Система розпізнавання емоцій людини з використанням алгоритмів глибокого навчання

Abstract

In our research, we introduce and implement a flexible framework for constructing convolutional neural networks (CNNs) that are capable of real-time processing. We validate the effectiveness of our models by creating a real-time vision system that combines face detection and emotion classification tasks using our proposed CNN architecture. We provide details of the training procedure setup and evaluate our models on standard benchmark datasets. Our results show an accuracy of 66% on the FER-2013 emotion dataset.