Real-Time Detection of Student Engagement: Deep Learning-Based System

Due to the spread of COVID-19, E-learning has become the only option for the teachers and students. However, it is difficult for a teacher to monitor each student’s engagement while teaching online. This paper aims to develop an automated real-time video-based system to detect student engagement during online classes effortlessly and efficiently. The MobileNet model is trained on an eye images dataset from Kaggle and achieved 99% accuracy on data validation. The result obtained from training the model is used with the Viola–Jones algorithm and OpenCV. By using the built-in camera of the laptop, the system can detect whether the student is engaged or disengaged from his/her eye gaze. In the disengagement state detection, a buzzer sound starts.