Machine Vision & Industrial Inspection AI worked example

Defect Classification Accuracy at 99% required classification accuracy per quality plan: a worked example

What does the result look like when required classification accuracy per quality plan reaches 99%? The full calculation is worked below with real intermediate numbers. Use it when validating an AI defect classification system and you need to confirm that the model correctly labels defect types (scratch, crack, void, contamination) at the accuracy level required by your quality plan.

The inputs for this scenario

  • Defects correctly assigned to the right defect class: 186 defects (unchanged)
  • Total defects in the classification validation sample: 200 defects (unchanged)
  • Required classification accuracy per quality plan: 99 % (raised for this scenario; the documented default is 95)

Working through the calculation

  • Applying the documented formula (Defect classification accuracy = correctly classified / total defects x 100) to the inputs above produces each figure below.
  • At this operating point the engine returns 93 % for defect classification accuracy, the number this scenario is built around.
  • At this operating point the engine returns 6 points for gap to required classification accuracy.
  • At this operating point the engine returns 186 count for defects correctly classified.
  • At this operating point the engine returns 200 count for total defects in validation sample.

How this compares with the baseline

  • Against the tool's baseline example, where required classification accuracy per quality plan sits at 95% and the headline result is 93 %, this scenario lands almost exactly on the baseline at 93 %.
  • A figure at this level is achievable when required classification accuracy per quality plan is genuinely sustained, not just peaked for a shift. A single overall accuracy can mask one weak class; a confusion matrix is needed to see which defect types are being mixed up.

Results at a glance

  • Defect classification accuracy: 93 % (headline result)
  • Gap to required classification accuracy: 6 points
  • Defects correctly classified: 186 count
  • Total defects in validation sample: 200 count

Run it with your numbers

  • Every input above is editable in the live Defect Classification Accuracy calculator, which recalculates instantly and can be shared with the inputs intact.

Last reviewed 2026-05-12.