Application of Facial Emotion Recognition Based on Multi-Scale Fusion in Smarter Education

The traditional teaching mode has little interaction between teachers and students, and teachers can not grasp students' emotional changes in time. Therefore, how to timely and effectively feed back students' emotional state to teachers in class has become an urgent problem to be solved in current classroom teaching. This paper takes the classroom as the application scenario and uses the deep learning technology to propose a residual attention network and multi-scale fusion student emotion recognition model. The model uses image preprocessing to improve the quality of the input image, uses residual attention module to guide the network to focus on local salient features, uses multi-scale fusion module to strengthen the information interaction between different levels and expand the receptive field, and obtains high accuracy on FER2013 data set. Using this method to identify the students' attention state in class and judge the teacher's teaching effect and students' classroom listening attitude can correct students' listening attitude, improve teachers' teaching efficiency and promote the improvement of teaching quality.