Introducing TUNA SCOPE into Dalian: Comparing quality inspection results from an AI and a master Japanese artisan

ISSUED : 2019.11.16

Dalian, China. With one of the world’s largest international trading ports and as a major hub for seafood processing, Dalian is a city that serves a vital function for China.

Together with Mr. Ishii, an artisan from Misaki Megumi Suisan who has mastered the craft of tuna inspection through years of experience in Japan, we carried out a quality inspection in Dalian at a processing facility belonging to Dalian Global Food Corporation, a subsidiary of the general trading company Sojitz Corporation, ahead of the facility’s implementation of the AI inspection system TUNA SCOPE.

An AI examiner and a skilled master artisan. How consistent will their quality inspection results be?

While reverifying the accuracy of the AI system, we pondered the possibilities of TUNA SCOPE and its potential to digitally transform the tuna business.

Almost 20 years an artisan, but still in training

For almost 20 years, artisan Mr. Ishii has determined tuna quality day in and day out as a tuna buyer for Misaki Megumi Suisan.

Inspecting quality just by examining the cross section of a tail is not the easiest of tasks, and even Mr. Ishii states that he still has a lot to learn. It is thought that the quality of a tuna can be greatly affected by the amount of time it takes to haul the tuna, carry out the various on-board bleeding and cleaning processes and flash freeze it.

“The tuna tail’s freezes first and it does so faster than the rest of its body, meaning that the closer you get to the edible parts around the stomach, the slower it takes to freeze. So sometimes the tuna is already damaged which makes determining its overall quality just by looking at the differences in the tiniest parts of its tail’s cross section, extremely challenging,” said Mr. Ishii. He explained that artisans draw from their extensive experience to identify the types of cross sections that suggest poor quality and the parts that are indicative of a quality tuna. This enables artisans to gradually develop their own unique set of inspection standards.

Mr. Ishii has spent years developing his artisan’s intuition, so it does not come as a surprise that he is confident in his eye for tuna inspection.
One after another, the tuna was brought out from a -60℃ deep freezer for the AI and the artisan to inspect.

“At first, I was determined to somehow get one over on the AI” said Mr. Ishii, but with a staggering number of tuna laid out in front of him, tensions began to run high between both the artisan and the TUNA SCOPE development team.

How will their results compare?

Around 90% of the AI’s results for the tuna that was graded as the highest quality matched those of the artisan

An artisan who has spent years at markets devoting himself to mastering the craft of tuna inspection and an AI examiner born out of deep learning. Standing shoulder-to-shoulder, the pair approached the task ahead - inspecting the quality of around 5,000 kilograms of frozen yellowfin tuna.

The results were surprising.

When we compared the results for the tuna that was graded as the highest quality, it was revealed that approximately 90% of the AI's results matched those given by the artisan.

After the grueling few hours jointly inspecting tuna quality with the AI, Mr. Ishii shared his thoughts about the experience, admitting to being awed by TUNA SCOPE’s ability to inspect and giving it his seal of approval.

Even artisans who have years of experience can be inconsistent in their inspections.

“Inspecting the cross section of a tail is a task that takes time and concentration. Sometimes disparities can occur depending on how we’re feeling that day, the lighting and other surrounding environmental factors. In other words, if this AI system keeps evolving, it might end up becoming a much faster and more consistent examiner than me,” he said.

Before long, AI may be training the future generation of human artisans

After experiencing what it is like to inspect tuna quality in collaboration with an AI system, Mr. Ishii began to see a potential future in which the roles may be reversed. A future where AI will assume the role of a teacher, passing on the tricks of the trade to human artisans and helping them hone their skills as opposed to a future where AI is acquiring the skill from humans.

“AI is reliable in the way that it can pick up on small details that the human eye can miss. It’s interesting to see what it can detect, and there’s a lot I can learn from it” he said.

“The number of successors to this craft is dwindling by the minute, so I can’t help but approach my work without giving thought to how I can help raise the younger generation of artisans. I can imagine a future where TUNA SCOPE becomes a widely used app, and all the merchants going to the markets are buying tuna with a smartphone in hand. That doesn’t sound too bad actually (laughs).”

Having spent years inspecting tuna quality on the ground, Mr. Ishii has keenly felt the shortage of successors more than anyone. TUNA SCOPE quality inspections may help artisans by dispelling their anxiety surrounding this issue, and that future may be closer than we think.

TEXT BY Ryo Sasaki

PHOTOGRAPHS BY Takafumi Shindo / Ryo Sasaki