Yunus Emre KARA, Ph.D.
Computer Vision and Machine Learning Researcher/Engineer
yekara@yekara.com

Summary

I am a multidisciplinary, versatile researcher and developer with both academic and industrial experience. I have been actively involved in computer vision and machine learning since 2008. My masters thesis focused on keypoint tracking based human action recognition and my PhD thesis focused on Bayesian approaches to crowd-sourced label collection problem. I am also keen on using my background on mathematics in research problems. I am proficient in a multitude of programming languages and can quickly adapt to new languages as well.

Research Interests

  • Computer Vision
  • Image Processing
  • Pattern Recognition
  • Machine Learning
  • Bayesian Methods
  • Statistical Modeling
  • Action Recognition
  • Pose Estimation
  • Crowd-labeling and Annotator Modeling

Experience

  • Ottoman Digitalization Framework
    Transliteration of Ottoman script to modern Latin-script based Turkish. Part of the end-to-end framework for digitalization of Ottoman era documents.
    Research and development of a high accuracy and fast OCR application from scratch. Developed for IRCICA Library for the digitalization of Ottoman era documents written in cursive Ottoman Turkish script (abandoned a century ago). The documents consist of books, journals, periodicals, and handwritten manuscripts dating back to the 14th century. The project is developed in C++ and it is the first of its kind.
  • Ozyegin University, Istanbul, Turkey
    A joint government project with ASELSAN A.S. and Hacettepe University, developed for the General Directorate of Security of Turkey. Our team was responsible for the research and development of traffic analysis (vehicle tracking, traffic density, trespassing, etc.) and object analysis (detection and tracking of suspicious or lost objects, etc.) parts of the project. The system will work in real-time on 1000 urban security surveillance cameras in Ankara. My responsibilities include devising computer vision solutions that can work on limited processing power of the embedded system in real-time; overseeing the code integration to the mainframe; and scientific reporting of the research and development process.
  • Bogazici University, Istanbul, Turkey
    The concept of the wisdom of the crowds was applied to the problem of continuous-valued ground truth label collection. The aim of the project was to deduce robust crowd consensus along with individual annotator behaviors. As part of my PhD dissertation studies, I proposed Bayesian models that enable budget friendly and high quality label collection in an online manner. The label assessment system was developed in C++ using Eigen library. Two journal papers in JAIR (SCI) and Neurocomputing (SCI-E) were published.
    Worked on mathematical analysis and statistical modeling of particle based molecular communication scenarios. Developed a simulator using CUDA for nanoparticle diffusion in liquid medium. Two journal papers in CommLetters (SCI) and NanoComNet (SCI-E) were published.
    Research and development of a Kinect camera based real-time 3D hand pose estimation and hand shape classification application. The method is based on classification of depth pixels using Random Decision and Regression Forests. The project was developed in C++ using Qt, OpenCV, OpenNI, and Kinect SDK. Published one journal paper, one book chapter, 2 national and 3 international conference/workshop papers (CVPR, ECCV, ICCV).
    This research was the focus of my masters studies. I developed a real-time keypoint tracking system using C++ and OpenCV. Video segments were classified using the tracked keypoint trajectories for recognizing the action of the subject. The work was part of the "Gesture Based HCI in Emergency Management Systems" project. In addition, I developed computer vision-based fall detection part of the "Multimodal Wireless Sensor Networks for Elderly Care" project. 3 conference papers were published.
    A graduate level course focused on binary operations; edge detection; segmentation; feature extraction; object recognition; motion and tracking.
    Undergraduate senior elective focused on color quantization; signal theory; convolution, linear and nonlinear filters; sampling and reconstruction; edge detection; image compression; warping and morphing.
  • Idiap Research Institute, Martigny, Switzerland
    Participated in the "Multimodal Computational Modeling of Nonverbal Social Behavior in Face to Face Interaction" project. Preliminary research was conducted on continuous-valued ground truth label estimation using crowd-sourced personality trait annotations from videos under the supervision of Dr. Oya Aran.
  • Sigma Research and Development, Istanbul, Turkey
    Developed an augmented reality application for an interactive medical assistance system. Also, developed a pillow recommendation tool that makes use of the user's upper body structure. Both projects were developed in C++ using data from a depth camera.
  • CC Software and Consultancy, Istanbul, Turkey
    Back-end and front-end development for a gift card processing system, including POS terminal software, gift card printing back-end software, database design and web applications.

Education

Skills

  • Languages: Turkish (Native Language), English (Full professional proficiency)
  • Programming Languages: Actively developing in C, C++, Python, MATLAB, CMake; Experience in R, Java, Pascal, Various BASIC dialects, Prolog
  • Libraries and APIs: Actively using OpenCV, Qt, Boost, Eigen, OpenMP, Intel TBB, CUDA, PyTorch; Experience in Ogre, OpenNI, Kinect SDK, wxWidgets, OpenGL, OpenCL
  • Web Development: Experience in ASP, PHP, HTML, XML, CSS, Javascript, SQL
  • Database: Experience in MSSQL, PostgreSQL, MYSQL, Access

Publications

Crowd-Labeling and Annotator Modeling
Molecular Communications
Matrix/Tensor Factorization
Real-time 3D Hand Pose Estimation and Hand Shape Classification
Human Action Recognition

Datasets