When it comes to training machine vision and AI-powered retail checkout systems, packaged goods and locally created food items are treated very differently.
That’s at least according to Mashgin, a maker of touchless checkout systems. Company spokesperson Toby Awalt said that’s because another store on the network has likely already added that bag of chips or candy bar to their 10 thousand plus item database.
Not so when it comes to locally made food items.
“CPG items, we have to do less and less because there’s enough overlap,” said Awalt, who gave us a walkthrough of the system at CES 2022. “But for dishes, we’ll do every time.”
According to Awalt, adding a new food menu item for a restaurant doesn’t take that since most cafeterias or restaurants only serve between 15 and 50 items.
“You can do that relatively quickly,” he said.
Still, a new packaged good has to be entered into the system now and then. Whenever that happens, the operator has to position the package in several different positions to give the system enough info to recognize the product whenever it shows up under the camera.
“I actually do dice rolls with the product,” said Awalt, throwing a Häagen-Dazs ice cream bar onto the tray.
According to Mashgin, the company recommends the system capture 20 to 50 total positions of a product so it can recognize the product from various angles and also distinguish between different variations within the same product line (such as two different flavors of ice cream or potato chips).
You can watch a walkthrough of the Mashgin system below.
