Facial expression recognition using densely connected convolutional neural network and hierarchical spatial attention

•The proposed method can adaptively locate salient regions.•The proposed method can adaptively focus on the emotional related features.•The proposed method can make facial expressions being represented more efficiently. This paper is dedicated to eliminating the impact of redundant information from...

Full description

Saved in:
Bibliographic Details
Published in:Image and vision computing 2022-01, Vol.117, p.104342, Article 104342
Main Authors: Gan, Chenquan, Xiao, Junhao, Wang, Zhangyi, Zhang, Zufan, Zhu, Qingyi
Format: Article
Language:English
Subjects:
Densely connected convolutional neural network
Facial expression recognition
Facial image
Spatial attention
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:•The proposed method can adaptively locate salient regions.•The proposed method can adaptively focus on the emotional related features.•The proposed method can make facial expressions being represented more efficiently. This paper is dedicated to eliminating the impact of redundant information from emotional-unrelated regions on facial expression recognition (FER). To this end, a densely connected convolutional neural network with hierarchical spatial attention is proposed. Specifically, it can adaptively locate salient regions and focus on the emotional related features so that the facial expressions can be represented more efficiently. This superior performance is also verified by some experiments. Experimental results reveal that the proposed method can distinguish facial expression more accurately than existing state-of-the-art methods.
ISSN:0262-8856
1872-8138
DOI:10.1016/j.imavis.2021.104342