Skip to content
LexDataLexData
PlatformIndustriesCustomers
DocsThe Field GuideBlogWhy models drift
AboutCareersSecurityContact
Log inStart now

// MANUFACTURING

Pallet Scanning

Pallet Scanning. What it takes to run it.

The shipping label boxed on whichever of four faces the wrap left it, at the dock door, held for a person when the film crosses it.

Start with your footageBook a demo →All of manufacturing →

// 01 · The question

“Which face carries the label on this pallet, and what does it say, as it crosses the dock door?”

// 02 · Why it is hard

A pallet built at one supplier carries its label on the short side at chest height, and a pallet from another has it low on the long side under the film. A mixed pallet from a cross-dock carries three labels from three legs of its journey, and only one of them is the one this dock needs today. None of that is a rule that holds from the last pallet to the next, which is why four cameras look at every face rather than one camera watching where the label is supposed to be.

// 03 · What the labels have to be

Bounding box

Decided before the first frame

The label can land on any of four faces at any angle under the wrap, so each frame needs a box found rather than assumed at one fixed region. A second class, wrapped label, is kept separate because a label the film has crossed is a different problem from one that has not.

This is the decision that is expensive to reverse. The geometry has to match what the answer contains, and finding out it does not means labelling the set a second time.

// 03a · What it finds

The shipping label

boxed on whichever face the wrap left it, from cameras covering all four sides of the door

A label the film has crossed

boxed as its own class, wrapped label, so the dock knows which pallets need the digits read by eye rather than decoded off the bars

The barcode and the printed text once the box is found

read by asking a question of the crop, the same way a shelf tag's price is read off its own box

A read the system doubts

a label at a slant, half in shadow or under a heavy fold, held for the clerk with the frame rather than logged either way

// 03b · What it checks

  1. 01

    A label is found on at least one of the four faces the cameras cover before the pallet is logged through the door.

  2. 02

    The decoded number, or the text read when the bars will not decode, against the receipts the dock expects today; a number nothing matches goes to the dock office with the frame.

  3. 03

    A read the system doubts, through film, at a slant or half in shadow, is held for the clerk with the crop rather than scored either way.

  4. 04

    Every pallet through the door keeps the frames from all four cameras with its record, for the dispute that comes three weeks later.

// 03c · What the line provides

  • Two cameras at label height on whichever sides the security cameras do not already cover; everything else is the dock's own cameras and recorder.
  • Frames from this door's own pallets, not a generic label set, since the label lands wherever each shipper's dock crew left it.
  • The day's receipts or advance ship notices, in the dock's own system, for the match to run against.
  • The clerk who already walks the floor, now looking at the frames the system held rather than every pallet that comes through.

// 04 · How we run it

Three parts of one loop, on this job.

LexAnnotate

Defect Annotation

Pixel-level defect annotation for surface scratches, cracks, dents, weld faults, and assembly misalignments. Multi-sensor alignment for multi-camera inspection systems.

LexInsight

Line Performance

Ask which shift or which line is producing the most of one defect class, and get an answer with the frames behind it, graded against the plant's own written standard.

LexAlert

Quality Alerts

Monitor defect detection accuracy across changing production conditions. Detect when lighting shifts, new materials, or equipment aging degrades model performance. Auto-retrain to maintain 99%+ accuracy.

How the loop fits together →

// 05 · What breaks it after launch

Covariate shift

Something is now in the way

→Covariate shift →

A wrapper that winds one more turn than usual puts a fold of film across the barcode on the next pallet through, and the frame is fine, the label is found, and the bars will not decode. A model retrained on frames of wrapped labels only learns to expect the fold; it does not read a crease any better for having seen more of them.

It is not the only one that can get this use case, it is the one that usually gets it first. All ten conditions.

// 06 · In manufacturing

Defect Detection at Scale

Full write-up ↓Every case study →

The Challenge

Vision models performed well in pilots but lost accuracy as factory conditions shifted.

Our Execution

Built pixel-level annotated datasets and multi-sensor alignment for reliable defect detection.

The Result

Maintained 99%+ defect detection accuracy across changing production environments. Reduced unplanned downtime.

99%+

Accuracy maintained in production

// 06a · Questions we get asked

Does this replace the handheld scanner?

No. It takes the pallets that would otherwise each stop at the door for a walk around with a handheld scanner. The handheld stays in the dock office for the exceptions: the pallet with three labels where the rule picked wrong, or the one that came in through the personnel door.

Can it read the barcode through the stretch wrap?

Through clear, flat film, usually. Through a fold or a crease across the bars, no, and nothing reads a barcode through a crease. The printed digits beside the bars survive a crease more often than the bars do, and those are read instead; what neither the camera nor a barcode reader can make out goes to the clerk with the crop.

Which label counts on a pallet that carries three?

The one whose number matches a receipt the dock expects today. The other two are logged with the frame, because the dispute three weeks from now is sometimes about the leg of the journey the dock did not know it was on.

Does the dock need new cameras?

Usually two, one each side of the door at label height, because the security cameras were mounted high to see faces and a label at knee height on the far side reads as a smear from there. Everything else is the cameras and the recorder the dock already has.

Read next

Pallet scanning at the dock door, the full walk-through →The same find-then-decide shape, on the packing line instead of the dock →What a fold of stretch film does to a barcode the camera can still see fine →
Start with your footageBook a demo →

// 07 · Also in manufacturing

Surface Defect Detection→Weld Quality Inspection→Assembly Verification & Error Proofing→Package & Label Inspection→Date & Lot Code Inspection→Predictive Maintenance→Multi-Sensor Monitoring→Tablet Defect Inspection→Syringe Barrel Defect Detection→Foreign Material Detection→Glass Defect Detection→Semiconductor Wafer Inspection→Strip Surface Defect Inspection→Ceramic Defect Detection→Wood Defect Detection→Pipe & Tube Defect Inspection→Body-in-White Inspection→Contact Lens Defect Inspection→Coffee Bean Defect Inspection→Parts Count Verification→All forty-one→
Surface Defect DetectionWeld Quality InspectionAssembly Verification & Error ProofingPackage & Label InspectionDate & Lot Code InspectionPredictive MaintenanceMulti-Sensor MonitoringTablet Defect InspectionSyringe Barrel Defect DetectionForeign Material DetectionGlass Defect DetectionSemiconductor Wafer InspectionStrip Surface Defect InspectionCeramic Defect DetectionWood Defect DetectionPipe & Tube Defect InspectionBody-in-White InspectionContact Lens Defect InspectionCoffee Bean Defect InspectionParts Count VerificationAll forty-one →

Send us a week of this footage. We’ll show you what comes back.

Start with your footageBook a demo →
LexData
LexData

Product

  • Platform
  • Industries
  • Use cases

Resources

  • Docs
  • The Field Guide
  • Blog
  • Why models drift

Industries

  • Energy & utilities
  • Oil & gas
  • Agriculture
  • Manufacturing
  • Insurance
  • Retail
  • Robotics

Company

  • About
  • Customers
  • Careers
  • Contact

Trust

  • Security
  • Privacy
  • Terms

Stay updated

What we learn running vision models in production.

See everything.
Miss nothing.

Stay updated

What we learn running vision models in production.

Terms of use & Privacy policy

© 2026 LexData Labs · All rights reserved