Computer Vision for Quality Control.

Computer vision for quality control means using AI vision models to inspect products visually — catching missing welds, surface defects, wrong components, mis-orientations, packaging errors. The technology has matured to the point where useful QC inspection runs on phone-camera photos, not £20k industrial vision rigs.

The two flavours of computer vision in industry: classical machine vision (geometric measurement, dimensional checks, calibrated rigs) and AI vision (visual judgement, pattern matching, robustness to variation). They're complementary. Classical vision measures the part to 0.01mm; AI vision tells you whether the right hardware is fitted, the welds are present, and the finish looks right.

For SMB manufacturers and fabricators, the gap between current practice and modern AI vision is significant. Most SMBs do visual QC by eye, on dispatch, by a tired person at the end of a busy day. The defects that should never leave the building leave the building. Customers expect photo evidence anyway, so the input photos already exist — they just aren't being used as inspection, only as archive.

Production deployments are typically: define what "right" looks like per product type (reference photos + structured checklist); inspect every dispatch photo against the schema; pass clean parts straight through, flag uncertain ones for human review. Adoption is fast because the workflow already involves taking photos — the change is just that the photos now drive inspection rather than archive.