DURIP Development of Minimalist Cameras for Lightweight Vision
Abstract
This DURIP proposal describes our plan to develop a new breed of imaging systems for AI that we refer to as minimalist vision sensors. A minimalist camera uses the minimum number of pixels needed to accomplish any given vision task. This approach has two major benefits. First, such a camera can function with power that is many orders of magnitude less than that consumed by a traditional camera. This, in turn, paves the way for fully self-powered cameras that can use the light they receive to harvest the energy needed to operate the camera. In short, the camera can function without an external power supply (battery or power source) and can be deployed in the wild without being tethered. This significantly broadens the application domains in which a camera can be used. The second major advantage of minimalist cameras is that, since they capture minimal information (often a few dozen measurements), they implicitly respect privacy. That is, the information they capture, while sufficient to solve a specific vision task, is inherently insufficient to extract personal information such as the identities of people in the scene. We describe our plan for fabricating and field-testing several prototypes of minimalist vison sensors and explain how the requested funds will be used to achieve our milestones. The equipment and hardware we plan to acquire will not only facilitate the development of the proposed imaging systems but also help up develop a cutting-edge facility for both teaching and research related to imaging.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 24, 2025
- Source ID
- N000142512139
Entities
People
- Shree Nayar
Organizations
- Office of Naval Research
- Trustees of Columbia University in the City of New York
- United States Navy