A Facial Recognition Approach for Real-Time Student Attendance Tracking

Abstract

Attendance tracking remains a fundamental component of academic administration, serving as a basis for evaluating student engagement, verifying participation, and ensuring institutional accountability. Traditional methods—such as manual roll calls, paper-based sign-in sheets, or RFID card systems—are often inefficient, susceptible to human error, and unsuitable for environments requiring contactless interactions. In response to these limitations, this study presents a real-time student attendance tracking system that leverages facial recognition technology built using Google Teachable Machine. The proposed solution integrates a web-based interface with a lightweight machine learning model capable of processing live video streams, identifying students, and automatically logging attendance with minimal human intervention. The system was designed to balance accessibility, accuracy, and performance through a structured model development pipeline. A curated dataset containing variations in lighting, facial orientation, and expression was prepared to enhance the robustness of the Convolutional Neural Networks (CNN) during real-world operation. The trained model was deployed using TensorFlow.js to enable real-time inference directly within a browser environment, eliminating the need for extensive backend computational resources and supporting seamless integration with existing academic workflows. Experimental evaluation demonstrated strong performance, achieving recognition accuracy exceeding 90% under controlled settings while maintaining reliable functionality in more challenging conditions, including suboptimal lighting and partial facial occlusion. The system effectively mitigated common administrative challenges by reducing manual workload, improving the precision of attendance records, and ensuring efficient real-time data capture. Nonetheless, limitations associated with environmental variability highlight opportunities for further refinement in model training and system architecture. Overall, the findings indicate that CNN-based facial recognition systems offer a practical and scalable solution for modernizing attendance management in educational contexts. The proposed framework contributes to ongoing advancements in AI-supported educational technologies and underscores the potential for expanding automated administrative tools within broader institutional ecosystems.