Biologically Inspired Computer Vision
Our main area of research is the development of machine learning techniques for image recognition tasks. To date, the major problem of computer vision is the so called "semantic gap": While humans use high level concepts for describing reality, computers can do no more than extract low level features such as edges or color from the pixel content of images. In the forseeable future, computers will not have an "understanding" of images in the way of humans.
Our group tries to bridge this gap by integrating human knowledge interactively into computer vision in the following projects:
- Visual Analytics for video data: Human-Computer-Interaction and Visualization as a link between human and machine processing
- Semi-automatic learning for reducing the manual effort in creating data sets for training recognition systems
- Computer Vision in cooperation and interaction with the user
- Vision 2.0: efficient development of specific image recognition systems for large scale extraction of context from images