BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Shared Segmentation of Natural Scenes using Dependent Pitman-Yor P
 rocesses - Dr Erik Sudderth (UC Berkeley)
DTSTART:20081008T123000Z
DTEND:20081008T133000Z
UID:TALK14193@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:We explore statistical frameworks for the simultaneous\, unsup
 ervised segmentation and discovery of visual object categories from image 
 databases.  Examining a large set of manually segmented scenes\, we show t
 hat object frequencies and segment sizes both follow power law distributio
 ns\, which are poorly captured by standard methods.  Motivated by this\, w
 e develop an alternative family of models based on the Pitman-Yor (PY) pro
 cess\, a generalization of the Dirichlet process.  This nonparametric prio
 r distribution leads to learning algorithms which discover an unknown set 
 of objects\, and segmentation methods which automatically adapt their reso
 lution to each image.  Generalizing previous applications of PY priors\, w
 e use non-Markov Gaussian processes to infer spatially contiguous segments
  which respect image boundaries.  Using a novel family of variational appr
 oximations\, our approach produces segmentations which compare favorably t
 o state-of-the-art methods\, while simultaneously discovering categories s
 hared among natural scenes.
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR