Jane Wu
UC Berkeley
janehwu@berkeley.edu
Bio
Jane is a postdoctoral fellow in EECS at UC Berkeley, advised by Professor Jitendra Malik and funded by the NSF Mathematical Sciences Postdoctoral Research Fellowship and the UC President’s Postdoctoral Fellowship. She received her Ph.D. in CS from Stanford University, advised by Professor Ronald Fedkiw. During her Ph.D., Jane interned at Google and NVIDIA to investigate 3D generative models, novel view synthesis, ML for fluid simulation, and autonomous vehicles perception. She received her B.S. in CS and Mathematics (double major) from Harvey Mudd College, where her research in robotics was advised by Professor Chris Clark and Professor Jim Boerkoel. Jane served as a co-chair for the women in CS groups at Harvey Mudd and Stanford. She is a recipient of the Stanford Gerald J. Lieberman Fellowship and the CRA Outstanding Undergraduate Researcher Award, and in 2022 she was named a Rising Star in Computer Graphics by WiGRAPH.
Areas of Research
- Computer Graphics and Vision
Perception for Physical Reasoning and Interaction
My research is focused primarily on designing machine learning models that reason about the 3D geometry and dynamics of the physical world from visual data. In particular, my recent work investigates reconstructing cloth worn on the body and hand-object interactions. Cloth is a challenging physical structure to model due to the presence of high-frequency wrinkles and complex self-collisions, and datasets capturing hand-object interactions are currently limited and subject to frequent/large occlusions. The ability to reason about humans and objects in 3D is central to interaction, with applications in AR/VR, manipulation, human-to-robot handovers, teleoperation, etc. Since RGB cameras are widely available in consumer-grade devices, the problem of lifting 2D data to 3D applies to both the training and deployment of perception algorithms for such applications. More broadly, my research interests lie at the intersection of computer vision, computer graphics, and physical simulation. The overlap between these three areas is rapidly growing in CS, and I believe there is a synergy between incorporating graphics/physics principles in computer vision algorithms and deep learning-based simulation and rendering to accelerate real-time graphics systems.