In DeepDeSRT, they use semantic segmentation for parsing, what actually are the segmented objects and what are the different types?
