Dynamics 365 Commerce & Machine Learning

Automation of the purchasing process – product recognition function

The modern running of retail stores is not as obvious as it was couple of years ago. Strong competition on the market, which does not stand still and constantly introduces innovations to attract customers, is a big challenge for brands operating stores, which often do not keep up with trends. Consumers nowadays are much more demanding, and their expectations are not only a good product, but also a high comfort of shopping. Therefore, in order not only to retain existing customers, but also to attract new ones, retailers must respond to the needs of a modern consumer by introducing innovative solutions in their stationary stores.

One such solution is the automatic placement of products taken from the shelf in the shopping cart of Dynamics 365 system.

Dynamics 365 helps ensure that consumers are satisfied at every stage of the purchasing process

Enable hassle-free shopping

Deliver personalised and intuitive digital trading environments.

Stand out from the competition and automate actions using AI

Bring analysis to all areas of your retail and sales data.

Give yourself more flexibility and the ability to adapt faster

Meet your redefined needs on a scalable trading platform.

Image recognition – Azure Custom Vision

The modern functionalities of Dynamics 365 Commerce are complemented by Azure services. We use Azure Custom Vision to automatically update the basket in the POS upon the product removed from the store shelf.

Azure Custom Vision is an image recognition service that allows you to create, deploy, and improve your own image ID models. Image ID labels according to their visual characteristics. Each label represents a classification or object. Unlike an image processing service, Custom Vision allows you to define your own labels and train custom models to detect them.

Custom Vision uses a machine learning algorithm to analyse images. We send groups of images with specific properties. Images are tagged with their own custom labels (tags) at the time of upload. The algorithm then trains on this data and calculates its own accuracy by testing itself on the same images. After the model is trained, you can test, retrain, and finally use it in an image recognition application to classify images or detect objects. You can also export the model itself for offline use.