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

Deep Label Distribution Learning With Label Ambiguity.

Bin-Bin Gao, Chao Xing, Chen-Wei Xie

    IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
    |April 4, 2017
    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

    Mitochondria directly interact with the nuclear pore complex.

    Nature·2026
    Same author

    Membrane-separated electrodes enable high-rate low-energy electrochemical carbon capture.

    Science advances·2026
    Same author

    Protein Response to ACL Injury in Humans Show Early Cartilage Remodeling and Differences by Sex.

    bioRxiv : the preprint server for biology·2026
    Same author

    The Transcription Factor 12 of Basic Helix-Loop-Helix Plays an Essential Role in Retinal Health.

    Investigative ophthalmology & visual science·2026
    Same author

    Linkage between HLA-B8 and HLA-DQ2.5 Contributes to Ancestry-Dependent Genetic Risk for Celiac Disease.

    medRxiv : the preprint server for health sciences·2026
    Same author

    Dual-Masked and Discriminative Reconstruction for Unified Vision Anomaly Detection.

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

    Style-Aware Contrastive Test-Time Adaptation: A Dual-Cache Model for Robust Vision-Language Alignment.

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

    Semantic Frame Interpolation.

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

    Physics-Guided Cross-Modal Decoupling with Test-Time Adaptation for Hyperspectral Image Restoration.

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

    Deep label distribution learning (DLDL) addresses limited training data in visual recognition. This method uses ambiguous labels to improve deep ConvNets, preventing overfitting and enhancing performance in tasks like age estimation.

    Area of Science:

    • Computer Science
    • Artificial Intelligence
    • Machine Learning

    Background:

    • Convolutional neural networks (ConvNets) excel in visual recognition but require large labeled datasets.
    • Collecting sufficient labeled data is challenging for tasks like age estimation, head pose estimation, multi-label classification, and semantic segmentation.
    • Ambiguous label information in these tasks differs from traditional classification problems.

    Purpose of the Study:

    • To propose a novel method, deep label distribution learning (DLDL), to effectively utilize label ambiguity in visual recognition tasks.
    • To enhance the performance of ConvNets when training data is limited.
    • To improve recognition accuracy in domains with inherent label uncertainty.

    Main Methods:

    • Converted image labels into discrete label distributions.

    Related Experiment Videos

  • Employed deep ConvNets to learn these label distributions.
  • Minimized Kullback-Leibler divergence between predicted and ground-truth label distributions.
  • Leveraged label ambiguity in both feature and classifier learning to prevent overfitting.

    • Main Results:

    • The DLDL method demonstrated significant performance improvements over state-of-the-art methods in apparent age estimation and head pose estimation.
    • The approach also enhanced recognition performance in multi-label classification and semantic segmentation tasks.
    • The method effectively prevents network overfitting, particularly beneficial with small training datasets.

    Conclusions:

    • Deep label distribution learning is an effective strategy for visual recognition tasks with ambiguous labels and limited data.
    • The DLDL method offers a robust solution for improving the accuracy and generalization of ConvNets in challenging domains.
    • This approach provides a valuable advancement for computer vision applications requiring precise recognition under data constraints.