Facial Expression Recognition Model Based on CNN and Data Augmentation Method

Abstract

Face expressions are essential in expressing human emotions, and it is accomplished by separating features and categorizing them. Facial expression recognition technology has been widely employed in human-computer interaction, telemedicine, mental health, and criminal investigation detection. In recent years, significant advances in deep learning have facilitated the development of facial expression recognition, making it increasingly accessible. Convolutional neural networks (CNN) are the foundation of the face expression recognition model presented in this article. In order to maximize the final accuracy, an abundance of sample photographs are required for training and optimization. As a result, the 35,886 facial expression photos from the FER2013 dataset, which contains all seven emotions, were used for both training and testing. The photos were scaled down to 4848 pixels during data preparation, and data augmentation was carried out. To get accurate face expression recognition results, various optimizers and parameters were chosen for the training network, which was a bespoke structure based on the VGG network design. The model constructed in this study achieved an accuracy of 73.16% during prediction.