Automated Self-Checkout

Artificial Intelligence revolutionizes the store experience — from automating inventory tasks to improving the efficiency of retail processes — ultimately providing deeper insights into retailers’ decision-making.

Across the retail sector, businesses struggle to keep up with labor shortages, rising costs, and customer expectations. AI solutions can have a significant impact in overcoming these challenges, helping businesses thrive in a competitive market.

In this article, we explore how AI-powered solutions are leveraged to optimize retail operations today, and a kit that we created to help you get started.

Bringing AI-Powered Solutions to the Store

To keep up with changing customer demands, retailers leverage innovative AI-powered solutions that target customer-centered experiences and operational efficiency. Let’s explore these in detail:

Customer-Centered Experiences

For customers, AI streamlines the checkout experience with solutions such as Intelligent Queue Management, enabling store managers to open new registers and deploy cashiers. The shared theme across these kinds of AI solutions is for customers to get in and out of the store as smoothly and quickly as possible.

Another example is optimizing the customer self-checkout process to make it much faster. While automated self-checkout has become mainstream, it’s often plagued with challenges such as incorrectly reading barcodes and triggering an alert for a store associate to come help — creating a headache for customers and a bottleneck for the business. By leveraging AI to uplevel and transform these solutions, retailers offer automated self-checkout that streamlines the whole process, provides a better user experience, and frees up store associates to handle other meaningful tasks. By applying sensors and AI to video cameras, customers can pick up items and simply leave the store, automatically and accurately charging items without the checkout process.

The benefits of these types of solutions can extend beyond queue management and automated self-checkout to the operational opportunities AI can unlock.

Operational Efficiency

By leveraging AI toward operational efficiency, retailers can:

• More precisely keep track of items and inventory so they know when shelves need to be restocked.
• Track foot traffic to enable them to strategically place items across their stores.

AI can perform other vital functions like theft detection, workforce management — ensuring employees are allocated to the most urgent tasks at each moment — and customer behavior analysis to better understand purchasing trends in real-time.

Building an End-to-End Retail AI Solution

Along with AI, computer vision is a popular technology to make all this happen. Computer vision models can be used to detect, track, and analyze items, inventory, and customers. Developers are key to making these opportunities a reality.

Thanks to existing infrastructure, many stores are already equipped with hardware, cameras, and other devices to apply these AI-based functions. All a developer needs is their laptop and the right software tools to bring these types of solutions to life.

With the latest Jupyter Notebook, we showcase exactly how developers can build on their models and solutions once they are familiar with the capabilities.

In this object detection example, we use software solutions like OpenVINO™,Roboflow’s supervision library, and Ultralytics YOLOv8 — a state-of-the-art object detection model. YOLOv8 enables the application to track and detect objects in real-time. These software tools provide the foundational building blocks necessary to develop an automated self-checkout system, showing developers how they can build a real-time object detection and tracking application that provides retailers with valuable analytics.

We use the OpenVINO toolkit to optimize the YOLOv8 models into a smaller footprint. This makes the models able to run efficiently on Intel® hardware and edge devices with less latency and increased runtime. Finally, we use the Roboflow supervision library to define zones for objects, which enables retailers to follow an item or customer and gain insights into what items are most popular and how inventory moves across the store. This information can then be used to create new and innovative applications for inventory management, self-checkout kiosks, and barcode scanning.

Building on the Automated Self-Checkout Solution

Developers can take this application to the next level by fine-tuning their models, creating custom datasets, and combining multiple models to achieve more than just detection and classification. Let’s look at two sample scenarios exploring how the application can be built on.

Recognizing Products Faster

One example scenario is a model that can distinguish between individual products, such as a peanut butter jar and a soda bottle.

We created the Automated Self-Checkout Retail Reference Implementation as the next step from the Edge AI Reference Kit that developers can use to deploy AI models for automated self-checkout that can:

• Recognize non-barcoded items faster.
• Recognize the product SKU and items placed in transparent bags.
• Reduce the steps involved in identifying products when there is no exact match.

Reliable AI Models for Retail

Another example scenario is with models that can compensate for one another. For example, if you extend your automated self-checkout pipeline to include a barcode detection model, barcodes often can be smudged, peeled off, or partly covered by some other sticker, to the point that the barcode detection model fails to recognize it. In this case, another object detection model can potentially recover that failure by looking at the shape, weight, or other features of the item to provide reliable input.

Extending AI Models Beyond Retail Operations

The Jupyter Notebook is just a starting point for exploring the possibilities of OpenVINO, Ultralytics YOLOv8, and Roboflow. This example can be easily adapted to other industries, such as counting cars in a parking lot, tracking inventory and supplies in a healthcare setting, or ensuring worker safety on a factory floor.

We look forward to seeing what other applications and industries developers are helping to transform with these solutions. To learn more, please check out our Edge AI Reference Kits, and join the OpenVINO discussion on GitHub.

Additional Resources

OpenVINO™ Documentation

OpenVINO™ Notebooks

Provide Feedback & Report Issues

Notices & Disclaimers

Intel technologies may require enabled hardware, software, or service activation.

No product or component can be absolutely secure.

Your costs and results may vary.

© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.