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

Global remapping of the sensory homunculus emerges early in childhood development.

Nature communications·2026
Same author

Toward embodied and action-oriented studies and models of peripersonal space: Comment on "Computational models of peripersonal space representations" by Bertoni et al.

Physics of life reviews·2026
Same author

The body is not there to compute: Comment on "Informational embodiment: Computational role of information structure in codes and robots" by Pitti et al.

Physics of life reviews·2025
Same author

PreCNet: Next-Frame Video Prediction Based on Predictive Coding.

IEEE transactions on neural networks and learning systems·2023
Same author

Tactile training facilitates infants' ability to reach to targets on the body.

Child development·2023
Same author

CIVIT dataset: Integral microscopy with Fourier plane recording.

Data in brief·2023

Related Experiment Video

Updated: Jan 18, 2026

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

9.8K

Automatic infant 2D pose estimation from videos: Comparing seven deep neural network methods.

Filipe Gama1, Matěj Mísař1, Lukáš Navara1

  • 1Czech Technical University in Prague, Faculty of Electrical Engineering, Department of Cybernetics, Prague, Czech Republic.

Behavior Research Methods
|September 10, 2025
PubMed
Summary
This summary is machine-generated.

This study evaluates seven human pose estimation methods for infant movement analysis. ViTPose performed best, offering a viable tool for motor development research and early disorder diagnosis.

Keywords:
Body keypoint estimation from videosHuman pose estimation methods comparisonInfant pose estimationInfants in supine positionMarkerless pose estimation

More Related Videos

Integration of Animal Behavioral Assessment and Convolutional Neural Network to Study Wasabi-Alcohol Taste-Smell Interaction
06:19

Integration of Animal Behavioral Assessment and Convolutional Neural Network to Study Wasabi-Alcohol Taste-Smell Interaction

Published on: August 16, 2024

829
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
06:37

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention

Published on: December 15, 2023

5.3K

Related Experiment Videos

Last Updated: Jan 18, 2026

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

9.8K
Integration of Animal Behavioral Assessment and Convolutional Neural Network to Study Wasabi-Alcohol Taste-Smell Interaction
06:19

Integration of Animal Behavioral Assessment and Convolutional Neural Network to Study Wasabi-Alcohol Taste-Smell Interaction

Published on: August 16, 2024

829
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
06:37

Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention

Published on: December 15, 2023

5.3K

Area of Science:

  • Computer Vision
  • Developmental Pediatrics
  • Machine Learning

Background:

  • Markerless human pose estimation is crucial for in-the-wild movement studies.
  • Existing methods are trained on adult datasets, limiting infant application.
  • Early diagnosis of motor disorders can be facilitated by accurate infant motion analysis.

Purpose of the Study:

  • To test and compare the performance of seven popular human pose estimation methods on infant videos.
  • To evaluate method accuracy beyond standard metrics, including torso length errors and detection reliability.
  • To identify methods suitable for real-time infant motion analysis.

Main Methods:

  • Seven popular computer vision methods (AlphaPose, DeepLabCut, Detectron2, HRNet, MediaPipe, OpenPose, ViTPose) were evaluated.
  • Videos of infants in supine and complex positions were used for testing.
  • Performance was assessed using standard metrics, torso length error ratio, detection analysis, and confidence ratings.

Main Results:

  • ViTPose demonstrated the best performance among the tested methods without fine-tuning.
  • AlphaPose achieved near real-time performance (27 fps) among competitive methods.
  • DeepLabCut and MediaPipe showed lower performance compared to others.

Conclusions:

  • Several pose estimation methods show promise for infant movement analysis, with ViTPose being the top performer.
  • Accurate infant pose estimation can significantly aid motor development research and early disorder detection.
  • Further research can leverage these findings to develop specialized tools for pediatric movement analysis.