pixel-to-reward mapping

pixel to reward (1) natural language guidance (1) sparse reward (1) robot manipulation (1) reward learning (1) policy learning (1) multimodal learning (1) sample efficiency (1) reinforcement learning (1)