In this study, we propose an enhanced student attendance system leveraging palmprint biometrics combined with advanced deep learning models, specifically EfficientNet and Vision Transformer (ViT). Traditional attendance methods, such as manual roll calls, are prone to errors and inefficiencies. By integrating palmprint recognition, known for its high accuracy and non-intrusive nature, with EfficientNet optimized performance and Vision Transformer self-attention mechanisms, we aim to improve the reliability and scalability of attendance systems. Data augmentation techniques are applied to enhance model robustness against variations in palm orientations and lighting conditions. Our system is evaluated using the CASIA Palmprint Image Database, where EfficientNet outperforms ViT B32 across key metrics, achieving a high accuracy of 97.88%. These results demonstrate the system's effectiveness for real-time applications in educational settings, offering a scalable and accurate solution for tracking student attendance.