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 10, 2026

Automatic Laser-based Geometry Capture for Finite Element Analysis of Weld Beads
07:58

Automatic Laser-based Geometry Capture for Finite Element Analysis of Weld Beads

Published on: July 25, 2025

740

WELD-DETR: A Real-Time Welding Defect Detection Framework with Multi-Scale Feature Fusion and Multi-Kernel Perception

Yu Liang1, Bojian Yu1, Mengyu Ding1

  • 1School of Marine Engineering Equipment, Zhejiang Ocean University, Zhoushan 316022, China.

Sensors (Basel, Switzerland)
|November 27, 2025
PubMed
Summary
This summary is machine-generated.

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

Physics-Informed Side-Scan Sonar Perception: Tackling Weak Targets and Sparse Debris via Geometric and Frequency Decoupling.

Sensors (Basel, Switzerland)·2026
Same author

Prevalence, patterns, and geographical distribution of cardiometabolic multimorbidity and its association with unhealthy behaviors among Chinese adults: Evidence from the China National Nutrition Health Survey (2015).

Journal of clinical lipidology·2025
Same author

3D printed near-infrared high-numerical aperture achromatic metalens.

iScience·2025
Same author

NIR-triggered in situ self-supplied H<sub>2</sub>O<sub>2</sub> for boosting photothermal disinfection via CuO<sub>2</sub>-mediated catalytic cascade reaction.

Colloids and surfaces. B, Biointerfaces·2025
Same author

DNA methylation-mediated low expression of ZNF582 promotes the proliferation, migration, and invasion of clear cell renal cell carcinoma.

Clinical and experimental nephrology·2022
Same author

A 9-gene prognostic signature for kidney renal clear cell carcinoma overall survival based on co-expression and regression analyses.

Chemical biology & drug design·2022

A new deep learning framework, WELD-DETR, enhances real-time welding defect detection by fusing multi-scale features and using multi-kernel perception. It achieves high accuracy for small defects in complex industrial settings.

Area of Science:

  • Materials Science and Engineering
  • Computer Science
  • Artificial Intelligence

Background:

  • Welding quality is critical for safety in manufacturing, aerospace, and construction.
  • Existing deep learning defect detection methods struggle with real-time performance, small defects, and complex conditions.

Purpose of the Study:

  • To develop an advanced, real-time welding defect detection framework.
  • To overcome limitations of current methods in accuracy and adaptability.

Main Methods:

  • Proposed WELD-DETR framework utilizing multi-scale feature fusion and multi-kernel perception.
  • Introduced a hierarchical feature pyramid (HFPS) for improved detection of micron-level defects.
  • Developed a multi-kernel perception wavelet convolution (MPWC) module for enhanced edge and texture analysis.
Keywords:
X-ray images analysisautomated defect recognitiondeep learningwelding defects

More Related Videos

Surrogate Model Development for Digital Experiments in Welding
09:17

Surrogate Model Development for Digital Experiments in Welding

Published on: March 28, 2025

1.8K
Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control
05:47

Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control

Published on: August 29, 2025

406

Related Experiment Videos

Last Updated: Jan 10, 2026

Automatic Laser-based Geometry Capture for Finite Element Analysis of Weld Beads
07:58

Automatic Laser-based Geometry Capture for Finite Element Analysis of Weld Beads

Published on: July 25, 2025

740
Surrogate Model Development for Digital Experiments in Welding
09:17

Surrogate Model Development for Digital Experiments in Welding

Published on: March 28, 2025

1.8K
Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control
05:47

Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control

Published on: August 29, 2025

406
  • Created an industrial-grade welding dataset and employed transfer learning for cross-condition training.

    • Main Results:

    • WELD-DETR achieved 98.2% mAP@0.5-0.95 and 96.8% precision.
    • Demonstrated a real-time inference speed of 58 FPS on an RTX 2060 GPU.
    • Showcased superior accuracy and real-time performance in challenging industrial environments with high noise and reflections.

    Conclusions:

    • WELD-DETR significantly improves real-time welding defect detection accuracy and adaptability.
    • The framework shows strong potential for intelligent welding quality assurance and process optimization.
    • Outperforms existing state-of-the-art methods in complex industrial scenarios.