Self-checkout lanes are usually a quick way to buy a few things at the grocery store, but what if you want to pick up some fruit or any other item without a barcode? Toshiba is happy to help, using a new automatic checkout system with built-in cameras that can tell exactly what you’re buying. No more flipping through a list of twenty different types of pears; this system can identify different variations in the variety of fruit just by looking at it. When first installed, the system checks the image of the product against a built-in database, asking the user to select which item they’re trying to buy, with the most likely results on top. While going through this with a large number of shoppers, the recommendations get better and better.
This kind of item recognition is no simple task. According to Keiji Yanai, a scientist at the University of Electro-Communications in Tokyo quoted in an article in New Scientist, a system like this is much more difficult than facial recognition. Because the objects are so similar (a Pink Lady apple versus a Golden Delicious, for example), it introduces that much more difficulty in choosing between them.
Toshiba hopes to be saving you the valuable 30 seconds it would have otherwise taken you to page through a menu by bringing the system to market within the next three years.