Amazon Go is Amazon’s cashier-less shopping platform. Contrary to Amazon’s online website, the Go is a brick and mortar store- similar to your regular convenience store. What’s unique about the Go is that you can pick up an item and just walk out of the store without having to wait in line to pay for it. The store sells ordinary grocery items like bread, dairy and packaged ready to eat meals.
A customer has to hold up a QR code on the Amazon Go app onto scanning devices, similar to subway turnstiles. After that they just start shopping as they normally would. Cameras mounted on the ceiling track each person and what item they pick up or put down and updates that on their cart. When done, he/she just has to walk out of the store and is billed only for the items picked up.
Naturally, Amazon is pretty secretive about the technology used. However, what they have revealed is that it’s a combination of Artificial Intelligence, Computer Vision, Deep Learning and data pulled up from multiple sensors that enables this cashier-less experience. In Amazon’s own words,
“It’s similar to the same technology you’d find in self driving cars. We’re calling it ‘Just Walk Out’ technology”
AI is pretty much the rage these days and everyone’s heard more than enough about it.
Computer Vision however, is something that’s a little vague. It deals with teaching computers to make educated guesses based on visual cues like images or video feeds. It aims at automating the human visual system.
Computer Vision deals with acquiring information from digital images, and then processing them to arrive at a digital output, like numeric or symbolic information in the form of decisions.
Using the horde of cameras on the ceiling, the software is able to uniquely classify and track each individual. In fact, the software is so good that it was able to uniquely distinguish between an army of customers in Pikachu costumes.
The Cameras
After that, the program follows each person around virtually and keeps track of what they pick up and put down. The program analyzes which shelves a customer interacts with to determine what item they chose. In order to aid the system, there are weight sensors on the shelves that help the program narrow down on an item.
The Shelves with items
In addition to this, Amazon also keeps a track of each person’s previous shopping history to zero-in on an item in the event of the program being confused. For example- if the program is unable to determine whether someone picked up a bottle of mustard or ketchup, it simply looks at the person’s previous choices. If the person buys mustard more often than not, then that’s what they’re more likely to have picked up this time as well.
Like any new software, Amazon Go has its fair share of problems. Some customers have been able to leave the store without being charged for certain items. The program also has trouble when there are more than 20 people in the store at once. It also runs into issues when there is a group of people shopping together- like families. There is the added issue of dealing with grabbers- kids that impulsively pick up items only to have their parents put it back immediately. It is still unknown as to how the system would deal with such issues.
Furthermore, many people feel vulnerable about their shopping habits being monitored. A lot of people also feel uncomfortable being tracked by cameras while they shop. Another issue is that the technology facilitates impulse buying.
It is yet to be seen how Amazon will deal with these issues. However, the technology is still impressive and is completely unparalleled. It goes to show just how far Artificial Intelligence and Deep Learning will go on to influence our daily lives and how the future of retail shopping is shaping up.