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

Maplets for correspondence-based object recognition.

Junmei Zhu1, Christoph von der Malsburg

  • 1Computer Science Department, University of Southern California, Los Angeles, CA, USA. jzhu@memphis.edu

Neural Networks : the Official Journal of the International Neural Network Society
|November 24, 2004
PubMed
Summary

This study introduces a novel system for visual object recognition, achieving invariance to transformations like scale and orientation. The approach utilizes maplets for robust feature matching, improving upon existing dynamic link systems.

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

The Cooperative Network Architecture: Learning Structured Networks as Representation of Sensory Patterns.

Neural computation·2026
Same author

Active Peptides Derived from Snail Mucus Promoted Wound Healing by Enhancing Endothelial Cell Proliferation and Angiogenesis.

International journal of molecular sciences·2025
Same author

How are segmentation and binding computed and represented in the brain?

Cognitive processing·2024
Same author

Toward understanding the neural code of the brain.

Biological cybernetics·2021
Same author

ADH7, miR-3065 and LINC01133 are associated with cervical cancer progression in different age groups.

Oncology letters·2020
Same author

An applet for the Gabor similarity scaling of the differences between complex stimuli.

Attention, perception & psychophysics·2016

Area of Science:

  • Computer Vision
  • Artificial Intelligence
  • Computational Neuroscience

Background:

  • Current visual object recognition systems struggle with variations in position, orientation, and scale.
  • Existing dynamic link models face challenges in convergence speed and adaptability.

Purpose of the Study:

  • To develop a correspondence-based system for visual object recognition with enhanced invariance.
  • To introduce a novel representation intermediate between high- and low-dimensional correspondence methods.

Main Methods:

  • The system employs higher-order links, termed maplets, to represent object correspondences.
  • Maplets are specific to narrow ranges of mapping parameters (position, scale, orientation) and interact cooperatively.
  • The model is based on dynamic links but addresses limitations of previous formulations.

Related Experiment Videos

Main Results:

  • The proposed system demonstrates invariance to position, orientation, scale, and deformation.
  • Face recognition experiments show competitive performance compared to other published systems.
  • The system exhibits improved speed of convergence and a wider range of allowed variations.

Conclusions:

  • This work presents a significant step towards robust visual object recognition.
  • The findings suggest a new direction for reformulating neural dynamics, incorporating rapid network self-organization.
  • The maplet-based approach offers a promising alternative for modeling brain state organization.