An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition

Kwang-Uk Han; Song-Jun Yun; Chol Song; Kwang-Min Kim; Chol-Jun O

doi:doi:10.11648/j.ajnna.20251101.11

Research Article |

| Peer-Reviewed

An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition

Kwang-Uk Han

, Song-Jun Yun

, Chol Song

, Kwang-Min Kim

, Chol-Jun O^*

Published in American Journal of Neural Networks and Applications (Volume 11, Issue 1)

Received: 10 February 2025 Accepted: 19 March 2025 Published: 17 April 2025

Views: Downloads:

Download PDF

Share This Article

Twitter
Linked In
Facebook

Abstract

In recent years, there have been growing interests on deep learning based face recognition which currently produces state of the art standards in face detection, recognition and verification tasks. As is well known, loss function for extracting face feature plays a crucial role in deep face model. In this regards, margin-based loss functions which apply a fixed margin between the feature and the weight have attracted many interests. However, such margin-based losses have a somewhat limitation in enhancing the discriminative power and generalizability of the face model, since the intra-class and inter-class variations in the real face training sets are often imbalanced. In particular, the embedding feature whose angle between the feature and the weight is distributed around 90° or 180° on the hypersphere reflects the difficult embedding feature in the process of classes. These phenomena occur when one considers those class which contains few number of embedding data. In order to address this problem, in this paper we propose an improved adaptive angular margin loss that incorporates the adaptive and robust angular margin on the angular space between the feature and the corresponding weight instead of constant margin. Our new margin loss function is constructed by incorporating adaptive and more robust angular margin constraint on angular space between the embedding feature and the corresponding weight. The proposed loss function improves the feature discrimination by minimizing the intra-class variation and maximizing the inter-class variation simultaneously. We present some experimental result on LFW, CALFW, CPLFW, AgeDB and MegaFace benchmarks, which demonstrate the effectiveness of the proposed approach.

Published in	American Journal of Neural Networks and Applications (Volume 11, Issue 1)
DOI	10.11648/j.ajnna.20251101.11
Page(s)	1-10
Creative Commons	This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited.
Copyright	Copyright © The Author(s), 2025. Published by Science Publishing Group

