- Scientists have developed AI that identifies individual southern rock lobsters by the unique patterns on their shells.
- Matching lobster photos with supply-chain records could help verify where seafood came from and how it reached buyers.
- Researchers hope to expand the technology's use across the industry and eventually let diners trace their lobster with a photo.
Facial recognition is getting a rather shell-covered counterpart.
Scientists in Tasmania are developing a system that uses photographs to distinguish one southern rock lobster from another. The unusual application of artificial intelligence has a practical purpose: helping fishers, exporters and buyers keep track of individual animals as they move from the ocean toward the dinner table.
Quinn Fitzgibbon, a professor at the University of Tasmania who specializes in aquaculture and the physiology of aquatic organisms, says the ability to trace lobsters will create transparency and strengthen trust across the seafood sector.
A shell of its own
To most of us, one lobster looks a lot like the next. Yet the patterns on a southern rock lobster’s shell give AI something distinctive to work with, and is compared by researchers to a fingerprint.
“Lobsters have unique patterns and the AI model recognizes individual lobsters based on those features,” Tara Kelly, a junior researcher in aquatic animal physiology at the University of Tasmania’s Institute for Marine and Antarctic Studies, said in an official press release. “It’s essentially a fingerprint which allows us to digitally tag lobsters.”
A photograph lets the system compare features on a lobster’s shell with images in a database and look for a match. If no match is found, the lobster can then be added as a new entry.
To create the database, Fitzgibbon led the project with two industry partners, Fiordland Lobster Company and the South Australian Lobster Company. The team photographed more than a thousand southern rock lobsters to train and test the model.
The university says the model works on photos taken three years apart, including after the lobsters had grown and molted several times, were covered in barnacles and algae, or had already been cooked.
“We’ve created a noninvasive traceability system for live southern rock lobsters that is incredibly robust, scoring 100% accuracy on the test set and real-world data, with processing times of less than a second,” Dean Giosio, an aquaculture engineer at IMAS, said in the same press release.
The journey to dinner
A lobster can pass through plenty of hands before it reaches a dinner plate. Linking its shell’s digital “fingerprint” to records of where it was caught and how it moved through the supply chain could help buyers check that the lobster they received is the one they were promised.
In China, southern rock lobster is a status food. People buy it to honor guests, and for years buyers there have paid far more than anyone else. That high price is also what makes fraud worth trying.
Once a lobster is out of the water and in a box, nothing on the animal proves where it was caught. Physical tags don’t work for the industry because they can be removed or forged, making it possible for someone to pass off a cheaper lobster as the prized one and pocket the difference.
The AI system is meant to close that gap by making the shell pattern a form of ID that can’t be removed or faked. In principle, a buyer at any point could photograph a lobster and check it against the record made when it was caught.
“Being able to verify where a lobster has come from and how it has been handled and transported creates reassurance for provenance claims and improves confidence in southern rock lobster as a premium product for Australia and New Zealand,” Fitzgibbon said.
iMOVE, the project’s funder, also listed protecting against seafood fraud and building consumer trust in key export markets as primary goals of the program.
Coming to a phone near you
Someday, checking your lobster’s backstory could start with snapping a photo of your dinner. The AI is already built and tested, and researchers are actively using it on boats. The public app that fishers, citizen scientists and diners could download is still in development.
“The lobster identification technology will be deployed via a website and phone app, which has the potential to support recreational fishers and citizen science,” Fitzgibbon said. “It would enable the public to contribute to lobster monitoring.”
Scientists plan on developing the model so it can be used by the processing sector and the fishery alike. Eventually, diners could get a look at that journey with a quick photo revealing a little more about the lobster before they crack its shell.