RESNET-50 FOR FACE AND IRIS-BASED AUTHENTICATION - A COMPARATIVE STUDY USING SVM AND NAIVE BAYESIAN CLASSIFIERS

This study explores a deep learning-based approach to enhance the performance of biometric recognition systems, specifically focusing on authentication using face and iris images in the well known ResNet-50 deep learning model. Traditional feature extraction methods often rely on handcrafted features and shallow models, which struggle with real-world variations such as lighting, pose, and occlusions, leading to limited generalization and accuracy. To address these challenges, the proposed method utilizes the ResNet50 convolutional neural net- work architecture for extracting robust and discriminative features from biometric images. ResNet50’s deep residual connections enable the capture of complex patterns often missed by conventional techniques. The extracted features are then classified using two traditional machine learning algorithms—Support Vector Machines (SVM) and Naive Baye’s (NB) for both face and iris images—chosen for their effectiveness in high-dimensional data classification and complementary strengths in pattern recognition. The findings suggest that the face images perform better com- pared to iris images for authentication in both classifiers, however, the Naive Bayesian classifier is found to outperform SVM in both modalities. The conclusion reached is by combining deep learning models with classical machine learning classifiers, the reliability and scalability of bio- metric identification systems are enhanced. Moreover, a multimodal approach for authentication may yield better performance as each modality adds to the performance of the identification task.