Co-attentive multi-task convolutional neural network for facial expression recognition

作者：

Highlights：

• Unlike previous hard parameter sharing methods, we introduce FLD as the auxiliary task of FER based on soft parameter sharing methods.

• Based on the cross-stitch network, we propose an end-to-end co-attentive convolutional neural network.

• We conduct detailed experiments to analyze all aspects of our method, which validate the effectiveness and generalizability of our approach.

摘要

•Unlike previous hard parameter sharing methods, we introduce FLD as the auxiliary task of FER based on soft parameter sharing methods.•Based on the cross-stitch network, we propose an end-to-end co-attentive convolutional neural network.•We conduct detailed experiments to analyze all aspects of our method, which validate the effectiveness and generalizability of our approach.

论文关键词：Facial expression recognition,Facial landmarks detection,Multi-task learning

论文评审过程：Received 30 June 2020, Revised 5 September 2021, Accepted 24 October 2021, Available online 26 October 2021, Version of Record 6 November 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108401