BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Turkey BEGIN:DAYLIGHT DTSTART:20380119T061407 TZOFFSETFROM:+0300 TZOFFSETTO:+0300 RRULE:FREQ=YEARLY;BYDAY=3TU;BYMONTH=1 TZNAME:+03 END:DAYLIGHT BEGIN:STANDARD DTSTART:20160907T000000 TZOFFSETFROM:+0300 TZOFFSETTO:+0300 RRULE:FREQ=YEARLY;BYDAY=1WE;BYMONTH=9 TZNAME:+03 END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20200222T131838Z UID:C3FCB5EB-CF2D-4B4C-A727-C42B6360A951 DTSTART;TZID=Turkey:20200221T133000 DTEND;TZID=Turkey:20200221T153000 DESCRIPTION:21 February 2020 (13:40): IEEE AP/MTT/EMC/ED Turkey Seminar Ser ies (S.65)\n\nSpeaker: Prof. Pınar Duygulu Şahin\, Hacettepe University\ n\nTopic: "Recognizing and Transferring the Styles of Artists Who Illustra te Children’s Books"\n\nLocation: Middle East Technical University\, Ank ara\, Turkey\n\nAbstract: In this talk\, I will present our recent works t o explore illustrations in children’s books as a new domain in classific ation of artists and unpaired image-to-image translation. Our work is moti vated from a young boy’s capability to recognize an illustrator’s styl e in a totally different context. The boy’s enthusiasm let us to start t he journey to explore the capabilities of machines to recognize the style of illustrators.\n\nFirst\, we collected pages from children’s books to construct a new illustrations dataset consisting of about 9500 pages from 24 artists. We exploited deep networks for categorizing illustrators and w ith around 94% classification performance our method over-performed the tr aditional methods by more than 10%.\n\nGoing beyond categorization we expl ored transferring style. We show that although the current state-of-the-ar t image-to-image translation models successfully transfer either the style or the content\, they fail to transfer both at the same time. We propose a new generator network to address this issue and show that the resulting network strikes a better balance between style and content.\n\nThere are n o well-defined or agreed-upon evaluation metrics for unpaired image-to-ima ge translation. So far\, the success of image translation models has been based on subjective\, qualitative visual comparison on a limited number of images. To address this problem\, we propose a new framework for the quan titative evaluation of image-to-illustration models\, where both content a nd style are taken into account using separate classifiers. In this new ev aluation framework\, our proposed model performs better than the current s tate-of-the-art models on the illustrations dataset.\n\nBio: Pınar Duygul u has received her BSc\, MSc and PhD degrees from Department of Computer E ngineering at Middle East Technical University\, Ankara\, Turkey in 1996\, 1998 and 2003 respectively. During her PhD\, she was a visiting scholar a t University of California at Berkeley under the supervision of Prof. Davi d Forsyth. After being a post-doctoral researcher at Informadia Project at Carnegie Mellon University\, she joined to Department of Computer Enginee ring at Bilkent University\, Ankara\, Turkey in 2004. During 2014 and 2015 she was at Carnegie Mellon University as a research associate. Currently\ , she is a faculty member at Department of Computer Engineering at Hacette pe University\, Ankara\, Turkey. She received Science Academy’s Young Sc ientist Award (BAGEP) in 2015\, Fulbright scholarship in 2013\, TUBITAK Ca reer award in 2005\, and the best paper in Cognitive Vision award at Europ ean Conference on Computer Vision in 2002. Her current research interests include computer vision and multimedia data mining\, specifically object\, face and action recognition in large image and video collections and anal ysis of historical documents.\n\nSpeaker(s): Prof. Pinar Duygulu Sahin\, \ n\nAnkara\, Ankara\, Türkiye LOCATION:Ankara\, Ankara\, Türkiye ORGANIZER:ozergul@metu.edu.tr SEQUENCE:0 SUMMARY:IEEE AP/MTT/EMC/ED TURKEY CHAPTER SEMINAR SERIES -- SEMINAR 65 URL;VALUE=URI:https://events.vtools.ieee.org/m/224527 X-ALT-DESC:Description: <br /><p><strong>21 February 2020 (13:40): &nbsp\;I EEE AP/MTT/EMC/ED Turkey Seminar Series (S.65)</strong></p>\n<p>Speaker: P rof. Pınar Duygulu Şahin\, Hacettepe University</p>\n<p>Topic: "Recogniz ing and Transferring the Styles of Artists Who Illustrate Children&rsquo\; s Books"</p>\n<p>Location:&nbsp\;Middle East Technical University\, Ankara \, Turkey</p>\n<p>Abstract: In this talk\, I will present our recent works to explore illustrations in children&rsquo\;s books as a new domain in cl assification of artists and unpaired image-to-image translation. Our work is motivated from a young boy&rsquo\;s capability to recognize an illustra tor&rsquo\;s style in a totally different context. The boy&rsquo\;s enthus iasm let us to start the journey to explore the capabilities of machines t o recognize the style of illustrators.</p>\n<p>First\, we collected pages from children&rsquo\;s books to construct a new illustrations dataset cons isting of about 9500 pages from 24 artists. We exploited deep networks for categorizing illustrators and with around 94% classification performance our method over-performed the traditional methods by more than 10%.</p>\n< p>Going beyond categorization we explored transferring style.&nbsp\;&nbsp\ ;We show that although the current state-of-the-art image-to-image transla tion models successfully transfer either the style or the content\, they f ail to transfer both at the same time. We propose a new generator network to address this issue and show that the resulting network strikes a better balance between style and content.</p>\n<p>There are no well-defined or a greed-upon evaluation metrics for unpaired image-to-image translation. So far\, the success of image translation models has been based on subjective \, qualitative visual comparison on a limited number of images. To address this problem\, we propose a new framework for the quantitative evaluation of image-to-illustration models\, where both content and style are taken into account using separate classifiers. In this new evaluation framework\ , our proposed model performs better than the current state-of-the-art mod els on the illustrations dataset.</p>\n<p>Bio: Pınar Duygulu has received her BSc\, MSc and PhD degrees from Department of Computer Engineering at Middle East Technical University\, Ankara\, Turkey in 1996\, 1998 and 2003 respectively. During her PhD\, she was a visiting scholar at University o f California at Berkeley under the supervision of Prof. David Forsyth. Aft er being a post-doctoral researcher at Informadia Project at Carnegie Mell on University\, she joined to Department of Computer Engineering at Bilken t University\, Ankara\, Turkey in 2004. During 2014 and 2015 she was at Ca rnegie Mellon University as a research associate. Currently\, she is a fac ulty member at Department of Computer Engineering at Hacettepe University\ , Ankara\, Turkey. She received Science Academy&rsquo\;s Young Scientist A ward (BAGEP) in 2015\, Fulbright scholarship in 2013\, TUBITAK Career awar d in 2005\, and the best paper in Cognitive Vision award at European Confe rence on Computer Vision in 2002. Her current research interests include c omputer vision and multimedia data mining\, specifically object\, face and action recognition in large image and video collections and analysis of h istorical documents.</p> END:VEVENT END:VCALENDAR