Informative Multimodal Unsupervised Image-to-Image Translation

<!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>Informative Multimodal Unsupervised Image-to-Image Translation</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="Informative Multimodal Unsupervised Image-to-Image Translation"> <meta name="description" content="We propose a new method of multimodal image translation, called InfoMUNIT, which is an extension of the state-of-the-art method MUNIT. Our method allows controlling the style of the generated images and improves their quality and diversity. It learns to maximize the mutual information between a subset of style code and the distribution of the output images. Experiments show that our model cannot only translate one image from the source domain to multiple images in the target domain but also explore and manipulate features of the outputs without annotation. Furthermore, it achieves a superior diversity and a competitive image quality to state-of-the-art methods in multiple image translation tasks"/> <meta name="keywords" content="Multimodal Image-to-Image Translation, Mutual Information, GANs, Manipulating Features, Disentangled Representation."/>   <meta name="dc.title" content="Informative Multimodal Unsupervised Image-to-Image Translation"> <meta name="citation_author" content="Tien Tai Doan"> <meta name="citation_author" content="Guillaume Ghyselinck"> <meta name="citation_author" content="Blaise Hanczar"> <meta name="dc.type" content="Article"> <meta name="dc.source" content="Computer Science & Information Technology (CS & IT)"> <meta name="dc.date" content="2021/04/24"> <meta name="dc.identifier" content="10.5121/csit.2021.110503"> <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="title" content="Informative Multimodal Unsupervised Image-to-Image Translation"> <meta name="description" content="We propose a new method of multimodal image translation, called InfoMUNIT, which is an extension of the state-of-the-art method MUNIT. Our method allows controlling the style of the generated images and improves their quality and diversity. It learns to maximize the mutual information between a subset of style code and the distribution of the output images. Experiments show that our model cannot only translate one image from the source domain to multiple images in the target domain but also explore and manipulate features of the outputs without annotation. Furthermore, it achieves a superior diversity and a competitive image quality to state-of-the-art methods in multiple image translation tasks."/> <meta name="dc.subject" content="Multimodal Image-to-Image Translation"> <meta name="dc.subject" content="Mutual Information"> <meta name="dc.subject" content="GANs"> <meta name="dc.subject" content="Manipulating Features"> <meta name="dc.subject" content="Disentangled Representation">   <meta name="prism.publicationName" content="Computer Science & Information Technology (CS & IT)"> <meta name="prism.publicationDate" content="2021/04/24"> <meta name="prism.volume" content="11"> <meta name="prism.number" content="5"> <meta name="prism.section" content="Article"> <meta name="prism.startingPage" content="37">   <meta name="citation_journal_title" content="Computer Science & Information Technology (CS & IT)"> <meta name="citation_publisher" content="AIRCC Publishing Corporation"> <meta name="citation_author" content="Tien Tai Doan, Guillaume Ghyselinck and Blaise Hanczar"> <meta name="citation_title" content="Informative Multimodal Unsupervised Image-to-Image Translation"> <meta name="citation_online_date" content="2015/04/24"> <meta name="citation_volume" content="11"> <meta name="citation_issue" content="5"> <meta name="citation_firstpage" content="37"> <meta name="citation_author" content="Tien Tai Doan"> <meta name="citation_author" content="Guillaume Ghyselinck"> <meta name="citation_author" content="Blaise Hanczar"> <meta name="citation_doi" content="10.5121/csit.2021.110503"> <meta name="citation_abstract_html_url" content="https://aircconline.com/csit/abstract/v11n5/csit110503.html"> <meta name="citation_pdf_url" content="https://aircconline.com/csit/papers/vol11/csit110503.pdf">   <meta property="og:site_name" content="AIRCC" /> <meta property="og:type" content="article" /> <meta property="og:url" content="https://aircconline.com/csit/abstract/v11n5/csit110503.html"/> <meta property="og:title" content="Informative Multimodal Unsupervised Image-to-Image Translation"/> <meta property="og:description" content="We propose a new method of multimodal image translation, called InfoMUNIT, which is an extension of the state-of-the-art method MUNIT. Our method allows controlling the style of the generated images and improves their quality and diversity. It learns to maximize the mutual information between a subset of style code and the distribution of the output images. Experiments show that our model cannot only translate one image from the source domain to multiple images in the target domain but also explore and manipulate features of the outputs without annotation. Furthermore, it achieves a superior diversity and a competitive image quality to state-of-the-art methods in multiple image translation tasks."/>   <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 05, April 2021</font></h2> <h4 style="text-align:center;height:auto;"><a>Informative Multimodal Unsupervised Image-to-Image Translation</a></h4> <h3>  Authors</h3> <p class="#left right" style="text-align:">Tien Tai Doan<sup>1,2</sup>, Guillaume Ghyselinck<sup>1</sup> and Blaise Hanczar<sup>2</sup>, <sup>1</sup>Dental Monitoring, France, <sup>2</sup>University of Evry Val d鈥橢ssonne, France</p> <h3>  Abstract</h3> <p class="#left right" style="text-align:justify">We propose a new method of multimodal image translation, called InfoMUNIT, which is an extension of the state-of-the-art method MUNIT. Our method allows controlling the style of the generated images and improves their quality and diversity. It learns to maximize the mutual information between a subset of style code and the distribution of the output images. Experiments show that our model cannot only translate one image from the source domain to multiple images in the target domain but also explore and manipulate features of the outputs without annotation. Furthermore, it achieves a superior diversity and a competitive image quality to state-of-the-art methods in multiple image translation tasks. </p> <h3>  Keywords</h3> <p class="#left right" style="text-align:justify">Multimodal Image-to-Image Translation, Mutual Information, GANs, Manipulating Features, Disentangled Representation.</p><br> <button type="button" id="button"><a target="blank" href="/csit/papers/vol11/csit110503.pdf">Full Text</a></button>   <button type="button" id="button"><a href="http://airccse.org/csit/V11N05.html">Volume 11, Number 05</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

Informative Multimodal Unsupervised Image-to-Image Translation