An Evaluation of Virtual Classroom Performance with Artificial Intelligence Components

Abstract

The recent advancement of information technology allows educators and students to interact with Artificial Intelligence (AI) through smart classroom channels. This channel is one of the latest technology-enhanced learning (TEL) that provides a learning environment with educators and students interaction during the learning process. Currently, smart classrooms are believed to change current dull teaching methods and enhance the students’ learning experience. Hence, this paper shows a comprehensive investigation of applying AI components to an intelligent classroom system (a.k.a virtual classroom system) that provides hand gestures and face detection through e-learning classrooms. Machine Learning libraries are implemented and compared on three machines with varying hardware specifications and capabilities. As a result of this study, Tensorflow Handpose provides more accuracy than MediaPipe Hands, although it requires higher computational capabilities. Face-api.js outperforms TensorFlow and MediaPipe when it comes to executing face detection functions. In addition to the study, the presented face and hand APIs can be adopted in a real time implementation for smart classroom systems.