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 16, 2026

End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
03:31

End-To-End Deep Neural Network for Salient Object Detection in Complex Environments

Published on: December 15, 2023

Edge-Optimized Semi-Supervised Deep Learning for Power Line Component Inspection.

Nico Surantha1, Hanfei Zhang1, Daiki Watanabe1

  • 1Department of Electrical, Electronics, and Communication Engineering, Faculty of Science and Engineering, Tokyo City University, Tokyo 158-8557, Japan.

Sensors (Basel, Switzerland)
|July 15, 2026
PubMed
Summary

This study introduces an efficient semi-supervised deep learning framework for drone-based power line inspection. It reduces data labeling needs and optimizes performance on edge devices for reliable electrical infrastructure maintenance.

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

Combined torso and foot heat exposure reduces acute <i>ad libitum</i> energy intake in young adults.

Temperature (Austin, Tex.)·2026
Same author

Excess of post-resistance exercise energy expenditure: a comparison between morning and evening.

European journal of applied physiology·2026
Same author

Association between body mass index and frailty for middle-aged and older adults in Japan: a cross-sectional study of the Osaka health disparity solution program.

BMC public health·2026
Same author

UNC45B Reduction With Aging: A Myofiber-Intrinsic Promoting Factor for Sarcopenia.

Aging cell·2026
Same author

[Effectiveness of Prescription Inquiries Using 2D Codes of Laboratory Data by Community Pharmacies].

Yakugaku zasshi : Journal of the Pharmaceutical Society of Japan·2026
Same author

Shear flow during reciprocal shaking does not disrupt protein films at the solid-liquid interface.

International journal of pharmaceutics·2026

Area of Science:

  • Electrical Engineering
  • Computer Vision
  • Artificial Intelligence

Background:

  • Power line inspection is crucial for electrical infrastructure reliability.
  • Current drone and deep learning methods face challenges with large labeled datasets.
  • Resource-constrained edge devices limit the deployment of complex AI models.

Purpose of the Study:

  • To propose an edge-optimized semi-supervised deep learning framework for power line component inspection.
  • To reduce the reliance on extensive labeled data for AI model training.
  • To enable efficient real-time inspection on embedded systems.

Main Methods:

  • Implemented a semi-supervised learning (SSL) strategy using pseudo-labeling and confidence-based sample selection.
  • Applied hardware-software (HW-SW) co-optimization, including quantization, for edge deployment.
Keywords:
Jetson Orin NanoRaspberry Pi 5YOLOedge-AIsemi-supervised learning

Related Experiment Videos

Last Updated: Jul 16, 2026

End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
03:31

End-To-End Deep Neural Network for Salient Object Detection in Complex Environments

Published on: December 15, 2023

  • Developed a generalized edge-AI deployment score (GEADS) for performance evaluation.

    • Main Results:

    • Debiased semi-supervised learning (DeSSL) improved detection performance (mAP@0.5, F1-score) compared to standard SSL.
    • YOLOv7-Tiny (INT8) on Raspberry Pi 5 achieved the highest GEADS (0.657), indicating balanced performance.
    • The proposed GEADS metric proved more interpretable and stable than existing metrics.

    Conclusions:

    • The proposed framework effectively reduces annotation effort and maintains robust recognition performance for power line inspection.
    • HW-SW co-optimization enables real-time AI deployment on resource-limited edge devices.
    • The GEADS metric offers a reliable method for evaluating edge AI deployment.