Research Interests

I am interested in developing machine learning algorithms to obtain high level understanding of images and video, with particular emphasis on building models of shape and appearance. I am largely interested in building and training probabilistic models, which I have applied to human pose estimation and segmentation tasks. However, I have also taken a non-probabilistic approach to shape analysis using mathematical morphology, which I have applied to microscope image analysis.

Biography

I am currently a Research Fellow in the Vision Group of the School of Computing at the University of Leeds. I graduated in 2005 with a first class masters MMath (Hons) degree in Mathematics from Bangor University, after which I continued my studies and undertook a PhD in Computer Science at the same university, supervised by Prof Ludmila Kuncheva. During my PhD I developed an automatic classification and segmentation system for analysis of microfossil images, this system is now being used in industry for oil discovery.

 

On completion of my PhD I began post doctoral work at the University of Bradford in 2010 where I worked on two European Framework 7 projects, HERMES and ECOGEM. I developed methods of shape representation for use in image and video retrieval systems and also investigated the application of machine learning tools to increase efficiency of electrically powered vehicles.

 

I joined the University of Leeds in 2011 and now work with Dr Mark Everingham on two EPSRC funded projects. The first project concerns training automatic human pose estimation models from weakly annotated images or video. The second project is in joint partnership with the University of Oxford and the University of Surrey and the goal is to allow temporal visual content to be learnt from standard TV broadcast signals using annotations in the form of subtitles and sign language signers.