Graphics Seminar

— 2:00pm

Location:
Virtual Presentation - ET - 50% / 50% - Content/CTA

Speaker:
ERIC RYAN CHAN , Ph.D. StudentComputational Imaging GroupComputer Science DepartmentStanford University
https://ericryanchan.github.io

Efficient Geometry-aware 3D Generative Adversarial Networks

Generative adversarial networks can synthesize 2D images of impressive quality but typical approaches are unable to capture the structure of 3D scenes. 3D-aware GANs combine adversarial training with breakthroughs in neural rendering to synthesize 3D scenes while relying on easily obtainable 2D image training data. However, such methods require computationally expensive neural rendering, which makes them more expensive and lower quality than state-of-the-art 2D image GANs. With this work, we significantly improve the efficiency and quality of 3D-aware synthesis, which allows us to generate photorealistic renderings (in real-time!) and detailed geometry without requiring 3D ground-truth or multi-view training data. Our method enables a plethora of interesting applications, including photorealistic 3D avatar generation, and single-image 3D reconstruction.   — Eric Chan is a Ph.D. student at Stanford where he is currently working with Prof. Gordon Wetzstein’s Computational Imaging group. During his childhood in Oakland, CA, a family full of architects and many years spent in robotics competitions embedded an appreciation for design, robotic locomotion, and spatial understanding. After studying mechanical engineering and computer science at Yale, he began learning the basics of computer vision in the hope of teaching his robots and algorithms how to better understand the world around them. Over the last couple of years, his focus has shifted to the intersection of 3D graphics and vision—to generalization across 3D representations and 3D generative models. Zoom Participation. See announcement.

For More Information:
igkioule@andrew.cmu.edu


Add event to Google
Add event to iCal