EmoStudent: Developing a Dataset to Analyse Students' Emotional Well-Being

Abstract

This article introduces an initial version of a dataset designed to educate and assess models that concentrate on studying of the emotional condition of students throughout the remote learning process. This dataset comprises short video clips showing the faces of individuals from diverse ethnic backgrounds and age groups in front of the computer screen. No dataset specialising in the proctoring systems problem solving task was found (process of working in front of the computer, emotions during the educational process). As a result, existing datasets for solving problems in related fields were analysed: emotion classification, emotion recognition, and face recognition. Building on this analysis and the specifics of the chosen data source (YouTube videos with Creative Commons license), the previously established criteria for creating the dataset were modified and expanded. A more adaptable approach was introduced concerning the categorization based on age and ethnicity. A path for future endeavors was also delineated, proposing an enhancement of the current implementation to encompass a broader spectrum of emotions and individuals with various forms of disabilities in subsequent iterations.

Keywords

• dataset
• computer vision
• emotion understanding
• artificial intelligence-based proctoring systems