Call for Participation

                       DESIGNING VISION SYSTEMS
                   http://www.bmva.ac.uk/meetings/

                 One-day BMVA Symposium to be held at
                    The Royal Statistical Society
                       12 Errol Street, London

                                  on

                          27th November 2002

 Chairs:
         Adrian Clark (University of Essex)
         Aaron Sloman (University of Birmingham)

Machine vision systems are usually constructed from components using a
combination of experience and "suck it and see".  Moreover, they are
often aimed at solving a particular practical problem, requiring a
limited set of techniques for image analysis and interpretation.

In contrast, animal vision, especially human vision, forms part of a
complex information processing architecture in which visual
capabilities serve a very wide range of different functions, including
recognition of individual objects, understanding complex scenes,
perceiving causal relationships, learning associations, planning
routes and complex actions, interpreting facial expressions, seeing
what people intend to do, predicting collisions and other happenings,
and controlling complex manipulations of 3-D objects in tasks as
varied as picking berries, building nests in trees, assembling a
meccano model, painting pictures, and making sculptures.

Emulating that diversity, flexibility and sophistication in the context
of a complete agent architecture may be a long way off, but this
symposium aims to identify requirements for achieiving that and to
report preliminary moves in that direction.

We seek papers that analyse task requirements for machines with
animal-like visual capabilities or present architectures,
representations, algorithms or methodologies by which vision
systems can be constructed from components in principled ways or which
provide information that support such approaches, especially in the
context of a complete autonomous human-like robot.

Relevant topics include:

  statistically-motivated approaches
  optimization-oriented approaches
  automatic vision system construction
  visual learning
  the use of vision in planning and controlling intricate actions
  forms of organisation of visual information
  understanding moving objects and machinery
  characterizing vision system performance
  testing vision algorithms
  producing a taxonomy of visual affordances
  analysing relationships between visul and other systems in a complete
    architecture.
  surveying the scope and limits of current techniques and theories
  how human vision evolved.

This is an illustrative list, not an exhaustive list.

Submission date: 2nd September. Please submit an extended summary that
is no longer than two A4-sized pages in length (PDF or PostScript
preferred) and which includes links or pointers to to web-based
illustrations, demonstration material or papers giving more details.

Please submit papers by email attachment to Adrian Clark
([log in to unmask]) by 17:00 on Monday 2nd September 2002.