Robotic Attendance Scheme in the Classroom Using Artificial Intelligence and Internet of Things

Abstract

Facial recognition is one example of the ways in which computer vision is being applied in modern, people-oriented utility apps. Facial recognition technology has improved in recent years, but it still lags behind other biometric authentication methods such as fingerprint and iris scanning or Radio Frequency Identification (RFID) cards in terms of accuracy. However, it is still frequently utilised due to the fact that the recognition procedure does not involve touching the device in any way. This research takes advantage of advances in facial recognition technology to offer an embedded device integrated AI and IoT-based automated attendance solution for use in smart classrooms. This study uses the ResNet architecture to explore a deep learning classification method for identifying student attendance. In this research, we used facial photos from students to train ResNet-18 and ResNet-50. Models educated to differentiate regular from irregular student attendance. Our photos were evaluated on 20%, 25%, and 40% of simulated and real-world datasets, respectively. Results from three types of testing data show that ResNet-50 outperforms ResNet-18 in rapports of accuracy, F1-measure, and mean accuracy value. This study makes use of the FER2013 dataset for labelling faces, as well as real-time student faces. This study uses a deep learning approach to show how dependable and repeatable student image analysis