Keywords

Face Recognition, Class Imbalance, Angular Margin, Softmax Loss

References

[1]	Boutros F, Damer N, Kirchbuchner K, Kuijper A (2022) ElasticFace: Elastic margin loss for deep face recognition. preprint arXiv:2109.09416v4.
[2]	Cao Q, Shen L, Xie W, et al (2018) Vggface2: a dataset for recognizing faces across pose and age. In: 2018 13th IEEE international conference on automatic face & gesture recognition (FG 2018), IEEE, p 67-74 https://doi.org/10.1109/FG.2018.00020
[3]	Deng J, Guo J, Xue N, et al (2019) Arcface: additive angular margin loss for deep face recognition. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, p 4690-4699 https://doi.org/10.1109/CVPR.2019.00482
[4]	Deng J, Guo J, Yang J, et al (2022), ArcFace: additive angular margin loss for deep face recognition, preprint arXiv: 1801.07698v4.
[5]	Duan Y, Lu J, Zhou J (2019) Uniformface: learning deep equidistributed representation for face recognition. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, p3415~3424.
[6]	Guo Y, Zhang L, Hu Y, et al (2016) Ms-celeb-1m: a dataset and benchmark for large-scale face recognition. In: European conference on computer vision, Springer, p 87-102.
[7]	Huang GB, Mattar M, Berg T, et al (2008) Labeled faces in the wild: a database forstudying face recognition in unconstrained environments. In: Workshop on faces in’Real-Life’Images: detection, alignment, and recognition, p 1-15.
[8]	Huang GB, Ramesh M, Berg T, Learned-Miller E (2007) Labeled faces in the wild: A database for studying face recognition in unconstrained environments. Technical Report 07-49, University of Massachusetts, Amherst, Tech. Rep.
[9]	Jiao J, Liu W, Mo Y et al (2021) Dyn-arcface: dynamic additive angular margin loss for deep face recognition. Multimed Tools Appl 80(17): 25741-25756.
[10]	Kemelmacher-Shlizerman I, Sitz SM, Miler D, Brossard E (2016) The megaface benchmark: 1 million faces for recognition at scale. In: proceedings of the IEEE conference on computer vision and pattern recognition.
[11]	Khalifa A, Al-Hamadi A (2023) JAMsFace: joint adaptive margins loss for deep face recognition. Neural Computing and Applications, 35: 19025-19037. https://doi.org/10.1007/s00521-02308732-5
[12]	Kim Y, Park W, Roh M, et al (2020) Groupface: learning latent groups and constructing group-based representations for face recognition, In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, p 5620~5629 https://doi.org/10.1109/CVPR42600.2020.00566
[13]	Kim Y, Park W, Shin J (2020) Broadface: looking at tens of thousands of people at once for face recognition, In: European conference on computer vision https://doi.org/10.48550/arXiv.2008.06674
[14]	Liu H, Zhu X, Lei Z, et al (2019b) Adaptiveface: adaptive margin and sampling for face recognition. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, p 11947-11956. https://doi.org/10.1109/CVPR.2019.01222
[15]	Liu W, Wen Y, Yu Z, et al (2017) Sphereface: deep hypersphere embedding for face recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition, p 212-220. https://doi.org/10.1109/CVPR.2017.713
[16]	Liu Y, Li H, Wang X (2017) Rethinking feature discrimination and polymerization for large-scale recognition, preprint arXiv:1710.00870.
[17]	Meng Q, Zhao S, Huang Z, Zhou F (2021) Magface: a universal representation for face recognition and quality assessment, In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. https://doi.org/10.1109/CVPR46437.2021.01400
[18]	Moschoglou S, Papaioannou A, Sagonas C, et al (2017) Agedb: the first manually collected, in-the-wild age database. In: proceedings of the IEEE conference on computer vision and pattern recognition workshops, p 51-59. https://doi.org/10.1109/CVPRW.2017.250
[19]	Schroff F, Kalenichenko D, Philbin J (2015) Facenet: a unified embedding for face recognition and clustering. In: Proceedings of the IEEE conference on computer vision and pattern recognition, p 815-823. https://doi.org/10.1109/CVPR.2015.7298682
[20]	Sun Y, Chen Y, Wang X, et al (2014) Deep learning face representation by joint identification-verification, In: NeurIPS, p 1988~1996.
[21]	Sun Y, Cheng C, Zhang Y, et al (2020) Circle loss: a unified perspective of pair similarity optimization. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, p 6397~6406. https://doi.org/10.1109/CVPR42600.2020.00643
[22]	Taigman Y, Yang M, Ranzato M, et al (2014) Deepface: closing the gap to human-level performance in face verification, In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. https://doi.org/10.1109/CVPR.2014.220
[23]	Wang H, Wang Y, Zhou Z, et al (2018) Cosface: large margin cosine loss for deep face recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition, p 5265-5274. https://doi.org/10.48550/arXiv.1801.09414
[24]	Wang M, Deng W (2021) Deep face recognition: a survey. Neurocomputing 429: 215-244. https://doi.org/10.1016/j.neucom.2020.10.081
[25]	Wen Y, Zhang K, Li Z, et al (2016) A discriminative feature learning approach for deep face recognition. In: European conference on computer vision, Springer, p 499-515.
[26]	Xu J, Liu X, Zhang X, et al (2023) “X2-softmax: margin adaptive loss function for face recognition”, preprint arXiv:2312.05281v2.
[27]	Yi D, Lei Z, Liao S, Li S. Z. (2014) Learning face representation from scratch. preprint arXiv:1411.7923.
[28]	Zhang K, Zhang Z, Li Z et al (2016) Joint face detection and alignment using multitask cascaded convolutional networks. IEEE Signal Process Lett 23(10): 1499-1503. https://doi.org/10.1109/ICISCE.2017.95
[29]	Zhang W, Chen Y, Yang W, et al (2020) Class-variant margin normalized softmax loss for deep face recognition, IEEE Trans on Neural Networks and Learning Systems. https://doi.org/10.1109/TNNLS.2020.3017528
[30]	Zhang X, Zhao R, Qiao Y, et al (2019) Adacos: adaptively scaling cosine logits for effectively learning deep face representations. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, p 10823~10832. https://doi.org/10.1109/CVPR.2019.01108
[31]	Zhang Y, Deng W, Wang M, et al (2020) Global-local gcn: large-scale label noise cleansing for face recognition. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. https://doi.org/10.1109/CVPR42600.2020.00775
[32]	Zheng T, Deng W (2018) Cross-pose lfw: a database for studying cross-pose face recognition in unconstrained environments. Beijing Univ Posts Telecommun Tech Rep 5:7. https://doi.org/10.48550/arXiv.1708.08197
[33]	Zheng T, Deng W, Hu J (2017) Cross-age lfw: a database for studying cross-age face recognition in unconstrained environments. arXiv preprint arXiv:1708.08197.

