Seeing in 3D: Towards Generalizable 3D Visual Representations for Robotic Manipulation - Robotics Institute Carnegie Mellon University