Introduction

This module is used to detect and segment the defect areas in the image.

Applicable Scenarios

Renewable energy: Suitable for detecting various types of defects. This module is capable of working under complicated situations, such as when the defects are tiny, the background is complex, or the positions of the workpieces are not fixed, etc. For example, this module can be used for weld seam inspection or appearance inspection of lithium batteries.

../../../_images/defection_11.png

Electronics manufacturing: Suitable for detecting the surface defects, such as stains, bubbles, scratches, etc., of functional modules and electronic components.

../../../_images/defection_21.png

PCB manufacturing, printing, daily necessities manufacturing, and other industries: Suitable for detecting surface defects, such as scratches or foreign object debris, of PCB, connectors, printing products, daily necessities, and other objects.

../../../_images/defection_3.png

General Workflow

../../../_images/application_flow1.png

Tips

The performance of the trained model depends on the followings.

  1. The quality of labeling.

  2. Selecting a proper ROI.

  3. Selecting a suitable dataset.