Creation of color recipes using AI « Future Center KI NRW

The company Robert GmbH (Exact Colors) from Emsdetten has been producing colors and color mixtures since the 1950s. One of the company’s complex tasks is to find out and produce the recipe for a specific color mixture. With around 10,000 colors per year, the company has so far been able to rely on its employees with many years of experience and “a good eye”. Software also supported the process. In view of the shortage of skilled workers and the foreseeable departure of many employees into their well-deserved retirement, the company is forced to test the possibilities of artificial intelligence (AI). To this end, managing director Benjamin Robert turned to the AI ​​NRW Future Center to ask to what extent a model trained on machine learning can take on this complex task and whether the model learns from past mistakes.

Approach and solution

During the 10-day intensive consultation of the “Zukunftszentrum KI NRW” project, the goal was to investigate how well an AI can predict a specific color mixture based on the “reflection curve” measured by the company.

There is not always just one right solution for remixing a customer’s color. It can happen that different color pigments are used for the same target color, which can lead to the same result. In order to enable the process of color mixing, a responsible person previously had to have the knowledge to be able to say which colors should be considered for a color mixture of a target color and which of the different color mixtures should be preferred. Here, an AI that is able to predict a color mixture makes work significantly easier for newcomers, but also for current experts.

In order to test whether the existing data was sufficient for the case described and what had to be done to enable AI support, the existing data was analyzed, cleaned and expanded. Various models were then tested using machine learning to try to best represent the rules for generating a specific color mixture.

The first prediction goal was to predict which color pigments would actually be needed for a mixture. The second goal was to investigate whether an AI could predict how much of a color pigment should be used for a color mixture. Since the total amount of a color pigment depends on the general target amount, the goal was not to predict the amount in grams, but rather the percentage of each pigment for a mixture.

In addition to two individual AIs, it is also possible to combine both AIs. The aim is to predict whether a color pigment is needed and in what quantity.

Live test and conclusion

During the final discussion, the developed prototypes were subjected to a live test. A reflection curve was measured for colors with which the AI ​​was not trained, for which the various approaches were supposed to make a prediction as to which pigments were needed and in what quantities. The prediction of the mixture was then mixed together and compared with the target color by the expert. The goal of this test was to determine how well the AI ​​can apply what it has learned to new colors and how satisfactory the results are from an expert perspective.

“We have seen that it works with artificial intelligence.”

Ben Robert, Gmanaging director Robert GmbH

It can be concluded that an AI that determines pigments and their required quantity represents a work support in everyday life that can particularly benefit people who do not recognize at first glance which color pigments are needed for a mixture and which of the various possible mixtures represents the optimal first mixture. But such a forecast can also make work easier for experts in the field, as an initial mixing can be carried out more quickly, which means that the fine-tuning of the mixture starts more quickly, which requires a certain level of color expertise.

Due to the satisfactory results that were achieved during the intensive consultation, the company’s motivation is even greater to delve further into the topic of AI and to actively use AI as tested in everyday business life. In order to implement this, contact with another funded project was established at the end of the consultation by the Future Center’s pilot function.

In addition, further measures are planned with the Future Center to impart further knowledge on the topic of AI and digitalization. We look forward to future collaboration and are excited to see how the project we were able to start together develops!

Back