Imprecise Probability as a Linking Mechanism between Deep Learning, Symbolic Cognition and Local Feature Detection in Vision Processing

Abstract

A novel approach to computer vision is outlined, involving the use of imprecise probabilities to connect a deep learning based hierarchical vision system with both local feature detection based preprocessing and symbolic cognition based guidance. The core notion is to cause the deep learning vision system to utilize imprecise rather than single-point probabilities, and use local feature detection and symbolic cognition to affect the confidence associated with particular imprecise probabilities, thus modulating the amount of credence the deep learning system places on various observations and guiding its pattern recognition/formation activity. The potential application to the hybridization of the DeSTIN, SIFT and OpenCog systems is described in moderate detail. The underlying ideas are even more broadly applicable, to any computer vision approach with a significant probabilistic component which satisfies certain broad criteria.