How do you calculate defect detection rate?

Divide the defects the vision system correctly detected by the total actual defects in your validation sample, then multiply by 100. With 94 of 100 known defects caught, that is 94 / 100 x 100 = 94%.

Is detection rate the same as recall?

Yes. In machine vision and quality, detection rate, recall, sensitivity, and true-positive rate all describe the same thing: the fraction of real defects the system finds. The complement, the missed fraction, is your escape rate.

What is a good defect detection rate?

Most production vision specs require 95% to 99%+ depending on defect criticality. The example's 94% falls 4 points short of a 98% spec, which would typically block release until the model is retrained or thresholds are adjusted.

What does the gap to specification mean?

It is your required detection rate minus your current rate. A 98% requirement against a 94% actual leaves a 4-point gap, meaning the system misses 4 in every 100 defects more than your spec allows — a clear action item before production.

Why use a validation sample instead of live data?

Live production rarely gives you ground truth on how many defects actually existed. A validation sample with a known, confirmed defect count lets you measure true detection rate, because you know the denominator exactly.

Detection rate vs false accept rate — how do they relate?

They are complements on the defect population: a 94% detection rate means a 6% miss rate among the defects in your sample. Detection rate looks at how many defects you caught; false accept rate (and false accept cost) looks at the consequences of the ones you missed.

Machine Vision & Industrial Inspection AI calculator

Vision Defect Detection Rate Calculator

Defect detection rate is the recall of your machine vision system — the share of genuine defects it actually catches in a controlled validation sample. Vision engineers and quality teams use it as the primary acceptance metric when validating a new inspection model, because it measures the one failure mode customers care about most: missed defects. Reporting it against a required specification turns a raw model score into a pass/fail decision the plant can sign off on. It is the number that gates whether a vision cell goes into production or back to the lab.

What this calculator does

Calculate the defect detection rate (recall) of a machine vision or AI inspection system by comparing the number of defects correctly detected to the total number of actual defects in the inspected population.
Use it when validating a vision system against a known-defect sample set and you need to calculate detection rate and compare it to your system specification or customer requirement.
It computes the percentage of known defects the vision system correctly detected and the point gap to your required detection specification.

Formula used

Defect detection rate = defects detected / total actual defects x 100
Gap to specification = required detection rate - current detection rate

Inputs explained

Defects correctly detected by vision system:
Total actual defects in validation sample:
Required detection rate specification:

How to use the result

Use it during model validation, periodic vision-system audits, or after retraining to confirm the system still meets its detection target.
Detection rate ignores false rejects entirely — a system can hit 100% detection by flagging everything, so it must be read alongside a false-reject or precision metric.

Common questions

How do you calculate defect detection rate? Divide the defects the vision system correctly detected by the total actual defects in your validation sample, then multiply by 100. With 94 of 100 known defects caught, that is 94 / 100 x 100 = 94%.
Is detection rate the same as recall? Yes. In machine vision and quality, detection rate, recall, sensitivity, and true-positive rate all describe the same thing: the fraction of real defects the system finds. The complement, the missed fraction, is your escape rate.
What is a good defect detection rate? Most production vision specs require 95% to 99%+ depending on defect criticality. The example's 94% falls 4 points short of a 98% spec, which would typically block release until the model is retrained or thresholds are adjusted.
What does the gap to specification mean? It is your required detection rate minus your current rate. A 98% requirement against a 94% actual leaves a 4-point gap, meaning the system misses 4 in every 100 defects more than your spec allows — a clear action item before production.
Why use a validation sample instead of live data? Live production rarely gives you ground truth on how many defects actually existed. A validation sample with a known, confirmed defect count lets you measure true detection rate, because you know the denominator exactly.

Last reviewed 2026-05-12.