GAN-Based Data Augmentation and Anonymization for Mask Classification

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml"> <head> <title>GAN-Based Data Augmentation and Anonymization for Mask Classification</title>  <meta charset="UTF-8"> <meta name="viewport" content="width=device-width, initial-scale=1.0"> <meta http-equiv="X-UA-Compatible" content="ie=edge"> <meta name="title" content="GAN-Based Data Augmentation and Anonymization for Mask Classification"> <meta name="description" content="Deep learning methods, especially convolutional neural networks (CNNs), have made a major contribution to computer vision. However, deep learning classifiers need large-scale annotated datasets to be trained without over-fitting. Also, in high-data diversity, trained models generalize better. However, collecting such a large-scale dataset remains challenging. Furthermore, it is invaluable for researchers to protect the subjects' confidentiality when using their personal data such as face images. In this paper, we propose a deep learning Generative Adversarial Networks (GANs) which generates synthetic samples for our mask classification model. Our contributions in this work are two-fold that the synthetics images provide. First, GANs' models can be used as an anonymization tool when the subjects' confidentiality is matters. Second, the generated masked/unmasked face images boost the performance of the mask classification model by using the synthetic images as a form of data augmentation. In our work, the classification accuracy using only traditional data augmentations is 93.71 %. By using both synthetic data and original data with traditional data augmentations the result is 95.50 %. It is shown that the GAN-generated synthetic data boosts the performance of deep learning classifiers"/> <meta name="keywords" content="Convolutional Neural Network, Data Anonymization, Data Augmentation, Generative Adversarial Network, Mask classification"/>   <meta name="dc.title" content="GAN-Based Data Augmentation and Anonymization for Mask Classification "> <meta name="citation_authors" content="Mustafa 脟elik"> <meta name="citation_authors" content="Ahmet Haydar脰rnek"> <meta name="dc.type" content="Article"> <meta name="dc.source" content="Computer Science & Information Technology (CS & IT) Vol. 11, No.23"> <meta name="dc.date" content="2021/12/21"> <meta name="dc.identifier" content="https://doi.org/10.5121/csit.2021.112315"> <meta name="dc.publisher" content="AIRCC Publishing Corporation"> <meta name="dc.rights" content="http://creativecommons.org/licenses/by/3.0/"> <meta name="dc.format" content="application/pdf"> <meta name="dc.language" content="en"> <meta name="dc.description" content="Deep learning methods, especially convolutional neural networks (CNNs), have made a major contribution to computer vision. However, deep learning classifiers need large-scale annotated datasets to be trained without over-fitting. Also, in high-data diversity, trained models generalize better. However, collecting such a large-scale dataset remains challenging. Furthermore, it is invaluable for researchers to protect the subjects' confidentiality when using their personal data such as face images. In this paper, we propose a deep learning Generative Adversarial Networks (GANs) which generates synthetic samples for our mask classification model. Our contributions in this work are two-fold that the synthetics images provide. First, GANs' models can be used as an anonymization tool when the subjects' confidentiality is matters. Second, the generated masked/unmasked face images boost the performance of the mask classification model by using the synthetic images as a form of data augmentation. In our work, the classification accuracy using only traditional data augmentations is 93.71 %. By using both synthetic data and original data with traditional data augmentations the result is 95.50 %. It is shown that the GAN-generated synthetic data boosts the performance of deep learning classifiers."