CRC 1233 "Robust Vision"

The collaborative research center “Robust Vision – Inference Principles and Neural Mechanisms” (CRC 1233) deals with basic principles of biological and machine vision, and is a close collaboration between scientists from the University and the Max Planck Institute for Intelligent Systems. Human visual perception is amazingly robust: Even in highly variable environments, we are able to make reliable inferences about the spatial arrangement of the world from limited visual information. To achieve this, our brain must perform complex computations. Artificial vision systems, in turn – as used, for example, in self-driving cars – are making steep progress in reproducing the visual skills of humans. The goal of this joint initiative will be to better understand the principles and algorithms that enable robust visual inference both in humans and machines.


Topics, aims and projects

Aim 1a: Generative and causal modeling

Aim 1b: Feedback and neural representations

  • Project 5: Task–dependend top down modulation of visual processing (von LuxburgFranz)
  • Project 6: Top-down control of visual inference in sensory representations in early visual cortex (NienborgMackeWichmann)
  • Project 7: Large-scale neuronal interactions during natural vision (Siegel)
  • Project 8: Integration of bottom-up and top-down processing in perceptual learning during sleep (RaussNienborg)

Aim 2: Dynamic input

  • Project 9: Natural dynamic scene processing in the human brain (BartelsBlack)
  • Project 10: Natural stimuli for mice: environment statistics and neural representations in the early visual system (BusseSchaeffelEuler)
  • Project 11: Stable vision in the presence of fixational eye movements: where and how is the retinal image perceptually stabilized? (SchaeffelHafed)

Aim 3: Precortical image transformations

  • Project 12: Image processing within a locally complete retinal ganglion cell population (EulerBethge)
  • Project 13: Visual processing of feedforward and feedback signals in the dLGN (BusseBerens)
  • Project 14: Image-processing computations in artificial vision (StinglMackeZeckZrenner)

Mercator project

Learning representations for efficient, visual inference in dynamic scenes (Bohg)

Early Career Research Group

 Holistic Scene Modeling (Tang)