International Workshop on Semantic Knowledge in Computer Vision
http://www.research.ge.com/vision/skcv05

October 16, 2005, Beijing, PRC.
In association with ICCV 2005.

Paper submission: 		July 18, 2005 (after ICCV decisions)
Notification of acceptance: 	August 12
Receipt of camera ready copy: 	August 22


The use of semantic knowledge in computer vision is rapidly becoming
more widespread and significant.  In areas such as event recognition,
object recognition and content-based image retrieval, context and
common-sense knowledge are being used to achieve performance that is
not attainable by purely bottom-up, data-driven approaches.  In many
applications, meaningful visual recognition is not possible without
contextual, semantic support.  However, the use of computational
knowledge in computer vision is still in its infancy and many
fundamental challenges remain.

This workshop will bring together an interdisciplinary group of
researchers in computer vision, knowledge representation and
ontologies, machine learning, natural language and other areas to
examine the issues and recent results in using semantic knowledge for
vision problems.  Recent progress in machine learning has enabled the
rigorous management of uncertainty in large-scale reasoning problems,
and this has re-kindled the use of semantic methods and reasoning in
computer vision.  Simultaneously, the natural language and artificial
intelligence communities have developed large computational models and
databases of semantic knowledge, such as CYC, OMCSNet and WordNet,
that can be used for intelligent reasoning about real-world,
common-sense knowledge.  The multimedia and information fusion
communities are using both evidential reasoning methods and semantic
knowledgebases to fuse multiple data sources for intelligent object
and event recognition.

Papers are solicited in all disciplines related to the central theme,
including but not limited to:

o use of existing, large-scale knowledgebases/ontologies for vision problems
o new ontologies for visual objects, video events, etc.
o user-centric ontologies
o unsupervised learning of event ontologies
o automatic concept detection
o semantic representations of spatio-temporal data
o context-based recognition
o high-level event recognition
o semantic image and video annotation
o semantic event-based retreival of video
o content-based queries and use cases
o integration of vision and natural language
o visual learning vs. prior, structured knowledge
o probabilistic models for dynamic systems
o temporal logic in vision
o multi-agent multi-threaded representations
o situational awareness through visual perception
o MPEG-7


PROGRAM

The program will emphasize invited talks from researchers outside of
CV, as well as those using high-level semantics to solve vision and
perception problems.  Approximately half of the program will consist
of open submission papers.

A follow-on workshop is planned for a related venue outside of
computer vision, such as AAAI or IJCAI.


ORGANIZATION

General Chairs:

Anthony Hoogs, GE Research				hoogs@crd.ge.com
Mubarak Shah, University of South Florida		shah@cs.ucf.edu
Tom Huang, University of Illinois at Urbana-Champaign	huang@ifp.uiuc.edu

Program Committee:

Jake Aggarwal, University of Texas
Kobus Barnard, University of Arizona
Michael Chan, General Electric
Rama Chellappa, University of Maryland
John Kender, Columbia University
Kevin Murphy, University of British Columbia
Ram Nevatia, University of Southern California
Jens Rittscher, General Electric
Chris Town, Cambridge University
James Wang, Pennsylvania State University

PAPER SUBMISSION

In keeping with the spirit of a workshop, submitted papers may
emphasize intellectual risks and argue for ideas that do not yet have
comprehensive experimental support.  Hence papers may not need
describe fully developed algorithms, methods, or results as would
normally be required for acceptance at ICCV.

Papers describing novel, unpublished research are solicited in the
areas listed above and closely related topics.  Reviewing will be
double-blind by members of the program committee.  Each paper will
receive three reviews.  Acceptance will be based on relevance to the
workshop, novelty, and technical quality.

Papers should be at most 8 pages in length, in the same style format
as ICCV, and encoded as pdf.  Please ftp your pdf file to  using
the first author's last name as the filename (e.g. mylastname.pdf).

One supplemental file may be included, up to a size of 10MB.  Please
send these via ftp also, using the same filename as the paper with
"_supp" appended (e.g. mylastname_supp.pdf).

All accepted papers will be included in the electronic ICCV
proceedings.  There will not be a hardcopy proceedings for this
workshop.