Embodied vision systems that enable robots to perceive and understand environments like humans. Provides efficient, robust 3D understanding, open-set object detection, semantic segmentation, and language-grounded perception for precise, generalizable robot actions.