AI for the Pastry AisleBRAIN Co., Ltd. IP

Website: BRAIN Co., Ltd.External site: a new window will open.
Category: Otrers

Image recognition technology identifies and calculates the total price of unpackaged breads

BakeryScan employs image recognition technology to identify and quickly calculate the total price of unpackaged breads, simplifying staff training and reducing time at the register – no barcode or radio frequency identification (RFID) required! The underlying technology has potential applications in any field that requires rapid and accurate sorting and identification.

Japanese bakeries offer a wide assortment of fresh, unpackaged breads. Yet identifying the various types of rolls and pastries – and their corresponding prices – can be time-consuming and difficult for new staff at the register.

Enter BakeryScan, point-of-sale (POS) image recognition technology developed by Hyogo Prefecture-based BRAIN Co., Ltd. BakeryScan scans a customer’s items, identifies the various breads, and calculates the total in mere seconds.

BakeryScan is a POS system that instantly scans and calculates prices for unpackaged breads; courtesy of BRAIN Co., Ltd.

IT in the Bakery

Since its founding in 1982, BRAIN has amassed tremendous experience in computer system research and development. Long before computer graphics could be overlaid on television footage, this 20-person software company developed a graphic system to display sports scores onscreen, and even developed graphics technology to aid local textile businesses.

So when a local bakery approached BRAIN about designing a system to recognize and tabulate different varieties of bread, the company jumped into action. 10 prototypes were developed over three years, adjusting factors such as the camera height and angle, as well as adding a patented backlighting design. The first commercial model debuted in 2013, and customers in Japan have steadily increased since.

BRAIN Co., Ltd. developed 10 BakeryScan prototypes over three years.

Sophisticated AI

BakeryScan’s patented bread identification method is built around a sophisticated artificial intelligence (AI) that’s capable of distinguishing between similar bread types while still accommodating day-to-day variation.

Capable of learning like a human, when the system is unsure about a particular item, it highlights the item and proposes possible answers to the operator. After the operator selects the correct option, this feedback is incorporated into future scans.

The more BakeryScan is used, the more accurate it gets, with identification rates approaching – and even exceeding – 97 percent over time. Beyond time savings, the system also provides valuable POS feedback, which can help significantly improve sales efficiency.

The company’s headquarters in the mountains of Hyogo Prefecture; courtesy of BRAIN Co., Ltd.

Expanding Applications

The learning and analysis technology BRAIN uses to power BakeryScan is called AI-Scan. This technology has potential far beyond bakeries, and is already being used in Japan to identify prescription medicines such as tablets and pills that have been mixed together.

BRAIN hopes to move into overseas markets once a strong domestic foundation has been established. Shinnosuke Hara, executive officer and sales department general manager at BRAIN, sees potential in places such as U.S. and European supermarkets or cafeterias, and hopes to license AI-Scan technology directly to retailers or even POS system manufacturers. “We know that more R&D will be needed, so we’re looking for partners,” he explains.

With a track record of accuracy and efficiency, the importance of AI-Scan’s image-recognition technology seems destined only to rise.

Based on interview in September 2016

BRAIN Co., Ltd. CEO Hisashi Kambe; courtesy of BRAIN Co., Ltd.

Website: BRAIN Co., Ltd.External site: a new window will open.