Address: Beytepe Campus, Ankara, Turkey TR-06800
e-mail: erkut at cs dot hacettepe dot edu dot tr
Phone: +90 312 297 7500/149
Fax: +90 (312) 297 7502
My research centers on the areas of computer vision and machine learning. I believe the right algorithms and representations are the ones that take into account the contextual influences. Thus, the research objective that my students and I pursue is to incorporate different kinds of context (spatial, temporal and/or cross-modal) into all levels of visual processing from low to intermediate and high-level vision.
Current research interests: Visual Saliency Prediction, Automatic Image Description, Video/Photoset Summarization, Image Filtering, Image Editing
Supérieure des Télécommunications
Middle East Technical University
University of California
Oct. 2007 - Dec. 2007
Virginia Bioinformatics Institute, Virginia Tech
Jul. 2004 - Aug. 2004
[December 2016]: Our work on analysis of automatic evaluation metrics for image captioning is accepted to EACL 2017 as a long paper: "Re-evaluating Automatic Metrics for Image Captioning".
[December 2016]: Our paper on using GANs to generate outdoor scenes from attributes and semantic layouts is out on arXiv.
[November 2016]: Our work on learning dynamic saliency will be published in Signal Processing: Image Communication: "A Comparative Study for Feature Integration Strategies in Dynamic Saliency Estimation".
[July 2016]: Our paper on deep models for dynamic saliency estimation is out on arXiv.
[May 2016]: Slides from our "Deep Learning in Computer Vision" tutorial at SIU 2016 are now available online.
[May 2016]: Our paper on Turkish image captioning won the Alper Atalay Best Student Paper Award (First Prize) at SIU 2016.
[May 2016]: Our paper on photo collection summarization won the IEEE Best Student Paper Award (Second Prize) at SIU 2016.
Project Duration: 3 years (10/01/2012-10/01/2015)
This project will explore the influences of visual context and multiple cues on a number of computer vision problems. First, a novel visual saliency or attention model will be developed towards a direction that combines information coming from multiple cues with the contextual knowledge. The goal is to come up with a model that can effectively predict where people look in an image. In the second part of the project, we will investigate the problem of image filtering with a focus on devising appropriate ways of extracting high-level contextual knowledge for filtering and using them to guide the ongoing image smoothing process. The third part of the project will be about developing adaptive approaches to image segmentation that integrates information obtained from multi cues. A novel and effective segmentation algorithm will be developed that adaptively combines high-level prior knowledge with the information obtained from different visual cues at different scales.
Sponsors: The Scientific and Technological Research Council of Turkey (TUBITAK) Career Development Program (Award# 112E146)
Project web site
Project Duration: 3 years (04/01/2014-04/01/2017)
This project will explore the connection between vision and language from different directions in which we will integrate computer vision and natural language processing methods. By using these two fields together, automatic systems that transcribe the visual content of images with vivid descriptions which are very alike to human language will be obtained. Similarly, in the context of this project, retrieval systems that describe the sentence or paragraph-based textual queries visually via related images or image sets will be constructed.
Sponsors: The Scientific and Technological Research Council of Turkey (TUBITAK) The Support Program for Scientific and Technological Research Projects (Award# 113E116) and European Union under European Cooperation in Science and Technology (COST) Programme (ICT COST IC1037 Action)
Project web site