Getting Out of Line: AI Lets Shoppers Avoid Long Waits at Checkout

Shopping in the future may feel a lot like shoplifting does today — without the risk of getting nabbed — if two artificial intelligence startups have their way.

New Zealand’s IMAGR and Silicon Valley’s Mashgin aim to make checking out of grocery stores and company cafeterias a walk in the park. Almost literally.

Many supermarkets offer self-checkout to save shoppers time. IMAGR founder William Chomley wants to skip the checkout altogether, so you can just walk right out the door. It’s similar to the idea behind Amazon Go, being tested in a grocery store in downtown Seattle, which lets customers shop without ever stopping at a cashier on the way out.

IMAGR makes SmartCart, an ordinary grocery cart with an AI computing video camera attached. The device tracks what goes into the cart, tallies the total along the way and syncs that with payment information on the shopper’s mobile phone.

“We want to give people the ability to shop as they normally would, and then just walk past the cashier and out of the store,” Chomley said.

High Noon at the Checkout Counter

Mashgin was born out of frustration over lunch breaks spent waiting in lines rather than chatting with friends. It’s installed its automated checkout system, also called Mashgin, in several Silicon Valley company cafeterias, including NVIDIA’s. Using GPU deep learning and computer vision, it recognizes your soup, salad or soda faster than you can gulp.

The elegant Mashgin self-checkout station features a very simple user interface. Customers simply place their lunch on the device, where five 3D cameras examine it from different angles to identify and price each item. To pay, customers swipe a credit card.

Demonstration of a future version of the Mashgin AI cafeteria checkout.
This animation depicts a future version of the Mashgin AI cafeteria checkout. Currently the device detects packaged goods, soups, salads and takeout containers, but is still being trained to identify foods on a plate. Animation courtesy of Mashgin.

The startup trained its system on a dataset of common items found in cafeterias, using the CUDA parallel computing platform, NVIDIA GeForce GTX 1080 GPUs and cuDNN with the Caffe deep learning framework. Mashgin customizes its system for each company’s cafeteria, and its deep learning algorithm learns new items as more people use it.

“It’s a huge market and there’s this big problem,” said Abhinai Srivastava, who founded the company with Mukul Dhankhar. “Everyone wants to eat at 12 o’clock.”

Catching Rays, Not Delays

IMAGR’s Chomley created SmartCart because he wasn’t getting enough sunshine. Stuck behind his computer screen at an investment fund most days, he yearned to spend a few minutes soaking up rays during lunch. Instead, the line for food at a small grocery near his office ate up his entire break.

Chomley quit his job and began work on what is now SmartCart. After several false starts — at one point, he had to take a job moving furniture to keep the company afloat — he and the IMAGR team set their course on deep learning and computer vision to enable SmartCart.

Using our TITAN X GPU and the TensorFlow deep learning framework, IMAGR initially trained its algorithms on images of grocery store products. Next, it used the SmartCart video camera to learn to recognize products put into or removed from the cart — say you reconsidered that half-gallon of chocolate chocolate-chip ice cream over a second bunch of kale. Finally, the team trained the algorithm on barcodes to learn prices.

IMAGR is planning a small SmartCart trial at a New Zealand grocery chain within the next couple of months. Chomley said several of the world’s largest supermarket chains have expressed interest in SmartCart.

“People just don’t want to be standing in huge lines,” he said. “They want to get in and get out.”

To learn more about how AI computing is changing industries, subscribe to NVIDIA’s AI Podcast on iTunes http://nvda.ws/2hQ4Leb or Google Play Music http://nvda.ws/2hQaIrh.

Similar Stories