Segmentation Outputs used for task learning

Has anyone got a chance to work with network or task learning which dealt with segmentation outputs as input and how have they used them?