In our industry, printing packaging in a color correct way is very important. Package printing is often done with other inks than just CMYK. These inks are called spot inks. The red on a Coca-Cola package is printed with a dedicated red spot-ink. When two or more of these spot-inks overprint each other in specific percentages, a new color is created. For spot-inks, it is not possible to measure with a spectrophotometer the overprints of all possible spot colors because there are thousands of combinations. To predict the color that results from overprinting spot colors we currently have a mathematical/physical model. However, these models are not always close enough to the printing reality of the customer. The aim of the internship is to build a neural network that can predict how the overprint of spot colors will look like.

PyTorch knowledge is welcome