Abstract
Printing conditions ensure that prints are produced consistently and to a high standard by describing all relevant parameters of the printing process. This is particularly important in industries where print quality is critical, such as in the production of packaging and labels.
In this paper, we present a novel approach to the classification of printing conditions using machine learning models. Our method involves the analysis of spectral measurement data to determine technical metadata such as the printing process, substrate type and print sequence. The accuracy of this data is crucial for setting up physical models, which can be used to predict overprints.
Previously, this data had to be manually entered which was error prone. However, our models can support the user by predicting this information accurately and quickly to ultimately ensure a high quality of prints.
We demonstrate the effectiveness of our method through automated and manual experimentation. Additionally, we describe the setup of an evaluation in a production environment.
