Face Recognition Based Attendance System

Attendance recording is a routine academic task, yet manual roll calls, paper registers and card-based methods are time-consuming, error-prone and open to proxy attendance. This paper presents a Face Recognition Based Attendance System that automates attendance using facial biometrics on low-cost, locally hosted infrastructure. The system is built with Python and Flask, uses the face_recognition library, based on a pretrained deep-learning face-embedding model from dlib, for 128-dimensional face encodings, OpenCV for camera processing and SQLite for local storage. It provides two workflows: administrator-controlled enrollment, in which front, left and right facial references are stored for each student and duplicate identities are rejected, and continuous webcam-based verification, in which each detected face is compared with the enrolled encodings using Euclidean distance. A face within a predefined threshold is marked present once per day, while an unmatched face is labelled unrecognized rather than assigned to a student. An administrator dashboard supports viewing, manual correction and reporting. The prototype is aimed at classroom entrances and small institutions. Dedicated liveness detection and large-scale quantitative evaluation are not yet included and are identified as future work.