You can only win real freedom by taking risksBRAIN Co.,Ltd.
<Series 2 (Total 2 series)>
"BakeryScan" instantly identifies the type and quantity of bread on a tray. The idea for this product development was proposed to Brain in 2007. What steps were taken during the six years until release and what was the deciding factor for the breakthrough?
Research and development of "BakeryScan" was started in order to survive
The company is headquartered in Nishiwaki, Hyogo, a textile city. In 2007, while conducting research and development to automate the trial weaving process for fiber, the idea for bread image identification was proposed.
"At that time, we felt the domestic textile market had reached a plateau. By comparison, bakeries are everywhere in Japan. Some research revealed a total of up to 13,000 companies. Immediately after the collapse of Lehman Brothers, the business environment was tough. There was obviously the fun of trying something new, but more importantly, we went with this idea in order to survive."
Prior to development, a test store was created to conduct marketing research. This produced data such as "tripling the types of bread increases the sales per unit area of store space by about 1.5 times," and "selling the products side by side without packaging increases sales threefold."
"In other words, when running a bakery, it is better to offer many types of bread. But since the bread can't be packaged, it is difficult to identify with a barcode."
The focus of development was the identification of unpacked bread. However, there were two problems unique to bread.
"The first is the similarity between different types with similar shapes and colors. The second is the individual differences among the same type. The finish is not exactly the same for the same type. Identifying them correctly was a very difficult challenge."
The AI doesn't help humans, humans help the AI. Thinking from a different angle is the key to breakthroughs
After about 3 years, 50 kinds of bread can be distinguished with 98% accuracy in the laboratory, but bakeries have differing brightness, and the color of baked bread differs slightly depending on the heat and the dough condition. Fortunately, there was a subsidy from the Ministry of Economy, Trade and Industry, and thus the prospect for the funding needed to dive into technical development, but 100% accuracy remained out of reach. Looking at the problem from a different angle provided the breakthrough.
"We tend to think that AI helps people, but what if people could make up for the shortcomings of AI? So BakeryScan does not perfectly identify breads. The colors green, yellow, and red are displayed in order of decreasing accuracy, and the cashier does the final check. AI learns how to do the final check by the cashier, and accuracy improves steadily."
After that, improvements were made at the stores that had implemented the product on a test basis, and design modifications were made, culminating in the March 2013 release. The subsequent response and the application of the technology to cancer research are described in the first part of the series.
"BakeryScan" screen. The identification accuracy is indicated by green, yellow, and red colors.
Development and sales were driven by meeting people and numerous collaborations
To date, there have been a variety of meetings with people and collaborations.
"We have worked with several universities other than the University of Hyogo. A university with a wealth of ideas is a treasure trove. And in fact, the threshold to university is not high. It even happened that I would search for a related keyword, go straight to the research lab of the professor in the search results, and start a collaboration."
In terms of sales, the marketing power and sales network of large companies speak for themselves.
"The sales power of major companies is really on another level. Approximately 90% of the sales of "BakeryScan" are via major contract manufacturers."
On the other hand, it is necessary to pay attention to patents and rights during collaborations. The company has also had quite bitter experiences. Nevertheless, Mr. Kanbe insists on taking risks, and in the end quoted a US poem to encourage small and medium-sized manufacturing companies as follows.
"In general, education in Japan teaches safety, and parents want stable lives. The greatest hazard in life is to risk nothing. The person who risks nothing may avoid suffering and sorrow, but they cannot learn or grow. Only the person who risks something can win real freedom."
Series "You can only win real freedom by taking risks"
Series 1 Developed an AI cash register that instantly identifies bread. High expectations for application to cancer research
Series 2 You can only win real freedom by taking risks
Brain Co., Ltd.
President Hisashi Kanbe
Brain was founded as a PC shop in 1982 in Nishiwaki City, Hyogo Prefecture, the site of a thriving textile industry. Later, while developing phrase conversion software and NHK's character information display system, as well as developing technology for image identification for fibers, computer system R&D became a pillar of its business. "BakeryScan" was developed in 2013, and various solutions driven by this technology have been rolled out.
Coverage date November 15, 2019
Creating Holidays in Agriculture through the Power of IoT
Technological ability to respond to severe demands of a top pro, significant manufacturer Challenging the unknown possibilities of carbon fiber
GRAPHITE DESIGN Inc.
Aiming for a Full-scale Commercialization of a University-Initiated Handwritten Character Recognition Technology
iLabo Co., Ltd.