Our Spatial AI branch aims to research advanced AI solutions that are able to perceive and understand autonomously the physical world, both static and dynamic, with its corresponding spatial semantics. With sensory inputs of different modalities including visual, audio and range etc., we develop algorithms to digitalize and model the physical world either in a dense manner as point clouds or in a compact manner as scene graphs. We believe a complete scene representation encodes not only the geometrical attributes, but also scene semantics in its static or dynamic form, which empowers any AI system with answers or reactions to queries, such as “Where are things? What are their functions? What are the relations among them? And what are the dynamics of the scene?”.