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 3, 2026

Artificial Intelligence-Based System for Detecting Attention Levels in Students
06:37

Artificial Intelligence-Based System for Detecting Attention Levels in Students

Published on: December 15, 2023

Efficient spatio-temporal modeling for human action recognition from RGB streams using unitary temporal encoding and

Abdul Majid1, Yulin Wang1, Aqsa1

  • 1School of Computer Science, Wuhan University, Wuhan, 430072, Hubei, China.

Scientific Reports
|July 1, 2026
PubMed
Summary

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

DNA barcoding and phylogenetic insights into the selected endemic flora of the Western Himalayas.

Scientific reports·2026
Same author

Investigation of lead-free double perovskite Cs₂CeAgCl₆ for resistive switching and optoelectronic applications.

Scientific reports·2026
Same author

Activation of sarco/endoplasmic reticulum Ca²⁺-ATPase 2 (SERCA2) reduces brain injury and improves cognitive function following ischemic stroke.

Biomedicine & pharmacotherapy = Biomedecine & pharmacotherapie·2026
Same author

Letter to Editor on: "Trends and Racial/Ethnic Disparities in Pharmacologic and Surgical Management of Treated Female Patients With Urinary Incontinence".

Neurourology and urodynamics·2026
Same author

Molecular detection of Cyclospora cayetanensis in fresh produce and irrigation water in peri-urban settings: a cross-sectional study of the environmental-agricultural interface.

BMC microbiology·2026
Same author

Assessing pneumonia severity using neutrophil-to-lymphocyte and platelet-to-lymphocyte ratios: a cross-sectional comparison with CURB-65 score.

Therapeutic advances in infectious disease·2026

This study introduces a novel, lightweight Human Action Recognition (HAR) model that efficiently integrates spatial, temporal, and consistency modeling. The model achieves high accuracy with low computational cost, making it suitable for real-time edge applications.

Area of Science:

  • Computer Vision
  • Artificial Intelligence
  • Machine Learning

Background:

  • Human Action Recognition (HAR) is challenging due to complex temporal dynamics, redundant frames, and subtle visual differences.
  • Existing transformer and multimodal HAR models are computationally expensive, limiting their use in real-time or resource-constrained environments.

Purpose of the Study:

  • To develop a novel and lightweight Human Action Recognition (HAR) model.
  • To address the computational cost and real-time limitations of current HAR approaches.
  • To integrate spatial, temporal, and consistency modeling efficiently.

Main Methods:

  • Proposed a lightweight HAR model combining EfficientNet-B2 for spatial feature extraction.
  • Utilized a Unitary Temporal Encoder (UTE) for long-range temporal dependencies.
Keywords:
Computer visionHuman action recognitionTemporal modelingUnitary temporal encoder

Related Experiment Videos

Last Updated: Jul 3, 2026

Artificial Intelligence-Based System for Detecting Attention Levels in Students
06:37

Artificial Intelligence-Based System for Detecting Attention Levels in Students

Published on: December 15, 2023

  • Incorporated an Adaptive Temporal Consistency Module (ATCM) for local temporal consistency.

    • Main Results:

    • Achieved Top-1 accuracy of 97.10% on UCF101 and 87% on HMDB51 using only RGB input.
    • The model has 9.3 million parameters with an inference rate of 9.5 ms per 16-frame video clip.
    • Demonstrated competitive performance compared to existing RGB-based HAR methods while maintaining low computational cost.

    Conclusions:

    • The proposed HAR model is highly accurate and efficient, suitable for real-time edge inference.
    • Ablation studies confirmed the significant contribution of each proposed component to performance.
    • The model offers a practical solution for resource-constrained HAR applications.