is becoming challenging task until now. Many existing methods have been proposed to solve these kinds of problems. Metric learning is already becoming a hot research topic especially for the effectiveness in matching person images problem. For metric learning process, most of them utilize all sample pairs without considering the ratio between distance of each pair. However, not all pairs are useful for the training process. We consider that there are some outliers that can not give good effect in the learning process. For example, a distance too far or too close can influence other samples and mislead the training process, resulting in longer training process time, less accuracy, and other bad effects. In this paper, we propose a new method based on eliminating the outliers, which is called Outlier Sample Elimination. Our method divides negative pairs into three groups using some thresholds to find the proper sample for learning process. During the learning process, we only use the best sample pairs to take into account in our loss function. We eliminate other samples that are considered as outliers.We try to evaluate our proposed method using the challenging VIPeR dataset. Our experiment shows that our method achieves a competitive performance.

KW - Metric Learning

KW - Outlier

KW - Person Re-Identification

UR - http://www.scopus.com/inward/record.url?scp=85062437693&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062437693&partnerID=8YFLogxK

U2 - 10.1117/12.2326993

DO - 10.1117/12.2326993

M3 - Conference contribution

AN - SCOPUS:85062437693

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - 2018 International Conference on Image and Video Processing, and Artificial Intelligence

A2 - Su, Ruidan

PB - SPIE

ER -