Basler created a live demonstration of an automated retail check-out terminal equipped with AI software. The embedded vision system quickly identifies and classifies products in a shopping basket, and displays the pricing. This demo shows how combining AI and lean embedded vision systems can lead to innovative embedded applications that simplify our everyday life.
1. Overview
Embedded vision technology becomes smaller, more powerful and thus enables a wealth of new applications that rely on small size, low power consumption and low costs. One of the many areas where embedded technology can help to automate and simplify processes is retail.
One of the systems that will help to make the shopping experience easier are self check-out systems. Today those systems usually use 2D barcode scanners to detect and record products on conveyor belts.
More recent systems use traditional object classification methods (using features like color or type) to ensure precise identification of product features. These methods are not robust when deployed in unconstrained environments where different light and geometry conditions occur.
Those problems can be avoided when adding AI to the system: With latest AI technology it is possible to detect products without barcodes in a seamless environment and scale up the product portfolio easily on the fly.
2. Solution
For the embedded world tradeshow 2019 Basler created with the support of the close partnership with NXP Semiconductors a live demonstration of such an automated retail check-out terminal equipped with AI software. The embedded vision system quickly identifies and classifies products in a shopping basket, and displays the pricing. Customers can select what to put in their baskets, and the trained neural network then detects the products on the basis of a video stream – similar to how face recognition works – and finally the total pricing will be displayed.
2.1 Hardware
The AI retail solution is the upcoming new Basler Embedded Vision Kit. This state of the art edge processing system. Consisting of
– dart BCON for MIPI camera module from Basler with OnSemiconductor’s AR1335 sensor
– i.MX 8QuadMax SoC from NXP
– i.MX 8QuadMax based processing board from congatec
– SMARC 2.0 carrier board
The complete demo set-up is close to a real world application and incorporates all vision relevant factors.
2.2 Software
The software for this solution consists of two major parts: the system software and the application software
To make such an embedded system runs smoothly different elements of the system software need to be connected to create a coherent system. Basler used its long experience in different software fields to connect software elements like BSPs, NXP’s Yocto kernel and create a performant and lean embedded system.
The second part of the solution’s software is the application software. The software is based on a custom Convolutional Neural Network (CNN) by Irida labs. The model used is based on latest deep learning and edge processing techniques in order to provide fast and robust response in a seamless scenario. The training of the CNN takes place on the host side but the inference on the edge.
2.3 Advantages of this solution
This solution offers a robust system with a lot of advantages.
– This system is scalable. Products can be added easily to the system: a host-sided training allows for an easy deployment to the edge systems using IoT technology
– This robust system is optimized for vision, has a high accuracy and a fast inference time
– Hardware is Industry-proven and long-lasting
– Cost efficient, lean design due to embedded and IoT technology
– Small form factor allows for different integrating options
– Retail stores benefit from less labor costs as well as significantly improved customer experience through instant checkouts, minimized queues and 100% checkout capacity at all times, even when the shop is opened 24/7.
3. Conclusion
When creating this solution one of Basler‘s crucial tasks was to bring all relevant partners together and to combine the different elements to create a perfect solution. With Embedded Vision Solutions by Basler the customer only has one contact that takes care of all relevant coordination. In addition Basler takes care of optimizing each system for vision requirements and make different component work together seamlessly.
With the high flexibility that is necessary to create such a system Basler is the right partner to create complex embedded vision solutions.
