Industries / Food, farming and forestry / Food and beverage processing / Seafood processing
Sorting and grading
Sorting and grading is the work of dividing landed fish and shellfish by species, size and freshness, and checking each piece for flaws such as bones or worms.
5
Milestones achieved, from 1953 to 2024.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1953
Freshness gets a scorecard
Torry Research Station in Aberdeen, Scotland, published a method for judging raw fish by sight, smell and touch that scored each feature on its own instead of giving one overall impression. Later European grades built on the idea.
Credit
Torry Research Station (Shewan and colleagues); FAO calls it "the first modern and detailed method"
1976
Europe grades fish by law
The European Economic Community set common grades for fish at first sale. An inspector puts each lot into a freshness category, Extra, A or B, by sight, smell and touch, and into a size category by weight. The 1996 regulation that replaced the 1976 rule is still in force.
Credit
Council of the European Communities, Regulation (EEC) No 103/76; Council Regulation (EC) No 2406/96
1985
A points system for freshness
Researchers in Tasmania built a scoring sheet, the Quality Index Method, in which each sign of age earns demerit points that add up to one number. Danish and other European laboratories adapted it for cod, herring, saithe, redfish, sardines and flounder, giving inspectors a repeatable score.
Credit
Tasmanian Food Research Unit (Bremner and colleagues); Lyngby Laboratory, Denmark
1990
A grader that reads skin
Fish skin and flesh carry electric current differently as a fish ages. Rafagnataekni Electronics in Reykjavik built the sensor of Torry's handheld GR Torrymeter (1975) into a grading machine, and by 1990 a production model sorted 70 fish a minute into four lanes.
Lessonreadings change when fish are damaged, frozen, filleted or not bled, so FAO judged the method fit for grading batches, not single fish, and grading by the senses stayed the legal method.
Credit
Torry Research Station (Jason and Richards); Rafagnataekni Electronics
2024
Cameras find worms in cod
Plants find parasite worms (nematodes) in cod by "candling": a worker lays each fillet on a lit table, looks for dark shapes and cuts them out. Researchers at Nofima, a Norwegian food research institute, paired a hyperspectral camera, which records many bands of light beyond what the eye sees, in Maritech's Maritech Eye with a deep learning model. It found 73% of worms, against an industry average of 50% by hand on skinless fillets and 25% on fillets with skin, and a fish plant in Iceland installed it to sort cod fillets.
Lessonflagged fillets still go to people for trimming.
Credit
Nofima; Maritech; Norsk Elektro Optikk (camera)
Watch for
Cameras that sort by species
Hoseth Technology says its Flow A.I. ceiling cameras tell saithe from cod, estimate size, and spot wounds, cutting errors and badly filleted fish so they can be sorted out. It reports yield gains of 10 to 15 percent in early tests at several plants and a recent nine-camera installation at a cod plant.
Achieved whena party outside the company, such as Nofima, SINTEF or a processor publishing its own line data, reports species accuracy and the change in yield at a whitefish plant.
Claimed by
Hoseth Technology, in a Nor-Fishing innovation award nomination, July 2026
Open
01
An open worm image set
A public set of cod or hake fillet images, taken on a light table with an ordinary camera and a low-cost near-infrared camera, with every worm's position confirmed by the standard laboratory method, and a worm detector tested on it.
The briefCredit · unclaimed02
Testing auction cameras against graders
A public test of camera freshness grading against the official EU grades. At a fish auction, photograph boxes of iced fish next to the Extra, A or B grade the official grader gives each lot, record a second grader's call on the same boxes, and test whether an image model agrees with the graders as often as they agree with each other.
The briefCredit · unclaimed
Open 01 · the brief
An open worm image set
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of cod or hake fillet images, taken on a light table with an ordinary camera and a low-cost near-infrared camera, with every worm's position confirmed by the standard laboratory method, and a worm detector tested on it.
Why it is still open
A 2022 study measured hand candling's sensitivity, the share of worms it finds, at 31%. The system Nofima measured uses a hyperspectral camera, and the 2024 multispectral study of cod worms shares its 270 images only on request. Searches found no public, labelled fillet image set on which cheaper cameras can be tested.
Built with what exists
A light table, a colour camera, a low-cost near-infrared camera such as the Raspberry Pi Camera Module 3 NoIR with an infrared lamp, the UV-press method of ISO 23036-1:2021 (fillets pressed thin and frozen, then viewed under ultraviolet light, where the worms glow) run with a food laboratory for the reference count, and an object detector such as Ultralytics YOLO11, which trains on one graphics chip.
First step
Buy 20 skinless cod fillets, photograph each on a light box, have someone candle them and mark the worms they find, then ask a food laboratory to run the UV-press check on the same fillets to count what was missed.
Done looks like
300 fillets of at least two species imaged and then checked by UV-press; images and worm positions published under an open licence; the detector's share of worms found, and its false alarms per fillet, reported beside a trained candler's results on the same fillets.
Open 02 · the brief
Testing auction cameras against graders
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of camera freshness grading against the official EU grades. At a fish auction, photograph boxes of iced fish next to the Extra, A or B grade the official grader gives each lot, record a second grader's call on the same boxes, and test whether an image model agrees with the graders as often as they agree with each other.
Why it is still open
EU law still has freshness judged by a person's senses at first sale. FRESQO, an EU-funded Greek research project, built a prototype in 2023 that photographs every fish in an auction crate and scores its freshness, but searches found no published comparison of it, or of any similar camera system, with the official grades given lot by lot.
Built with what exists
A fixed camera over the auction conveyor with a polarising filter to cut glare from ice, the labelling tool Label Studio, and an open image model such as Meta's DINOv2 with a small classifier trained on one graphics chip.
First step
Ask a fish auction's management for permission to photograph one day's sales, and record the official grade on each lot's label beside its photo.
Done looks like
1,000 lots of at least three species, each with grades from two graders; agreement between the two graders, and between the model and each grader, reported as percentages; images and grades published.