Better copies of paintings using AI and 3D printing

"If you simply reproduce the color of a painting as it appears in the gallery, it may look different in your home. Our system works in all lighting conditions, which shows a much greater color reproduction capability than almost any previous work, "said researcher Changil Kim of the computer and computer lab. artificial intelligence of MIT (CSAIL).

The system was dubbed "RePaint", and would be "more than four times more accurate than the most modern physical models to create the exact color shades for different works of art," according to MIT , as a result of an oil test paints created by an artist collaborating in research.

Instead of the usual four printing inks (cyan, magenta, yellow and black), RePaint uses 10 different transparent inks stacked in thin layers by the 3D printer, plus the traditional halftone, using tiny ink dots instead of a continuous layer of color. This combination has been called "contonant color" and, apparently, it allows to better understand the nuances of the colors of the painting.

A deep learning algorithm has been formed to predict optimal ink overlays.

The printing takes so long that so far only a few inches of surface have been produced, although according to MIT, the team expects advanced commercial 3D printers to speed up the process.

Future projects include extending the available ink library (colors including "cobalt blue" that has so far deceived RePaint) and creating a paint-specific algorithm for selecting inks. The team also hopes to get better details to take into account aspects such as surface texture and reflection, in order to achieve specific effects such as glossy and matte finishes.

"The value of art objects has increased rapidly in recent years. It's becoming increasingly common to lock them up in warehouses, away from public view, "said Mike Foshey, mechanical engineer. "We are building technology to reverse this trend and create accurate, inexpensive reproductions that everyone can enjoy."

RePaint will be presented at ACM Siggraph Asia in December.

CSAIL worked with Professor Szymon Rusinkiewicz, Professor of Computer Science, Princeton University.