Talk Abstract: The current success in artificial intelligence relies heavily on internet-scale data with unified representations. However, such large-scale homogeneous data is not readily available for spatial computing applications involving 3D geometry. In this talk, I will present two approaches to building spatial intelligence systems with limited 3D data: leveraging existing mathematical models and utilizing available data in different modalities. I will share my works applying these approaches to develop machine learning methods that can synthesize and analyze 3D geometry. Finally, I will discuss the future opportunities and challenges of data-efficient spatial intelligence.

Speaker Biography: Guandao Yang is a postdoctoral scholar at Stanford, where he works with Professor Leonidas Guibas and Professor Gordon Wetzstein. He completed his Ph.D. at Cornell Tech in 2023, where he was advised by Professor Serge Belongie and Professor Bharath Hariharan. During his doctoral studies, Guandao gained experience working at various industry research labs, including NVIDIA, Intel, and Google. He received his bachelor’s degree from Cornell University in Ithaca, majoring in Mathematics and Computer Science.