High FCR in seabass and seabream — every kilo of feed that sinks to the bottom is margin thrown away.
In open-sea seabass and seabream cages, every pellet the school does not eat is feed cost, seabed fouling and a high FCR that shows up in the year's results. iLEAN Vision reads the behavior of the school underwater and proposes the feed setpoint for each feeding session — the person signs.
Feed is the biggest cost of the operation — and the hardest one to fine-tune.
At a Mediterranean grow-out site, feed accounts for between 50% and 70% of variable cost. Every tenth of FCR saved is money in the year's results, and every tenth lost is money going to the seabed. And even so, the feed setpoint is usually decided from three blocks of information that almost never line up at the same time:
- The manufacturer's table — designed for laboratory conditions. Your site is not a laboratory.
- The feeder's fixed program — calibrated at the start of the cycle and reviewed by the technician whenever they can.
- The cage technician's eye — the best instrument you have, but they cannot be at all 12 cages at once, nor at 06:30 when the first feeding session starts.
The result: underfeeding (you lose growth) or overfeeding (you lose feed to the bottom, you foul the cage, you raise FCR, you leave organic residue that drives oxygen consumption up). The classic system gets 90% of the feeding sessions right. That 10% is the year's FCR.
iLEAN does not replace the technician — it gives them eyes where they cannot reach.
The FCR problem is not a lack of information: it is information living in islands and arriving late — the camera records but nobody watches it at 06:30, the O2 probe data sits in the ERP, the last manual sampling was two weeks ago. iLEAN acts as the putty that fills those gaps without asking you to change the feeder, the probes or the team.
Vision reads the school underwater in real time. Connect captures probes, feeder and the technician's observations. The agent cross-references them with the satiation pattern and proposes a setpoint. The person signs — the feeder does not change its own program.
The iLEAN pieces applied to FCR control in seabass and seabream:
- iLEAN Vision (underwater Edge) — encapsulated cameras inside the cage with a CNN trained to read the behavior of the school: average speed, vertical spread, hunting aggressiveness, uneaten pellets sinking to the bottom. It works out the moment of satiation and proposes reducing or cutting the setpoint. It works with no network: if the site loses its satellite or 4G link, Edge keeps calculating and storing, and syncs when the link comes back.
- iLEAN Connect — captures data from the automatic feeder, the O2 and temperature probes and the biomass samplings, whether they live in the ERP or in the biologist's spreadsheet, and also captures what arrives from outside (a weather alert from the supplier, the retailer's specification on harvest weight). Everything enters the system at second zero, without anyone forwarding anything.
- Agent — cross-references the reading of the school, the growth curve expected from the table, the day's oxygen and temperature data, and the cage's history. If the theoretical setpoint does not match real behavior, it does not send an email at 22:00: it proposes an adjusted setpoint and alerts the cage technician. The person validates and the setpoint goes into the feeder.
Fixed feeder program vs. adaptive iLEAN setpoint
| Aspect | Fixed program + the technician's eye | With iLEAN Vision + Connect + Agent |
|---|---|---|
| Reading the school | Camera recording, nobody watches the 06:30 session | CNN analyzes the satiation pattern in every session |
| Reaction to a hunger or satiation peak | Noticed at the end of the cycle, at sampling time | Setpoint recalculated within the session itself |
| Closing FCR | Monthly or per cycle, with one-off samplings | Daily estimate with visual biomass + actual feed |
| Feed on the bottom | Only visible in diver footage | Detected on the line, setpoint cut within the session |
| Operation with no connectivity | The feeder runs its recipe regardless | Edge keeps calculating and storing, syncs on return |
| Per-cage file for the auditor | Rebuilt by hand across several systems | Automatic dossier per cage and cycle |
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 site. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- Mediterranean site with several seabass and seabream cages, an automatic feeder, and a target FCR missed in at least one of last year's cycles.
- Vision pilot on two cages (underwater camera + integration with the feeder and with the site's O2 probe). First value expected within a few weeks.
- Indicative payback between 4 and 9 months, depending on the starting FCR, current feed cost and the number of cages that would be added after the pilot.
- FCR reduction of ≥ 5% as a defensible floor; at sites where the setpoint is poorly tuned, that floor rises. The hard levers are tonnes of feed saved per cycle and kg of biomass recovered per cycle.
And the site biologist's reasonable doubt
“What if the AI proposes stopping the feed while the school is still hungry?” — the Agents have no hands on the critical work order: they propose, the person signs, the feeder executes. And hallucination is a problem of free generation, not of anchored tasks: when the AI simply recontextualizes the measured behavior of the school against a satiation pattern, the best models brought error below 1.5% [1]. If confidence drops because of turbidity or jellyfish, the system does not propose — it alerts.
[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.
What people ask about FCR and underwater vision in seabass and seabream
How do you measure the real FCR of a seabass and seabream cage?
An honest FCR (Feed Conversion Ratio) is kg of feed delivered divided by kg of biomass gained over the cycle. The numerator is usually reliable (the feeder logs it). The denominator is where almost the whole sector limps: manual samplings are one-off, noisy and sometimes taken in conditions that are hardly representative. iLEAN Vision estimates biomass per cage from the behavior of the school and its response pattern to feed, which lets you close FCR on daily data instead of monthly data.
Which signals from the school indicate satiation or stress?
Average school speed, vertical spread, density near the surface when the feeder sounds, how aggressively the fish rise, and pellets falling uneaten to the bottom. When seabass are hungry they rise in a column and hunt; when they are full, the feed starts dropping to the bottom and the school descends. iLEAN Vision detects that transition and proposes stopping or reducing the setpoint before uneaten feed fouls the cage and pushes FCR up.
Does underwater vision work with poor visibility or jellyfish?
The underwater camera does not need a perfect image: the CNN is trained to read movement patterns, not photographs. In turbid conditions or with jellyfish the network works with noisier signals and lowers its confidence — and when confidence falls below a configurable threshold, iLEAN Vision does not propose a setpoint and alerts the cage technician. Better not to propose than to propose wrong.
Can iLEAN Vision integrate with an existing automatic feeder?
Yes. iLEAN does not force you to change your feeder: it integrates through whatever channel is available — from a modern PLC to an old 4-20 mA output. If the feeder already has remote control, iLEAN Vision proposes the setpoint and the feeder executes once the technician signs. If it does not, the technician receives the proposal in the Connect app and enters it manually.
How much can FCR drop with AI feed control in seabass and seabream?
The sector reports FCR improvements of 5-15% when a fixed feeding program is replaced by an adaptive one based on reading the school. The exact figure depends on how well tuned your current program is and on the variability of your site (current, temperature, oxygen). Feed is the first variable cost of the operation: every tenth of FCR saved shows up in the year's results. We ask for your cage data and send you the estimated ROI in 48h.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your seabass and seabream site.
We work on your cages' real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See premium fishing and aquaculture