Advanced Fingerprint Recognition via Data Augmentation and Deep Capsule Networks

Authors- Mohanasudar V, Associate Professor Dr. C. Meenakshi

Abstract-Fingerprint recognition is an important application in biometric authentication systems that offers secure and robust identity verification. Variability challenges in fingerprint quality, environment, and noise, however, impact recognition performance. This project therefore proposes an advanced fingerprint recognition system based on deep learning models combined with conventional image augmentation methods. The system promotes improved accuracy and resilience of fingerprint classification through capsule networks with the addition of pre-trained convolution models like ResNet50, VGG16, and EfficientNetB0. The solution mitigates a modular approach with three major elements including data augmentation in creating diverse training samples, training of the model with capsule networks for enhanced learning of spatial features, and prediction using an interactive user interface to obtain real-time results. The structure offers better generalization, reduced overfitting, and greater user interaction and thus constitutes a sound solution to secure, scalable fingerprint verification under diverse real-world applications.

DOI: /10.61463/ijset.vol.13.issue3.139