Precise Adjacent Margin Loss for Deep Face Recognition

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

Abstract

Softmax loss is arguably one of the most widely used loss functions in CNNs. In recent years some Softmax variants have been proposed to enhance the discriminative ability of the learned features by adding additional margin constraints, which significantly improved the state-of-the-art performance of face recognition. However, the ‘margin’ referenced in these losses does not represent the real margin between the different classes in the training set. Furthermore, they impose a margin on all possible combinations of class pairs, which is unnecessary. In this paper we propose the Precise Adjacent Margin loss (PAM loss), which gives an accurate definition of ‘margin’ and has precise operations appropriate for different cases. PAM loss has better geometrical interpretation than the existing margin-based losses. Extensive experiments are conducted on LFW, YTF, MegaFace and FaceScrub datasets, and results show that the proposed method has state-of-the-art performance.
LanguageEnglish
Title of host publication2019 IEEE International Conference on Image Processing (ICIP)
ISBN (Electronic)978-1-5386-6249-6
Publication statusPublished - 26 Aug 2019
Event2019 IEEE International Conference on Image Processing (ICIP) - Taipei, Taiwan, Province of China
Duration: 22 Sep 201925 Sep 2019
https://ieeexplore.ieee.org/xpl/conhome/8791230/proceeding

Publication series

Name2019 IEEE International Conference on Image Processing (ICIP)
PublisherIEEE
ISSN (Print)1522-4880
ISSN (Electronic)2381-8549

Conference

Conference2019 IEEE International Conference on Image Processing (ICIP)
CountryTaiwan, Province of China
CityTaipei
Period22/09/1925/09/19
Internet address

Fingerprint

Face recognition
Experiments

Cite this

Wei, X., Wang, H., Scotney, B., & Wan, H. (2019). Precise Adjacent Margin Loss for Deep Face Recognition. In 2019 IEEE International Conference on Image Processing (ICIP) (2019 IEEE International Conference on Image Processing (ICIP)).
Wei, X ; Wang, H. ; Scotney, Bryan ; Wan, Huan. / Precise Adjacent Margin Loss for Deep Face Recognition. 2019 IEEE International Conference on Image Processing (ICIP). 2019. (2019 IEEE International Conference on Image Processing (ICIP)).
@inproceedings{1fcb4779525849938d84676a0f65919e,
title = "Precise Adjacent Margin Loss for Deep Face Recognition",
abstract = "Softmax loss is arguably one of the most widely used loss functions in CNNs. In recent years some Softmax variants have been proposed to enhance the discriminative ability of the learned features by adding additional margin constraints, which significantly improved the state-of-the-art performance of face recognition. However, the ‘margin’ referenced in these losses does not represent the real margin between the different classes in the training set. Furthermore, they impose a margin on all possible combinations of class pairs, which is unnecessary. In this paper we propose the Precise Adjacent Margin loss (PAM loss), which gives an accurate definition of ‘margin’ and has precise operations appropriate for different cases. PAM loss has better geometrical interpretation than the existing margin-based losses. Extensive experiments are conducted on LFW, YTF, MegaFace and FaceScrub datasets, and results show that the proposed method has state-of-the-art performance.",
author = "X Wei and H. Wang and Bryan Scotney and Huan Wan",
year = "2019",
month = "8",
day = "26",
language = "English",
isbn = "978-1-5386-6250-2",
series = "2019 IEEE International Conference on Image Processing (ICIP)",
publisher = "IEEE",
booktitle = "2019 IEEE International Conference on Image Processing (ICIP)",

}

Wei, X, Wang, H, Scotney, B & Wan, H 2019, Precise Adjacent Margin Loss for Deep Face Recognition. in 2019 IEEE International Conference on Image Processing (ICIP). 2019 IEEE International Conference on Image Processing (ICIP), 2019 IEEE International Conference on Image Processing (ICIP), Taipei, Taiwan, Province of China, 22/09/19.

Precise Adjacent Margin Loss for Deep Face Recognition. / Wei, X; Wang, H.; Scotney, Bryan; Wan, Huan.

2019 IEEE International Conference on Image Processing (ICIP). 2019. (2019 IEEE International Conference on Image Processing (ICIP)).

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

TY - GEN

T1 - Precise Adjacent Margin Loss for Deep Face Recognition

AU - Wei, X

AU - Wang, H.

AU - Scotney, Bryan

AU - Wan, Huan

PY - 2019/8/26

Y1 - 2019/8/26

N2 - Softmax loss is arguably one of the most widely used loss functions in CNNs. In recent years some Softmax variants have been proposed to enhance the discriminative ability of the learned features by adding additional margin constraints, which significantly improved the state-of-the-art performance of face recognition. However, the ‘margin’ referenced in these losses does not represent the real margin between the different classes in the training set. Furthermore, they impose a margin on all possible combinations of class pairs, which is unnecessary. In this paper we propose the Precise Adjacent Margin loss (PAM loss), which gives an accurate definition of ‘margin’ and has precise operations appropriate for different cases. PAM loss has better geometrical interpretation than the existing margin-based losses. Extensive experiments are conducted on LFW, YTF, MegaFace and FaceScrub datasets, and results show that the proposed method has state-of-the-art performance.

AB - Softmax loss is arguably one of the most widely used loss functions in CNNs. In recent years some Softmax variants have been proposed to enhance the discriminative ability of the learned features by adding additional margin constraints, which significantly improved the state-of-the-art performance of face recognition. However, the ‘margin’ referenced in these losses does not represent the real margin between the different classes in the training set. Furthermore, they impose a margin on all possible combinations of class pairs, which is unnecessary. In this paper we propose the Precise Adjacent Margin loss (PAM loss), which gives an accurate definition of ‘margin’ and has precise operations appropriate for different cases. PAM loss has better geometrical interpretation than the existing margin-based losses. Extensive experiments are conducted on LFW, YTF, MegaFace and FaceScrub datasets, and results show that the proposed method has state-of-the-art performance.

M3 - Conference contribution

SN - 978-1-5386-6250-2

T3 - 2019 IEEE International Conference on Image Processing (ICIP)

BT - 2019 IEEE International Conference on Image Processing (ICIP)

ER -

Wei X, Wang H, Scotney B, Wan H. Precise Adjacent Margin Loss for Deep Face Recognition. In 2019 IEEE International Conference on Image Processing (ICIP). 2019. (2019 IEEE International Conference on Image Processing (ICIP)).