Robust Visual Representation across modalities in Semantic Scene UnderstandingPublished in: PhD Thesis Paper Slides