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 Video

Updated: Mar 30, 2026

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

1.2K

Discriminative Unsupervised Feature Learning with Exemplar Convolutional Neural Networks.

Alexey Dosovitskiy, Philipp Fischer, Jost Tobias Springenberg

    IEEE Transactions on Pattern Analysis and Machine Intelligence
    |November 6, 2015
    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

    Treatment outcomes, symptoms of anxiety and depression, and satisfaction with life after orthognathic surgery.

    Frontiers in oral health·2026
    Same author

    Coronary Culprit Lesion Location and Intramyocardial Hemorrhage in STEMI.

    Circulation. Cardiovascular imaging·2026
    Same author

    A Retrospective Analysis of Thromboelastometry and Conventional Laboratory Values in Nonsurgical and Noncardiac Surgery Patients: Not What One Might Expect From Cardiac Surgery Data.

    Journal of cardiothoracic and vascular anesthesia·2026
    Same author

    E-wave propagation index predicts left ventricular thrombus in patients after ST-elevation myocardial infarction.

    European heart journal. Acute cardiovascular care·2026
    Same author

    Improving cosmological reach of a gravitational wave observatory using Deep Loop Shaping.

    Science (New York, N.Y.)·2025
    Same author

    Effects of marine heat waves and cold spells on a polar shallow water ecosystem.

    Scientific reports·2025

    This study introduces a novel method for training deep convolutional networks using only unlabeled data. The approach learns generic features robust to transformations, outperforming state-of-the-art unsupervised learning methods on several datasets.

    Area of Science:

    • Computer Vision
    • Machine Learning
    • Deep Learning

    Background:

    • Deep convolutional networks excel at learning task-specific features for computer vision.
    • Supervised learning requires large labeled datasets, posing a challenge for new tasks.
    • Unsupervised learning methods aim to learn from unlabeled data.

    Purpose of the Study:

    • To develop an approach for training convolutional networks using only unlabeled data.
    • To achieve generic feature learning that is robust to image transformations.
    • To evaluate the performance of learned features on classification and geometric matching tasks.

    Main Methods:

    • Training a convolutional network by discriminating between surrogate classes.
    • Generating surrogate classes by applying transformations to image patches.

    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

    10.1K

    Related Experiment Videos

    Last Updated: Mar 30, 2026

    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

    1.2K
    Deep Neural Networks for Image-Based Dietary Assessment
    13:19

    Deep Neural Networks for Image-Based Dietary Assessment

    Published on: March 13, 2021

    10.1K
  • Utilizing unlabeled data for feature representation learning.

    • Main Results:

    • The learned generic features outperform state-of-the-art unsupervised learning on STL-10, CIFAR-10, Caltech-101, and Caltech-256 datasets.
    • Features demonstrate robustness to applied transformations.
    • Outperforms SIFT descriptor on geometric matching tasks.

    Conclusions:

    • Unsupervised training of convolutional networks is feasible using transformation-based surrogate tasks.
    • Generic features learned through this method offer advantages in geometric matching.
    • The approach provides a valuable alternative when labeled data is scarce.