Michael Jamieson

Background Information

I am a Ph. D. student at the University of Toronto in the Computer Vision group.  I recieved my M.A.Sc. and B.A.Sc. degrees from the Department of Systems Design Engineering at the University of Waterloo in June of 2003 and June of 1999 respectively. 

In between degrees I've worked at a small radar consulting company in Ottawa designing and testing countermeasures for Synthetic Aperture Radar (SAR) systems and at Idée Inc., a company that develops new techniques for visual search. I still work there part-time to help develop new search algorithms.

This page is currently quite bereft of context, but I will add additional papers, links and general information as time permits. In the meantime, if you have any questions, please feel free to contact me at jamieson [at] cs [dot] toronto [dot] edu.

Research Interests

I'm interested in using techniques from computer vision, computational linguistics and machine learning to automatically discover some of the semantic structure within today's huge image and video collections. The more our tools can understand the connections images and videos have to meaningful concepts and to one another, the more useful these collections will be. While there is a lot of utility to be gained in this area without solving the general vision problem, progress is still intimately connected to progress on the difficult problems of object recognition.

My current research explores how we can use patterns of correspondence between the visual and linguistic elements of a large collection of image-caption pairs to help discover meaningful groups and configurations of features in both domains.

Recent Projects

Date Title Description Online
October 2007Learning Structured Appearance Models from Captioned Images of Cluttered ScenesOur ICCV paper describing a method for automatically discovering configurations of local features with meaningful correspondences with caption words.(pdf)
June 2006Using Language to Drive the Perceptual Grouping of Local Image FeaturesOur CVPR paper describing our earlier method for finding meaningful groups of local features based on language-vision correspondence.(pdf)