Learning locally-adaptive decision functions for person verification

Zhen Li, Shiyu Chang, Feng Liang, Thomas S Huang, Liangliang Cao, John R. Smith

Research output: Contribution to journalConference article

Abstract

This paper considers the person verification problem in modern surveillance and video retrieval systems. The problem is to identify whether a pair of face or human body images is about the same person, even if the person is not seen before. Traditional methods usually look for a distance (or similarity) measure between images (e.g., by metric learning algorithms), and make decisions based on a fixed threshold. We show that this is nevertheless insufficient and sub-optimal for the verification problem. This paper proposes to learn a decision function for verification that can be viewed as a joint model of a distance metric and a locally adaptive thresholding rule. We further formulate the inference on our decision function as a second-order large-margin regularization problem, and provide an efficient algorithm in its dual from. We evaluate our algorithm on both human body verification and face verification problems. Our method outperforms not only the classical metric learning algorithm including LMNN and ITML, but also the state-of-the-art in the computer vision community.

Original languageEnglish (US)
Article number6619307
Pages (from-to)3610-3617
Number of pages8
JournalProceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
DOIs
StatePublished - Nov 15 2013
Event26th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2013 - Portland, OR, United States
Duration: Jun 23 2013Jun 28 2013

Fingerprint

Learning algorithms
Computer vision

Keywords

  • Face Verification
  • Pedestrian Re-identification
  • Person Verification

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition

Cite this

Learning locally-adaptive decision functions for person verification. / Li, Zhen; Chang, Shiyu; Liang, Feng; Huang, Thomas S; Cao, Liangliang; Smith, John R.

In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 15.11.2013, p. 3610-3617.

Research output: Contribution to journalConference article

@article{00633ed06b8d4e41a11e34cf3b0917c6,
title = "Learning locally-adaptive decision functions for person verification",
abstract = "This paper considers the person verification problem in modern surveillance and video retrieval systems. The problem is to identify whether a pair of face or human body images is about the same person, even if the person is not seen before. Traditional methods usually look for a distance (or similarity) measure between images (e.g., by metric learning algorithms), and make decisions based on a fixed threshold. We show that this is nevertheless insufficient and sub-optimal for the verification problem. This paper proposes to learn a decision function for verification that can be viewed as a joint model of a distance metric and a locally adaptive thresholding rule. We further formulate the inference on our decision function as a second-order large-margin regularization problem, and provide an efficient algorithm in its dual from. We evaluate our algorithm on both human body verification and face verification problems. Our method outperforms not only the classical metric learning algorithm including LMNN and ITML, but also the state-of-the-art in the computer vision community.",
keywords = "Face Verification, Pedestrian Re-identification, Person Verification",
author = "Zhen Li and Shiyu Chang and Feng Liang and Huang, {Thomas S} and Liangliang Cao and Smith, {John R.}",
year = "2013",
month = "11",
day = "15",
doi = "10.1109/CVPR.2013.463",
language = "English (US)",
pages = "3610--3617",
journal = "Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition",
issn = "1063-6919",
publisher = "IEEE Computer Society",

}

TY - JOUR

T1 - Learning locally-adaptive decision functions for person verification

AU - Li, Zhen

AU - Chang, Shiyu

AU - Liang, Feng

AU - Huang, Thomas S

AU - Cao, Liangliang

AU - Smith, John R.

PY - 2013/11/15

Y1 - 2013/11/15

N2 - This paper considers the person verification problem in modern surveillance and video retrieval systems. The problem is to identify whether a pair of face or human body images is about the same person, even if the person is not seen before. Traditional methods usually look for a distance (or similarity) measure between images (e.g., by metric learning algorithms), and make decisions based on a fixed threshold. We show that this is nevertheless insufficient and sub-optimal for the verification problem. This paper proposes to learn a decision function for verification that can be viewed as a joint model of a distance metric and a locally adaptive thresholding rule. We further formulate the inference on our decision function as a second-order large-margin regularization problem, and provide an efficient algorithm in its dual from. We evaluate our algorithm on both human body verification and face verification problems. Our method outperforms not only the classical metric learning algorithm including LMNN and ITML, but also the state-of-the-art in the computer vision community.

AB - This paper considers the person verification problem in modern surveillance and video retrieval systems. The problem is to identify whether a pair of face or human body images is about the same person, even if the person is not seen before. Traditional methods usually look for a distance (or similarity) measure between images (e.g., by metric learning algorithms), and make decisions based on a fixed threshold. We show that this is nevertheless insufficient and sub-optimal for the verification problem. This paper proposes to learn a decision function for verification that can be viewed as a joint model of a distance metric and a locally adaptive thresholding rule. We further formulate the inference on our decision function as a second-order large-margin regularization problem, and provide an efficient algorithm in its dual from. We evaluate our algorithm on both human body verification and face verification problems. Our method outperforms not only the classical metric learning algorithm including LMNN and ITML, but also the state-of-the-art in the computer vision community.

KW - Face Verification

KW - Pedestrian Re-identification

KW - Person Verification

UR - http://www.scopus.com/inward/record.url?scp=84887324982&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84887324982&partnerID=8YFLogxK

U2 - 10.1109/CVPR.2013.463

DO - 10.1109/CVPR.2013.463

M3 - Conference article

AN - SCOPUS:84887324982

SP - 3610

EP - 3617

JO - Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition

JF - Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition

SN - 1063-6919

M1 - 6619307

ER -