By JUNICHI TAKITSUBO/ Staff Writer
Hisashi Kanbe, president of Brain Co., shows a demonstration model of BakeryScan in Nishiwaki, Hyogo Prefecture.
NISHIWAKI, Hyogo Prefecture--The brains behind a successful bakery are no doubt the developers of pastries and other delectable goods, but this brain makes checking them out a piece of cake.
BakeryScan, a bread image recognition system, is spreading among bakeries across Japan as the system instantly identifies various types of breads and calculates the bill for them at the register.
As products like freshly baked bread tend not to be wrapped, making it impossible to attach barcodes or price tags to them, it often takes time for cashiers to determine the sum, leading to long lines.
Based on artificial intelligence technology, the BakeryScan system was invented by Brain Co., a computer system developer here.
When bread products on a tray are put on the counter at the register, images are taken of them that are then analyzed using AI, with the contours and prices shown on the monitor within one second.
At large bakeries that sell as many as 100 types of products, the system allows even new part-timers to quickly handle a large number of customers.
Since its release in 2013, BakeryScan has been introduced at 400 Anderson Group outlets and other bakery operators across the country, receiving overseas news coverage.
Brain, which started as a small computer shop operator in 1982, began developing BakeryScan around the time of the collapse of U.S. investment bank Lehman Brothers in 2008.
THE RIGHT INGREDIENTS
While the company's system to display the results of professional baseball games and foreign exchange rates drew attention when it was adopted by Japan Broadcasting Corp. (NHK) in 1984, few development orders came in at the time, putting Hisashi Kanbe, Brain's president, at a loss as to the future of the company.
For that reason, when Kanbe, now 69, who formerly worked at the predecessor of Panasonic Corp., was asked if his company could create a device to quickly scan products at bakeries, he jumped at the offer.
“There are more than 10,000 bread stores across the nation,” said Kanbe. “I wanted to pitch our product to the huge market to move our business forward.”
Another factor that led Kanbe to the decision is the fact that Brain had accumulated know-how on image identification technology, as it had developed a camera-mounted system to check whether textiles made of Banshu-ori fabric--a local specialty in Hyogo Prefecture--are properly woven in accordance with design drawings.
The development of BakeryScan did not go smoothly initially, although the project was taken on with the future of the company at stake.
It was difficult for the system to precisely distinguish different baked goods with the same shape, such as those with ham and those with corn. To solve this issue, the system was improved to assess the position of ingredients and the shadow patterns in numerical form so the AI could make correct determinations.
After spending an entire year developing the system jointly with the University of Hyogo, a prototype was completed. But even the completed system could not identify bread products with 100 percent accuracy.
Brain then changed directions to leave some things to humans instead of clinging to a quest for perfection. Under BakeryScan, unidentified items are shown in yellow for cashiers to choose the correct product types from displayed candidates. The AI system learns staff’s choices to improve accuracy.
Brain has already begun taking the next step forward.
It received a surprise inquiry two years ago from a doctor who told a Brain official that “bread resembles cancer cells.” The physician went on to say that humans can detect only a limited number of cancer cells in a certain period of time.
Acting on the prospect that an image recognition system could more quickly identify tumors even if they are at early stages, Brain is proceeding with joint research with the Louis Pasteur Center for Medical Research in Kyoto to complete such a system.