Cite This Article

Plain Text BibTeX RIS

APA Style

Han, K., Yun, S., Song, C., Kim, K., O, C. (2025). An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition. American Journal of Neural Networks and Applications, 11(1), 1-10. https://doi.org/10.11648/j.ajnna.20251101.11

Copy | Download

ACS Style

Han, K.; Yun, S.; Song, C.; Kim, K.; O, C. An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition. Am. J. Neural Netw. Appl. 2025, 11(1), 1-10. doi: 10.11648/j.ajnna.20251101.11

Copy | Download

AMA Style

Han K, Yun S, Song C, Kim K, O C. An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition. Am J Neural Netw Appl. 2025;11(1):1-10. doi: 10.11648/j.ajnna.20251101.11

Copy | Download

@article{10.11648/j.ajnna.20251101.11,
  author = {Kwang-Uk Han and Song-Jun Yun and Chol Song and Kwang-Min Kim and Chol-Jun O},
  title = {An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition
},
  journal = {American Journal of Neural Networks and Applications},
  volume = {11},
  number = {1},
  pages = {1-10},
  doi = {10.11648/j.ajnna.20251101.11},
  url = {https://doi.org/10.11648/j.ajnna.20251101.11},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ajnna.20251101.11},
  abstract = {In recent years, there have been growing interests on deep learning based face recognition which currently produces state of the art standards in face detection, recognition and verification tasks. As is well known, loss function for extracting face feature plays a crucial role in deep face model. In this regards, margin-based loss functions which apply a fixed margin between the feature and the weight have attracted many interests. However, such margin-based losses have a somewhat limitation in enhancing the discriminative power and generalizability of the face model, since the intra-class and inter-class variations in the real face training sets are often imbalanced. In particular, the embedding feature whose angle between the feature and the weight is distributed around 90° or 180° on the hypersphere reflects the difficult embedding feature in the process of classes. These phenomena occur when one considers those class which contains few number of embedding data. In order to address this problem, in this paper we propose an improved adaptive angular margin loss that incorporates the adaptive and robust angular margin on the angular space between the feature and the corresponding weight instead of constant margin. Our new margin loss function is constructed by incorporating adaptive and more robust angular margin constraint on angular space between the embedding feature and the corresponding weight. The proposed loss function improves the feature discrimination by minimizing the intra-class variation and maximizing the inter-class variation simultaneously. We present some experimental result on LFW, CALFW, CPLFW, AgeDB and MegaFace benchmarks, which demonstrate the effectiveness of the proposed approach.
},
 year = {2025}
}

Copy | Download

TY  - JOUR
T1  - An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition

AU  - Kwang-Uk Han
AU  - Song-Jun Yun
AU  - Chol Song
AU  - Kwang-Min Kim
AU  - Chol-Jun O
Y1  - 2025/04/17
PY  - 2025
N1  - https://doi.org/10.11648/j.ajnna.20251101.11
DO  - 10.11648/j.ajnna.20251101.11
T2  - American Journal of Neural Networks and Applications
JF  - American Journal of Neural Networks and Applications
JO  - American Journal of Neural Networks and Applications
SP  - 1
EP  - 10
PB  - Science Publishing Group
SN  - 2469-7419
UR  - https://doi.org/10.11648/j.ajnna.20251101.11
AB  - In recent years, there have been growing interests on deep learning based face recognition which currently produces state of the art standards in face detection, recognition and verification tasks. As is well known, loss function for extracting face feature plays a crucial role in deep face model. In this regards, margin-based loss functions which apply a fixed margin between the feature and the weight have attracted many interests. However, such margin-based losses have a somewhat limitation in enhancing the discriminative power and generalizability of the face model, since the intra-class and inter-class variations in the real face training sets are often imbalanced. In particular, the embedding feature whose angle between the feature and the weight is distributed around 90° or 180° on the hypersphere reflects the difficult embedding feature in the process of classes. These phenomena occur when one considers those class which contains few number of embedding data. In order to address this problem, in this paper we propose an improved adaptive angular margin loss that incorporates the adaptive and robust angular margin on the angular space between the feature and the corresponding weight instead of constant margin. Our new margin loss function is constructed by incorporating adaptive and more robust angular margin constraint on angular space between the embedding feature and the corresponding weight. The proposed loss function improves the feature discrimination by minimizing the intra-class variation and maximizing the inter-class variation simultaneously. We present some experimental result on LFW, CALFW, CPLFW, AgeDB and MegaFace benchmarks, which demonstrate the effectiveness of the proposed approach.

VL  - 11
IS  - 1
ER  -

Copy | Download

Author Information

Kwang-Uk Han

Institute of Mathematics, State Academy of Sciences, Pyongyang, Democratic People’s Republic of Korea

http://orcid.org/0009-0007-5835-904X
Song-Jun Yun

Institute of Mathematics, State Academy of Sciences, Pyongyang, Democratic People’s Republic of Korea

http://orcid.org/0009-0005-0894-4596
Chol Song

Institute of Mathematics, State Academy of Sciences, Pyongyang, Democratic People’s Republic of Korea

http://orcid.org/0009-0009-7383-937X
Kwang-Min Kim

Institute of Mathematics, State Academy of Sciences, Pyongyang, Democratic People’s Republic of Korea

http://orcid.org/0009-0000-9350-2605
Chol-Jun O

Institute of Mathematics, State Academy of Sciences, Pyongyang, Democratic People’s Republic of Korea

