Queen Mary Vision Laboratory seminar
------------------------------------


A Prototype Face Recognition System for Indexing of Unconstrained Video

Dr Simon Clippingdale
NHK (Japan Broadcasting Corporation)
Science & Technology Research Labs
www.nhk.or.jp/strl/index-e.html

Time: 11am-12am, Wednesday, 17th of September
Place: Room PP1 (People's Palace Lecture Room 1 next to the College main
building)



A prototype system which tracks and recognises faces in unconstrained video
is described. The system updates multiple running hypotheses based on the
results of multiresolution deformable template matching, performing
recognition by integrating identity information while tracking changes in
face pose, size and position. A number of modifications which have
improved the robustness of the system are discussed.




Simon Clippingdale did his Ph.D. in Computer Science at Warwick in 1988,

and

worked at Warwick on image processing, neural network and vision-related
project before went to Japan. For the last 7 years he has been with the
Science & Technology Research Labs of NHK (Japan Broadcasting Corporation)
in Tokyo, working in the image recognition group. They currently have
projects on face recognition in unconstrained video, and image and video
sequence retrieval. Other groups at NHK are also working on compression,
synthesis of virtual views, chromakey-less virtual studio, digital
networks.