Deep Reinforcement Learning

In this work we propose OVRL, a two-stage representation learning strategy for visual navigation tasks in Embodied AI.