A Face Recognition Based Classroom Attendance

Traditional attendance management in educational institutions suffers from critical inefficiencies, including time consuming manual roll calls, vulnerability to proxy attendance (buddy punching), transcription errors, and difficulty in generating real-time analytics. This paper presents a Smart Attendance System that leverages face recognition technology to automate student attendance management with high accuracy and zero internet dependency. The proposed system employs OpenCV 4.8.0 for real-time face detection using Haar Cascade classifiers, the face_recognition library (backed by a dlib-based deep residual network) for generating 128-dimensional facial encodings, and Python 3.10 as the core development platform. Students are registered by capturing 30 facial images, from which unique mathematical feature vectors are extracted and stored in a local database. During attendance sessions, live camera feeds are analyzed frame-by frame; detected faces are compared against stored encodings using Euclidean distance with a threshold of 0.50, and verified identities are automatically logged with precise timestamps in CSV and Excel formats. The system has been tested with over 500 students, achieving 95%+ recognition accuracy in controlled environments at a processing speed of 10-20 frames per second, with a response time under one second per recognition event. The system operates entirely offline, ensuring institutional data privacy and compliance. Results demonstrate significant reduction in administrative overhead, complete elimination of proxy attendance, and reliable automated reporting, establishing this as a practical, scalable, and cost effective solution for modern educational institutions.