Smart Shopping:How Does Cashier-less Technology work

While the Researcher fraternity is still in the early stage of exploring XAI or explainable AI, the Amazon research team has utilized the power of Computer vision, Deep learning AI, video streaming, Big Data Analytics to bring the magic of smart shopping, into reality, the cashier-less Amazon Store. It works like an autonomous vehicle that recognizes the objects in its paths employing computer vision through deep learning algorithms and takes prompt and appropriate actions. Not only just Amazon, but there are also over 150 companies including Microsoft, that are working on human-free retail using machine learning, IoT, RFID, and more.  

The cashier-less retail stores are usually equipped with thousands of cameras, IoT sensors that are located across the store. The computer vision and underneath deep learning algorithms allow customers to shop, then leave the store without waiting in line to pay. However, buyers must use a smartphone application to enter the cashier-less store. The smart app prompts the user to scan the QR code for authentication. The shopping inside the store may be facilitated with a smart shopping cart equipped with barcode readers sensors and scales that scans groceries, links to online shopping lists. Customers after logging into the shopping cart place shopping bags in the basket. The cart can scan items with a bar code and weigh barcode-less products. Customers then exit through a specific sensor-enabled lane that automatically charges the credit card on their Amazon account. The app also records all the data about a user’s shopping and then through association mining a well-known concept from Big Data Analytics identifies the user shopping behavior for further recommendation and promoting its business.

The architecture of the cashier-less store provides services that include computer vision, streaming service, Entry/Exit Detection, Payment &Recievce, IoT, and Sensor fusion and collection services. The underneath Computer vision technology makes full use of deep learning algorithms for person detection, object recognition, pose estimation, human Activity detection, and analysis. In addition to this sensor fusion techniques are used to aggregate signals across the different sensors. A huge set of Labeled data is required for most computer vision problems such as activity detection, pose estimation. The research team at amazon has followed a similar technique used by Deep mind to train AlphaStar. They generated a massive training set through simulation and trained the deep learning algorithms utilizing the power of cloud computing.

While Amazon cashier-less store heavily employs machine learning, deep learning, and computer vision, a starter company “Caper” has introduced a smart AI featured shopping cart to make AI-powered shopping more accessible. The weight-sensitive shopping carts equipped with sensors and screens can identify what customers put into their carts. The screens provide interactive maps around the store, lead customers to promotions, and allow customers to pay directly from the cart. This form of AI shopping experience is more manageable to implement across already existing supermarkets.

AI with computer vision and Big data analytics is turning the shape of the business, however poses questions related to data privacy, ethics, security, validity, and more importantly will this technology be able to change the future of the high street. The AI research community at the University of East London is actively participating to find the answers to these problems. The technologies mentioned used in Amazon cashier-less framework are just some of the technologies that are taught in career-led courses offered by the University of East London. Please visit https://www.uel.ac.uk/subjects/subject-areas/computer-science-and-digital-technologies" for further details.