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 7, 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

Weakly Supervised PatchNets: Describing and Aggregating Local Patches for Scene Recognition.

Zhe Wang, Limin Wang, Yali Wang

    IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
    |February 17, 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

    Post-marketing safety surveillance of ado-trastuzumab emtansine: a real-world data retrospective cohort study using disproportionality analysis.

    Expert opinion on drug safety·2024
    Same author

    Toward stakeholders' understanding of media reporting on doctor-patient relationship issues: trust, unfamiliarity and uncertainty in the Chinese context.

    Frontiers in public health·2024
    Same author

    Large Language Model Enhanced Logic Tensor Network for Stance Detection.

    Neural networks : the official journal of the International Neural Network Society·2024
    Same author

    Correction to "A Double Network Composite Hydrogel with Self-Regulating Cu<sup>2+</sup>/Luteolin Release and Mechanical Modulation for Enhanced Wound Healing".

    ACS nano·2024
    Same author

    Comparison of robotic-assisted and laparoscopic partial nephrectomy based on the PADUA score and the predictive value of the PADUA score and the Mayo Adhesive Probability score for postoperative complications: a single-center retrospective study.

    Journal of cancer research and clinical oncology·2024
    Same author

    LSD1 Demethylates and Destabilizes Autophagy Protein LC3B in Ovarian Cancer.

    Biomolecules·2024

    This study introduces a hybrid visual representation for scene recognition, combining Convolutional Neural Networks (CNNs) with traditional methods. The new approach, VSAD, enhances patch-level feature extraction for improved image recognition accuracy.

    Area of Science:

    • Computer Vision
    • Machine Learning
    • Artificial Intelligence

    Background:

    • Traditional image recognition methods include Fisher vectors with local descriptors (e.g., SIFT) and Convolutional Neural Networks (CNNs).
    • These methods have shown success but often operate at different levels of detail or complexity.
    • A gap exists in effectively combining the strengths of both local feature encoding and deep learning for scene recognition.

    Purpose of the Study:

    • To propose a novel hybrid visual representation for image recognition, specifically targeting scene recognition.
    • To develop an end-to-end patch-level architecture (PatchNet) for modeling local patch appearances.
    • To create a unified representation (VSAD) that aggregates local patch features for global scene understanding.

    Main Methods:

    Related Experiment Videos

    Last Updated: Mar 7, 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
    • Developed PatchNet, a customized, weakly supervised CNN architecture for patch-level feature extraction guided by image-level supervision.
    • Introduced the Visual Scene Ảnh Description (VSAD) representation, integrating PatchNet's robust features and semantic probabilities.
    • Aggregated local patch descriptions into a global representation for scene recognition.

    Main Results:

    • Achieved state-of-the-art performance on two standard scene recognition benchmarks.
    • Attained 86.2% accuracy on the MIT Indoor67 dataset.
    • Reached 73.0% accuracy on the SUN397 dataset.

    Conclusions:

    • The proposed hybrid VSAD representation effectively leverages CNNs and descriptor encoding for superior scene recognition.
    • PatchNet enables effective weakly supervised learning of patch-level features.
    • The VSAD approach sets a new benchmark for scene recognition tasks.