Comparative Analysis of Fingerprint Authentication Using the Convolutional Neural Network (CNN) Method for Two Datasets

Fingerprint authentication is a vital component of biometric identification systems due to its inherent uniqueness, immutability, and ease of use. This study presents a comparative analysis of fingerprint authentication methods utilizing Convolutional Neural Networks (CNNs) across two datasets. Six CNN architectural models—Inception V3, Inception ResNet V2, MobileNet, ResNet-50, ResNet-152, and VGG16—were employed to evaluate their performance in classifying fingerprint patterns into four categories: arch, loop, double loop, and whorl. The datasets were processed through a robust training and validation pipeline involving 20 epochs, each comprising 10 iterations. The experimental process focused on identifying the most accurate architecture for fingerprint classification, considering factors such as processing speed, precision, and generalization capability. Results demonstrate notable variations in the performance of the models, highlighting the strengths and limitations of each architecture in fingerprint pattern recognition. This research contributes to the development of more effective and efficient fingerprint authentication systems, emphasizing the importance of model selection in enhancing system accuracy and reliability. By providing insights into the comparative performance of state-of-the-art CNN architectures, this study aims to support advancements in biometric technology and its applications in secure identification systems. 

Keywords: Finger Print Dataset, Image Processing Techniques, Deep Learning, classification and Accuracy.