Abstract: Visual understanding of 3-D objects is essential for robotic manipulation, autonomous navigation, and augmented reality. However, existing methods struggle to perform this task efficiently ...