Dyn-arcFace: dynamic additive angular margin loss for deep face recognition

Deep convolutional neural networks (CNNs) are widely used in face recognition, because they can extract features with higher discrimination, which is the basis for correctly identifying the identity of a face image. In order to improve the face recognition performance, in addition to improving the s...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications 2021-07, Vol.80 (17), p.25741-25756
Main Authors: Jiao, Jichao, Liu, Weilun, Mo, Yaokai, Jiao, Jian, Deng, Zhongliang, Chen, Xinping
Format: Article
Language:English
Subjects:
Algorithms
Artificial neural networks
Computer Communication Networks
Computer Science
Data Structures and Information Theory
Datasets
Face recognition
Feature extraction
Multimedia Information Systems
Neural networks
Object recognition
Special Purpose and Application-Based Systems
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Deep convolutional neural networks (CNNs) are widely used in face recognition, because they can extract features with higher discrimination, which is the basis for correctly identifying the identity of a face image. In order to improve the face recognition performance, in addition to improving the structures of convolutional neural networks, many new loss functions have been proposed to enhance the distinguishing ability of extract features. However, according to our research, it is found that when using the present loss functions in CNNs, there is overfitting of the training dataset and redeuces the effect of face recognition. Therefore, a new loss function called Dyn-arcFace(Dynamic Additive Angular Margin Loss for Deep Face Recognition) is proposed in this paper. In Dyn-arcFace, the traditional fixed additive angular margin is developed into a dynamic one, which can reduce the degree of overfitting caused by the fixed additive angular margin. To verify the effect of Dyn-arcFace, we tested on different layers of neural networks. The proposed algorithm achieved state-of-the-art performance on the most popular public-domain face recognition datasets.
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-021-10865-5