Replace manual and rule-based defect classification — sort inspection images automatically at production scale
Applied SmartFactory Defect Classification is a production-proven, AI-based solution that automatically detects and classifies inspection images into more than 100 defect categories. By replacing manual and rule-based defect review, it improves classification accuracy, reduces cycle time, and enables consistent, scalable defect decisions across development, ramp, and high-volume production. For yield teams, the constraint has rarely been inspection capacity — it has been review capacity. Automating classification removes the queue between an image being captured and a disposition being made.
Applied SmartFactory Defect Classification improves yield by enabling earlier and more consistent defect detection and disposition, recovering 0.3–1% yield that contributes directly to top-line revenue. It classifies more than 90% of inspection defects automatically at greater than 97.5% accuracy across all defect classes, with a reported 0% escape rate. Shorter review queues cut inspection-to-decision cycle time by roughly 26%.
Defect Classification is designed for production scale deployment and addresses the limitations of manual and rules based defect classification systems.
These results enable manufacturers to improve operational consistency while reducing reliance on manual review.
Proven production results include:
How SmartFactory Defect Classification automates defect classification at production scale:
Automated defect classification improves manufacturing yield and inspection throughput.
How defect classification data is operationalized across the factory:
It classifies inspection images into more than 100 defect categories. Because the approach is AI-based rather than rule-based, the category set is not limited to conditions someone wrote a rule for in advance.
In production environments it achieves greater than 97.5% accuracy across all defect classes, with greater than 90% automation and a reported 0% escape rate. The escape rate matters most: defects are consistently identified before downstream processing.
It reduces reliance on manual review rather than eliminating it, with up to 90% reduction in person-to-machine ratios. Engineers stay in the loop for the cases that need judgment instead of reviewing everything.
By improving inspection tool utilization, it allows manufacturers to defer or avoid additional scan tool capital expenditure. The existing inspection fleet does more because results are no longer waiting on review capacity.
Defect results feed directly into yield analysis, disposition, and engineering workflows rather than sitting in an inspection system. That is what turns a classification into a decision.
Yes. It integrates with SPC, FDC, and Run-to-Run Control to provide consistent defect context across quality and process control functions, which enables closed-loop process control and faster root cause analysis.
Defect Classification sorts inspection images into more than 100 categories automatically, so disposition happens at inspection speed rather than review speed.
