Industrial Anomaly Detection

Spectral clustering and transfer learning for fabric defect detection and localization

Automated fabric inspection and anomaly localization
Conceptual research workflow

This project developed methods for identifying abnormal fabric-product images and localizing defects for automated visual quality control.

Research period: September 2022 to February 2024
Funder: National Research Foundation (NRF), South Korea
Role: Project Lead

Objectives

  • Separate defective and non-defective fabric images.
  • Detect and localize abnormal regions to support image-based industrial inspection.

Contributions

  • Applied spectral clustering to group images for defect/non-defect binary classification.
  • Used transfer learning with deep convolutional neural networks (CNNs) for defect detection and localization.
  • Developed computer vision experiments for assessing image-based product quality.

Methods

Spectral clustering, anomaly detection, image classification, convolutional neural networks, transfer learning, defect localization, industrial visual inspection.