This study investigates the integration of transfer learning methods with fingerprint recognition technologies to enhance student attendance systems, providing a comparative analysis between two pre-trained models: ResNet50 and MobileNetV2. Given the critical importance of accurately and efficiently identifying students in educational settings, adopting advanced neural network architectures offers a promising avenue for improving the functionality and design of attendance systems. Our methodology involves fine-tuning these models on specific datasets of fingerprint images and assessing their performance using key metrics: accuracy, precision, recall, F1 score, and the Matthews Correlation Coefficient (MCC). The comparative analysis revealed that the ResNet50 model outperformed MobileNetV2 with an impressive 100% accuracy rate, while MobileNetV2 also demonstrated significant effectiveness, achieving an accuracy of 99.66%. This study underscores the potential of employing advanced neural networks through transfer learning to bolster educational attendance systems.