Conditional Generative Adversarial Network- Based Data Augmentation for Enhancement of Iris Recognition Accuracy

Min Beom Lee,Kang Ryoung Park,Yu Hwan Kim

doi:10.1109/access.2019.2937809

Min Beom Lee, Kang Ryoung Park + Show 1 more

Open Access

PDF Available

https://doi.org/10.1109/access.2019.2937809

Copy DOI

Export

Save

Cite

Journal: IEEE Access	Publication Date: Jan 1, 2019
Citations: 49	License type: CC BY 4.0

Affiliation: Dongguk University

Abstract
Highlights/Summary
Full-Text PDF
Similar Papers

Abstract

Listen

Presently, lots of previous studies on biometrics employ convolutional neural networks (CNN) which requires a large amount of labeled training data. However, biometric data are considered as important personal information, and it is difficult to obtain large amounts of data due to individual privacy issues. Training with a small amount of data is a major cause of overfitting and low testing accuracy. To resolve this problem, previous studies have performed data augmentation that are based on geometric transforms and the adjustment of image brightness. Nevertheless, the data created by these methods have high correlation with the original data, and they cannot adequately reflect individual diversities. To resolve this problem, this study proposes iris image augmentation based on a conditional generative adversarial network (cGAN), as well as a method for improving recognition performance that uses this augmentation method. In our method, normalized iris images that are generated through arbitrary changes in the iris and pupil coordinates are used as input in the cGAN-based model to generate iris images. Due to the limitations of the cGAN model, data augmentation, which uses the periocular region, was found to fail with regard to the improvement of performance. Based on this information, only the iris region was used as input for the cGAN model. The augmentation method proposed in this paper was tested using NICE.II training dataset (selected from UBIRS.v2), MICHE database, and CASIA-Iris-Distance database. The results showed that the recognition performance was improved compared to existing studies.

Highlights

Over the last decade, deep learning technology has achieved excellent performance in a variety of fields in computer vision, such as image classification and object detection
This study has proposed a new iris recognition method, where convolutional neural network (CNN) networks are trained with training data generated by the pix2pix GAN model, and recognition is performed
The iris and pupil coordinates were adjusted to normalize the iris images, and these images are entered as a training dataset in the pix2pix GAN, which has a conditional generative adversarial network (cGAN) structure in order to perform the training

Summary

Introduction

Deep learning technology has achieved excellent performance in a variety of fields in computer vision, such as image classification and object detection. Zhang et al proposed dual model learning combined with multiple feature selection for accurate visual tracking by fusing the handcrafted features with the multi-layer features extracted from the convolutional neural network (CNN) [69]. In other research, they proposed the method of spatially attentive visual tracking using multi-model adaptive response fusion [70]. Researchers have used methods that create additional data by applying various geometric transformations to the existing training data. Applying typical geometric transform-based data augmentation methods to small data sets, rather than larger datasets like ImageNet [4], is not sufficient for resolving these issues, because it produces very limited diversities in the existing data [3], [5]

Methods

Results

Conclusion