Dear Colleague,

please distribute this to your PhD students/researchers/colleagues who
might be interested.

=================================================================
			CALL FOR PARTICIPATION


            8th VIPS ADVANCED SCHOOL ON COMPUTER VISION,
            PATTERN RECOGNITION, AND IMAGE PROCESSING

                   June 18-21, 2007, Verona, Italy

This school is the eighth of a series of intensive courses,
aimed at PhD students and researchers in the areas of Computer Vision,
Image Processing and Pattern Recognition. It is organized and
sponsored by the Vision, Image Processing and Sound (VIPS) Laboratory
of the Department of Computer Science, University of Verona. The
course is residential, spanning a week, so that attendees can install
a more productive interaction with the lecturer.



Invited speaker
===============
Dr. NEBOJSA JOJIC
Microsoft Corp.


Course Title
============
MODELING NATURAL SIGNALS WITH STATISTICAL HIERARCHICAL MODELS

The course will cover probabilistic inference techniques and modeling
strategies that allow machine learning approaches for automatic
extraction of medium level representations of natural signals. By
structuring statistical generative models to mimic the structure of
the real world, the models should be able to automatically adapt to
audio, visual or multimodal signals during the unsupervised model
fitting (learning) stage, thus providing a medium-level representation
suitable for compression, transmission, search, editing, enhanced
viewing experience, etc. These models are object-based, where an
object can produce sounds, have a changing appearance, move and be
exposed to attenuation in audio domain, illumination in video domain,
and, when other objects are present, to occlusion or additive mixing
in both domains. Adaptivity is the main requirement to these models.
For example, the same model should be applicable to tracking a person
in front of a cluttered background, and to tracking a flock of birds.
The tracking task, as well as many other tasks performed jointly, such
as de-noising, dynamic mosaic building or object removal as well as
separating audio sources and associating them to object appearances,
are all achievable as probabilistic queries, i.e., inference of the
hidden variables associated to the world structure. All this should
be doable using the data itself, without special application-specific
initialization procedure or the separate supervised training stage.

The course will also cover modeling biological data, such as biological
sequences, binding energy data, and crystal structure data, as well
as one example of probabilistic inference applied to a highly refined
example of sequence data: human-generated machine code.


Course program
==============
1st  Day: Monday, 18th of June 2007 - Graphical Models
	* Bayesian networks,
	* Markov random fields and factor graphs.
   	* Simple inference techniques.
	* Generative models.
   	* Case study: Computer Vision.

2nd Day: Tuesday, 19th of June 2007 - Advanced Inference and Learning
	* Parameterized models, parameters as variables, models
	  for classification, regression and clustering.
	* Learning partially unobserved graphical models, free energy,
	  iterative conditional modes, sampling methods, variational
           methods and the EM algorithm.
	

3rd Day: Wednesday, 20th of June 2007 - Some General-Purpose Graphical 
Models.
	* Mixtures of Gaussians.
         * HMMs.
         * The multivariate Gaussian.
         * Factor analysis.
         * Linear dynamic systems, Kalman filtering and smoothing,
           learning linear dynamic systems.


4th Day: Thursday, 21th of June 2007 - The Art of Modeling
	* Occlusion models for visual and auditory data.
	* Epitomes for vision, audio and biology applications.
	* Modeling molecular binding.
         * Deformable spectrograms for audio representation.



************************************************************
**** Registration deadline is June 6 2007 ****
************************************************************

Check out the following website for full details:

http://vips.sci.univr.it/html/VIPSschool2007_8.html


Directors:
      Prof. Vittorio Murino
      Prof. Andrea Fusiello



Local Organizers:
      Dr.  Marco Cristani,
      Dr. Michela Farenzena
      Dong Seon Cheng
      Riccardo Gherardi
      Davide Moschini
      Alessandro Perina