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: Jan 17, 2026

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
09:41

Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping

Published on: April 21, 2023

2.2K

NanoHTNet: Nano Human Topology Network for Efficient 3D Human Pose Estimation.

Jialun Cai, Mengyuan Liu, Hong Liu

    IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
    |September 17, 2025
    PubMed
    Summary

    This study introduces NanoHTNet, an efficient 3D human pose estimation (HPE) network, and PoseCLR, a pre-training method. Together, they enable accurate 3D HPE on resource-constrained edge devices like Jetson Nano.

    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

    Development and characterization of monoclonal antibodies to spring viraemia of carp virus.

    Veterinary immunology and immunopathology·2008
    Same author

    Electrophysiological characterization of a novel Kv channel blocker N,N'-[oxybis(2,1-ethanediyloxy-2,1-ethanediyl) ]bis(4-methyl)-benzenesulfonamide found in virtual screening.

    Acta pharmacologica Sinica·2008
    Same author

    Cloning and characterization of an rRNA methyltransferase from Sorangium cellulosum.

    Biochemical and biophysical research communications·2008
    Same author

    Reconstruction of anophthalmic orbits and contracted eye sockets with microvascular radial forearm free flaps.

    Ophthalmic plastic and reconstructive surgery·2008
    Same author

    Using low intensity ultrasound to improve the efficiency of biological phosphorus removal.

    Ultrasonics sonochemistry·2008
    Same author

    [Study on the spectrum of the flocculent conformation of polymer ferric sulfate flocculants].

    Guang pu xue yu guang pu fen xi = Guang pu·2008

    Area of Science:

    • Computer Vision
    • Machine Learning
    • Robotics

    Background:

    • Resource-constrained edge devices like Jetson Nano limit the application of 3D human pose estimation (HPE).
    • Existing HPE models struggle to efficiently utilize structural priors from human skeletal data.
    • Developing efficient models tailored for edge deployment is crucial for widespread HPE adoption.

    Purpose of the Study:

    • To develop a highly efficient 3D HPE network suitable for edge devices.
    • To leverage both explicit and implicit spatio-temporal human body priors for improved accuracy.
    • To introduce a general pre-training method to enhance 3D HPE model performance.

    Main Methods:

    • Proposed Nano Human Topology Network (NanoHTNet): A compact 3D HPE network utilizing stacked Hierarchical Mixers for explicit feature extraction.

    More Related Videos

    Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring
    06:32

    Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring

    Published on: July 14, 2023

    1.8K

    Related Experiment Videos

    Last Updated: Jan 17, 2026

    Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
    09:41

    Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping

    Published on: April 21, 2023

    2.2K
    Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring
    06:32

    Author Spotlight: Automated Deep Brain Stimulation for Parkinson's Disease - Exploring the Possibilities and Challenges of Home Monitoring

    Published on: July 14, 2023

    1.8K
  • Spatial Hierarchical Mixer: Learns human physical topology across semantic levels.
  • Temporal Hierarchical Mixer: Captures local movements and global action coherence using discrete cosine transform and low-pass filtering.
  • Efficient Temporal-Spatial Tokenization (ETST): Enhances spatio-temporal interaction while reducing computational complexity.
  • PoseCLR: A contrastive learning-based pre-training method for extracting implicit human topology representations by aligning 2D poses.

    • Main Results:

    • NanoHTNet demonstrates superior efficiency, making it ideal for edge devices like Jetson Nano.
    • The combination of NanoHTNet and PoseCLR significantly improves 3D HPE performance.
    • Experimental results show outperformance compared to state-of-the-art methods in efficiency and accuracy.

    Conclusions:

    • NanoHTNet, enhanced by PoseCLR pre-training, offers an efficient and effective solution for 3D HPE on edge devices.
    • The proposed methods successfully leverage explicit and implicit spatio-temporal priors for robust human pose estimation.
    • This work paves the way for practical 3D HPE applications in resource-limited environments.