Face Recognition-Based Smart Attendance System Using Raspberry Pi and InsightFace

– Conventional attendance recording in academic settings depends on repetitive, error-prone manual procedures that invite proxy marking and consume valuable instructional time. This paper presents a contactless, AI-driven attendance framework deployed on a Raspberry Pi 4B edge device. A USB webcam streams live classroom video; each frame is analysed by the InsightFace ArcFace model, which generates a compact facial embedding matched against a MongoDB Atlas database using cosine similarity. Verified identities trigger automated attendance entries through a Node.js and Express RESTful API, while a React.js front-end delivers role-based dashboards for administrators, faculty, and students. Evaluation under varied indoor lighting conditions yielded recognition accuracy exceeding 95% with an average end-to-end latency of 1.1 seconds. Total hardware cost per terminal is approximately INR 6,500, representing a 75 to 90 percent reduction compared to commercial biometric systems. The proposed system eliminates fraudulent proxy registration, reduces administrative overhead, and establishes a scalable IoT platform for broader institutional digitisation.