Learning Geometry From Vision For Robotic Manipulation