[ChimeText] 18 July 10.30 am (next week Friday): 3 seminars on 1) Real-Time Document Image Retrieval with LLAH 2) Large-Scale and Real-Time Specific Object 3) Pattern recognition with supplementary information

[Min-Yen Kan]

Dear CHIME text processing members:

A jointly organized seminar among SoC, NUS and PREMIA on pattern
recognition techniques may be of interest to you.

Cheers,

Min
--------------------------------

JOINTLY ORGANIZED SEMINAR BY SOC, NUS & PREMIA

TITLE : (1) Real-Time Document Image Retrieval with LLAH (2)
       Large-Scale and Real-Time Specific Object
       Recognition (3) Pattern recognition with
       supplementary information -  an overview and
       recent results

SPEAKER : (1) & (2) Prof. Koichi Kise;  (3) Dr. Masakazu Iwamura
         (1) & (2) Department of Computer Science and Intelligent
         Systems, Osaka prefecture University
         (3) Graduate School of Engineering, Osaka Prefecture
         University

TIME : July 18, 2008, 10:30am - 12:00pm, Fri
(Tea Reception at The Venue)

VENUE : SR2, COM1 #02-04
       School of Computing, National University of Singapore

Chaired by Dr Tan Chew Lim (tancl at comp.nus.edu.sg)

ABSTRACTs:

(1) Real-Time Document Image Retrieval with LLAH:
In this talk I describe our image retrieval method named Locally
Likely Arrangement Hashing (LLAH). LLAH has following excellences: (1)
fast and accurate retrieval even under perspective distortion and
occlusion, (2) the captured region and skew of the document can be
estimated from correspondences of feature points subsidiarily obtained
during retrieval process. In this talk, I show a demonstration of a
real-time document image retrieval system with a DB up to 20,000 pages
in 1/7 second per query. Reference: http://imlab.jp/LLAH/.

(2) Large-Scale and Real-Time Specific Object Recognition: What will
happen if we have a tool for "linking" real objects to the cyberspace?
Will it be effective for solving some problems of our daily life? How
can we realize such functionality? --- Barcodes or RFIDs?  In my talk,
I would like to discuss with you the possibility of realizing such
functionality by "object recognition"  technologies.  As an example
case, I introduce a method of large-scale (100,000 objects) and
real-time (200ms) recognition of planar objects with a cheap web
camera. A demo is also planned.

(3) Pattern recognition with supplementary information - an overview
and recent results : In pattern recognition, unavoidable error is
called Bayes error, which is caused by almost same features derived
from different classes. The royal road to decrease the Bayes error is
to devise new good features. However, it is not  always possible. For
example, in character recognition, ``I'' and ``l'' in some fonts have
the identical shape, and they cannot be distinguished by only their
appearances. In order to cope with the problem, we have proposed to
introduce a ``hint'' on the true class, which is called
``supplementary information.''  For the problem of ``I'' and ``l,'' a
small hint that ``the answer is capital letter'' is enough to
recognize them.  In this talk, we present an overview and recent
results of a new pattern recognition framework which employs the
supplementary information.


BIO-DATAs:

Prof Koichi Kise received his B.E., M.E. and Ph.D. degrees in
communication engineering from Osaka University, Japan, in 1986, 1988,
and 1991, respectively. In 1990, he joined Osaka Prefecture
University, Japan, where he is now a  Professor of Department of
Computer Science and Intelligent Systems.  From 2000 to 2001, he was a
Visiting Professor at German Research Center for Artificial
Ingelligence (DFKI), Germany. His research interests include document
image analysis, object recognition and information retrieval. He
received several domestic and international awards including IEICE
Best Paper Award in 2006, IAPR/ICDAR2007 The Best Paper Award.  He is
a member of the editorial board of IJDAR, an associate editor of
pattern recognition. He has organized several international workshops,
such as CBDAR2005, CBDAR2007 and DAS2008.

Dr Masakazu Iwamura received his B.E., M.E. and Ph.D Engineering
degrees from Tohoku University, Japan, in 1998, 2000 and 2003,
respectively. He is an assistant professor at Graduate School of
Engineering, Osaka Prefecture  University. His research interest
includes pattern recognition, object recognition and information
retrieval. He received IEICE Best Paper Award in 2006, IAPR/ICDAR2007
The Best Paper Award, Best Poster Award in MIRU2005, Best
Demonstration Award in MIRU2006 and MIRU2007, and Osaka Pref. Univ.
President Prize in 2006, 2007 and 2008. He is an organizing committee
member of CBDAR2005, CBDAR2007 and DAS2008, and a program committee
member of DAS2008.