Automated Student Attendance using Facial Biometrics

Abstract

In most academic environments, attendance continues to be recorded manually despite the increasing digitalisation of campus activities. These conventional procedures, which rely heavily on roll calls or handwritten signature sheets, tend to interrupt the flow of lessons, introduce errors, and create opportunities for proxy sign-ins. With the rise of computer vision and deep neural networks, facial-recognition technology has emerged as a realistic alternative capable of operating automatically and unobtrusively. This study revisits the progression of facial- recognition algorithms from classical statistical techniques such as PCA, DWT, DCT, Haar cascades and Eigenfaces to modern embedding-driven architectures including MTCNN, FaceNet and ArcFace. A prototype system was developed and deployed in a classroom setting to assess performance in real-world conditions. While the model delivered reliable recognition under stable lighting, it also revealed predictable limitations when faces were partially obscured or when illumination varied sharply. Alongside the technical analysis, this paper examines the privacy, fairness, and consent requirements that accompany the use of biometric data, referencing both India’s Digital Personal Data Protection Act (DPDP) 2023 and international guidelines such as those issued by the UK Information Commissioner’s Office. The findings suggest that automated biometric attendance can be highly effective, provided that institutions employ robust privacy safeguards, transparent governance, and regular audits of accuracy and demographic consistency.