Classificação de emoções a partir de expressões faciais com deep learning

Abstract

Mental health is a subject that is increasingly on the rise, and it is of paramount importance to know and talk about. Problems such as stress, anxiety and depression are considered as the diseases of the century, affecting more and more people around the world. Identifying and treating such disorders is expressly necessary since, in severe cases such as depression, they can even lead to suicide. In a classroom, the situation is no different, learning difficulties, lack of motivation and fear of not being able to meet expectations are just a few examples of what can lead a student to develop a mental health problem. Students with negative emotional states have greater learning difficulties, which impacts their school performance and can further worsen their emotional state, becoming a vicious cycle. Identifying emotions in the classroom can help in identifying learning difficulties. In this context, diagnostic systems, based on machine learning (machine and deep learning), have been used to identify cases of anxiety, distress and depression. Such diagnoses are possible by obtaining classification models, using learning mechanisms. As an example, a model based on a deep learning architecture, that identifies and classifies students' emotions, can help the teacher with feedback about their class. For the purposes of this conclusion work, the objective is to design and implement a classification model based on deep learning methods for classifying emotions through images of human faces. This would be the first step towards the development of a system that can infer difficulties in learning in the classroom, thus enabling the educator to approach new teaching strategies that can improve the performance of the class and their emotional states. [resumo fornecido pelo autor]