<![CDATA[In educational institutions, manual outpass verification commonly fails because it is inefficient, relies on people, and is prone to counterfeit. This leads to delays in administration and security breaches. Standard digital substitutes that use barcodes or RFID technology only partially automate the process and don't include biometric validation. To solve these problems, an intelligent web-enabled framework is being built that uses deep learning-based face recognition and connects to the Internet of Things infrastructure. The system uses OptimizedEdgeNet, a new, lightweight convolutional neural network that is developed for embedded edge devices like the Raspberry Pi. A mathematical optimization approach is proposed that takes into account recognition accuracy, decision threshold, computing latency, and energy use. The proposed methodology employs Gaussian distance modelling of facial embeddings to analytically find the ideal matching threshold (τ*) that lowers the total recognition error. Additionally, equations for latency and energy are developed as factors of operation count, processor frequency, and hardware efficiency, creating a multi-objective optimization function that enhances recognition performance while reducing delay and power consumption. The analytical model is confirmed by experimental testing, which shows that the Raspberry Pi 4B can recognize 96.2% of the time with an average delay of 1.64 seconds and an energy use of 2.4 J per inference. The suggested deep learning architecture and mathematical formulation offer a thorough, real-time, and resource-efficient approach for automating secure digital outpasses in regulated settings.]]>