Contact Email

http://orcid.org/0009-0006-9913-8094

Download PDF

Sections

Plain Text BibTeX RIS

APA Style

Han, K., Yun, S., Song, C., Kim, K., O, C. (2025). An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition. American Journal of Neural Networks and Applications, 11(1), 1-10. https://doi.org/10.11648/j.ajnna.20251101.11

Copy | Download

ACS Style

Han, K.; Yun, S.; Song, C.; Kim, K.; O, C. An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition. Am. J. Neural Netw. Appl. 2025, 11(1), 1-10. doi: 10.11648/j.ajnna.20251101.11

Copy | Download

AMA Style

Han K, Yun S, Song C, Kim K, O C. An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition. Am J Neural Netw Appl. 2025;11(1):1-10. doi: 10.11648/j.ajnna.20251101.11

Copy | Download

@article{10.11648/j.ajnna.20251101.11,
  author = {Kwang-Uk Han and Song-Jun Yun and Chol Song and Kwang-Min Kim and Chol-Jun O},
  title = {An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition
},
  journal = {American Journal of Neural Networks and Applications},
  volume = {11},
  number = {1},
  pages = {1-10},
  doi = {10.11648/j.ajnna.20251101.11},
  url = {https://doi.org/10.11648/j.ajnna.20251101.11},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ajnna.20251101.11},
  abstract = {In recent years, there have been growing interests on deep learning based face recognition which currently produces state of the art standards in face detection, recognition and verification tasks. As is well known, loss function for extracting face feature plays a crucial role in deep face model. In this regards, margin-based loss functions which apply a fixed margin between the feature and the weight have attracted many interests. However, such margin-based losses have a somewhat limitation in enhancing the discriminative power and generalizability of the face model, since the intra-class and inter-class variations in the real face training sets are often imbalanced. In particular, the embedding feature whose angle between the feature and the weight is distributed around 90° or 180° on the hypersphere reflects the difficult embedding feature in the process of classes. These phenomena occur when one considers those class which contains few number of embedding data. In order to address this problem, in this paper we propose an improved adaptive angular margin loss that incorporates the adaptive and robust angular margin on the angular space between the feature and the corresponding weight instead of constant margin. Our new margin loss function is constructed by incorporating adaptive and more robust angular margin constraint on angular space between the embedding feature and the corresponding weight. The proposed loss function improves the feature discrimination by minimizing the intra-class variation and maximizing the inter-class variation simultaneously. We present some experimental result on LFW, CALFW, CPLFW, AgeDB and MegaFace benchmarks, which demonstrate the effectiveness of the proposed approach.
},
 year = {2025}
}

Copy | Download

TY  - JOUR
T1  - An Improved Adaptive Angular Margin Loss Function for Deep Face Recognition

AU  - Kwang-Uk Han
AU  - Song-Jun Yun
AU  - Chol Song
AU  - Kwang-Min Kim
AU  - Chol-Jun O
Y1  - 2025/04/17
PY  - 2025
N1  - https://doi.org/10.11648/j.ajnna.20251101.11
DO  - 10.11648/j.ajnna.20251101.11
T2  - American Journal of Neural Networks and Applications
JF  - American Journal of Neural Networks and Applications
JO  - American Journal of Neural Networks and Applications
SP  - 1
EP  - 10
PB  - Science Publishing Group
SN  - 2469-7419
UR  - https://doi.org/10.11648/j.ajnna.20251101.11
AB  - In recent years, there have been growing interests on deep learning based face recognition which currently produces state of the art standards in face detection, recognition and verification tasks. As is well known, loss function for extracting face feature plays a crucial role in deep face model. In this regards, margin-based loss functions which apply a fixed margin between the feature and the weight have attracted many interests. However, such margin-based losses have a somewhat limitation in enhancing the discriminative power and generalizability of the face model, since the intra-class and inter-class variations in the real face training sets are often imbalanced. In particular, the embedding feature whose angle between the feature and the weight is distributed around 90° or 180° on the hypersphere reflects the difficult embedding feature in the process of classes. These phenomena occur when one considers those class which contains few number of embedding data. In order to address this problem, in this paper we propose an improved adaptive angular margin loss that incorporates the adaptive and robust angular margin on the angular space between the feature and the corresponding weight instead of constant margin. Our new margin loss function is constructed by incorporating adaptive and more robust angular margin constraint on angular space between the embedding feature and the corresponding weight. The proposed loss function improves the feature discrimination by minimizing the intra-class variation and maximizing the inter-class variation simultaneously. We present some experimental result on LFW, CALFW, CPLFW, AgeDB and MegaFace benchmarks, which demonstrate the effectiveness of the proposed approach.

VL  - 11
IS  - 1
ER  -

Copy | Download