NRF Vision 2020–Shekel Scales, the leader in AI-powered advanced weighing solutions, introduced Shekel Visual Recognition (SVR),embedded with Edgify’s machine-learning training framework. To significantly reduce loss at the point-of-sale (POS) and improve customers’ experiences at checkout, the solution leverages EdgeX to provide retailers access to data for real-time analytics. Powered by the serverless and cloudless framework of Edgify, the solution is plug and play without any additional infrastructure or hardware costs. The EdgeX compatible solution, enhanced by Shekel and Edgify will be on display at Intel’s booth #4637 at the NRF 2020 Vision show at the Javits Center.
Edge-X, a Linux Foundation solution, has been adapted for the retail sector, by collaborating between different sensors, including Shekel’s SVR and Edgify’s edge training solution for cloudless and serverless AI to allow for true autonomous self-checkout at the edge.
A recent SurveyMonkey study, sponsored by Shekel, determined nearly 75 percent of consumers frustrations with self-checkout stem from needing store personnel to intervene with transactions and difficulty entering produce and baked goods. And, nearly 70 percent of consumers would use self-checkout more frequently if the overall technology (ease-of-use, more accurate technology and category entry) was improved.
Shekel’s SVR solution improves the customer’s shopping experience by reducing errors and time at checkout. Partnering with Edgify, Shekel introduces the world’s first distributed, cloudless machine-learning solution that automatically recognizes products and allows the existing retailer hardware to teach itself how to recognize new products. The Edgify solution, introduced to the world of retail through Shekel, allows retailers to start using computer vision-based checkout processes without the headache of new infrastructure or any privacy concerns involved with cloud-based training.
“We’re honored to collaborate on Open Retail Initiative,” said Larry Susman, Vice President Sales and Marketing-Retail, at Shekel. “Our combined solution enhances the customer’s overall retail shopping experience and is a game-changer for retail self-checkout accuracy, efficiency and loss prevention and expands our footprint and presence.”
The Edgify Framework reduces near 95 percent of time at till and increases the accuracy of product recognition to 100 percent, which will reduce loss by close to 45 percent. Edgify can train on the edge, meaning that there is no need to transfer the data anywhere and the model can be trained as the data is generated, so retailers can train on 100 percent of their data instead of 0.1 percent data. The Edgify framework offers the following benefits to retailers:
- No extra hardware needed – all you need is a simple USB camera
- The entire Edgify framework can be downloaded through an API
- A complete cloudless solution – minimal infrastructure investment from the retailer side as everything is done on the edge devices that the retailer has in store
- No data collection – the Edgify framework uses the data generated at the store level only as its training data to create the model, so there’s no need to collect data in databases (or pay for databases as is done in most cases)
“This is the first time that a computer vision-based machine learning model will be trained live on the edge without any use of the cloud or servers, but only using the CPU of the self-checkouts (SCO). We also have developed the ability to distribute the model to other SCOs in the chain’s estate,” said Ofri Ben-Porat, CEO and Co-founder of Edgify. “There are currently no solutions in the market today that can combine models trained on different machines, using different classes, and different batch sizes, in order to create one optimized model without the use of any cloud provider/server.”
