QualityVerified Real deployment at a named or precisely described company, with a measured result from a primary source.

AI reads the date code and checks the label on every pack at line speed, so the wrong label or code never ships (Ilchester Cheese, Del Monte Foods)

Added 2026-11-16 · Last updated 2026-11-16

Manufacturers in this case: Ilchester Cheese · Del Monte Foods · all manufacturers using AI

Role

Quality assurance and packaging (label, date code and lot code checks) · Typical users: quality assurance manager, packaging line supervisor, food safety or technical manager, operations director, continuous improvement engineer

The problem

On a packaging line, the wrong label, the wrong film or an unreadable date code can turn into a recall, retailer fines or a consumer health risk, and manual checks cannot keep up with fast lines. At Ilchester Cheese (UK, part of the Tine Group), the slicing line runs at 200 packs per minute; the operations director said that at that speed the operator simply cannot check packs individually, which they had been doing before. The business had over 200 SKUs, 75 formats and 10 production lines, exported to 21 countries, and relied on paper-based checks. At Del Monte Foods (US), more than a thousand cans of vegetables pass on the conveyor lines every minute, and the continuous improvement manager said the real danger lies in a labelling error reaching the consumer. At a seafood cannery packing several brands on the same lines, a code stamped on the can bottom, the UPC on the label and the carton barcode must all agree, and the existing code reader was so inconsistent that operators could not trust it. For context, undeclared allergens caused 15 of the 65 recalls of meat and poultry products by the US Food Safety and Inspection Service (USDA FSIS) in 2023, and mislabeling 1 more.

How AI helps

Cameras on the line photograph every pack, can or case, and a deep learning model reads the printed codes even when the font, position, print technology or contrast varies, then compares them with what the current job should carry. At Ilchester Cheese, OAL's APRIL Eye uses AI to verify date codes in real time, including international formats such as Canadian date codes, while cameras check that the right film and label are on every pack; every pack gets a digital record with a photo for audits, and operators get alerts and prompts on a simple dashboard. According to an OAL article, APRIL Eye was trained on half a million pictures of date codes in various fonts and print formats and stops the line when a pack carries an incorrect date code. At Del Monte Foods, the system built by Tensor ID (TensorVision360 software with Zebra cameras and fixed industrial scanners) checks in three stages: the variety code on each can before labelling, then each label, then that multipacks carry matching codes and labels; it is integrated with the plant's manufacturing execution system so a product changeover calls up the right codes and labels. At the seafood cannery, an integrator ran a deep learning reader trained on just three days of the plant's own can codes side by side with a conventional rules-based OCR camera for three weeks on the live line, then designed one system where the can code, label UPC and carton barcode are checked against each other from a recipe per product.

Step by step

  1. List every code and label element that must be right on each pack (date, lot, product code, allergen label, barcode, film) and where it is applied on the line.
  2. Pull the last one or two years of labelling and coding incidents, holds, complaints and retailer fines, and rank them by cost and risk.
  3. Start on the line with the worst record: run the camera in parallel with the manual check for a few weeks and sort every reject into real defect, unreadable print or false reject.
  4. Link the system to the production order or recipe so the expected codes load automatically at changeover instead of being typed by hand.
  5. Decide what happens on a fail (stop the line, reject the pack, alert the supervisor) and keep a photo record of every pack for traceability and audits.
  6. Review false rejects and missed reads every week with the supplier until they are stable, then extend to the other lines.

Tools

  • OAL APRIL Eye (AI date code verification) with label and film verification cameras: used at Ilchester Cheese
  • Tensor ID TensorVision360 with Zebra CV60 machine vision cameras, FS40 fixed industrial scanners and Aurora software: used at Del Monte Foods
  • Deep learning OCR reader with barcode verification, PLC and HMI integration (SolVIS): trialled at a seafood cannery
  • Other tools to evaluate: deep learning OCR tools from machine vision suppliers, the code verification option of your coder or labeller supplier, inspection modules of your line integrator
  • To start without new software: a list of your recent labelling and coding incidents and any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below

In the AI tools index: OAL APRIL Eye · Tensor ID TensorVision360 · all AI tools used by manufacturers

Result

Ilchester Cheese (OAL case study, the supplier): 50% increase in slicing line speed and capacity, the traceability exercise cut from about three hours to 10 to 15 minutes, and zero labelling or coding errors recorded since going live, with audit-ready records and pack-level photos. Del Monte Foods (Zebra success story, the hardware supplier): the system inspects 100% of labels and all the OCR print on cans; after its pilot year the company reported fewer labelling issues, reduced retailer fines and less paperwork and manual refocusing, without published figures. Seafood cannery (SolVIS case study, the integrator; plant not named): on one day of live production during the trial, 269,190 cans read with a false-reject rate of 0.0784% and no errors or skips detected; the deep learning reader handled the plant's print variation better than the rules-based camera, and the remaining rejects were real printing problems (codes rotated out of view or smeared ink).

As reported by the source; most figures are vendor or self-reported and not independently verified.

Human review: cautions

  • The figures come from the suppliers and integrator (OAL, Zebra, SolVIS) and are not independently audited; Del Monte published no figures, and the seafood cannery is not named.
  • The cannery result is one day of a three-week trial, not a long-term production average.
  • Ilchester's capacity gain also came from automating the checks and case packing, not from AI alone.
  • A camera only checks what it is told to expect: the expected codes must come from the production order or recipe, not from manual typing, or the system will confirm the wrong code.
  • Verification does not replace your allergen and label change control; keep artwork approval and line clearance procedures.

Try it yourself

You are a quality assurance manager in a food or consumer goods plant. I want to decide where automatic label and date code verification would remove the most risk.

I will paste: (1) a list of our packaging lines with speed (packs per minute), number of SKUs and how labels and codes are checked today; (2) our labelling and coding incidents for the last one or two years [date, line, product, what was wrong, how it was found, quantity held or recalled, cost or retailer fine if known].

Do this:
1. Group the incidents by type (wrong label, wrong film, missing or wrong date code, unreadable code, wrong barcode, mixed product) and by line, with count and cost.
2. Rank the lines by risk and say for each which checks a camera could take over and which still need a person.
3. For the top line, list what the system must check, what data it needs from production orders or recipes at changeover, and what should happen on a fail.
4. Draft a parallel trial plan: how long, how to classify each reject (real defect, unreadable print, false reject) and what false-reject rate we would accept.
5. List the questions to ask vision suppliers before asking for a quote.

Use only the data I paste in; if something is missing, write 'not found' or list it as a question instead of guessing. Do not send, order, post or change anything in any system: produce drafts for my review and approval.

Want more like this? Join the newsletter

Related cases

All 19 Quality use cases →

Sources

Weekly: 1–2 practical AI use cases for your plant

Newsletter in

Free, one email a week. By subscribing you agree to receive the weekly newsletter (and its sponsor message) from AI Agents for Manufacturing. Every email includes an unsubscribe link, and you can unsubscribe anytime. See our Privacy Policy.