Today's success of state of the art methods for semantic segmentation is
driven by large datasets. Data is considered an important asset that needs to
be protected, as the collection and annotation of such datasets comes at
significant efforts and associated costs. In addition, visual