Document Actions

You are here: Home / The Network / NNCN / D - J Collaborations / Frankfurt am Main - Komatsu

Frankfurt am Main - Komatsu

German - Japanese Collaboration in Computational Neuroscience

Autonomous learning of active depth perception: from neural models to humanoid robots
(funded by BMBF and JST)

Perceiving the three-dimensional structure of the environment is among the most fundamental tasks that biological and artificial vision systems must solve. In biological vision systems, depth perception is an active process that utilizes movements of the eyes, head, or body. Importantly, biological vision systems learn this task autonomously and they can self-calibrate. It is unclear what the underlying learning mechanisms are and how such autonomous learning and self-calibration can be mimicked in

The partners of this project propose a solution based on a generalization of the efficient coding hypothesis to active perception. The project will develop robots with an autonomously learning and fully self-calibrating binocular vision system that will utilize vergence eye movements, motion parallax and optic flow information for depth perception. Furthermore, a neurally plausible model of the development of binocular depth perception in human infants and other mammals will also be developed. It will be used to model normal and abnormal visual development in binocular vision pathologies. Thus, the project may inspire improved forms of clinical intervention.


The following scientists take part in this collaboration:

  • Prof. Jochen Triesch, Frankfurt Institute for Advanced Studies (FIAS), Frankfurt am Main
  • Prof. Sungmoon Jeong, Japan Advanced Institute of Science and Technology (JAIST), Komatsu, Japan


German Coordinator:

Prof. Jochen Triesch

Johanna Quandt Research Professor
Senior Fellow, Frankfurt Institute of Advanced Studies
J.W. Goethe University
Ruth-Moufang-Straße 1
60438 Frankfurt am Main

Tel.: +49 (0)69 798 47531