Estimating the pose of hand-held objects is a critical and challenging problem in robotics and computer vision. While leveraging multi-modal RGB and depth data is a promising solution, existing ...