The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
O recente desenvolvimento de modelos generativos baseados em aprendizagem profunda intensificou drasticamente o interesse na síntese de dados e suas aplicações. A síntese de dados assume uma importância adicional especialmente para algumas tarefas de reconhecimento de padrões nas quais algumas classes de dados são raras e difíceis de coletar. Em um conjunto de dados de íris, por exemplo, as amostras de classes minoritárias incluem imagens de olhos com óculos, pupilas superdimensionadas ou subdimensionadas, localizações desalinhadas da íris e íris obstruída ou contaminada por pálpebras, cílios ou reflexos de iluminação. Esses conjuntos de dados desequilibrados de classe geralmente resultam em desempenho de classificação tendencioso. As redes generativas adversárias (GANs) são uma das estruturas mais promissoras que aprendem a gerar dados sintéticos por meio de um jogo minimax para dois jogadores entre um gerador e um discriminador. Neste artigo, utilizamos a rede adversária geradora condicional Wasserstein de última geração com penalidade de gradiente (CWGAN-GP) para gerar a classe minoritária de imagens de íris, o que economiza enorme quantidade de custos de trabalho humano para coleta de dados raros. Com nosso modelo, o pesquisador pode gerar quantas imagens de íris de casos raros desejar e isso ajuda a desenvolver qualquer algoritmo de aprendizado profundo sempre que for necessário um grande tamanho de conjunto de dados.
Yung-Hui LI
National Central University,Hon Hai Research Institute
Muhammad Saqlain ASLAM
National Central University
Latifa Nabila HARFIYA
National Central University
Ching-Chun CHANG
University of Warwick
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Yung-Hui LI, Muhammad Saqlain ASLAM, Latifa Nabila HARFIYA, Ching-Chun CHANG, "Conditional Wasserstein Generative Adversarial Networks for Rebalancing Iris Image Datasets" in IEICE TRANSACTIONS on Information,
vol. E104-D, no. 9, pp. 1450-1458, September 2021, doi: 10.1587/transinf.2021EDP7079.
Abstract: The recent development of deep learning-based generative models has sharply intensified the interest in data synthesis and its applications. Data synthesis takes on an added importance especially for some pattern recognition tasks in which some classes of data are rare and difficult to collect. In an iris dataset, for instance, the minority class samples include images of eyes with glasses, oversized or undersized pupils, misaligned iris locations, and iris occluded or contaminated by eyelids, eyelashes, or lighting reflections. Such class-imbalanced datasets often result in biased classification performance. Generative adversarial networks (GANs) are one of the most promising frameworks that learn to generate synthetic data through a two-player minimax game between a generator and a discriminator. In this paper, we utilized the state-of-the-art conditional Wasserstein generative adversarial network with gradient penalty (CWGAN-GP) for generating the minority class of iris images which saves huge amount of cost of human labors for rare data collection. With our model, the researcher can generate as many iris images of rare cases as they want and it helps to develop any deep learning algorithm whenever large size of dataset is needed.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2021EDP7079/_p
Copiar
@ARTICLE{e104-d_9_1450,
author={Yung-Hui LI, Muhammad Saqlain ASLAM, Latifa Nabila HARFIYA, Ching-Chun CHANG, },
journal={IEICE TRANSACTIONS on Information},
title={Conditional Wasserstein Generative Adversarial Networks for Rebalancing Iris Image Datasets},
year={2021},
volume={E104-D},
number={9},
pages={1450-1458},
abstract={The recent development of deep learning-based generative models has sharply intensified the interest in data synthesis and its applications. Data synthesis takes on an added importance especially for some pattern recognition tasks in which some classes of data are rare and difficult to collect. In an iris dataset, for instance, the minority class samples include images of eyes with glasses, oversized or undersized pupils, misaligned iris locations, and iris occluded or contaminated by eyelids, eyelashes, or lighting reflections. Such class-imbalanced datasets often result in biased classification performance. Generative adversarial networks (GANs) are one of the most promising frameworks that learn to generate synthetic data through a two-player minimax game between a generator and a discriminator. In this paper, we utilized the state-of-the-art conditional Wasserstein generative adversarial network with gradient penalty (CWGAN-GP) for generating the minority class of iris images which saves huge amount of cost of human labors for rare data collection. With our model, the researcher can generate as many iris images of rare cases as they want and it helps to develop any deep learning algorithm whenever large size of dataset is needed.},
keywords={},
doi={10.1587/transinf.2021EDP7079},
ISSN={1745-1361},
month={September},}
Copiar
TY - JOUR
TI - Conditional Wasserstein Generative Adversarial Networks for Rebalancing Iris Image Datasets
T2 - IEICE TRANSACTIONS on Information
SP - 1450
EP - 1458
AU - Yung-Hui LI
AU - Muhammad Saqlain ASLAM
AU - Latifa Nabila HARFIYA
AU - Ching-Chun CHANG
PY - 2021
DO - 10.1587/transinf.2021EDP7079
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E104-D
IS - 9
JA - IEICE TRANSACTIONS on Information
Y1 - September 2021
AB - The recent development of deep learning-based generative models has sharply intensified the interest in data synthesis and its applications. Data synthesis takes on an added importance especially for some pattern recognition tasks in which some classes of data are rare and difficult to collect. In an iris dataset, for instance, the minority class samples include images of eyes with glasses, oversized or undersized pupils, misaligned iris locations, and iris occluded or contaminated by eyelids, eyelashes, or lighting reflections. Such class-imbalanced datasets often result in biased classification performance. Generative adversarial networks (GANs) are one of the most promising frameworks that learn to generate synthetic data through a two-player minimax game between a generator and a discriminator. In this paper, we utilized the state-of-the-art conditional Wasserstein generative adversarial network with gradient penalty (CWGAN-GP) for generating the minority class of iris images which saves huge amount of cost of human labors for rare data collection. With our model, the researcher can generate as many iris images of rare cases as they want and it helps to develop any deep learning algorithm whenever large size of dataset is needed.
ER -