BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:dsu21's list
SUMMARY:Graph algorithms for more efficient inference in 1
 st-order and higher-order MRF's - Professor Ramin 
 Zabih\, Cornell University
DTSTART;TZID=Europe/London:20150811T140000
DTEND;TZID=Europe/London:20150811T150000
UID:TALK60361AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/60361
DESCRIPTION:Efficient inference is a major challenge for the M
 RF's that arise in computer vision.\nMost such MRF
 's are 1st-order\, and are typically solved with m
 ethods like message \npassing or graph cuts. I wil
 l present a new preprocessing technique for 1st-or
 der\nMRF's that makes widely used graph cut method
 s an order of magnitude more\nefficient. Higher-or
 der MRF's are very powerful\, but present a much m
 ore difficult \nchallenge\; I will describe techni
 ques based on a variant of submodular flow that\nc
 an perform efficient inference over some important
  higher-order priors.\n
LOCATION:Engineering Department - Lecture Room - LR4
CONTACT:
END:VEVENT
END:VCALENDAR