Skip to ContentSkip to Navigation
About us Latest news News News articles

Colloquium Computer Science, Prof. Justus Piater

25 February 2013

Date:                         Monday, February 25th 2013

Speaker:                   Prof. Justus Piater
Intelligent and Interactive Systems
Institute of Computer Science, University of Innsbruck, Austria
Room:                         5161.0267 (Bernoulliborg)
Time:                           16.00

Title: Hierarchical Models and Visual Inference in Humans and Machines


While the computational details of human vision remain largely obscure,
it is clear that at least some important aspects of vision proceed by
successive abstraction over multiple levels from retinotopic feature
representations to high-level concepts.  While this general idea has a
great deal of appeal to computational scientists, it is far from clear
how to turn it into practical systems.  One may argue that this lack of
understanding slows progress in both computer vision and high-level
visual neuroscience.  In this talk I will review key aspects of
biological vision that are relatively well understood, and will describe
our own attempt at building a hierarchical vision system.  At each
level, it constructs spatial distributions of juxtapositions of features
or parts.  Each level presents to the level above an abstraction of the
level below, while presenting to the level below its own idea of what
should be seen in the scene given the level above's scene
interpretation.  Together, these spatial distributions form a layered
Markov network that models visual recognition as a bidirectional
probabilistic inference procedure that seeks consensus between stored
models and perceived stimuli at all levels of the hierarchy.

Colloquium coordinators are Prof.dr. M. Aiello (e-mail : M.Aiello and

Prof.dr. M. Biehl (e-mail: M.Biehl

Last modified:07 June 2018 10.59 a.m.

More news