This seminar was held in person in Wu and Chen Auditorium as well as virtually…
Reconstructing depth and motion of every pixel for arbitrary scenes is a core problem in 3D vision with many downstream applications. In this talk, I will describe some of our recent efforts toward this goal, including various strategies to obtain effective training data for single-image 3D reconstruction, and new neural architectures that advance the state of the art of multiview 3D reconstruction.