Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

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

Fusion sensing for urban resilience.

Innovation (Cambridge (Mass.))·2026

Same author

Switching Photocatalytic NH<sub>3</sub> Oxidation Pathways from NO<sub><i>x</i></sub> to N<sub>2</sub> via Synergistic Kineto-Thermodynamic Modulation on TiO<sub>2</sub>.

Environmental science & technology·2026

Same author

A novel strategy of "Separation Surgery Combined with Vertebroplasty and Interstitial Implantation of <sup>125</sup>I Seeds (SSVPI)" in managing thoracic metastases from lung adenocarcinoma with spinal cord compression.

European spine journal : official publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society·2026

Same author

Targeting Cuproptosis and Ferroptosis via a ROS-Responsive Nanoplatform for Enhanced Synergistic Therapy Against Hepatocellular Carcinoma.

Antioxidants (Basel, Switzerland)·2026

Same author

Efficient Glare Suppression Network for Nighttime Images with Lightweight Parallel Attention and Ghost Convolution.

Sensors (Basel, Switzerland)·2026

Same author

Noninvasive PPROM risk stratification with explainable AI using routine antenatal CRP and albumin.

NPJ digital medicine·2026

Search research articles

Related Experiment Video

Updated: May 15, 2025

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

Siamese network with change awareness for surface defect segmentation in complex backgrounds.

Biyuan Liu¹, Sijie Luo¹, Huiyao Zhan²

¹School of Resources and Environment, University of Electronic Science and Technology of China, Chengdu, 611731, China.

Scientific Reports

|April 8, 2025

Summary

This summary is machine-generated.

This study introduces a Transformer-based Siamese network for precise pixel-wise surface defect detection, mimicking human inspection. The novel approach improves accuracy in complex backgrounds, outperforming existing methods.

Keywords:

Change-aware decoder Contrastive learning Siamese network Surface defect segmentation Transformer-based encoder

More Related Videos

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

A Label-Free Segmentation Approach for Intravital Imaging of Mammary Tumor Microenvironment

A Label-Free Segmentation Approach for Intravital Imaging of Mammary Tumor Microenvironment

Published on: May 24, 2022

Related Experiment Videos

Last Updated: May 15, 2025

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

A Label-Free Segmentation Approach for Intravital Imaging of Mammary Tumor Microenvironment

A Label-Free Segmentation Approach for Intravital Imaging of Mammary Tumor Microenvironment

Published on: May 24, 2022

Area of Science:

Computer Vision
Machine Learning
Artificial Intelligence

Background:

Deep visual networks excel at regional defect detection but struggle with pixel-wise accuracy.
Varied defect appearances and limited data hinder high-quality defect segmentation.
Existing methods often over-rely on defect appearance, limiting generalization.

Purpose of the Study:

To enhance pixel-wise defect segmentation accuracy, especially in complex backgrounds.
To develop a model that mimics human inspection processes for defect detection.
To address the limitations of current deep learning approaches in surface defect analysis.

Main Methods:

Proposed a Transformer-based Siamese network with change awareness for defect segmentation.
Formulated defect segmentation as a change detection problem.
Introduced a multi-class balanced contrastive loss for class-agnostic defect encoding.
Developed a synthetic dataset of multi-class liquid crystal display (LCD) defects.

Main Results:

The proposed model outperforms leading semantic segmentation methods on multiple datasets.
Achieved state-of-the-art performance compared to semi-supervised approaches.
Demonstrated effective pixel-wise defect localization through a distance map and change-aware decoder.
Maintained a relatively compact model size.

Conclusions:

The Transformer-based Siamese network offers a robust solution for accurate pixel-wise surface defect detection.
Mimicking human inspection via change detection proves effective for complex background scenarios.
The developed model advances the field of automated visual inspection and defect analysis.