Computer vision solutions have for long been determined by expert imagination and tools for parameter tuning. Perception has been seen as passive analysis of sensor readings using features conceived by human understanding. These systems often represented highly specialized solutions that required permanent service and adjustment.
Recent developments derived from findings in biology and cognitive psychology suggest that applied principles of learning and uncertainty treatment make living systems not only less sensitive to unexpected deviations but also more precise and efficient due to their complex capabilities to adapt to specific task, machine and environment conditions. Particularly tasks in real world environments with a multitude of stimuli and varying conditions require intelligent mechanisms for interpretation, information fusion, and attention control for reliable extraction of relevant information.

The computational perception group pursues the objective to derive advanced computer vision methods from state-of-the-art understanding of human and animal perception, giving space for inspiration from biological systems that enable fast, robust and efficient operations. Perception of living agents takes place in a sensorimotor feedback loop with the environment. These systems are performing visual behaviors that integrate evidence over time within a dynamic world, that appropiately and timely react to task specific requirements, and that act to attain useful information sources.

Broadening the view and exploring alternative concepts is the action line to pace towards innovative methods, keeping contact with institutions of similar focus on an interdisciplinary level. Joanneum Research represents the innovative interface between research and industry - the intention of the computational perception group is to move at the frontier of emergent vision technologies, having an eye on long term exploitation plans.