LCFC-Laptop: A Benchmark Dataset for Detecting Surface Defects in Consumer Electronics.

Area of Science:

• Computer Vision
• Materials Science
• Manufacturing Engineering

Background:

• Consumer electronics face reputational risks from surface defects.
• Existing datasets lack specificity for consumer electronics, hindering model performance.
• Automated defect detection is crucial due to industry automation and short product cycles.

Purpose of the Study:

• Introduce a specialized dataset for consumer electronics surface defect detection.
• Address the limitations of existing datasets in this domain.
• Facilitate the development of advanced defect detection algorithms.

Main Methods:

• Developed a specialized optical sampling system with six lighting configurations.
• Collected 14,478 high-resolution images of six defect types (scratches, particles, dirt, etc.).
• Established standardized annotation criteria and applied bounding boxes and pixelwise masks.

Main Results:

• Created the first comprehensive, multiclass, multi-defect dataset for consumer electronics surface defects.
• Dataset includes pixel-level ground-truth annotations for object detection and semantic segmentation.
• Benchmarked common semantic segmentation methods using the new dataset.

Conclusions:

• The publicly available dataset supports research in automated surface defect detection.
• Enables development and testing of algorithms under realistic production conditions.
• Aims to improve quality control and reduce reputational damage in the consumer electronics industry.