AI now allows you to assess the quality of a tuna, and this is only the beginning!

As it develops, artificial intelligence is embedded everywhere. A few days ago, we told you that Google planned an AI model capable of analyzing Dolphins communication. The news of the day also has a relationship with artificial intelligence and the sea, but in a very different register. Indeed, four Japanese companies: Fujitsu, Sonofai Inc., Ishida Tec Co., Ltd. And Tokai University) have just announced the launch of a new machine, called Sonofai T-01. Combining ultrasound and artificial intelligence, this device is capable of assessing the fat content of albacore tuna frozen in a dozen seconds, according to its designers. If the Sonofai T-01 proves itself, it could well shake habits in the fishing and food industry.

AI at the service of waves

How does Sonofai T-01 work? The device sends ultrasound at low frequency that cross the still frozen fish. Physics behind this method is simple: the waves do not behave in the same way depending on the environment they encounter. Here, the fat content of the tuna changes their journey. The lean parts reflect less ultrasonic waves than oily areas. It is at this moment that artificial intelligence comes into play. Designed by Fujitsu, its mission is to interpret the echoes, the signals returned after the passage of the waves in the tuna. In particular, it must sort out and eliminate the “noise”, these parasitic signals generated by the very structure of frozen fish. By filtering these interference, the AI ​​can draw up a precise cartography of the distribution of fats.

Fujitsu indicates that he has relied on his “Kozuchi” house platform to develop analysis algorithms. It is a set of software solutions that the Japanese group offers for various industrial applications. The adaptation of these tools to ultrasound analysis forms the technological heart of Sonofai T-01. One can imagine that the partnership with the Tokai University, whose work on the taste qualities of the tuna is recognized, made it possible to adjust the models of AI to better match the ultrasonic signatures with the characteristics sought on the market.

The end of the systematic cutting?

Traditionally, to judge the quality of a tuna and its richness in fat, the most widespread method is manual. Inspectors in charge of evaluation scient the tail of the frozen fish to visually examine the transverse cut. However, although an anchored in practices, this method suffers from several limitations. First, time: the operation takes about a minute per fish. Then, security: manipulating a saw on frozen blocks of several tens of kilos is not without risk for the staff. Finally, the invasive character: the fish is damaged by cutting.

The Sonofai T-01 machine erases these friction points, since twelve seconds are enough to obtain an assessment, a potential time reduction of 80 %. We would go to a much higher rate of treatment, announced at five tuna per minute for a single operator.

Profits for the tuna sector

Such a device should obviously not fail to interest professionals. First of all thanks to its ability to accelerate the control rate, as explained at the moment. Then, its measuring precision announced at 78.9 % is better than that of the visual inspection of a human. This should have a direct impact on the valuation of the tuna. Indeed, the pieces will be better assessed and, when you know that the “bintoro”, this very oily and therefore very expensive ventral part of the albacore, can be worth up to ten times the price of classic pieces, it is better to identify it correctly. under penalty of selling a value product at low prices.

Not having to cut the fish is another weight advantage. The tuna remains whole, which is preferable for its conservation and marketing, especially for large parts. Fewer interventions on the product, it is also potentially fewer losses. By refining the classification and avoiding superfluous cutouts, we can hope for better use of each fished tuna, an argument which is important in a sector concerned about resource management.

Industrial equipment soon on the market

For the rest, the Sonofai T-01 is not a small laboratory apparatus. It measures 2.7 m long and 1.45 m deep and 1.6 m high, dimensions which therefore destine it to industrial environments. It is planned to analyze tuna from 10 to 20 kg. The project matured quickly enough: started in March 2024, he saw the birth of a first prototype marketed at the end of 2024. The know-how of Ishida Tec in matters of food processing equipment was undoubtedly precious for the physical design of the machine.

Its marketing will start in June 2025 in Japan, for a price set around 30 million yen (around 182,250 €). At first, it is the transformation factories and Japanese fishing cooperatives that are targeted. But ambition is global. Fujitsu and its partners intend to offer their machine in the United States and other markets fond of quality tuna, carried by the popularity of sushi and sashimis.

Functions extended in the future

The fat assessment would however only be a first step. The designers of the Sonofai T-01 are already considering fleshing its capacities via software updates. The machine could learn in the future to assess other qualitative aspects such as the freshness, texture or firmness of tuna flesh. Such multiplication of analysis parameters would give an even more complete vision of the quality of each fish.

The designers of the device also evoke the possibility of expanding the use of the device to other species. Yellow tuna and bonitis are already mentioned as potential candidates. They hope to be able to analyze a hundred species in the next five years. In addition, if technology is adaptable, it could interest an even wider range of companies in the seafood sector. In all cases, this Sonofai T-01 seems to launch a tendency to automation and data analysis for quality control in the food industry. And more broadly, it is not so absurd to imagine that we could all have a small device capable of assessing the quality of most of our foods. But with regard to the size of the machine launched by Fujitsu and its partners, it will probably have to wait a few years.