Classification of Attentiveness on Virtual Classrooms using Deep Learning for Computer Vision

Gapi, K. T.; Magbitang, R. M. G.; Villaverde, J. F.

Nowadays, virtual classrooms are highly encouraged due to the COVID-19 pandemic. This could be a disadvantage because some students might not really be engaged with this kind of setup. This study presents a system for classifying level of attentiveness on virtual classrooms using deep learning for computer vision. The study confined in the development of the technology for classifying attentiveness itself, the integration of the system to virtual classrooms is not included in the scope. The criteria for the classification include the prediction of droopy corners of mouth facial cue, hanging eyelid facial cue, eye state, and eye gaze. The software of the system used the combinations of Convolutional Neural Network (CNN) models, Dlib, and OpenCV library. After evaluation, the system was able to successfully classify attentiveness of three classes with an overall accuracy of 83.33%. © 2021 ACM.