Description:
Kevin Murphy of the University of British Columbia examines the problem of finding instances of visual object categories in cluttered, real-world environments. By proposing a hierarchical approach, Murphy and his team categorize a scene, utilize global image statistics for prediction purposes and run an object detector. This top-down approach is much faster than standard “brute force” approaches and also reduces error-rates.
Speaker(s):
Kevin Murphy, University of British Columbia
|