> <meta name="dc.subject" content="Convolutional Neural Network"> <meta name="dc.subject" content="Data Anonymization"> <meta name="dc.subject" content="Data Augmentation"> <meta name="dc.subject" content="Generative Adversarial Network"> <meta name="dc.subject" content="Mask classification">   <meta name="prism.publicationName" content="Computer Science & Information Technology (CS & IT)"> <meta name="prism.publicationDate" content="2021/12/21"> <meta name="prism.volume" content="11"> <meta name="prism.number" content="23"> <meta name="prism.section" content="Article"> <meta name="prism.startingPage" content="191">   <meta name="citation_journal_title" content="Computer Science & Information Technology (CS & IT)"> <meta name="citation_publisher" content="AIRCC Publishing Corporation"> <meta name="citation_authors" content="Mustafa 脟elik,Ahmet Haydar脰rnek"> <meta name="citation_title" content="GAN-Based Data Augmentation and Anonymization for Mask Classification "> <meta name="citation_online_date" content="2021/12/21"> <meta name="citation_issue" content="23"> <meta name="citation_firstpage" content="191"> <meta name="citation_authors" content="Mustafa 脟elik"> <meta name="citation_authors" content="Ahmet Haydar脰rnek"> <meta name="citation_doi" content="https://doi.org/10.5121/csit.2021.112315"> <meta name="citation_abstract_html_url" content="https://aircconline.com/csit/abstract/v11n23/csit112315.html"> <meta name="citation_pdf_url" content="https://aircconline.com/csit/papers/vol11/csit112315.pdf">   <meta property="og:site_name" content="AIRCC" /> <meta property="og:type" content="article" /> <meta property="og:url" content="https://aircconline.com/csit/papers/vol11/csit112315.pdf"> <meta property="og:title" content="GAN-Based Data Augmentation and Anonymization for Mask Classification"> <meta property="og:description" content="Deep learning methods, especially convolutional neural networks (CNNs), have made a major contribution to computer vision. However, deep learning classifiers need large-scale annotated datasets to be trained without over-fitting. Also, in high-data diversity, trained models generalize better. However, collecting such a large-scale dataset remains challenging. Furthermore, it is invaluable for researchers to protect the subjects' confidentiality when using their personal data such as face images. In this paper, we propose a deep learning Generative Adversarial Networks (GANs) which generates synthetic samples for our mask classification model. Our contributions in this work are two-fold that the synthetics images provide. First, GANs' models can be used as an anonymization tool when the subjects' confidentiality is matters. Second, the generated masked/unmasked face images boost the performance of the mask classification model by using the synthetic images as a form of data augmentation. In our work, the classification accuracy using only traditional data augmentations is 93.71 %. By using both synthetic data and original data with traditional data augmentations the result is 95.50 %. It is shown that the GAN-generated synthetic data boosts the performance of deep learning classifiers."/>   <meta name="google-site-verification" content="t8rHIcM8EfjIqfQzQ0IdYIiA9JxDD0uUZAitBCzsOIw" /> <meta name="yandex-verification" content="e3d2d5a32c7241f4" />  <link rel="icon" type="image/ico" href="../img/ico.ico"/> <link rel="stylesheet" type="text/css" href="../main1.css" media="screen" /> <style type="text/css"> a{ color:white; text-decoration:none; line-height:20px; } ul li a{ font-weight:bold; color:#000; list-style:none; text-decoration:none; size:10px;} .imagess { height:90px; text-align:left; margin:0px 5px 2px 8px; float:right; border:none; } #left p { font-family: CALIBRI; font-size: 16px; margin-left: 20px; font-weight: 500; } .right { margin-right: 20px; } #button{ float: left; font-size: 14px; margin-left: 10px; height: 28px; width: auto; background-color: #1e86c6; } </style> </head> <body> <div class="font"> <div id="wap"> <div id="page"> <div id="top"> <form action="https://airccj.org/csecfp/library/Search.php" method="get" target="_blank" > <table width="100%" cellspacing="0" cellpadding="0" > <tr class="search_input"> <td width="665" align="right"> </td> <td width="236" > <input name="title" type="text" value="Enter the paper title" class="search_textbox" onclick="if(this.value=='Enter the paper title'){this.value=''}" onblur="if(this.value==''){this.value='Enter the paper title'}" /> </td> <td width="59"> <input type="image" src="../img/go.gif" /> </td> </tr> <tr> <td colspan="3" valign="top"><img