Iris Recognition: A Study of Various Pretrained Networks Approach

Abstract

A recent development in deep learning techniques. Iris biometrics is the superior and most refined of all biometrics systems. Iris trait is declared as the most powerful tool for human identification because of the strong and unique textual information. This paper concentrate on the importance of the iris feature, which is one of the unique biometric modality. This unique feature of the iris pattern can be extracted effectively using various pre-trained CNN models and analyzed its performance in this work. In iris recognition, the important step is to extract the most prominent features of the inputted iris image. Extracted features of the input iris image learned from the pre-trained network such as VGG16, Inception, ResNet, etc. are evaluated. The input image is initially localized, normalized, and then enhanced, then it will be classified into N classes. The performance of the proposed system is tested on two public datasets viz IITD and CASIA. A high accuracy rate is obtained with this proposed system.