Reinforcement learning for instance segmentation with high-level priors | IEEE Conference Publication | IEEE Xplore