Facial emotion recognition based on deep transfer learning approach

Facial expressions play a major role in the communication of emotions through nonverbal channels. In recent years, the topic of automatic facial expression recognition (FER) has become very popular. Researchers are looking at how it may be used in education, security surveillance, smart healthcare system, and to understand the behavior of a community or a person. As long as there are variations in images, such as different poses and lighting, accurate and robust FER remains a challenge using computer models. We developed an approach to automatically classifying facial expressions based on deep transfer learning. The approach was constructed with convolutional neural networks (CNN) and VGG19, which is a transfer learning model. To train the model, we employed contemporary deep learning techniques such as optimal learning rate finder, fine-tuning, and data augmentation. On both the Extended Cohn-Kanade (CK+) and the Japanese Female Facial Expression (JAFFE) datasets, the proposed model achieved accuracy values of 94.8% and 93.7%, respectively. The developed system has already been tested on a vast database and can be used to assist online education systems, surveillance systems, and smart healthcare systems in their daily activities.