KOMPASS brings label verification to every label on the line, checking it against the approved artwork for the product being run. AI reads the text, codes and graphics, confirms they belong to the active order and signals the line when a wrong label or outdated version appears.
A label for a sister product has the same shape, colours and layout. On a fast line nobody can read every word, and without label verification a barcode scan alone can pass a label that belongs to something else.
Our AI Product Inspection system, KOMPASS™, performs label verification on every pack: it reads text and codes, recognises artwork and compares the result with the approved label for the active order. Each decision reaches the line controls with an image of the label.
Jidoka's labelled-bottle demonstration shows custom AI models detecting label defects and reading printed text and artwork with OCR and OCV, so each label is checked for application quality and for compliance with legal and marketing content.
The expected artwork, text and codes for the running product come from the work order or recipe, so changeovers update the check.
The model reads text and codes and recognises artwork on every label, then compares them with the approved version.
Wrong or incomplete labels are rejected or held, and each result is stored with its image for audits and investigations.
Send us your label variants, the content that must be checked and how orders change on the line. We will assess the imaging and data link your label verification needs.
Get In TouchEach method needs a clear view of the label. The difference is how much label verification each one performs, and whether it knows which label should be there.
Diageo uses KOMPASS on Tigris vision hardware to check that label text, logos and orientation are right on every bottle at line speed.
Diageo's published results cover text, orientation and label defects across its labelled-bottle inspection, which includes the label verification described here.
Yes. Label inspection looks for physical faults such as smears, wrinkles or a misplaced label. Label verification confirms the content is right: the product name, variant, allergen statement, regulatory text, barcode content and the artwork itself, compared with the approved artwork for the product being run. A label fails if any element is missing, wrong or belongs to another product, even when it is perfectly applied. Each element can have its own rule, so a minor print variation does not fail a correct label.
The expected label comes from the production order, a recipe on the line controller or your manufacturing system. When the order changes, the reference changes with it. That removes the risk of an operator forgetting to switch an inspection template, which is one of the ways label mix-ups get through a line that already has a camera. Where no data link exists yet, a supervisor confirms the reference at the start of each run.
A scanner confirms that a barcode can be read. It does not confirm the barcode belongs to the product in the pack unless the decoded value is compared with the order. Even then, a label can carry the right barcode with the wrong product name or allergen line. Checking the decoded value alongside text and artwork closes that gap. Together, those checks confirm the label is right for the product inside.
Yes, when versions differ visibly or in their text. A changed regulatory statement, new logo or revised layout is part of what the model learns for the current artwork version. Old labels that remain in the warehouse after a redesign is a common source of errors, and it fails the check once the new version is the reference. Keeping one approved reference per variant also makes artwork changes easier to control.
Most begin at a roll changeover. The wrong roll of label stock is loaded, a partly used roll from the previous order stays on the applicator, or a few labels remain on the backing paper. A wrong label at the start of a run can be repeated on every pack until someone notices, so the first packs after a product switch deserve the closest check.
It can report gross problems such as a missing or badly misplaced label. Detailed placement, skew, wrinkles and print defects are covered by the packaging and label inspection use cases, and they can share a station with content checks. Combining them avoids adding another camera when one view can serve several checks. One camera can often serve both jobs where the label is clearly in view.
Food, beverage, pharmaceutical, medical device and personal care manufacturers carry the highest risk, because a wrong allergen line, dose or regulatory claim can trigger a recall. Artwork verification is also valuable wherever many variants share similar packaging, such as flavours, strengths or market-specific labels on one line. Contract packers running many brands on shared lines face the same risk every day.
Send examples of each label variant, including near-identical ones, the elements that must be checked and how orders and changeovers are managed today. Add the line speed and where a reject can happen. With that, we can assess camera placement, the order data link and training for label content verification on your line. Near-identical variants are the most useful samples, because they set how fine the distinction must be.
Share your label variants, critical content, order data and line speed. We will scope label verification for your line and set out timing and cost. Most KOMPASS projects are live within 6 to 8 weeks.
