TY - GEN
T1 - Local and global feature learning for subtle facial expression recognition from attention perspective
AU - Wang, Shaocong
AU - Yuan, Yuan
AU - Feng, Yachuang
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2019.
PY - 2019
Y1 - 2019
N2 - Subtle facial expression recognition is important for emotion analysis. In the field of subtle facial expression recognition, there are two intrinsic characters. Firstly, subtle facial expression usually exhibits very small variations in different facial areas. Secondly, those small variations are closely correlated, and they together form an expression. Inspired by these two characteristics of facial expression, a model focus on local variations and their correlations is proposed in this paper. We utilize several attention maps to automatically attend to distinct local regions and extract local features. And then, a self-attention operation is ensembled to extract global correlation feature over the whole image. The global and local features are further fused in an efficient way to classify the facial expression. Extensive experiments have been carried out on LSEMSW and CK+ datasets.
AB - Subtle facial expression recognition is important for emotion analysis. In the field of subtle facial expression recognition, there are two intrinsic characters. Firstly, subtle facial expression usually exhibits very small variations in different facial areas. Secondly, those small variations are closely correlated, and they together form an expression. Inspired by these two characteristics of facial expression, a model focus on local variations and their correlations is proposed in this paper. We utilize several attention maps to automatically attend to distinct local regions and extract local features. And then, a self-attention operation is ensembled to extract global correlation feature over the whole image. The global and local features are further fused in an efficient way to classify the facial expression. Extensive experiments have been carried out on LSEMSW and CK+ datasets.
KW - Attention
KW - Subtle facial expression recognition
UR - http://www.scopus.com/inward/record.url?scp=85076981025&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-31723-2_57
DO - 10.1007/978-3-030-31723-2_57
M3 - 会议稿件
AN - SCOPUS:85076981025
SN - 9783030317225
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 670
EP - 681
BT - Pattern Recognition and Computer Vision 2nd Chinese Conference, PRCV 2019, Proceedings, Part II
A2 - Lin, Zhouchen
A2 - Wang, Liang
A2 - Tan, Tieniu
A2 - Yang, Jian
A2 - Shi, Guangming
A2 - Zheng, Nanning
A2 - Chen, Xilin
A2 - Zhang, Yanning
PB - Springer
T2 - 2nd Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2019
Y2 - 8 November 2019 through 11 November 2019
ER -