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

Structured max-margin learning for inter-related classifier training and multilabel image annotation.

Jianping Fan1, Yi Shen, Chunlei Yang

  • 1Department of Computer Science, University of North Carolina, Charlotte, NC 28223, USA. jfan@uncc.edu

IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
|September 14, 2010
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

Classification of Systems-I01:26

Classification of Systems-I

Linearity is a system property characterized by a direct input-output relationship, combining homogeneity and additivity.
Homogeneity dictates that if an input x(t) is multiplied by a constant c, the output y(t) is multiplied by the same constant. Mathematically, this is expressed as:

You might also read

Related Articles

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

Sort by
Same author

RNA-binding protein hnRNPD induces epithelial-mesenchymal transition in Wilms' tumor via facilitating MAP4K4 mRNA stability.

Molecular genetics and genomics : MGG·2026
Same author

Learning curve for endoscopic retrograde appendicitis therapy: a multicenter retrospective study with different levels of trainees.

Gastroenterology report·2026
Same author

PufCB-Auth: A lightweight continuous multi-factor authentication scheme integrated PUF with charging behavior features for EV charging.

PloS one·2026
Same author

Vacancy-Engineered Interfacial Electrons Modulation in NiCo Hydroxide/MoS<sub>2</sub> Heterostructures for Boosted OER Electrocatalysis.

Small (Weinheim an der Bergstrasse, Germany)·2026
Same author

FEDRANN: effective long-read overlap detection based on dimensionality reduction and approximate nearest neighbors.

GigaScience·2026
Same author

Effects of dietary zinc fulvate supplementation on egg quality, serum biochemical indices, and intestinal health of aged laying hens.

AMB Express·2026
Same journal

Change-Prior-Guided Unsupervised Change Detection of Heterogeneous Remote Sensing Images.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

AgonicDreamer: Enhancing Multi-View Consistency in Text-to-3D Generation via Rectified Score Distillation.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

BiCM-Prompt: Bidirectional Cross-Modal Prompt Tuning for Class-Incremental Learning on Multisource Remote Sensing Images.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

GoP-based Quality Enhancement on Video Compression.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

Align then Tensorize: Multi-Level Consistent Anchor Graph Learning for Scalable Multi-View Clustering.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

Beyond Fidelity: Diverse Image Synthesis via Retrieval-Augmented Diffusion.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
See all related articles

This study introduces a structured max-margin learning algorithm for multilabel image annotation. The method effectively trains inter-related classifiers by leveraging visual concept networks and parallel computing, enhancing discrimination power.

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Machine Learning

Background:

  • Multilabel image annotation presents challenges due to the complexity of inter-related classifiers and visual diversity.
  • Existing methods struggle to effectively leverage visual similarity contexts for training numerous classifiers.

Purpose of the Study:

  • To develop a structured max-margin learning algorithm for more effective multilabel image annotation.
  • To enhance the training of inter-related classifiers by incorporating visual concept networks and parallel computing.

Main Methods:

  • Partitioning multilabel images into instances and developing an automatic instance label identification algorithm.
  • Employing a K-way min-max cut algorithm for instance clustering and kernel weight determination.

Related Experiment Videos

  • Constructing a visual concept network to characterize inter-concept visual similarity in the feature space.
  • Developing a structured max-margin learning algorithm integrating visual concept networks, max-margin Markov networks, and multitask learning.

    • Main Results:

    • The proposed algorithm significantly enhances the discrimination power of inter-related classifiers by leveraging inter-concept visual similarity.
    • Experiments demonstrated highly positive results across a large number of object classes and image concepts.
    • The developed parallel computing platform enables effective learning of numerous inter-related classifiers.

    Conclusions:

    • The structured max-margin learning algorithm offers a robust solution for multilabel image annotation.
    • Effectively utilizing visual concept networks and parallel computing improves classifier performance.
    • The approach demonstrates significant potential for advancing image analysis and annotation tasks.