Face box shape and verification

Research output: Contribution to journal › Conference article › Research › peer-review

Standard

Face box shape and verification. / Christiansen, Eric; Kwak, Iljung S.; Belongie, Serge; Kriegman, David.

In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), No. PART 1, 2013, p. 550-561.

Research output: Contribution to journal › Conference article › Research › peer-review

Harvard

Christiansen, E, Kwak, IS, Belongie, S & Kriegman, D 2013, 'Face box shape and verification', Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, pp. 550-561. https://doi.org/10.1007/978-3-642-41914-0_54

APA

Christiansen, E., Kwak, I. S., Belongie, S., & Kriegman, D. (2013). Face box shape and verification. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), (PART 1), 550-561. https://doi.org/10.1007/978-3-642-41914-0_54

Vancouver

Christiansen E, Kwak IS, Belongie S, Kriegman D. Face box shape and verification. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2013;(PART 1):550-561. https://doi.org/10.1007/978-3-642-41914-0_54

Author

Christiansen, Eric ; Kwak, Iljung S. ; Belongie, Serge ; Kriegman, David. / Face box shape and verification. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2013 ; No. PART 1. pp. 550-561.

Bibtex

@inproceedings{12a98545a6bb4a4baa3c9c3da9234588,
title = "Face box shape and verification",
abstract = "Successful face verification and recognition require matching corresponding points in a pair of images, and it is commonly acknowledged that alignment is a critical step prior to matching. Once aligned, a portion of the image can be compared or features can be extracted. This portion of the image, which we will call the face box, is often just the output of a face detector. While a good deal of effort has been devoted to alignment, the choice of face box has been largely neglected. This paper presents the first systematic study of the shape and size of the face box on face verification accuracy. We use representative algorithms on a dataset that allows for experimentation with differing 3-D pose, blur, noise, misalignment, and background clutter. The experiments lead to clear conclusions and recommendations that can improve the accuracy of other face recognition methods and guide future research.",
author = "Eric Christiansen and Kwak, {Iljung S.} and Serge Belongie and David Kriegman",
year = "2013",
doi = "10.1007/978-3-642-41914-0_54",
language = "English",
pages = "550--561",
journal = "Lecture Notes in Computer Science",
issn = "0302-9743",
publisher = "Springer Verlag",
number = "PART 1",
note = "9th International Symposium on Advances in Visual Computing, ISVC 2013 ; Conference date: 29-07-2013 Through 31-07-2013",

}

RIS

TY - GEN

T1 - Face box shape and verification

AU - Christiansen, Eric

AU - Kwak, Iljung S.

AU - Belongie, Serge

AU - Kriegman, David

PY - 2013

Y1 - 2013

N2 - Successful face verification and recognition require matching corresponding points in a pair of images, and it is commonly acknowledged that alignment is a critical step prior to matching. Once aligned, a portion of the image can be compared or features can be extracted. This portion of the image, which we will call the face box, is often just the output of a face detector. While a good deal of effort has been devoted to alignment, the choice of face box has been largely neglected. This paper presents the first systematic study of the shape and size of the face box on face verification accuracy. We use representative algorithms on a dataset that allows for experimentation with differing 3-D pose, blur, noise, misalignment, and background clutter. The experiments lead to clear conclusions and recommendations that can improve the accuracy of other face recognition methods and guide future research.

AB - Successful face verification and recognition require matching corresponding points in a pair of images, and it is commonly acknowledged that alignment is a critical step prior to matching. Once aligned, a portion of the image can be compared or features can be extracted. This portion of the image, which we will call the face box, is often just the output of a face detector. While a good deal of effort has been devoted to alignment, the choice of face box has been largely neglected. This paper presents the first systematic study of the shape and size of the face box on face verification accuracy. We use representative algorithms on a dataset that allows for experimentation with differing 3-D pose, blur, noise, misalignment, and background clutter. The experiments lead to clear conclusions and recommendations that can improve the accuracy of other face recognition methods and guide future research.

UR - http://www.scopus.com/inward/record.url?scp=84888243416&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-41914-0_54

DO - 10.1007/978-3-642-41914-0_54

M3 - Conference article

AN - SCOPUS:84888243416

SP - 550

EP - 561

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

SN - 0302-9743

IS - PART 1

T2 - 9th International Symposium on Advances in Visual Computing, ISVC 2013

Y2 - 29 July 2013 through 31 July 2013

ER -

ID: 302047181