This painting below looks almost exactly like the work of the famous Dutch painter Rembrandt, and as far as science and mathematics is concerned, it is a Rembrandt. But instead of oils and brushes, researchers from Microsoft and other organizations used data points and 3D printing.
To create this new Rembrandt, the team meticulously scrutinized 346 Rembrandt masterpieces to get a sense of not only the painter’s style, but types of subjects, positions, facial features, and typical outfits. After analyzing these paintings down to the very pixel, they determined that their new Rembrandt should be a portrait of “a caucasian male with facial hair between 30 or 40 years old in dark clothing and a collar wearing a hat with his face pointing to the right,” according to the video below.
With that framework in mind, influenced by a huge database built with statistical analysis and algorithms analyzing Rembrandt’s technique, the team set to work comparing single features, like an eyeball or the slight tilt of the head. To add a 3D dimension—after all, paintings are layers upon layers of paint—a height map created the final painting using a 3D printer.
Essentially, this painting is what you’d get if you mixed together almost every Rembrandt portrait to create one single painting, and the result is impressively realistic. But this Rembrandt, while a testament to the work of a genius, also denies the common idea that data analysis is meant only for business analytics or weird lab experiments. Just like a paint brush, in the right hands it can make a masterpiece.
The Next Rembrandt is a collaboration between:
ING / Microsoft / TU Delft / Mauritshuis / Rembrandthuis