Title: An improved convolutional neural network based on centre loss for facial expression recognition

Authors: Fangfei Zu; ChangJun Zhou; Xiao Wang

Addresses: Department of Mathematics and Computer Science, Zhejiang Normal University, Jinhua, Zhejiang, China ' Department of Mathematics and Computer Science, Zhejiang Normal University, Jinhua, Zhejiang, China ' Department of Mathematics and Computer Science, Zhejiang Normal University, Jinhua, Zhejiang, China

Abstract: In recent years, due to the development of deep learning and convolutional neural network (CNN), the field of computer vision has been booming. As a subdivision of face recognition technology, facial expression recognition (FER) has made a lot of progress. In this paper, a facial expression prediction model based on CNN is proposed, which uses multiple small-scale convolution layers to jointly extract features, and centre loss is introduced to further improve generalisation ability. The method has a shorter training time and a higher accuracy rate. After validation and testing on the Fer2013 dataset, the JAFFE dataset, and the CK+ dataset, the average accuracy of the proposed network architecture is 71.39%, 96.64%, and 99.39%, respectively.

Keywords: convolutional neural network; CNN; deep learning; centre loss; facial expression recognition; FER.

DOI: 10.1504/IJAIS.2021.117903

International Journal of Adaptive and Innovative Systems, 2021 Vol.3 No.1, pp.58 - 73

Received: 22 Jan 2021
Accepted: 29 Mar 2021
Published online: 04 Oct 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article