IEEE AP/MTT/EMC/ED TURKEY CHAPTER SEMINAR SERIES -- SEMINAR 65 : vTools Events

IEEE.org | IEEE Xplore Digital Library | IEEE Standards | IEEE Spectrum | More Sites

IEEE AP/MTT/EMC/ED TURKEY CHAPTER SEMINAR SERIES -- SEMINAR 65

#Pinar #Duygulu #Sahin #Machine #learning #Classification

21 February 2020 (13:40): IEEE AP/MTT/EMC/ED Turkey Seminar Series (S.65)

Speaker: Prof. Pınar Duygulu Şahin, Hacettepe University

Topic: "Recognizing and Transferring the Styles of Artists Who Illustrate Children’s Books"

Location: Middle East Technical University, Ankara, Turkey

Abstract: In this talk, I will present our recent works to explore illustrations in children’s books as a new domain in classification of artists and unpaired image-to-image translation. Our work is motivated from a young boy’s capability to recognize an illustrator’s style in a totally different context. The boy’s enthusiasm let us to start the journey to explore the capabilities of machines to recognize the style of illustrators.

First, 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 with around 94% classification performance our method over-performed the traditional methods by more than 10%.

Going beyond categorization we explored transferring style. We show that although the current state-of-the-art 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.

There are no well-defined or agreed-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 models on the illustrations dataset.

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 of California at Berkeley under the supervision of Prof. David Forsyth. After being a post-doctoral researcher at Informadia Project at Carnegie Mellon University, she joined to Department of Computer Engineering 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 Hacettepe University, Ankara, Turkey. She received Science Academy’s Young Scientist Award (BAGEP) in 2015, Fulbright scholarship in 2013, TUBITAK Career award in 2005, and the best paper in Cognitive Vision award at European 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 analysis of historical documents.