MENU



The demonstration consisted of a small gated shop floor with a footprint of about six square meters, packed with shelves, monitored by a set of four cameras mounted on the ceiling feeding their video streams to a local rack sporting Intel’s edge-AI hardware.

In the demonstration, pre-registered shoppers first had to identify themselves by scanning a phone ID QR-code to unlock the gate. Once in the shop, they could pick and choose whatever items sitting on the shelves, or even change their mind and place back some items on the shelves.

AI-enabled computer vision algorithms running across the four cameras’ video streams automatically identified and tracked the items, adding them to the shopper’s virtual shopping cart and charging users the right amount upon exit through the mobile payment method of their choice. The AI algorithms were developed by 2017 startup CloudPick using Intel’s OpenVINO (Open Visual Inference & Neural network Optimization) toolkit, itself specifically designed to take computer vision and deep-learning inference to vision applications at the edge.

Shailesh Chaudhry, product manager for retail at Intel explained eeNews Europe that while CloudPick licenses its algorithms to retail managers, its business model is to provide Fulfillment-as-a-Service (FaaS), enabling efficient order fulfillment picking operations for brands and retailers. As well as supporting automated check-out, Cloudpick’s Intelligent Retailing Platform (C-IRP) also automates inventory analysis and monitors shoppers’ traffic which can help retailers optimize store layout for example.


According to Chaudhry, the four camera setup in a small shop alleyway worked reliably with up to 25 shoppers at a time, without inducing AI-inferencing errors from vision occlusion. Catering for more people at a time is just a matter of increasing the computing power, he said.

Such cashier-less grab-and-go grocery retail stores could deliver 24/7 service in hotels, high-rise apartment blocks and office complexes, but Chaudhry even envisions that such autonomous shops could become mobile. This could be implemented as a container-type mobile shop operated at different fixed locations and replenished by the retailer’s logistics. In the future, one could even imagine that a fleet of autonomous driving shops would drive to your door step, offering on-demand corner shops.

CloudPick has already established some partnerships with large retail corporations and it has 30 installations operating in China. The company is now talking to retailers in the US and in Korea. The retail market has always been a big vertical for Intel, and it is getting bigger and bigger, Chaudhry concluded.

Related articles:

Deep-learning promises fully automated check-out

Neural Computing on a stick: deep learning hardware becomes affordable

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News

Share:

Linked Articles
10s