Visual Robotic Manipulation with Depth-Aware Pretraining