A Graph Convolutional Network for Emotion Recognition in Context
Hanxin Zeng, Gen Li, Tong Tong, Qinquan Gao · 2020
In this paper, we propose a novel framework based on graph convolutional neural network for emotion recognition by utilizing the region-based semantic information. To extract informative sematic region, we first utilized a bottom-up attention module to extract salient images. After that, these images are used for establishing graph features. This graph convolutional neural network extracts emotion features from graph node features for emotion recognition. Additionally, to remove redundant features, this paper utilizes Gate Recurrent Unit that consists of gate and memory units. In the end, we conduct some emotion recognition experiments to prove the superior performance of our model on CEAR dataset compared to the state-of-the-art approaches.