A Comparative Analysis of Diverse Deep Learning Techniques for Facial Emotion Recognition

In this paper a different methodology of deep learning algorithms is presented in facial emotion recognition. In this paper, we jump right into computer vision and affective computing by studying how different deep learning approaches applied to a dataset composed of facial images annotated with emotional expressions will work. We employ algorithms such as AlexNet, LeNet, Convolutional Neural Networks (CNN), Long Short-Term Memory Networks (LSTM), VGG16, and Residual Networks (ResNet). By leveraging the intrinsic properties of these algorithms, we conduct a comprehensive study on facial features to enhance emotion recognition. For this investigation, we use a dataset that includes the data necessary to train and test models. We selected a variety of algorithms to enable an accurate comparison of their performance in recognizing facial emotions, utilizing metrics like precision, recall, and F1 score. Following the advent of AlexNet and LeNet, we explored the general application of CNNs and specifically adapted LSTM for facial recognition. We also investigated VGG16 and ResNet to determine how each algorithm excels in identifying emotional expressions on faces. This work makes a significant step forward in emotion recognition in computer vision and it also explores the algorithm selection for similar tasks.