BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:LOCUS: Learning Object Classes with Unsupervised Segmentation - Jo
 hn Winn\, Microsoft Research
DTSTART:20050622T140000Z
DTEND:20050622T150000Z
UID:TALK4340@talks.cam.ac.uk
CONTACT:Phil Cowans
DESCRIPTION:LOCUS (Learning Object Classes with Unsupervised Segmentation)
  is a system for learning object class models and object segmentations fro
 m unannotated images. LOCUS uses a generative probabilistic model to combi
 ne bottom-up cues of color and edge with top-down cues of shape and pose. 
 A key aspect of the model is that the object appearance is allowed to vary
  from image to image\, allowing for significant within-class variation. \n
 \nI will show that LOCUS successfully learns class models from unlabelled 
 images\, whilst also giving segmentation accuracies that rival existing su
 pervised segmentation methods. LOCUS also infers the position and pose of 
 the object in each image. \n\nThe intention is to use these class models f
 or simultaneous recognition and segmentation of objects. I will present so
 me preliminary results for this task and discuss some promising extensions
  to the LOCUS model. Finally\, I will demonstrate that LOCUS can be used t
 o perform motion segmentation and object tracking in video\, despite chang
 es in illumination\, pose and background clutter. \n\nThis is joint work w
 ith Nebojsa Jojic of MSR Redmond\n
LOCATION:Ryle Seminar Room\, Cavendish Laboratory
END:VEVENT
END:VCALENDAR