Fingerprint Identification from Digital Images Using Deep Learning

Authentication methods, particularly those based on biometrics, are becoming increasingly popular due to their superior security, cost-effectiveness, and user-friendliness compared to conventional methods. The spread of contagious illness has led to the urgent need for contactless fingerprint verification in various sectors. However, challenges arise in fingerprint categorization in touchless systems due to factors like image clarity, background interference, besides external conditions. This research introduces an initial work on a deep learning system for touchless fingerprint identification, utilizing a comprehensive dataset of 2,143 images from 175 participants. The task is presented as a classification problem. Our proposed solution combines preprocessing strategies with deep and transfer learning models, incorporating various preprocessing methods to boost the classification efficacy in touchless fingerprint identification. Based on our tests, the InceptionResnet model emerged as the top performer, registering an accuracy rate of 89%.