Case study
Automated label inspection
A poorly readable label can make it all the way to the customer. A camera on the line checks every single item automatically.
The problem
Manufacturing and distribution companies work with dozens to hundreds of products that need labels. Every label must contain specific information and a barcode or QR code. These labels are printed in high volumes on the packaging line, but print quality fluctuates — the code or text can come out blurred, hard to read or damaged.
The problem escalates with production volume: the more labels are printed, the higher the chance a defective one appears among them. An employee has to check the labels manually to verify that they are legible and that the code can be scanned.
What this causes
- Random and inconsistent inspection — only a sample of labels is checked, not every item.
- Defective labels in circulation that only surface when scanned at the customer or in the warehouse.
- Delays in production, because a defective batch has to be traced and reprinted afterwards.
- Complaints and returns when the customer cannot scan the barcode or QR code.
- Dependence on an employee's attention rather than on a systematic check of every label.
What the client gains
- An estimated 85–95% time saving on label inspection compared to manual processing.
- A check of every printed label, not just a random sample.
- A defective label is caught right on the line, not at the customer or in the warehouse.
- Fewer complaints and returns caused by unscannable barcodes and QR codes.
- A shift from sample checker to a role where the system does the checking and the employee only handles flagged cases.
How we solve it
A camera on the packaging line
We install a camera on the packaging line that automatically captures the label on the product right after it is applied.
Assessing code readability
From the image, the system evaluates whether the barcode or QR code is legible and can be scanned.
Signalling the result
If the label is fine, the system gives a green signal and stores an OK record. If it detects an error, it triggers a red signal and a sound so the operator can check or remove the product.
Recording to a digital log
Every result (both OK and FAIL) is continuously stored in a digital log for traceability and auditing.
Further uses
- Extending the check to other label elements (expiry date, batch, serial number)
- Print error statistics by printer, line or shift
- Automatic reprinting or replacement of a defective label without operator intervention
- Extending the camera system to check other types of codes (RFID tags, serial numbers on other packaging)
Technologies
Python
OpenAI
SharePoint
Azure
Case studies
You might be interested
Znalostní asistent pro finanční oddělení
Odpovězení na dotazy z předpisů s citací a vždy z aktuální verze.
Claims handling
AI drafts the decision rationale and verifies every reference to the contractual terms.
Invoice approval automation
Invoices that match the purchase order are approved automatically, with no manual intervention.
