Enhancing Student Attendance System through Fingerprint Recognition Using Transfer Learning Techniques
Slimane Ennajar, Walid Bouarifi · 2024
Biometric recognition systems have revolutionized student attendance tracking in educational institutions, offering a reliable alternative to traditional methods. This paper explores the integration of fingerprint recognition technology with transfer learning techniques to enhance attendance systems. Specifically, it employs two pre-trained convolutional neural networks, InceptionV3 and NASNetMobile, for fingerprint recognition. The models are evaluated based on accuracy, precision, recall, F1 Score, and Matthews Correlation Coefficient. InceptionV3 demonstrates exceptional performance with an accuracy score of 99.86%, indicating near-perfect classification capabilities, while NASNetMobile exhibits competent performance but lags behind InceptionV3. Analysis of loss, accuracy trends, and Receiver Operating Characteristic curves further elucidates model behaviors. The study underscores the potential of deep learning models, particularly InceptionV3, in bolstering precision-critical applications like fingerprint identification in student attendance systems. Future research avenues include real-time system integration, multimodal biometric fusion, and scalability assessment in diverse environments.