Deep Learning-Based Recognition of Facial Expressions

Abstract

Abstract: Convolutional neural networks were used in deep learning to maintain a system for recognizing face expressions of emotion. We were created as two distinct models. The first model was a suggested CNN architecture that was trained on the FER-2013 dataset. The model could classify expressions into 7 different categories with an accuracy rate of 67.18%. Using the FER-2013 dataset and a transfer learning strategy, the second model was produced. The model was able to categories the expressions into 4 groups with an accuracy of 75.55%. A mobile web application that quickly executes our FER models on a device is also provided by us. We introduce generic assessment standards, general face recognition databases, and face recognition research for real-world scenarios. We present a prospective analysis of facial recognition. Face recognition has emerged as the field's most promising area for future advancement.