Edge attention-based transformer for metal surface defect segmentation

Key Points

• Defect segmentation accuracy improved with an edge attention-based transformer model, achieving notable results.
• The model exhibited precision and recall rates above 85%, indicating its effectiveness in identifying defects.
• Analysis utilized a novel transformer architecture, leveraging edge attention mechanisms for enhanced accuracy in segmentation tasks.
• This approach may enable better quality assessments in manufacturing, though further validation across diverse metal types is needed.