Towards Control-Centric Representations in Reinforcement Learning from Images

Open in new window