Machine Learning & Computer Vision

Smart Data Processing for Modern Processes

Would you like to reduce your scrap rates and implement modern monitoring for your production processes? Are you looking for ways to automate and digitize your production workflows? I would be happy to help you develop a customized machine learning solution tailored to your specific needs.

What impact can artificial intelligence have on manufacturing?

Neural networks form the basis of intelligent computer vision models. They are trained using large amounts of data and can be used to make predictions on additional (new) data. In image processing, image data is used as the input format for the networks, and the expected output is classification, object detection, or image masks. An often underestimated alternative is classic image processing algorithms, for example for edge detection or to identify quality defects based on brightness differences in the image.

Classification

Detection

Segmentation

In many applications, automation using neural networks for quality control and process monitoring can reduce scrap, minimize safety risks in production, and cut down on downtime. This saves time, costs, and effort. Feel free to contact me—together we’ll find the best solution for your application!

Impact in Production

Production monitoring not only allows safety-critical incidents to be tracked, but also actively prevents them. For example, a crane system is automatically shut down as soon as a person is detected beneath the crane hook. AI-powered monitoring also provides valuable support in quality control by reducing the scrap rate and preventing product recalls through early and automated inspections. Other applications include level monitoring, seal inspection, and predictive maintenance. This helps prevent breakdowns and allows for targeted planning of maintenance downtime!

Further information

Our collaboration – Your result

Once we have defined the problem, please send me a few sample recordings or data. I will then develop various approaches based on these, which we will discuss together in the next step.

With this iterative method of programming and review, we get closer and closer to finding the right solution for your process. Working together and consulting with each other ensures that the software ultimately meets your expectations and requirements. It doesn't help anyone if I present you with a solution for something you didn't want in the first place.

Clear Communication

Individual Software

Collaborative reviews

For reviews, we can meet online or directly at your premises. Especially when it comes to lighting and image capture issues, on-site appointments offer unbeatable advantages. We can look at the problem together and consider where cameras, light sources, or other sensors need to be added. For software adjustments and discussions about the layout of the user interface, online appointments are also a great option and often save time and stress.

On-site commissioning, as well as knowledge transfer and training for your team on-site, are of course included—please feel free to contact me about this!

Ultimately, what matters most to me is that you’re satisfied with the solution and always have someone to turn to for bug reports, suggestions for improvements, or urgent support. I look forward to long-term partnerships and exciting challenges!

Free initial consultation!

Frequently Asked Questions

The initial consultation is always free! Depending on your requirements, a feasibility study takes 1–2 weeks and costs approximately €3,000. Afterward, you’ll know whether a computer vision system is a worthwhile investment for your application and what parameters should be considered during development. The cost of a complete system depends heavily on the development effort and starts at €15,000.

There must be sufficient usable data available to serve as the basis for a machine learning or computer vision system. Alternatively, methods such as data augmentation, simulated data, or transfer learning can be used.

Most neural networks are trained on GPUs (graphics cards), but in production they run in inference mode on CPUs. Therefore, there is no need to purchase special graphics cards to operate intelligent computer vision systems. The infrastructure required for training is already included in the price.

Fully developed systems can be integrated via desktop applications or web interfaces. Communication via OPC UA or TCP/IP protocols is also possible to implement a smart factory approach. Dashboards and reports summarize the system’s results and present them in a consistent manner to the user.