BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Fast Visual Tracking By Temporal Consensus - Carl Scheffler (Unive
 rsity of Cambridge)
DTSTART:20071015T101500Z
DTEND:20071015T111500Z
UID:TALK8047@talks.cam.ac.uk
CONTACT:Philip Sterne
DESCRIPTION:A.H. Gee and R. Cipolla\, ``Fast visual tracking by temporal c
 onsensus''\, _Image and Vision Computing_\, 14(2):105-114\, 1996.\n\nThis 
 is a somewhat dated paper with interesting points on doing visual tracking
  of head pose. They use a very simplistic and inaccurate feature (pixel/co
 rner) tracker but nevertheless obtain very good pose recovery results.\n\n
 The paper can be downloaded here:\n\n  http://mi.eng.cam.ac.uk/reports/svr
 -ftp/gee_tr207.ps.Z\n\n*Abstract* At the heart of every model-based visual
  tracker lies a pose estimation routine. Recent work has emphasized the us
 e of least-squares techniques which employ all the available data to estim
 ate the pose. Such techniques are\, however\, susceptible to the sort of s
 purious measurements produced by visual feature detectors\, often resultin
 g in an unrecoverable tracking failure. This paper investigates an alterna
 tive approach\, where a minimal subset of the data provides the pose estim
 ate\, and a robust regression scheme selects the best subset. Bayesian inf
 erence in the regression stage combines measurements taken in one frame wi
 th predictions from previous frames\, eliminating the need to further filt
 er the pose estimates. The resulting tracker performs very well on the dif
 ficult task of tracking a human face\, even when the face is partially occ
 luded. Since the tracker is tolerant of noisy\, computationally cheap feat
 ure detectors\, frame-rate operation is comfortably achieved on standard h
 ardware.
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR