KOMPASS brings label wrinkle inspection to every applied label, along with checks for skew, bubbles, lifted edges and flagging. AI models trained on your labels separate real faults from glare and print on glossy film, and the line is signalled before poorly applied labels ship.
A crooked, creased or lifting label makes a good product look faulty. On a fast line, these faults are subtle in a single glance and repeat until someone stops the labeller.
Our AI Product Inspection system, KOMPASS™, combines lighting designed for glossy labels with AI models trained on your own containers, so label wrinkle inspection separates real creases, bubbles and lifting from print and glare. Each container result is stored with the image that triggered it.
Jidoka's labelled-bottle demonstration shows KOMPASS detecting critical visual defects on labelled bottles in-line, faults that are hard to filter out manually at high speed, with automated ejection of rejected bottles.
Multiple lights or angles make creases, bubbles and lifted edges visible on glossy and matte labels. Round containers are viewed from several sides.
The model checks label angle and surface, then classifies skew, wrinkles, bubbles, tears and lifting against your acceptance standard.
Failing containers are ejected, defect trends point to the labeller cause and every result is stored with its image.
Send us containers with typical label faults and your acceptance standard. We will assess the lighting and views your label wrinkle inspection needs.
Get In TouchEach method needs a view of the label surface. The difference is how label wrinkle inspection separates real faults from print and glare.
For Diageo, KOMPASS on Tigris hardware checks label condition, text and orientation in-process and ejects faulty bottles automatically.
Diageo's labelled-bottle results cover damaged and defective labels as part of the complete inspection.
Label wrinkle inspection detects the faults that come from application: skew, wrinkles, creases, air bubbles, lifted edges, flagging, folds and tears. Each is classified separately, so you can set different limits for each one and see which fault is most common. Print defects on the label itself are covered by the misprint inspection use case and can share the same station. The same image can also confirm the label is present and positioned correctly.
Wrinkled labels and bubbles create only faint changes in brightness, while the printed artwork around them has strong contrast. Comparing every pixel with a reference image confuses the two. Multi-angle lighting makes surface changes stand out, and an AI model trained on real examples learns what a crease looks like across different artwork, which reduces both misses and false rejects. That combination matters most on premium packs, where small creases are the main complaint.
Glossy film behaves like a mirror, so a point light creates highlights that can look like creases. Diffuse or carefully angled lighting reduces those reflections, and training includes acceptable labels with normal highlights so the model learns to ignore them. Matte and textured labels are lit differently, and both can run on the same line. Lighting is fixed once set, so every container is seen under the same conditions.
Label skew is the angle between the label edge and the container's reference line, such as its base or shoulder. The station measures it on every container and compares it with your limit. Crooked labels often come from container rotation during application or a misaligned guide, and the trend shows whether skew is constant or growing over a shift. A growing trend usually points to wear, while a constant offset points to a setup error.
Label flagging happens when a label end lifts away from the container, usually on small diameters, cold containers with condensation or labels with stiff material. Low adhesive coverage and worn wipe-down brushes or rollers contribute too. Because flagging can appear after the label leaves the labeller, the camera is placed far enough downstream to see it develop. Moving the camera a short distance downstream often reveals faults that appear only after application.
Yes. Torn labels, scuffs and missing pieces are visible surface defects and are classified alongside wrinkles and bubbles. They often come from handling after labelling, such as guide rails, starwheels or container contact. Images of each failure help locate the point on the line where damage happens, which is often different from where wrinkles start. Separating handling damage from application faults sends maintenance to the right part of the line.
It should not, if the acceptance standard is set with your quality team. Small wrinkles from normal material flex can be acceptable, while a crease across text is not. Training includes borderline examples labelled by your team, and validation compares the station's decisions with theirs before automatic rejection is switched on. Agreed examples also help new operators understand the standard the station applies.
Send containers with typical label defects, examples of acceptable labels and your acceptance standard. Add the label materials, container shapes, line speed and where a reject can happen. With that, we can recommend lighting, camera views and training for label wrinkle inspection on your line. Containers from cold storage are worth including if labels are applied to chilled product.
Share containers with typical faults, your acceptance standard, materials and line speed. We will design label wrinkle inspection around your labels and explain timing and cost. KOMPASS usually goes live in 6 to 8 weeks.
