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

Recognition invariance obtained by extended and invariant features.

Shimon Ullman1, Evgeniy Bart

  • 1Department of Computer Science and Applied Mathematics, Weizmann Institute of Science, Rehovot 76100, Israel. shimon.ullman@weizmann.ac.il

Neural Networks : the Official Journal of the International Neural Network Society
|August 4, 2004
PubMed
Summary

This study introduces extended features for view-invariant object recognition, enabling systems to identify novel objects from new angles without 3D data. This approach enhances generalization capabilities in visual 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

Neural correlates of minimal recognizable configurations in the human brain.

Cell reports·2025
Same author

Human-like scene interpretation by a guided counterstream processing.

Proceedings of the National Academy of Sciences of the United States of America·2023
Same author

Gaze following requires early visual experience.

Proceedings of the National Academy of Sciences of the United States of America·2022
Same author

Author Correction: View-tuned and view-invariant face encoding in IT cortex is explained by selected natural image fragments.

Scientific reports·2021
Same author

Oculo-retinal dynamics can explain the perception of minimal recognizable configurations.

Proceedings of the National Academy of Sciences of the United States of America·2021
Same author

View-tuned and view-invariant face encoding in IT cortex is explained by selected natural image fragments.

Scientific reports·2021

Area of Science:

  • Computer Vision
  • Machine Learning
  • Cognitive Science

Background:

  • Visual systems excel at recognizing objects despite changes in appearance.
  • Recognizing novel objects from new viewing directions remains a significant challenge due to image variations.

Purpose of the Study:

  • To develop a method for view-invariant object recognition that generalizes across different viewing directions.
  • To enable recognition of novel objects from a single view without relying on 3D information.

Main Methods:

  • Utilized extended features, defined as equivalence classes of informative image fragments representing object parts under varying views.
  • Extracted these features during learning from images of moving objects.
  • Implemented and tested the model on natural face images.

Related Experiment Videos

Main Results:

  • The approach demonstrated effective generalization across significant changes in viewing direction.
  • The model compensated for novel viewing angles without explicit 3D object representations.
  • Performance was evaluated against alternative recognition methods.

Conclusions:

  • Extended features provide a robust mechanism for view-invariant recognition.
  • Efficient and flexible generalization in recognition systems relies on extracting diverse informative features.
  • The approach shows biological plausibility and potential for broader applications in visual recognition.