AI and fish farming in Japan

Known locally as madai, red sea bream is a Japanese delicacy that is traditionally eaten at celebrations and on feast days.

Intensive fishing has seen wild stocks fall over the years and producers must rely on aquaculture to maintain supplies. Kindai University’s Aquaculture Research Institute plays a crucial role in ensuring red sea bream stocks can meet demand.

The institute’s Shirahama Station supplies about 12 million red sea bream fingerlings – juvenile fish that have developed fins and scales but are only about the length of a human finger – to fish farms around the country where they grow to full size.

The fish are hand-sorted twice a year before distribution. Counting the fingerlings and culling deformed fish is laborious and requires a highly-skilled workforce. 

“Some members of our staff have as much as 30 years of experience,” says Naoki Taniguchi – who manages the Institute’s Larval Rearing Division and is Deputy General Manager of the Aquaculture Technology and Production Center. “Sorting work is very time-consuming and takes up to eight hours a day during the busy shipping periods.”

Taniguchi wants to automate as it is getting harder to attract talented people for the task. The Institute is currently developing an automated sorting system using Microsoft’s Azure Machine Learning Studio and Azure IoT Hub. The same Microsoft technologies are addressing another pain point: the need for a person to constantly adjust the pump flow that delivers the fingerlings to the conveyor belt. If the flow is too high, sorters will fall behind. If the flow is too low, production times will suffer.

Source: Microsoft Japan. Workers hand-sort red sea bream fingerlings.

Microsoft Japan and Toyota Tsusho Corporation have co-developed an artificial intelligence (AI)-driven sorting system that calculates optimum flows and makes automatic adjustments. Image analysis is used to count the fish on the conveyor belt and the spaces between them. It also uses machine learning to understand how the sorters complete their task at different volumes of fingerlings.

This means that the institute can have fewer people working on the same task, allowing them to take more breaks or focus on other work, said Taniguchi.

Explore: 

Read the Microsoft feature story