A Demonstration of Tracking using Dynamic Neural Fields on a Programmable Vision Chip

Abstract

A demonstration is made of a programmable vision chip, containing an array of photosensors collocated with a processing circuitry and memory, implementing a Dynamic Neural Field (DNF) over a simple saliency map. The system detects and tracks moving objects with strong contrasts. The computation of the DNF's dynamics is performed entirely on the vision chip. The system solely outputs relevant information that has been processed on the array: the activity of the DNF and/or address-events corresponding to the coordinates of its bumps of activity. ABSTRACTA demonstration is made of a programmable vision chip, containing an array of photosensors collocated with a processing circuitry and memory, implementing a Dynamic Neural Field (DNF) over a simple saliency map. The system detects and tracks moving objects with strong contrasts. The computation of the DNF's dynamics is performed entirely on the vision chip. The system solely outputs relevant information that has been processed on the array: the activity of the DNF and/or address-events corresponding to the coordinates of its bumps of activity.