Grading Skrei into A and B cannot depend on how tired the last shift of the season is.
Grading Skrei cod fillets into A and B means cross-referencing color, firmness and surface integrity at line speed — and doing it every year inside a short window. iLEAN Vision reads every fillet on the belt with a CNN, proposes the grade, and an experienced operator signs off the batch that goes to premium European foodservice.
Three visual parameters, two shifts, one short season window.
Skrei concentrates the bulk of the revenue into a few weeks a year. Every fillet that comes off the belt has to be graded A (premium foodservice) or B (mid-range foodservice / processing) against three realities that are hardly ever measured with the same rigor from one shift to the next:
- Color — Skrei flesh has a characteristic pearly white. The veteran operator reads it at a glance; the operator on the last shift of the week, more tired, lets nuances slip past.
- Firmness — a freshly cut fillet behaves differently from one with another 24 hours in the chiller. Firmness is what shows up most on the palate, and what is hardest to inspect by eye at belt speed.
- Surface integrity — bruises from the catch, tears, badly removed skin. Small defects that change the declared grade and, with it, the margin.
Sending grade B at grade A prices gets charged back fast by the Michelin customer. Sending grade A at grade B prices gets charged quietly, on your own invoice. The classic system works 99% of the time. That 1% is the season's margin.
iLEAN Vision over the belt — the veteran operator of the last fillet, on every fillet.
The Skrei grading problem is not a lack of judgment: it is human variability sustained across a short, high-volume window. iLEAN acts as the putty between the veteran's eye and a belt running at season speed, without asking you to change your seafood ERP or your filleting line.
Edge sees every fillet before packing. Connect captures the batch and vessel traceability. An agent cross-references the proposed grade against the foodservice destination and, if there is doubt, calls the operator. The person signs — never the other way round.
The three iLEAN pieces applied to Skrei fillet grading:
- iLEAN Vision (Edge) — a terminal with a CNN over the belt in the chilled room, with calibrated lighting. It reads color by pixel, infers firmness from the specular response of the fillet and detects surface defects. It proposes an A/B grade in milliseconds. It works with no network: if the plant loses WiFi, the belt keeps being graded.
- Connect — captures the season recipe, the batch, the catching vessel, the date and the foodservice customer's details, whether they come from a modern ERP or from the season manager's spreadsheet. And it captures what comes from outside (an email from the European importer with a new rule, a WhatsApp message from the port about a change in the catch) at second zero.
- Agents — they cross-reference the proposed grade against the end customer and the season's commitments; if a batch ends up compromised (too much B where A was promised, for instance), they alert the plant manager so it can be reallocated before closing, not after.
Manual grading by the veteran vs. assisted grading with iLEAN Vision
| Aspect | Manual grading | With iLEAN Vision + Connect + Agent |
|---|---|---|
| Consistency across shifts | Depends on the operator's eye and fatigue | The same criterion on every fillet, every shift |
| Belt speed | Drops at the end of the shift whenever there is doubt | Holds season speed without losing grade accuracy |
| Reading the color | Subjective, with variable lighting | Pixel calibrated against the light box |
| Detecting a faint bruise | Slips through when the line is rushed | Flagged by the CNN, grade B proposed |
| Operation without network | n/a | Edge keeps grading on the panel's own power |
| Per-batch declaration for the importer | Rebuilt by hand, days | Automatic dossier with the fillet photo and traceability |
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.
- Skrei plant with one seasonal filleting belt and exports to premium European foodservice. Volume concentrated in a short window.
- Edge pilot over the belt (terminal + camera + calibrated lighting + integration with the seafood ERP/vertical system). First value expected within a few weeks, inside the season itself.
- Indicative payback between 4 and 9 months, depending on volume and on the documented margin lost to inconsistent grading between shifts.
- Reduction in grading errors between shifts ≥30% (a conservative estimate). The hard lever is margin recovered through consistency, not "faster" — the belt does not need to run faster, it needs not to get it wrong on the last shift.
And the plant manager's reasonable doubt
"What if the AI gets it wrong and sends a grade B fillet to the premium customer?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely reads what a sensor sees and compares it with a learned pattern (color, surface defect), the best models brought error below 1.5% [1]. And even so, what is critical is never decided alone: iLEAN proposes the grade and the operator signs. 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 grading Skrei with vision
What separates a grade A Skrei fillet from a grade B?
Skrei is migratory North Atlantic cod caught in a short winter window, when it reaches the Norwegian coast to spawn — a premium product with an origin seal and a very narrow quality window. Grade A calls for pearly white flesh, high firmness to the touch, no bruising or tearing, a whole loin and uniform thickness: it is the fillet that travels to a European Michelin restaurant. Grade B is perfectly wholesome product with one deviation (a small bruise from the catch, slightly lower firmness because of the days since landing, a marginally irregular cut): it goes to mid-range foodservice or to processed products. The line between them is drawn by three visual parameters — color, firmness and surface integrity — read in seconds by an expert. The classic system works; what changes with AI is that consistency across operators and across shifts stops depending on fatigue.
Why does the Skrei season demand such fast grading?
The Skrei season lasts only a few weeks a year and concentrates the bulk of the plant's revenue. If you lose belt speed while grading, you lose volume; if you lose precision, you send grade B at grade A prices (margin lost with the premium customer) or grade A at grade B prices (margin lost on your own invoice). And the big European buyers are auditing the traceability of the declared grade in ever finer detail. iLEAN Vision reads every fillet on the belt — color, firmness inferred from the specular response, and surface integrity — and proposes the grade instantly; the experienced operator signs. It holds season speed and consistency across shifts.
How does iLEAN Vision read color and firmness without touching the fillet?
Edge (iLEAN Vision) is a terminal with a camera and a CNN trained for the case, mounted over the belt in the chilled filleting room. Color is read straight from the pixel, calibrated against the lighting of the light box itself. Firmness is not measured directly — it is inferred from the specular response of the fillet (a fresh fillet reflects differently from one that has spent two more days in the chiller) and from how the relief behaves. Surface integrity (bruises, tears, badly removed skin) is exactly what a CNN does best. The system proposes a grade; the person validates and signs. It works with no network: if the plant loses connectivity, the terminal keeps grading, because what is critical cannot depend on a router.
Can grading be integrated with the seafood ERP and the foodservice customer?
Yes — that is the piece iLEAN Connect brings. Every fillet graded and signed off by the operator is linked to the batch, the catching vessel, the catch date and the end customer through the ERP or the vertical seafood system you already run. If the European foodservice importer asks for a per-batch declaration (grade A with a certified Skrei seal), the system generates it with the photo of the fillet and the traceability, with nobody rebuilding anything by hand. And if the customer changes the rules mid-season by email, Connect captures it at second zero and the Agents adjust the threshold.
What does a vision-based Skrei grading pilot cost in a mid-sized plant?
The order of magnitude of an Edge pilot over a filleting belt is that of any food-industry Edge pilot: an initial investment covering the terminal, camera, calibrated lighting and integration with the seafood ERP/MES, plus an annual licence. The indicative payback (an estimate to be validated) is between 4 and 9 months, depending on season volume and on what grading errors cost you today. The hard lever is twofold: margin recovered through consistency across shifts, and a per-batch declaration the foodservice buyer can audit. We ask for your plant's data and send you the estimated ROI in 48h, with your numbers.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your Skrei plant.
We work on your plant's real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See the food industry