Unifons Vision · machine vision

Zero wrong labels out the door.

A wrong pallet label means returns, angry customers and searching. Vision reads every label with neural networks trained on your own labels: and lets nothing wrong slip through unnoticed.

Unifons Vision: review queue with two engines, verdict and audit log
99.99%+ read reliability 0 misreads let through, validated < 20 s per pallet, two sides offline on your own network how we measure →

How it works

THE SCAN HEADA moving head with an industrial camera and polarised light scans a stationary pallet from two sides, in under 20 seconds. No glare, no foil reflection, no missed corner.
THE NETWORKNeural networks trained on your own labels and products: not on stock photos. The system knows your fonts, your printers, your wrinkles.
THE VERDICTEvery read passes five control gates (below). Certain is certain; anything uncertain goes to a human, image included.
THE RECORDSEvery decision, including human corrections, sits in a full audit trail. You can account for every pallet afterwards.

The five control gates

Our principle: a wrongly read barcode that gets through is worse than a missed one. So every read must pass five independent checks.

01

Within one image

If reads within the same image contradict each other, the read is uncertain.

02

Between images

The geometry of the rig is measured: a label must reappear between shots where it should be.

03

Two independent readers

Two different reading engines must agree, or a second shot must confirm.

04

Second read on doubt

If one engine misses something, a targeted re-read happens before any verdict.

05

Checked against the order list

Every read label must appear on the expected list. If the list itself is wrong, the system fails closed, not open.

A human where it matters

Anything uncertain goes into a queue: the operator sees the image, corrects or confirms, and the line keeps moving. A held read is not an error: it's exactly what the system is supposed to do. Both numbers, automatic and held, we discuss openly.

For the deep reader

Vision is a standalone machine: industrial camera on a moving axis, polarised lighting, its own computing. It runs fully offline on your own network: no internet connection needed and your production data never leaves the premises.

Validation happens on real production runs: every read compared afterwards with what was actually on the pallet. Result so far: zero wrong labels let through.

neural network on your own labels · five gates · audit trail · offline appliance · measurement method →

No more wrong labels out the door?

The Source Scan shows what there is to gain on your line. €950, fully credited if you continue. If we can't find any concrete improvements, the invoice is waived.

Book a Source Scan

Reinier Weerts

founder, hands-on technologist

r.weerts@unifons.nl

you get a reply from a human: no forms, no ticket numbers