Using image recognition and machine-learning, researchers have developed a new self-checkout system that can distinguish between objects with subtle color and shape differences.
Toshiba researchers have developed a new self-checkout system that can identify similar looking objects like fruits in supermarkets. By incorporating image recognition and machine-learning applications and a webcam, the system is able to distinguish between fruits with subtle color and shape differences.
Here’s how it works: at a self-checkout, the system shows a list of pictures by comparing the shopper’s product against its database of images, with the machine’s closest match at the top. The shopper then verifies if it matches his product and confirms his purchase by selecting the appropriate option.
The technology is flexible and gets more accurate with usage. It allows users to correct it in case its first choice doesn’t match the bought product. They can then choose the correct product name and the system will use the new photo for the particular product in future transactions.