Iberian ham slicing in a white room with AI — fat, bone and marbling slice by slice, pack weight intact.
In white room slicing every slice counts — a hidden bone punctures the pack, and badly judged marbling throws the grade off. iLEAN Vision detects fat, bone and marbling slice by slice and proposes the cut adjustment. The line keeps the operator's rhythm — the AI proposes, the person validates.
Every slice is a pack, and every pack is a customer.
White room slicing is the last process before Iberian ham reaches the shelf. It is where the margin is concentrated — and where a variable piece has to be turned into a product that must be uniform:
- Every ham is different — size, marbling, firmness, bed. The slicer operator adjusts by eye, and the customer's pack can come out a little thicker, a little fattier, a little off-weight depending on the piece and on how far into the shift it is.
- A hidden bone punctures the pack — a splinter the operator did not spot tears the packing film, spoils the modified atmosphere and forces the pack to be scrapped (or, worse, it reaches the customer).
- Marbling sells — or lies — a pack labeled “100% acorn-fed” with grain-fed marbling is a regulatory risk (RD 4/2014) and a blow to the brand. An acorn-fed pack with a poor slice lowers the average ticket and the retailer's rating.
- The line cannot be stopped — and sampling-based quality control lets through the deviations that happen between one sample and the next. The classic system works 99% of the time. That 1% is the complaint that arrives from the UK.
The problem is not knife skill — it is that every piece calls for a different cut and the operator cannot recalculate that for every slice without slowing the line down.
iLEAN Vision does not replace the master slicer — it puts another pair of eyes over the belt.
The white room is not the place for a lights-out factory experiment: it is exactly where it makes most sense to assist and simplify the operator's work, not automate it. iLEAN acts as the putty between the slicer, the packing system and the grade data in the ERP: it fills the gap where the decision used to be made “by eye” and is now made with data.
Edge sees the slice before packing. It proposes the adjustment. The operator validates without breaking rhythm. The agent cross-references marbling against the declared grade. If something does not add up, it holds the pack. The person signs.
The iLEAN pieces applied to white room slicing:
- iLEAN Vision (Edge) — a terminal with machine vision (CNN) over the belt, before the slicer and/or over the sealed pack. It detects fat/lean, hidden bone or cavities, and marbling slice by slice. It fires an actuator (stack light, rejector) in milliseconds if there is a critical defect. Sanitary IP69K enclosure. It works without a network: if the line loses WiFi, Edge keeps looking.
- iLEAN Connect — captures the batch recipe from the ERP, the commercial grade (acorn-fed / field grain-fed / grain-fed), the target pack weight and the SKU. And it captures what arrives from outside: an order change from the importer by email, a new rule from the UK retailer by WhatsApp, a recipe tweak from the R&D manager in a spreadsheet.
- Agent — cross-references the marbling actually measured with the declared grade and the piece of origin. If the piece is yielding marbling below the standard for its grade, it says so before the pack is sealed. It never shuts the line down on its own — it proposes a hold; the person validates and signs.
Slicing by eye vs. slicing assisted by iLEAN Vision
| Aspect | Manual slicing / sampling | With iLEAN Vision + Connect + Agent |
|---|---|---|
| Pack weight | Varies by piece and by shift | Steady: adjusted piece by piece |
| Hidden bone detection | The eye + a complaint afterwards | Before packing, in line |
| Marbling vs. grade consistency | Trust in the declared grade | Real marbling cross-referenced with the label |
| Quality control | Sampling | 100% of slices and packs |
| Packs scrapped for film puncture | Cost taken as a given | Reduced by detecting bone earlier |
| Traceability piece → pack → customer | By batch, not by piece | By piece and by pack |
Impact estimate for your plant — to be validated with your numbers.
The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- Iberian slicing white room, several slicers and a packing line, multi-SKU packs (60 g to 120 g), a mix of acorn-fed and grain-fed, exporting to UK/US retailers.
- Edge pilot on one slicer (camera over the belt + integration with the ERP/MES and the pack recipe + actuator). First value expected within a few weeks: the camera starts calibrating marbling and pack weight from day one.
- Indicative payback between 4 and 9 months, depending on your current cost of packs scrapped for bone/film puncture and on how much pack weight varies against the retailer's target.
- Hard levers: a reduction of ≥ 30% in scrapped packs, pack weight held inside a tighter tolerance (less overweight given away to the customer), marbling/grade consistency that stands up to an audit. Conservative figures, a defensible floor.
And the quality manager's reasonable doubt
“What if the camera mistakes good marbling for bad and throws valid slices away?” — hallucination is a problem of free generation, not of anchored tasks. When the AI limits itself to measuring fat/lean and the marbling pattern on a real slice and cross-referencing it with the declared grade, the best models brought error below 1.5%[1]. And even so, the camera proposes, the person signs: rejecting a slice or holding a pack is never done by the system alone. The three safety rings exist precisely for this.
[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.
What people ask about Iberian ham slicing in a white room with AI
What does AI vision detect when slicing Iberian ham — fat, bone, marbling?
iLEAN Vision detects, slice by slice, (1) the lean/fat ratio, so the look of the pack stays in line with its commercial grade; (2) the presence of bone or cavities before they puncture the pack or turn up in the customer's tray; (3) marbling (intramuscular fat infiltration), which is what tells an acorn-fed slice from a grain-fed one. And it returns, for each piece, a map of what every zone will yield once it is sliced.
How is the cut adjusted without stopping the slicing line?
The camera is mounted over the belt, before the slicer and/or over the sealed pack. In milliseconds it calculates the expected pack weight and the lean/fat/marbling mix, and proposes to the operator a cut adjustment (angle, thickness, rejection of a defective slice) that the packing system can receive. The line does not stop — the operator validates the proposal without breaking rhythm. The cut follows the operator's rhythm, not the other way round.
Does iLEAN Vision meet food-grade white room requirements?
Yes. The Edge terminals and the cameras are mounted in IP69K enclosures that withstand pressure washing and disinfectant, with no textile parts and no gaps where biofilm can build up. The wiring runs through sanitary conduit. And the electronics live outside the product flow, looking inward — exactly like any other sensor accepted in a food-grade slicing white room.
Is the pack weight held steady when the piece varies?
That is the whole point. Every Iberian ham is different: different size, different marbling, different bed. Without a system, the operator adjusts “by eye” and the customer's pack can come out at 80 g, 82 g or 78 g depending on the piece and on how tired the shift is. With iLEAN Vision, the system holds the pack weight because it adjusts the cut piece by piece, not by a fixed recipe.
How does it connect with grade labeling (acorn-fed/grain-fed)?
The real marbling measured by the camera is cross-referenced with the batch's declared commercial grade (100% acorn-fed, 50% acorn-fed, field grain-fed, grain-fed) in the ERP, and the agent verifies that the pack heading into packing is consistent with the label it carries. If the piece yields marbling below what is declared for that grade, the system holds the pack before sealing and alerts the quality manager. The person signs. Nothing reaches the market under the wrong grade.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your slicing white room.
We work on your plant's real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See food industry