Industrial Anomaly Detection

This project focused on abnormal image detection and defect localization for fabric products. The goal was to develop automated quality-control methods that can separate defective and non-defective images and localize abnormal regions using machine learning and deep learning.

Role: Project Lead

Key contributions:

• Applied spectral clustering to separate images into defect and non-defect groups for binary classification.
• Used transfer learning with deep convolutional neural networks for defect detection and localization.
• Developed computer vision workflows for automated industrial quality inspection.
• Contributed to anomaly detection research for real-time industrial applications.

Research focus:

• Image anomaly detection
• Defect localization
• Spectral clustering
• Transfer learning
• Industrial computer vision