UC-59 Analyzing Concentration Levels in Online Learning with Facial Values

Can deep learning models accurately predict whether an individual is focused or distracted on a task in order to improve learning efficiency? In the context of online learning with the use of a webcam, this project is aimed at detecting concentration levels of students to potentially assist with improving learning efficiency. Machine learning technologies have been utilized to evaluate students’ facial expression and eye movements to identify whether a student is focused or distracted. The machine learning branch that is employed is a supervised learning model. This supervised learning model makes predictions based on given input features. A total of 6 different models were employed. 4 of those models employed collected eye data. The other two models employed the use of facial and eye data to predict concentration. Ultimately, the eye model accuracy hovered between 50% and 56% accuracy in prediction, with a significant amount of loss. The eye models with attention provided the best accuracy and loss rates out of the four eye models. Secondly, the facial and eye models also hovered right around 50% accuracy with significant loss of around 3.8 and 3.7. The reported results suggest that the data was inaccurate or insufficient in some models to accurately predict concentration levels in an individual. Given a larger collection and more consistent data, the reported results would provide to be more accurate at predicting concentration.Advisors(s): Dr. Linh Le (Sponsor/Project Owner) Dr. Ying Xie (Sponsor/Project Owner)Topic(s): Artificial IntelligenceIT 498