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: Apr 22, 2026

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

4.4K

TinyAct: A framework for real-time action recognition in the cloud through distillation learning.

Yupaporn Wanna1, Kannika Wiratchawa1, Thanapong Intharah1

  • 1Visual Intelligence Laboratory, Department of Statistics, Faculty of Science, Khon Kaen University, Khon Kaen, Thailand.

Plos One
|April 20, 2026
PubMed
Summary

TinyAct offers a lightweight framework for real-time human action recognition by combining edge and cloud computing. This approach reduces computational load, making it ideal for resource-constrained devices in surveillance and IoT applications.

Related Concept Videos

Observational Learning01:12

Observational Learning

1.5K
Albert Bandura's observational learning, also known as imitation or modeling, occurs when a person observes and imitates another's behavior. It is a quicker process than operant conditioning. A well-known example is the Bobo doll study, where children who saw an adult acting aggressively towards the doll were more likely to act aggressively when left alone, compared to those who observed a nonaggressive adult. Many psychologists view observational learning as a form of latent learning...
1.5K

You might also read

Related Articles

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

Sort by
Same author

Democratizing social media for health research: LLM-powered data analytics platform for NCDs.

PloS one·2025
Same author

Machine learning-assisted detection of single-point mutations <i>via</i> DNA-templated gold nanoparticle growth.

Nanoscale·2025
Same author

AI-enhanced rapid diagnostic testing platform for mass opisthorchiasis screening.

Scientific reports·2025
Same author

Attacking the out-of-domain problem of a parasite egg detection in-the-wild.

Heliyon·2024
Same author

A Compact Differential Dynamic Microscopy-based Device (cDDM): An Approach Tool for Early Detection of Hypercoagulable State in Transfusion-Dependent-β-Thalassemia Patients.

ACS applied bio materials·2024
Same author

The in vitro assessment of rheological properties and dentin remineralization of saliva substitutes containing propolis and aloe vera extracts.

PloS one·2024

Area of Science:

  • Computer Vision
  • Artificial Intelligence
  • Machine Learning

Background:

  • Human action recognition is crucial for surveillance, healthcare, and smart environments.
  • Existing deep learning models are computationally intensive, hindering deployment on edge devices.
  • Resource-constrained environments necessitate efficient action recognition solutions.

Purpose of the Study:

  • To propose TinyAct, a lightweight framework for real-time human action recognition.
  • To enable deployment on resource-constrained edge devices by combining edge and cloud processing.
  • To reduce computational overhead while maintaining recognition accuracy.

Main Methods:

  • Utilized a 3D video autoencoder for compact spatiotemporal feature extraction.
  • Implemented an AIoT architecture with edge-based feature extraction and cloud-based classification.

Related Experiment Videos

Last Updated: Apr 22, 2026

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

4.4K
  • Employed knowledge distillation with a transformer teacher model (ILA-ViT-B/16) to train a compact student architecture.
  • Used classical machine learning classifiers (Random Forest, SVM, XGBoost) for action prediction.

    • Main Results:

    • TinyAct achieved competitive performance on the Kinetics-400 dataset with reduced computational requirements.
    • Autoencoder feature extraction significantly lowered computational overhead compared to end-to-end deep learning.
    • Knowledge distillation experiments showed non-pretrained student models yielded better results (e.g., 15.11% with SVM) than pretrained ones.
    • Baseline accuracies with 16-frame clips and 1024-D features: Random Forest (57.00%), SVM (55.00%), XGBoost (54.00%).

    Conclusions:

    • TinyAct provides an efficient solution for real-time human action recognition on edge devices.
    • The proposed AIoT architecture and knowledge distillation effectively balance performance and computational cost.
    • Modular design allows flexible deployment for various applications, including smart surveillance and IoT.
    • Effective action recognition is achievable without computationally intensive deep networks.