Design and Implementation of an Automated Real-Time Facial Recognition Attendance System with Advanced Liveness Detection

This paper describes an automated and non-contact infrastructure that combines deep learning algorithms with computer vision to verify user identity and track attendance in real time. Hardware camera modules are used to capture live video streams and separate the human facial matrix from the rest of the scene for real-time analysis. Biometric data collected from the user is converted into computable feature sets, which are then matched against a secure, local relational database to authenticate identity. A timestamped and validated user record is created in the local database once the confidence level of the target user profile reaches a minimum of 70%. This framework provides a means to reduce human error due to manual record-keeping and also addresses many of the inefficiencies associated with traditional physical time-keeping methods while simultaneously reducing the possibility of submitting false "buddy punching" records. Ultimately, this edge-computed framework provides a secure, reliable and high-volume identity verification method that can be deployed across multiple corporate/educational settings.