An Edge-Cloud Deep Learning Architecture for Secure Attendance Monitoring in Resource-Constrained Educational Environments
Existing attendance monitoring mechanisms in schools are facing several challenges especially in resource-constrained educational settings. The menace of truancy, absenteeism, proxy attendance and unfettered access to school premises by unauthorized persons pose serious security concerns at a time of spike in global terrorism and banditry activities. These challenges require technology-based proactive solutions to guarantee administrative responsibility and student safety. In this paper, a novel architecture of edge cloud deep learning is proposed which is tailored to address the problem of attendance tracking in real time in Nigerian educational sector. This system leverages Convolutional Neural Network (CNN) for biometric face recognition at the edge of the network, using a React web interface powered by TensorFlow.js. In place of the extremely hefty computational and bandwidth restrictions of the traditional centralized systems, a lightweight NoSQL cloud-based infrastructure (Firebase) is used for state management and real-time synchronization. Empirical evaluations were performed under dynamic real-world conditions with different light and physical occlusion. The model achieved a validation accuracy of 96.72% while on the edge, it had an inference latency of 185ms per frame and a cloud synchronization latency of 112ms per transaction. The system is essentially a transformation of how a basic administrative task is conducted into an active safeguarding tool which safeguards the wellbeing of students in traditionally very important and dynamic academic environments.
Download Paper