Attendance Management Optimization with Image Processing
Attendance Management Optimization with Image Processing
Authors:
Prof. Dr. Shubhra Mukherjee Mathur¹, Chanakya Marbate², Soham Kokate², Siddhant Dakhane², Venkat Tarun A.²
¹Assistant Professor, Department of Computer Engineering, MIT Art Design and Technology University, Pune, India
²Department of Computer Engineering, MIT Art Design and Technology University, Pune, India
- a) mukherjeeshubhra@gmail.com
b) chanakyamar26@gmail.com
c) sohamkokate16@gmail.com
d) siddhantdakhane23@gmail.com
e) appanitarun78@gmail.com
Abstract: Attendance management has traditionally been a manual , labor-intensive process in educational instituitions worldwide. As observed in many large-scale classroom settings, traditional methods are not only tedious but also highly susceptible to human error, resulting in lost registers and significant time loss. To address these inefficiencies , this paper proposes a facial recognition framework as a seamless alternative to roll calls. The framework evaluates four distinct combinations: RetinaFace + FaceNet, RetinaFace + ArcFace, MTCNN + FaceNet, and MTCNN + ArcFace. Following methdologies similar to the Image Based Attendance optimizer(IBA0) proposed by Mathur et al. (2025)[13], this study benchmark’s performance in realistic conditions. The results demonstrates that RetinaFace[2] is a superior detector , achieving 94.20% accuracy on easy cases. Ultimately , the combination of RetinaFace + FaceNet[4] emerged as optimal solution , yielding a Rank-1 Recognition Rate of 99.114%. This research contributes to the growing field of automated monitoring , complementing existing studies such as the Face Recognition Attendance Monitoring System at MIT-ADT UNI[14] .
Keywords: deep learning, attendance automation, face recognition, face detection.