Student Creates Ethical AI to Restore Damaged Artworks

MIT student's creation could help bring countless old paintings, currently stored in the back rooms of galleries with limited conservation budgets, to light.

Courtesy of the researchers at MIT

Restoring damaged works of art to their former glory can often take years, and some parts of the process are done by hand. But innovative tech, developed by an MIT mechanical engineering graduate student and described in a recent paper, may advance the process by producing faithful restorations in just a few hours.

That would save a fortune for cash-strapped institutions and enable hidden away artworks in need of restoration the chance to see the light of day (or, certainly, that of the gallery) again.

Alex Kachkine’s new method deploys AI to generate a digital version of the artwork in its prime, which is used to create a map of the required retouches. Then, the digital restoration is printed onto a thin polymer film, a “mask” that sticks to the painting and applies the retouches - and that can be easily removed. Plus, a digital version of the mask can be saved to help future conservators understand the specific changes that were made.

In a demonstration, Kachkine repaired 5,612 separate areas of a damaged 15th-century oil painting in 3.5 hours - about 66 times faster than traditional methods. “My hope is that conservators around the planet will be able to use these techniques to restore paintings that have never been seen by the general public,” he said, adding, “This is perhaps one of the rare cases where the application of AI to transfer styles and drawings from other artists … is actually very ethical.”

Here's Alex Kachkine demonstrating his invention...