Labels that print clean but apply crooked. Labels that apply straight but miss a copy. Codes that print but smudge. Each is a different failure with a different fix — and each is invisible to a single-threshold inspection. Multi-stage AI catches them as separate diagnostics.
An applicator misalignment isn't a printer problem. A smudge isn't an applicator problem. A missing copy isn't a quality problem — it's a feeder problem. Treating them all as 'label fail' means the wrong department gets the alert. Multi-class detection assigns the right diagnostic so the right fix happens fast.
Most existing label cameras do one thing: master-image comparison with a single threshold. The result is either too sensitive (false reject from a printer wear pattern) or too lax (real defects pass). We retrofit your camera with a multi-class AI model trained on your actual label and your actual defect patterns. Existing optics stay; only the analysis is upgraded.
The IX-AS/01 with our label inspection software is purpose-built for this. 12 MP resolves the smallest text and barcodes; coaxial diffuse lighting eliminates label glare; an integrated AI model is pre-trained on common substrate + ink combinations. Typical artwork onboarding takes one shift, not one week.
High-resolution capture at the moment of inspection — every label imaged.
Compare against current master. Position, rotation, and pixel-level print quality measured separately.
Multi-class AI: which kind of failure is this? Missing, crooked, smudged, faded, wrong artwork.
Verdict + class to PLC. Defects logged with image + class for the right department to action.
Tell us your line, your existing equipment, and your tolerance window. We'll come back with a retrofit assessment, a greenfield specification, and a rough cost estimate within two business days.