src="../img/top1.gif" alt="Academy & Industry Research Collaboration Center (AIRCC)" /></td> </tr> </table> </form> </div> <div id="font-face"> <div id="menu"> <a href="http://airccse.org">Home</a> <a href="http://airccse.org/journal.html">Journals</a> <a href="http://airccse.org/ethics.html">Ethics</a> <a href="http://airccse.org/conference.html">Conferences</a> <a href="http://airccse.org/past.html">Past Events</a> <a href="http://airccse.org/b.html">Submission</a> </div> <div id="content"> <div id="left"> <h2 class="lighter"><font size="2">Volume 11, Number 23, December 2021</font></h2> <h4 style="text-align:center;height:auto;"><a>GAN-Based Data Augmentation and Anonymization for Mask Classification</a></h4> <h3>  Authors</h3> <p class="#left right" style="text-align:">Mustafa 脟elik<sup>1, 2</sup>, Ahmet Haydar脰rnek<sup>1, 3</sup>, <sup>1</sup>Huawei Turkey R&D Center, Turkey, <sup>2</sup>Istanbul Technical University, Turkey, <sup>3</sup>Konya Technical University, Turkey</p> <h3>  Abstract</h3> <p class="#left right" style="text-align:justify">Deep learning methods, especially convolutional neural networks (CNNs), have made a major contribution to computer vision. However, deep learning classifiers need large-scale annotated datasets to be trained without over-fitting. Also, in high-data diversity, trained models generalize better. However, collecting such a large-scale dataset remains challenging. Furthermore, it is invaluable for researchers to protect the subjects' confidentiality when using their personal data such as face images. In this paper, we propose a deep learning Generative Adversarial Networks (GANs) which generates synthetic samples for our mask classification model. Our contributions in this work are two-fold that the synthetics images provide. First, GANs' models can be used as an anonymization tool when the subjects' confidentiality is matters. Second, the generated masked/unmasked face images boost the performance of the mask classification model by using the synthetic images as a form of data augmentation. In our work, the classification accuracy using only traditional data augmentations is 93.71 %. By using both synthetic data and original data with traditional data augmentations the result is 95.50 %. It is shown that the GAN-generated synthetic data boosts the performance of deep learning classifiers. </p> <h3>  Keywords</h3> <p class="#left right" style="text-align:justify">Convolutional Neural Network, Data Anonymization, Data Augmentation, Generative Adversarial Network, Mask classification.</p><br> <button type="button" id="button"><a target="blank" href="/csit/papers/vol11/csit112315.pdf">Full Text</a></button>   <button type="button" id="button"><a href="http://airccse.org/csit/V11N23.html">Volume 11, Number 23</a></button> <br><br><br><br><br> </div> <div id="right"> <div class="menu_right"> <ul> <li id="id"><a href="http://airccse.org/editorial.html">Editorial Board</a></li> <li><a href="http://airccse.org/arch.html">Archives</a></li> <li><a href="http://airccse.org/indexing.html">Indexing</a></li> <li><a href="http://airccse.org/faq.html" target="_blank">FAQ</a></li> </ul> </div> <div class="clear_left"></div> <br> </div> <div class="clear"></div> <div id="footer"> <table width="100%" > <tr> <td width="46%" class="F_menu"><a href="http://airccse.org/subscription.html">Subscription</a> <a href="http://airccse.org/membership.html">Membership</a> <a href="http://airccse.org/cscp.html">AIRCC CSCP</a> <a href="http://airccse.org/acontact.html">Contact Us</a> </td> <td width="54%" align="right"><a href="http://airccse.org/index.php"><img src="/csit/abstract/img/logo.gif" alt="" width="21" height="24" /></a><a href="http://www.facebook.com/AIRCCSE"><img src="/csit/abstract/img/facebook.jpeg" alt="" width="21" height="24" /></a><a href="https://twitter.com/AIRCCFP"><img src="/csit/abstract/img/twitter.jpeg" alt="" width="21" height="24" /></a><a href="http://cfptech.wordpress.com/"><img src="/csit/abstract/img/index1.jpeg" alt="" width="21" height="24" /></a></td> </tr> <tr><td height="25" colspan="2"> <p align="center">All Rights Reserved ® AIRCC</p> </td></tr> </table> </div> </div> </div> </div> </div> </div> </body> </html>

CINXE.COM

GAN-Based Data Augmentation and Anonymization for Mask Classification