Learning to segment images using dynamic feature binding

Authors:
Michael C. Mozer;Richard S. Zemel;Marlene Behrmann;Christopher K. I. Williams
Affiliations:
Department of Computer Science and Institute of Cognitive Science, University of Colorado, Boulder, CO 80309-0430 USA;Department of Computer Science, University of Toronto, Toronto, Ontario M5S 1A4;Department of Psychology and Faculty of Medicine and Rotman Research Institute of Baycrest Centre, University of Toronto, Toronto, Ontario M5S 1A1;Department of Computer Science, University of Toronto, Toronto, Ontario M5S 1A4
Venue:
Neural Computation
Year:
1992

Citing 8
Cited 1

A neural cocktail-party processor

Biological Cybernetics
Learning internal representations by error propagation

Parallel distributed processing: explorations in the microstructure of cognition, vol. 1
Computing with arrays of coupled oscillators: an application to preattentive texture discrimination

Neural Computation
Collective oscillations in the visual cortex

Advances in neural information processing systems 2
Synchronized oscillations during cooperative feature linking in a cortical model of visual perception

Neural Networks
Binding hierarchies: a basis for dynamic perceptual grouping

Neural Computation
The role of visual perception, selective attention, and short-term memory for symbol manipulation: a neural network model that learns to evaluate simple LISP expressions

Cognition and computer programming
Perceptual Organization and Visual Recognition

Perceptual Organization and Visual Recognition

Transformation equivariant Boltzmann machines

ICANN'11 Proceedings of the 21th international conference on Artificial neural networks - Volume Part I

Quantified Score

Hi-index	0.00

Visualization

Abstract

Despite the fact that complex visual scenes contain multiple, overlapping objects, people perform object recognition with ease and accuracy. One operation that facilitates recognition is an early segmentation process in which features of objects are grouped and labeled according to which object they belong. Current computational systems that perform this operation are based on predefined grouping heuristics. We describe a system called MAGIC that learns how to group features based on a set of presegmented examples. In many cases, MAGIC discovers grouping heuristics similar to those previously proposed, but it also has the capability of finding nonintuitive structural regularities in images. Grouping is performed by a relaxation network that attempts to dynamically bind related features. Features transmit a complex-valued signal (amplitude and phase) to one another; binding can thus be represented by phase locking related features. MAGIC's training procedure is a generalization of recurrent backpropagation to complex-valued units.