Iris Recognition using Convolutional Neural Network

The design of a pragmatic user authentication system is vital to provide accurate detection of personal identity. Iris recognition as a form of biometric identification technology has been actively researched for decades and is gaining wider popularity considering the increasing awareness of personal privacy. The rise of artificial intelligence provides a great opportunity to further elevate the penetration of iris recognition in safeguarding one's private data. Convolutional neural network is a practical algorithm that is highly suitable for image processing and pattern recognition, its effectiveness and flexibility have seen it being applied in many fields. This study focuses on the development of an iris recognition system based on convolutional neural network with high precision and efficiency. A total of iris samples from 20 individuals with both sides of the eyes included are used to train the deep recognition system. The model shows an early sign of underfitting and little convergence with inadequate number of training epoch. However, as the training epochs are increased, the trained model managed to achieve a testing accuracy of 99%.