Student Recognition and Activity Monitoring in E-Classes Using Deep Learning in Higher Education

Abstract

Monitoring student activity manually constantly is a laborious endeavor. Over the past few years, there has been a rapid expansion in the usage of cameras and the automatic identification of odd surveillance behavior. Different computer vision algorithms have been used to observe and monitor realworld activities. Most educational institutions are already offering online programs to lessen the impact of this epidemic on the education industry. However, ensuring that students are correctly identified, and their behaviors are monitored is crucial to make these online learning sessions dynamic and equivalent to the conventional offline classroom. In this study, we have introduced brand-new deep learning-based algorithms that continuously track a student's mood, including rage, contempt, happiness, sorrow, fear, and surprise. The effectiveness of student identification and activity monitoring in online classrooms was also studied using deep learning and a CNN model that reaches 99% accuracy. Our approach was superior because of its many convolutional layers, dropout regularization, and batch normalization. It caught crucial properties and decreased overfitting. By identifying them more frequently, deep learning techniques can enhance student engagement and learning outcomes in e-learning situations, according to the research. With these techniques, educators and instructors may support students more effectively by better comprehending their behavior and offering specialized and individualized support, improving academic performance and student activity evaluation.