Comparative Analysis of Facial Recognition Based Attendance System Using Machine Learning

Department of Data Science and Business Systems, SRM Institute of Science and Technology, Kattankulathur, India

^* Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

Automated attendance systems benefit from robust, scalable facial recognition models capable of handling real-world classroom variability illumination, pose, and occlusion. This study conducts a systematic comparison of four pipelines ArcFace, SFace, GhostFaceNet, and Dlib using the DeepFace and face recognition frameworks on a custom dataset of approximately 13,000 images representing 1,600 identities. Evaluation metrics include identification accuracy (Top-1), verification reliability (AUC, TAR@FAR), latency, and throughput (FPS). Dlib achieved the best overall performance (Top-1 = 87.93%, AUC = 0.9952, 11.53 FPS), outperforming deeper CNN-based embeddings in both speed and accuracy. The benchmark highlights trade-offs between accuracy, discriminability, and real-time efficiency, providing practical deployment insights for classroom automation.