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: Jul 14, 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

999

Using Multi-Layer Bidirectional Distillation to Enhance Local and Global Features for Action Recognition.

Shilu Kang1, Hua Huo1, Jiaxin Xu1

  • 1College of Information Engineering and Artificial Intelligence, Henan University of Science and Technology, Luoyang 471000, China.

Sensors (Basel, Switzerland)
|November 27, 2025
PubMed
Summary

Related Concept Videos

Parallel Processing01:20

Parallel Processing

The brain processes sensory information rapidly due to parallel processing, which involves sending data across multiple neural pathways at the same time. This method allows the brain to manage various sensory qualities, such as shapes, colors, movements, and locations, all concurrently. For instance, when observing a forest landscape, the brain simultaneously processes the movement of leaves, the shapes of trees, the depth between them, and the various shades of green. This enables a quick and...

You might also read

Related Articles

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

Sort by
Same author

DCANet: Disentanglement and Category-Aware Aggregation for Medical Image Segmentation.

Sensors (Basel, Switzerland)·2026
Same author

Rotation-Sensitive Feature Enhancement Network for Oriented Object Detection in Remote Sensing Images.

Sensors (Basel, Switzerland)·2026
Same author

EEG Sensor-Based Parkinson's Disease Detection Using a Multi-Domain Feature Fusion Network.

Sensors (Basel, Switzerland)·2025
Same author

Key-Frame-Aware Hierarchical Learning for Robust Gait Recognition.

Journal of imaging·2025
Same author

Segmentation-Assisted Fusion-Based Classification for Automated CXR Image Analysis.

Sensors (Basel, Switzerland)·2025
Same author

Parkinson's Disease Detection via Bilateral Gait Camera Sensor Fusion Using CMSA-Net and Implementation on Portable Device.

Sensors (Basel, Switzerland)·2025

This study introduces a Multi-Layer Bidirectional Distillation Model (MBD) for action recognition. The MBD model effectively balances local and global features, improving accuracy in both short and long videos.

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Machine Learning

Background:

  • Action recognition accuracy depends on balancing local and global video features.
  • Long-video understanding presents challenges in dynamically integrating these features.

Purpose of the Study:

  • To propose a Multi-Layer Bidirectional Distillation Model (MBD) for enhanced action recognition.
  • To explore synergistic enhancement between local and global features using a two-stream architecture.

Main Methods:

  • Utilized 3D CNN for local and Video Transformer for global spatio-temporal features.
  • Implemented bidirectional knowledge distillation at intermediate and final layers.
  • Employed an adaptive fusion strategy based on feature dominance for dynamic weighted summation.
Keywords:
3D CNNcomplementary mechanismsfeature fusionhuman action recognitionknowledge distillationvideo Transformer

Related Experiment Videos

Last Updated: Jul 14, 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

999

Main Results:

  • The MBD model demonstrated superior performance across four classic action recognition benchmarks.
  • Achieved improved accuracy in both short-video recognition and complex long-video scenarios.
  • Effectively suppressed noise from non-dominant features while maximizing dominant feature advantages.

Conclusions:

  • The MBD model offers a novel approach to action recognition by effectively integrating local and global features.
  • Bidirectional distillation and adaptive fusion strategies overcome limitations of traditional methods.
  • The model shows significant potential for real-world video understanding applications.