Face box shape and verification
Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt
Standard
Face box shape and verification. / Christiansen, Eric; Kwak, Iljung S.; Belongie, Serge; Kriegman, David.
I: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Nr. PART 1, 2013, s. 550-561.Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt
Harvard
APA
Vancouver
Author
Bibtex
}
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