Cosmetic defect detection
Colour deviation, finish variation, print quality, contamination.
Cosmetic quality is judged by the customer, not by a tolerance table. A colour that drifts, a finish that turns matte in one area, a print that smears — none of these stop the product working, and all of them generate complaints and returns.
This is exactly where fixed colour thresholds fail. Acceptable colour is a range, that range shifts with material batch and lighting, and the difference between an acceptable and an unacceptable deviation is often smaller than the variation the system already tolerates.
FOCUS™ AI learns the acceptable appearance envelope from real production output, including the variation your process legitimately produces. Deviations are assessed against that envelope rather than against a single reference value, and the decision is returned in real time so the part can be routed before more value is added.
What we inspect
- Colour deviationShifts too subtle for threshold-based colour logic.
- Finish variationGloss, matte and texture inconsistencies across a surface.
- Print qualityRegistration, smearing, missing elements and legibility.
- ContaminationMarks, residue and foreign particles on the visible surface.
How we get there
The first three steps of our deployment process. Nothing is promised before feasibility has been assessed.
- 01
Show us the challenge
Share your products, defect types, production environment and inspection requirements. Physical samples, images or video — whatever best represents the problem.
- 02
Validate feasibility
Our engineers analyse whether FOCUS™ AI can reliably identify the required deviations at production speed. Honest assessment — including the cases where we recommend a different approach.
- 03
Train FOCUS™ AI
Build and optimise the inspection model using representative production data — real defects, real variation, real production conditions.
ZERO DEFECTS IS THE TARGET.
ZERO RECALLS IS THE RESULT.
ZERORECALLS.AI
Vision that never loses focus.
Have an inspection challenge? Put our Vision AI to the test. We will tell you whether we can solve it — and if we can, we will show you how fast.
