Model-Based Machine Vision

In February 2010, Prof. Kevin H. Knuth was awarded a University at Albany Faculty Research Award for the proposal titled “Tripping the Light Phantasma: Bayesian Model-Based Machine Vision”

The goal of this research project is to demonstrate a novel approach to machine vision called Bayesian model-based machine vision. This approach will enable a machine to generate a three-dimensional (3D) model of an imaged scene based on a library of 3D graphics objects, primitives and textures endowed with semantic labels. This approach will leverage three new technologies: agnostic videogame engines, graphics processing units (GPUs?) and Bayesian inference engines and integrate them to create a system that learns a 3D model of a visual scene from images recorded by a camera.