COMPARATIVE ANALYSIS AND EVALUATION OF FACIAL EXPRESSION RECOGNITION USING DEEP LEARNING

Abstract

Facial emotion recognition (FER) is vital for human-computer interactions such as clinical practice and behavioral description. Accurate and robust FER using computer models remains a difficulty, because of the non-uniformity of human features and visual differences such as diverse facial positions and illumination. Among all FER model strategies, Convolutional Neural Networks (CNNs) show considerable promise owing to their excellent automated feature extraction and computational efficiency. In this study, we obtained the greatest single-network classification accuracy on the FER2013 dataset. We take the VGG Net architecture, carefully tweak its hyperparameters, and experiment with several optimization strategies. According to our experiments, our model has accuracy of 88.62 percent on FER2013 without needing extra training data. In this study, we apply Resnet and Dense Net models and use the FER2013 dataset to show algorithmic differences and performance implications. The Resnet model produced an accuracy of 88.62 percent, while the Dense Net model produced an accuracy of 85.83 percent.