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

Zero-Shot Learning Using Synthesised Unseen Visual Data with Diffusion Regularisation.

Yang Long, Li Liu, Fumin Shen

    IEEE Transactions on Pattern Analysis and Machine Intelligence
    |October 14, 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

    DepMatch: Boosting Semi-Supervised Semantic Segmentation by Exploring Depth Difference Knowledge.

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

    Collaborated With Hallucination: Enhancing Egocentric Grounded Question Answering via Error Demonstrations.

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

    Generalizable Egocentric Task Verification via Cross-Modal Hybrid Hypergraph Matching.

    IEEE transactions on pattern analysis and machine intelligence·2026
    Same author

    Jo-SNC: Combating Noisy Labels Through Fostering Self- and Neighbor-Consistency.

    IEEE transactions on pattern analysis and machine intelligence·2025
    Same author

    MSFlow: Multiscale Flow-Based Framework for Unsupervised Anomaly Detection.

    IEEE transactions on neural networks and learning systems·2024
    Same author

    Hierarchical Graph Pattern Understanding for Zero-Shot Video Object Segmentation.

    IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2023
    Same journal

    Relation DETR+: Exploring Explicit Position Relation Prior for Dense Prediction.

    IEEE transactions on pattern analysis and machine intelligence·2026
    Same journal

    RBF++: Quantifying and Optimizing Reasoning Boundaries across Measurable and Unmeasurable Capabilities for Chain-of-Thought Reasoning.

    IEEE transactions on pattern analysis and machine intelligence·2026
    Same journal

    CAFE: Cross-View Adaptive Fusion and Cluster Center Enhancement for Robust Multi-View Clustering.

    IEEE transactions on pattern analysis and machine intelligence·2026
    Same journal

    DIVER: Reinforced Diffusion Breaks Imitation Bottlenecks in End-to-End Autonomous Driving.

    IEEE transactions on pattern analysis and machine intelligence·2026
    Same journal

    Ethics-Aware Safe Reinforcement Learning for Rare-Event Risk Control in Interactive Urban Driving.

    IEEE transactions on pattern analysis and machine intelligence·2026
    Same journal

    Learning Shape Anchors for Holistic Indoor Scene Understanding.

    IEEE transactions on pattern analysis and machine intelligence·2026
    See all related articles

    This study introduces Unseen Visual Data Synthesis (UVDS) to generate training data for novel classes using semantic attributes, overcoming limitations in zero-shot learning (ZSL) and outperforming existing methods.

    Area of Science:

    • Computer Science
    • Artificial Intelligence
    • Machine Learning

    Background:

    • Collecting labeled training data is expensive and time-consuming.
    • Zero-shot learning (ZSL) uses semantic attributes to bridge the gap between low-level features and high-level classes.
    • Existing ZSL methods face challenges in preventing overfitting and ensuring discriminative synthesized data.

    Purpose of the Study:

    • To synthesize training data for novel classes using only semantic attributes.
    • To address challenges of overfitting and discriminative power in synthesized data for ZSL.
    • To improve the performance of zero-shot learning by generating effective unseen visual data.

    Main Methods:

    • Proposed Unseen Visual Data Synthesis (UVDS) algorithm to project semantic features into high-dimensional visual feature space.

    Related Experiment Videos

  • Introduced a latent embedding space to reconcile structural differences between visual and semantic spaces while preserving local structure.
  • Developed Diffusion Regularisation (DR) to diffuse feature variances across dimensions and alleviate overfitting through orthogonal transformation.

    • Main Results:

    • Demonstrated the effectiveness of synthesized unseen data for zero-shot learning across four benchmark datasets.
    • The proposed UVDS algorithm with Diffusion Regularisation significantly outperforms state-of-the-art methods.
    • Showcased the ability of DR to remove redundant attributes and reduce overfitting.

    Conclusions:

    • Synthesizing training data from semantic attributes is a viable approach for zero-shot learning.
    • The UVDS algorithm effectively generates discriminative and generalizable data for unseen classes.
    • The proposed Diffusion Regularisation is crucial for improving the quality and utility of synthesized data in ZSL tasks.