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 Concept Videos

You might also read

Related Articles

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

Sort by
Same author

Transcriptomic Profiling of the Developing Cardiac Conduction System at Single-Cell Resolution.

Circulation research·2019
Same author

Deconvolution and network analysis of IDH-mutant lower grade glioma predict recurrence and indicate therapeutic targets.

Epigenomics·2019
Same author

CT Super-Resolution GAN Constrained by the Identical, Residual, and Cycle Learning Ensemble (GAN-CIRCLE).

IEEE transactions on medical imaging·2019
Same author

Prevention effect of low-temperature atomization inhalation for radiation induced oral mucositis in patients with head and neck cancer undergoing radiotherapy.

Cancer management and research·2019
Same author

Treatments for patients with advanced neuroendocrine tumors: a network meta-analysis.

Therapeutic advances in medical oncology·2019
Same author

Interleukin-6 production mediated by the IRE1-XBP1 pathway confers radioresistance in human papillomavirus-negative oropharyngeal carcinoma.

Cancer science·2019

Related Experiment Video

Updated: Jun 9, 2025

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

475

Real-time spatiotemporal action localization algorithm using improved CNNs architecture.

Hengshuai Liu1, Jianjun Li2, Jiale Tong1

  • 1School of Digital and Intelligent Industry, Inner Mongolia University of Science and Technology, Baotou, 014000, Inner Mongolia, China.

Scientific Reports
|October 21, 2024
PubMed
Summary
This summary is machine-generated.

This study introduces a novel network for human spatiotemporal action localization, outperforming the YOWO model in speed and accuracy. The new model enhances feature extraction and bounding box regression for improved action recognition and localization.

Keywords:
2D CNNs3D CNNsReal-timeSpatiotemporal action localization

More Related Videos

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
05:41

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis

Published on: February 6, 2020

9.4K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

370

Related Experiment Videos

Last Updated: Jun 9, 2025

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

475
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
05:41

A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis

Published on: February 6, 2020

9.4K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

370

Area of Science:

  • Computer Vision
  • Artificial Intelligence
  • Machine Learning

Background:

  • Spatiotemporal action localization is crucial for understanding human activities in videos.
  • Existing models like YOWO utilize Convolutional Neural Networks (CNNs) but face limitations in efficiency and accuracy.
  • There is a need for faster and more accurate methods for real-time human action analysis.

Purpose of the Study:

  • To propose a novel, faster, and more accurate network for human spatiotemporal action localization.
  • To improve upon the YOWO model by refining feature extraction and bounding box regression techniques.
  • To develop a more lightweight and efficient architecture for action localization tasks.

Main Methods:

  • Utilized 2D CNNs for spatial feature extraction and 3D CNNs for spatiotemporal feature extraction, omitting feature fusion.
  • Integrated a coordinate attention mechanism into the 2D CNNs.
  • Employed CIoU loss for bounding box regression instead of coordinate offset loss.

Main Results:

  • Achieved a speed of 39 fps with 16-frame input clips, reducing parameters by 21.76 million compared to YOWO.
  • Improved frame-mAP by 17.09% on UCF-Sports and 7.15% on JHMDB-21 datasets (at IoU 0.5).
  • Enhanced video-mAP by 2.7%, 8.7%, and 14.4% at IoU thresholds of 0.2, 0.5, and 0.75 on the JHMDB-21 dataset.

Conclusions:

  • The proposed network offers a significant advancement in spatiotemporal action localization.
  • The model demonstrates superior performance in both speed and accuracy over existing methods like YOWO.
  • The refined architecture provides a more efficient solution for real-time human action analysis.