RESEARCH OF FACIAL EXPRESSION RECOGNITION BY DEEP LEARNING USING RESNET ARCHITECTURE

Abstract

Facial recognition is the main method for nonverbal processing intentions. Research on facial expression recognition has been interested in research and application in many parts of the world. Therefore, this paper focuses on the problem of facial expression recognition by deep learning method using ResNet101 network architecture. The reliability of the model was assessed based on the sample data set available FER2013 for the highest recognition rate of 71.22%. From the detailed analysis of the accuracy of each type of expression, the author offers the solution to propose three main expressive groups to develop a service quality assessment program with three levels: satisfaction, normal and unsatisfactory.