Visual Components Detector - AI-powered product identification system to support error-free spare parts procurement
Use Case at a Glance
This specific use case has been identified within the Modernization of the procurement processes using AI-supported product identification.
The right product is the focus of every procurement process in companies in the manufacturing industry and trade, from the initial needs assessment to the actual use.
Our Key Learnings
We frequently sat together with our industrial customers via MS Teams and made the following key learnings: In every discrete step of this complex business process, companies discover various optimization potentials in order to realize their strategic goals, for example, the reduction of warehouse logistics through shorter delivery times and the integration of e-procurement solutions in the ordering process.
Challenge: Reliable recognition of objects
The procurement process demands high expectations on fast delivery and top quality. But this is exactly where the problem of wrong product delivery arises.
Often the incorrect identification of the required product turns out to be the cause of the problem, due to incorrect or missing identification options such as manufacturer-specific article numbers or labels. For many product categories, however, the provision of identification options by the manufacturer is not practical, especially for products with a large variety. The correct specification of the needed product variant is therefore essential for determining requirements.
At the beginning of our journey, Damir (@ddobric) and his MVP fellow Bahrudin Hrnjica (@bhrnjica) intensively tried to solve this problem with AI technologies like Cognitive Services. After a while, it was clear that the challenge was a very specific one.
Finally, they decided to use ML.Net Image Classification with TensorFlow to prototype a possible solution. Two months after the initial idea, they have defined what to do. It took us around nine months to deliver the first running solution.
We have built an image capturing, cropping and training system that enables customers to capture images of the object that will be used for training of the model. The training of the model is running in Azure as an Azure Function and published the trained model to the Azure Storage. All used Azure services are transparent for the user who is using the ASP.NET Core application to maintain the whole process. Additionally, the VCD Android and iPhone app is used for scanning and object detection.
Today the Visual Components Detector (VCD) is a platform that enables the unambiguous detection of products, components or any other object.
It is a multi-tenant solution hosted in Azure that enables a user to train and deploy multiple trained models, that recognize a specific subset of products (objects).
Moreover, the VCD platform provides an Android and iPhone app that is used for product recognition. The App is called Product Detector:
What is the benefit for our customers?
- Significant efficient savings in operational processes: Incorrect orders and replacement deliveries are significantly reduced because the VCD, as an integral part of the procurement process, determines the right product quickly and reliably using image recognition.
- On the procurement side as well as on the sales side, considerable cost and time saving effects can be realized: Sustainable cost reduction in the repair process through increased decision-making accuracy and integrated ordering mechanism
- Simplification of the workshop processes
- Ensuring the sustainability of the service partners
The right spare part for every vehicle
In cooperation with our customer, a family owned company and specialist in vehicle electrics and spare parts for Asian vehicle models, we have developed the first branded VCD app. This App is already being used successfully in the procurement process for vehicle spare parts.
Finally, we realized that this app technology can be used for a lot of daily products. We believe it is a rewarding and leading experience and the VCD journey goes on…..
Author: [Damir Dobric](Connect: https://damirdobric.me), CEO and lead architect of DAENET Corporation #VCD #Cloud #AI #IoT