Expression recognition with deep features extracted from holistic and part-based models

作者：

Highlights：

• Compares holistic and part-based deep models for facial expression recognition (FER)

• Proposes multi-face multi-part (MFMP) model for FER

• Proposes an effective data augmentation strategy for FER

• Analyzes the effect of skip connections in deep FER models

• Superior performance on individual databases and cross-database evaluation scenario

摘要

•Compares holistic and part-based deep models for facial expression recognition (FER)•Proposes multi-face multi-part (MFMP) model for FER•Proposes an effective data augmentation strategy for FER•Analyzes the effect of skip connections in deep FER models•Superior performance on individual databases and cross-database evaluation scenario

论文关键词：Facial expression recognition,Convolutional neural networks,Part-based face representation,Data augmentation

论文评审过程：Received 25 January 2019, Revised 5 September 2020, Accepted 23 September 2020, Available online 28 September 2020, Version of Record 12 January 2021.

论文官网地址：https://doi.org/10.1016/j.imavis.2020.104038