Why?
An Amazon Fresh store recently opened near me and I decided to try the scan-less shopping. I only needed some milk and bread but was curious about the concept’s efficiency. So I opened the amazon app scanned my Amazon Fresh barcode and proceeded. I managed to return an item on Amazon, and get all I needed within 2 and a half minutes. Then a few hours later an invoice/receipt popped up in my Amazon account. This got me wondering how does this all work? Is it cameras? Face recognition? Or are there people watching through the cameras what we but maybe? So I decided to do some digging and find out.
JWO Technology
Amazon calls the technology Just Walk Out technology or JWO. This is the entire system they use to consistently have cashier-less stores that work 99% time and have given them the opportunity to sell to other companies such as Sainsbury’s who have opened a JWO store in Holborn
Tech Used:
To have a fully working store Amazon use many different technologies including:
- Artificial Intelligence
- Machine Learning
- Image Recognition
- An array of “Fusion sensors”
- Decades of data on how humans shop
Fusion Sensors:
The fusion sensors is what amazon uses to work what items have been taken. This technology fuses the data from multiple sensors to make an overall decision on what has happened (i.e. a loaf of bread has been taken off the shelf)
They include pressure sensors and weight measurements to detect when an item has been taken off a shelf or placed back. Each item is uniquely identified using RFID tags. Hundreds of cameras are used to track your movement through the store this is so that the system can add items you take to your virtual basket. When entering the shop a person isn’t identified by their face (which would have definitely raised privacy issues) but rather through tracking from when they enter the shop, for example if someone were to wear a mask or remove a jacket during their shop this would all be tracked so the computer can still work who is buying what.
AI and machine learning:
AI algorithms have been taught to better understand how humans shop
For example how long it takes for someone to decide to take an item after picking it up, and the likelihood of someone putting the item back. This “self-learning” algorithm teaches the system the likely human interactions that take place to reduce the number of errors produced due to actions that seem random to the computer.
Summary:
Although it may seem useless to some JWO has it’s benefits in small convenience type stores or large supermarkets when the shopper is in a rush. However, I believe the main benefit for Amazon is to show their AI and machine learning technology to the general public and show it’s possibility to be used in situations that can benefit most.
