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 Concept Videos

Force Classification01:22

Force Classification

2.2K
Forces play a crucial role in the study of physics and engineering. They are essential in describing the motion, behavior, and equilibrium of objects in the physical world. Forces can be classified based on their origin, type, and direction of action.
Contact and non-contact forces are two of the most widely used categories of forces. As the name suggests, contact forces require physical contact between two objects to act upon each other. Examples of contact forces include frictional,...
2.2K

You might also read

Related Articles

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

Sort by
Same author

Protocolized Pain Assessment and Management in Pediatric Trauma Patients.

The Journal of surgical research·2026
Same author

Connecting the dots - recognition of artificial and natural shapes relies on representing points of high information.

Vision research·2026
Same author

Diagnostic Outcomes of Robotic Bronchoscopy in a Multidisciplinary, Multi-hospital Lung Nodule Program.

The Journal of surgical research·2026
Same author

Small Scale Dissection and Antibody Staining of Eye-Antennal and Wing Imaginal Discs from Drosophila melanogaster.

Journal of visualized experiments : JoVE·2026
Same author

Three-dimensional shape cues affect human and artificial recognition systems differently.

PloS one·2026
Same author

Triglyceride levels and all-cause mortality in patients with left main coronary artery disease undergoing percutaneous coronary intervention.

Frontiers in cardiovascular medicine·2026

Related Experiment Video

Updated: Dec 21, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

926

Local features and global shape information in object classification by deep convolutional neural networks.

Nicholas Baker1, Hongjing Lu2, Gennady Erlikhman1

  • 1Department of Psychology, University of California, Los Angeles, Los Angeles, CA, United States.

Vision Research
|May 16, 2020
PubMed
Summary

Deep convolutional neural networks (DCNNs) process object shape using local features, unlike humans who rely on global shape. DCNNs struggle with global shape recognition, focusing instead on contour details.

Keywords:
Deep learningGlobal and local featuresObject recognitionShape

More Related Videos

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.8K
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
05:41

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis

Published on: February 6, 2020

9.8K

Related Experiment Videos

Last Updated: Dec 21, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

926
Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.8K
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
05:41

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis

Published on: February 6, 2020

9.8K

Area of Science:

  • Computer Vision
  • Cognitive Science
  • Artificial Intelligence

Background:

  • Deep convolutional neural networks (DCNNs) exhibit similarities to the human visual system.
  • Recent studies indicate limitations in DCNNs' ability to recognize objects based on shape.
  • A hypothesis suggests DCNNs prioritize local contour features over global shape information.

Purpose of the Study:

  • To test if DCNNs are sensitive to local contour features but lack global shape access.
  • To compare DCNN shape processing with human object recognition.
  • To investigate DCNNs' ability to process abstract spatial configurations of elements.

Main Methods:

  • Utilized transfer learning with AlexNet, VGG-19, and ResNet-50.
  • Retrained networks to classify circles and squares.
  • Probed networks with stimuli featuring conflicting global and local shape cues.

Main Results:

  • Networks classified stimuli based on local contour features, ignoring global shapes.
  • Modifying training data improved global shape responses but did not confer sensitivity to spatial configuration.
  • DCNNs demonstrated no sensitivity to the spatial relationships between local elements.

Conclusions:

  • DCNNs' shape processing is an inversion of human perception.
  • DCNNs extract local contour fragments without representing their spatial relationships for global shape formation.
  • Human object recognition predominantly relies on abstract relations of elements, a capability lacking in current DCNNs.