Deutsch  Language Icon  |  Gebärdensprache  |  Leichte Sprache  |  Contact

   International Junior Research Groups

Visual Efficient Sensing for the Perception-Action Loop (VESPA)

TU Munich

International Junior Research Group
at the interface of neurosciences and engineering

Head: Dr.-Ing. Michael Dorr

Contact: michael.dorr@elitenetzwerk.de

Duration: 5 years

Affiliated Program within the Elite Network of Bavaria:

Elite Graduate Program „Neurosciences“


  • Institute for Human-Machine Communication, TU Munich
  • Institute for Neuro-and Bioinformatics, Lubeck
  • Schepens Eye Research Institute, Harvard Medical School
  • Center for Cognitive and Behavioral Brain Imaging, Ohio State University
  • Brain and Cognitive Sciences, MIT

The real world in which humans perceive and act is overwhelmingly complex. To cope with this complexity with limited information-processing resources, many biological organisms, including humans, have evolved efficient sensing strategies.  The interdisciplinary International Junior Research Group "Visual Efficient Sensing for the Perception-Action Loop", will develop novel models of these strategies. Fundamental problems in basic and clinical vision science will be addressed, and more efficient computer vision systems will be built.

Our visual resolution is highest only in small part in the centre of the retina, the fovea, and drops dramatically towards the periphery. Although the peripheral visual field has poor resolution, it is essential for survival because it covers the majority of our visual environment. Furthermore, peripheral vision is used to select the image regions that are scrutinized in more detail by the fovea as it is shifted around several times per second by eye movements. The junior research group will develop and empirically validate new models of peripheral perception and its interaction with foveal vision, providing a fundamental account of how we see and how a representation of the visual world is built and maintained during active exploration of dynamic natural scenes.

Compared to biological systems, computer vision systems still perform poorly in complex, unconstrained environments such as the real world. The junior research group will use results of this modelling process to develop improved computer vision algorithms. Although biological efficient strategies are often near-optimal, however, unattended visual information sometimes may not be processed.  This is particularly problematic during normal aging, in neuro-pathological disorders, or when new tasks are being learned by healthy subjects. The models and systems developed by the junior research group can automatically identify critical information without these limitations. Interactive devices will be developed that continuously monitor the user's eye movements and modify in real time the scene contents in order to guide the user to attend the critical locations. These devices will have applications in human-machine communication and in clinical rehabilitation.

Dr.-Ing. Michael Dorr

Funded by:
  • Bavarian State Ministry of Eduction, Science and the Arts
  • Elite Network of Bavaria:
  • since November 1, 2014