Noah Snavely
Cornell University, NY·Computer Science
When a person looks at a street, they understand it as a 3D space full of cars, people and buildings. Scene understanding systems try to give machines the same ability, rebuilding a 3D scene from images or video.
- Fields
Computer Science
Engineering
- Topics
computer vision
computer graphics
3D scene understanding
- Grants
- 1 known grant, through Sep 2027
What they work on
- Develop compositional implicit neural scene representations that pair a network's ability to capture fine detail with physical models of scene structure, such as light fields.
- Create methods that infer these 3D representations from raw images and video, using new mathematical formats for self-supervised learning that needs less data.
- Apply the learned representations to computer vision and graphics tasks such as interactive design and scene synthesis, and test how well they handle conditions not seen in training.
More about the research
These systems matter for computer vision, computer graphics and robotics, including self-driving cars. Many recent systems use neural networks to represent scenes, because networks can capture fine detail. However, a network trained on one set of conditions often fails on new ones. A car trained only on sunny-day images may misread a snowy street.
Why it matters
Self-driving cars, robots and graphics tools need to perceive 3D scenes reliably. This research could make their perception hold up under changing conditions, such as different weather, rather than only the conditions seen in training. It could also support tools for interactive 3D design and for synthesizing new scenes.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Carl Vondrick
Columbia University, NY·Computer Science·Early-career
Build open-world recognition models for satellite, aerial and ground imagery from vision-language models without language annotations, using cross-modal contrastive learning and reasoning about change over time.
Nianyi Li
Clemson University, SC·Computer Science·Early-career
Build a multi-sensor imaging system that combines RGB cameras, depth sensors, LiDAR and event cameras to capture diverse data from unstructured outdoor environments.
Xiaolong Wang
University of California-San Diego, CA·Computer Science·Early-career
Recognize objects in videos of everyday environments without a fixed category list, training with text supervision and self-supervision instead of detailed labels.
The school
Cornell University: 37 labs on PI Pages, $1.6B a year on research, 537 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.