Up to now, TUNA SCOPE has mastered the selection of yellowfin and bigeye tuna and has been introduced at various sites in Japan and overseas.
Next, the technology set its sights on becoming an expert on bluefin tuna.Unlike yellowfin and bigeye tuna, which contain relatively little fat, the value of bluefin tuna depends not only on the quality of the red meat but also on its fat content.Evaluating quality on these two different points is said to be an extremely difficult task, even for professionals with years of experience.Also, unlike quick-frozen tuna, the cross-section of a raw tuna tail is uneven, and the slightest difference in the surface can change the appearance of the cross-section.To develop a new AI model that could take on the challenge, a huge amount of training data is required.
This is why we focused on Toyosu Market, Japan's largest wholesale seafood market, where highly rare bluefin tuna selected by expert middlemen examiners is sold daily.With the cooperation of these Toyosu middlemen, we built a system that continuously collects and transmits cross-sectional images of the bluefin tuna tails that line the market every day, as well as the examiner results. This became our real-time AI training data.
After about two years of data collection, an AI quality judgment algorithm based on the two points of “fat content and red meat quality” was created through trial and error.