A multi-modal facial recognition framework for continuous attendance tracking: Comparative performance evaluation

Authors

DOI:

https://doi.org/10.18488/76.v14i1.5374

Keywords:

Classroom analytics, Cloud-based monitoring, Continuous authentication, Convolutional neural networks, Haar cascade classifier, Multi-face recognition.

Abstract

Facial recognition is a mature biometric technology. Nonetheless, the use of this biometric in continuous classroom attendance monitoring is very limited. The use of this biometric for attendance monitoring relies on isolated points of verification. Moreover, the technology is vulnerable to proxy attendance. This work proposes a real-time, multi-modal facial recognition framework for continuous attendance tracking using a deep neural network pipeline. The system performs periodic face scanning during live classroom sessions, records attendance at 30-minute intervals, enforces a minimum 75% presence criterion, and synchronizes all operational events to a scalable Firebase cloud backend for real-time administrative access. The framework is evaluated on a real-world dataset comprising over 1,000 authentic student videos captured in conventional classroom environments, inherently incorporating variations in illumination, pose, occlusion, and background noise. Comparative evaluation is conducted against SVM, a vanilla CNN, InceptionV3, MobileNet, and VGG-19. This research presents a deep-learning-based automated attendance system that employs Haar Cascade for real-time face detection and CNN-based architectures for face recognition. The framework records 98.1% training accuracy and 97.4% validation accuracy, outperforming every baseline model. The MobileNet model is the only one close to this achievement. Analysis through statistical validation by one-way ANOVA (F (5,24) = 1408.61, p<0.0001), Tukey’s HSD, and paired t-tests (t=69.33, p<10⁻⁶) shows robust results. The proposed system is reliable, scalable, and computationally efficient, enabling low-overhead, large-scale deployment for continuous attendance monitoring and institutional process automation.

Downloads

Download data is not yet available.

Downloads

Published

2026-11-20