Automatic Visual Recognition in Natural, Medical, and 3D Imagery

Darwin College Sciences Group
Jamie Shotton (Microsoft Research Cambridge & Darwin College)
Entertaining Room, Darwin College
Thursday, April 22, 2010 - 14:00 to 15:00

In this seminar I will present an overview of some recent advances in computer vision and machine learning that enable automatic visual recognition of objects and body parts. In natural photographs, colour, texture and context provide powerful cues for segmenting the photo into regions and labelling those regions with learned category labels, such as person, tree, sky, etc. Similar techniques can be applied to medical scans (MRI or CT scans, for example) to locate the organs within the body, or to map out the grey and white matter in the brain. Finally, I will show how new 3D depth sensing cameras coupled with learned recognition algorithms will enable you to control computer games with your whole body, without having to wear or hold anything special.

Close menu
Site navigation mobile menu