E-Learning Facial Emotion Recognition Using Deep Learning Models

In the rapidly evolving landscape of e-learning, understanding and responding to students’ emotional states can significantly enhance the educational experience. In this research, a novel Facial Emotion Recognition (FER) for improving our understanding of students during e-learning is proposed. We explored the pivotal role of Deep Learning (DL) models in FER, shedding light on their effectiveness within the e-learning context. Leveraging a custom dataset tailored for e-learning scenarios, we employ six distinct DL architectures: MobileNet, VGG19, Convolutional Neural Networks (CNN), ResNet50, VGG16 and CNN-LSTM (long short-term memory). Our comprehensive evaluation demonstrates the capacity of these models to discern facial emotions with varying degrees of accuracy. Notably, the MobileNet model achieved an accuracy of 83,78%, VGG19 at 82,89%, CNN at 81,50% ResNet50 at 82,79%, VGG16 at 81,70% and CNN-LSTM at 68,15%.