Classification Model for Multi-Classes Iris Image Using Deep Learning Neural Networks
F.O Aranuwa, O.B Fawehinmi · International Journal of Darshan Institute on Engineering Research and Emerging Technologies · 2023
Iris image has been adjudicated one of the most reliable biometric traits for authentication as a result of its unique patterns, stringency to spoof attack and reliability.However, studies have revealed that inefficient classification sorts and techniques leading to classification errors and inaccurate matching has characterized its processes in many authentication and identification applications.Additionally, most of the existing works in the domain are based on single image classification, and major drawbacks of these methodologies remains insufficient learning input and classification errors.The current work uses Deep learning neural networks (DLNN), a subset of artificial intelligence that are effective in learning complex features from data, such as images.Majorly, the work is focused at classifying both left and right human iris images.Data for the work was acquired from the CASIA-Iris-lamp dataset (http://biometrics.idealtest.org).The dataset contains 16,163 iris datasets, and 20 iterations were passed on the data during modeling to determine the accuracy of the model.Performance metrics such as sensitivity, specificity and accuracy were used to evaluate the performance of the model.Experimental results show that the model performed well with classification accuracy of 99.57% which is relatively an improvement over the existing model that was used as a benchmark with 93.35%.The model correctly classified and predicted all images belonging to the right category as right irises, while it wrongly predicted 14 images belonging to the left category.Connotionally, the right iris recorded higher accuracy compared to that of the left iris.