Image Information for Learning in Context

Useful cognitive solutions will need to be taught, and thus need to to learn in changing contexts.  Today we re-teach most systems to adjust them to changes in environments.  The ultimate kind of augmentation.   Images in context are good examples for training material.

Vision Teacher    Technical University of Darmstadt (Germany)

Researchers at the Technical University of Darmstadt (TU Darmstadt) in Germany are teaching intelligent algorithms to detect cars, pedestrians, and potentially dangerous objects in x-ray images from transportation security. The software reconstructs image information that may be hidden by blurred or out-of-focus images, and it also could assist users in tedious tasks. To develop the system, the researchers photographed a real scene and separated the individual objects from each other by tracing their outlines. The researchers enabled the algorithms to work with less data by relying on computer games that show deceptively realistic street scenes. "Based on the information contained in the computer game, we can detect which object that is already known re-appears at a later point in time," says TU Darmstadt professor Stefan Roth. He notes this technique means the object no longer needs to be re-annotated on each video frame.   ... "