The escalating demand for efficient and secure attendance management systems has driven innovation beyond traditional manual methods, which are inherently prone to inaccuracies, inefficiency, and fraud. Artificial intelligence (AI)-driven facial recognition emerges as a transformative solution, offering real-time, non-intrusive, and highly accurate identification. This review paper synthesizes the current state of research in real-time face detection and recognition, particularly for attendance systems, across various hardware platforms and algorithmic approaches. It critically examines existing review papers and individual studies to identify significant research gaps in achieving comprehensive, energy-efficient, and robust solutions for crowded environments. Specifically, this paper highlights the lack of holistic hybrid hardware-software co-optimization, integrated intelligent low-power management with hybrid wake-up mechanisms, and seamless real-time database integration for high-throughput, multi-face data in existing literature. We then position a novel hybrid Field-Programmable Gate Array (FPGA) and NVIDIA Jetson AGX architecture as a promising solution to these identified gaps. This proposed system leverages the FPGA for low-latency image pre-processing and motion-triggered power management, and the Jetson AGX for high-performance deep learning inference and real-time database updates, thereby addressing the complex challenges of real-time attendance in crowded, power-constrained environments.