Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Neural network for dynamic binding with graph representation: form, linking, and depth-from-occlusion

J R Williamson1

  • 1Center for Adaptive Systems, Boston University, MA 02215, USA.

Neural Computation
|August 15, 1996
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Association of pre-treatment somatic cell counts with bacteriological cure following diagnosis of intramammary infection.

Research in veterinary science·2022
Same author

Premature Burial.

The Hospital·2018
Same author

Small-Vessel Disease: Diabetic Microangiopathy.

Angiology·2016
Same author

Transfer of reducing equivalents across the mitochondrial membrane. I. Hydrogen transfer mechanisms involved in the reduction of pyruvate to lactate in isolated liver cells.

Biochimica et biophysica acta·2009
Same author

Recent advances in RNA structure determination by NMR.

Current protocols in nucleic acid chemistry·2008
Same author

Thermodynamics and kinetics of central domain assembly.

Cold Spring Harbor symposia on quantitative biology·2003

This study introduces a novel neural network that solves the binding problem using graph representations, not temporal coding. It effectively segments objects in depth using visual cues, advancing computer vision.

Area of Science:

  • Computer Vision
  • Artificial Intelligence
  • Cognitive Science

Background:

  • The binding problem in visual perception remains a challenge.
  • Existing neural networks often rely on temporal coding.
  • Explicitly representing form attributes and their relations is crucial for visual understanding.

Purpose of the Study:

  • To present a neural network that solves the binding problem without temporal coding.
  • To develop a system that explicitly represents form attributes and their relations.
  • To demonstrate object segmentation and 3-D variable recovery in visual scenes.

Main Methods:

  • A novel neural network architecture is proposed.
  • The network dynamically allocates nodes and establishes arcs to create graph representations.

Related Experiment Videos

  • It utilizes line junction information for object segmentation and depth perception.

    • Main Results:

    • The network successfully solves the binding problem by representing form attributes and relations.
    • It achieves selective grouping and segmentation of objects in depth.
    • Results align with findings from recent visual search experiments.

    Conclusions:

    • The proposed neural network offers a solution to the binding problem via explicit representation and graph structures.
    • It provides a framework for recovering various three-dimensional (3-D) variables, including depth-from-occlusion, edge slant, and convexity.
    • This approach advances capabilities in computer vision and artificial intelligence for visual scene understanding.