Artificial cognitive systems are devices which have the ability to take the information from the environment, and then use it for planning and execution of various tasks, as well as for learning and updating the internal knowledge. One of the most difficult and complex problems in this context is the process of perception, which has to transform a huge amount of data, captured by cameras, microphones, and other sensors, into representations that serve as a basis for performing higher cognitive processes. One image contains millions of picture elements (pixels) and image flow encompasses between 25 and 30 images per second. The question is, which representations are appropriate and which algorithms are needed to translate this data onto the level of individual objects, object categories, actions, activities, and other high-level scene descriptors. The course "Perception in cognitive systems" will provide a systematic study of these questions and the latest insights on the solutions in this technical and scientific area. The main topics that will be covered in the course include selected theories on computational perception and modeling of perceptual processes, as well as learning, recognition, categorization and abstraction of visual entities, active perception (mobile and humanoid robots), attention mechanisms, visual context and spatial perception (including spatial relations). Laboratory exercises will encompass experimental work with sensor and robotic systems focusing primarily on software for recognition and categorization of objects, robot localization, and active vision.
** The Course is not available in the academic year 2012/2013.