Project Grant 2211260

This four-year, $400,000 Project Grant from the National Science Foundation's Division of Information and Intelligent Systems supports research at the Massachusetts Institute of Technology under the Computer and Information Science and Engineering program.

The grant funds the development of new techniques for composing implicit representations to enable machines to perceive and reconstruct 3D scenes in a more generalizable manner. Investigators will explore methods to make representations capable of encoding more structure like light fields while grounding them in physics. Key innovations include a new class of scene representations that bridges implicit neural representations' ability to capture details with physical representations' modeling of structure. The grant also supports developing methods to infer representations from raw images/videos through new parameterizations enabling data-efficient self-supervised learning, as well as leveraging representations for downstream computer vision and graphics tasks such as interactive design and scene synthesis.