Bayesian Vision
Abstract
Despite impressive progress in machine learning techniques over the last decade, current approaches fall far short of the capabilities of primate vision. These approaches rely mainly on brute-force learning of massive training sets. This is not only costly but also results in brittle systems that can fail spectacularly. In contrast, perception in humans and other primates structures the input by learning object categories from a small number of observations. To gain insight how this is achieved, we will use probabilistic Bayesian methods to use context information to learn object categories and relations between objects and their parts.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 16, 2019
- Source ID
- N000142012067
Entities
People
- Ernst Niebur
Organizations
- Johns Hopkins University
- Office of Naval Research
- United States Navy