CALL…

                                                                  CALL for
                                                             PARTICIPATION

WORKSHOP
COGNITIVE HUMANOID VISION
To be held at IEEE-RAS Humanoids'2008
http://humanoids2008.org/huma5.html

Chairs: Prof. Eduardo Bayro-Corrochano (CINVESTAV, Guadalajara, Mexico) and Prof. Ales
Leonardis (Visual Cognitive Systems Laboratory, University of Ljubljana)

The organizers invite you to submit a 3-4 pages length extended abstract for review to
the Workshop on Cognitive Humanoid Vision.
The workshop will take place on Monday, December 1st ,2008, at KAIST, Daejeon, Korea

Important Dates:

submission deadline ­ September 15th
send extended abstracts to: edb@gdl.cinvestav.mx
notification of acceptance - October 15th
final version ­ 5 November
workshop - December 1
All accepted abstracts will be published in a DVD workshop proceedings



Special Issue of International Journal of Humanoid Robotics (IJHR)
selected full extended papers will be consider as part of an special IJHR issue on
Cognitive Human Vision, which will be printed out in December 2009


Scope

In the middle eighties pioneers brought into play the concept of active vision. This
accompanied the thoughts of researchers and developers of vision systems up to know. On
the other hand research in neuroscience, cognitive robotics and humanoids is showing the
need for more understanding about the brain function and the human vision system.
Recently at the Workshop "The Active Vision of Humanoid Robots" (Humnoids'2007,
Pittsburgh) the wording Action Vision was introduced to relate vision with an expected
action when we are looking for the foundations of a cognitive architecture. Definitively, it
appears that we need a broad conceptual framework to bring together all those new ideas
and efforts for vision to push forward the design and development of intelligent perception
action systems.
         The visual manifold carries enormous amount of information which can not be
exhaustively processed in real time with current technology. So, the obvious WWWW
questions are: what, where, when and with which procedures we should look at. The rapid
growing in research interests like to understand how the brain works by building humanoids,
shows clearly that our current thoughts, theory, algorithms and hardware are by far still
insufficient to cope with the high algorithmic complexity involved by the image processing
to tackle in real time fundamentals problems like the correspondence problem other our
naïve predisposition to stick with processing in Euclidean metric in apparently obvious
subspaces. The attempt of this workshop goes beyond Active Vision or Action Vision, it
proposes as study topic the Cognitive Humanoid Vision as a much more general
framework to develop a real time vision system for humanoids.
          The Workshop goal is to gather researchers confronted with the challenge to
develop real time algorithms for cognitive vision. The traditional robot architectures are
calling for a cognitive architecture to embodied the sophistication required for an intelligent
machine. In this regard, the traditional concept of active vision evolves to a concept the
action vision and further to cognitive humanoid vision, as this is the way how the visual
processing is tight with a reactive behaviour as by humans confronted with dynamic
changes of all sort in their environment. In this workshop we expect contributions and a
debate with focus on following key issues:

      -   visual architecture for a cognitive architecture
      -   Feature extraction, grouping, 3D contours, visuo-motor representations
      -   Segment the scene into surfaces?
      -   How build models of objects and events in space and time?
      -   How integrate cue information
      -   Visuo-motor representations: how build and use them
      -   Priming models to bottom up information
      -   Model base segmentation, prospective anticipation
      -   Conscious reality instead of physical reality
      -   Perception of spatial layout
      -   Software for a basic visual front-end for humanoids
      -   Nature of the information that the humanoid should extract from images and video
      -   A language for processing the semantics of the meaning of the visual manifold
      -   Information extracting and learning at middle level vision


Submissions:



All accepted abstracts will be published in a DVD workshop proceedings. New results are
welcome. We also encourage researchers with relevant work to retarget previously
published results to the workshop topic. Authors will be asked to present their work as a talk.
Selected papers in an extend form will be considered for an special issue IJHR on
Cognitive Humanoid Vision which will be printed out in December 2009.

All submissions should describe the material to be presented at the workshop. The
extended abstracts should be 3 to 4 pages length and comply with the standard IEEE
conference page layout.in PDF format using the IEEE template
(http://humanoids2008.org/huma3-2.html). The abstracts should be emailed to
edb@gdl.cinvestav.mx
by September 15 th . Notification of acceptance will be given on October 15th and the
submission of the final version at November 5 th.