Computer Vision News - June 2022

22 Congrats, Doctor! Humans accomplish everyday tasks by perceiving, understanding, and interacting with a wide range of 3D objects, with diverse geometry, rich semantics, and complicated structures. One fundamental goal of computational visual perception is to equip intelligent agents with similar capabilities. My Ph.D. research is motivated by exploring answers to one central research question -- what are good visual representations of 3D shapes for diverse downstream tasks and how do we develop general frameworks to learn them at scale? In my Ph.D., I explored along with two directions to tackle the huge complexity of 3D data and tasks. First, I worked on developing compositional approaches that smaller, simpler, and reusable subcomponents of 3D geometry, such as the parts of an object, are discovered and leveraged toward reducing the complexity of the 3D data. Next, I investigated learning visual actionable representations over 3D shapes for robotic manipulation applications, where large-scale self-supervised learning frameworks using simulated interaction are investigated and proposed for learning task-specific 3D shape semantics. Kaichun Mo has recently completed his Ph.D. at Stanford University, advised by Prof. Leonidas Guibas. Before that, he received his BS.E. degree fromtheACMHonoredClass at Shanghai JiaoTong University. He has interned at Adobe Research, Autodesk Research (AI Lab), and Facebook AI Research. His research interests focus on learning visual representations of 3D data for various applications in 3D vision, graphics, and robotics. Congrats, Doctor Kaichun! Figure 1