Faculty Candidate Seminar

Generative Computer Vision for the Physical World

Ruoshi LiuPh.D. CandidateColumbia University
WHERE:
3725 Beyster Building
SHARE:
Ruoshi Liu

Zoom link for remote attendees

Meeting ID: 999 4631 3276 Passcode: 123123

Abstract: Generative models are revolutionizing our world, with the ability to generate photorealistic visual content that fools human perception. Despite their overwhelming presence in the cyber world, they haven’t been very useful in the physical world that we live in. In this talk, I will present how the rich priors encoded in large-scale generative models—ranging from shape and geometry to motion and dynamics—can be harnessed for real-world perception and interaction tasks. I will showcase how these models can facilitate advanced 3D reconstruction and enhance robotic manipulation by incorporating the structure of the physical world. Moreover, I will discuss methods to further refine and adapt these priors through self-learning, enabling machines to continually improve as they explore new scenarios and environments. Together, these breakthroughs build the foundation for my vision of creating self-supervised machines with generative priors that can learn to interact with the physical world.

Bio: Ruoshi Liu is a doctoral candidate in computer science at Columbia University. His research focuses on developing computer vision systems that can intelligently interact with the physical world. His work received wide news coverage. Open-source models and datasets he developed have been downloaded and used more than a million times by other researchers and engineers in the field. For more details, please go to https://ruoshiliu.github.io/.

 

Organizer

Stephanie Jones

Student Host

Ryan Feng

Faculty Host

Stella Yu