MIT -graduate student Alex Kachkine once carefully spent nine months on repairing a damaged Baroque -Italian painting, giving him enough time to wonder if technology could accelerate things. Last week, MIT News announced his solution: a technique used by AI -generated polymer films to physically recover for damaged paintings in hours instead of months. The research appears in nature.
The Kachkine method works by printing a transparent “mask” with thousands of precisely colored regions that can apply curators directly to an original artwork. In contrast to the traditional restoration, which the painting changes permanently, these masks are reportedly removed when needed. It is therefore a reversible process that does not permanently change a painting.
“Because there is a digital record of which mask was used in 100 years, the next time someone works with this, they will have an extremely clear understanding of what was done with the painting,” Kachkine told Mit News. “And that has never been possible before in the preservation.”
Figure 1 from the paper.
Credit: MIT
Nature reports that up to 70 percent of the institutional art collections remain hidden from the public because of damage – a large amount of cultural heritage that is unseen in storage. Traditional restoration methods, in which curators fill a areas damaged by a time while mixing exact coloring competitions for each region, can last weeks up to decades for a single painting. It is skilled work that requires both artistic talent and deep technical knowledge, but there are simply not enough curators to tackle the backlog.
The Mechanical Engineering student came up with the idea during a Cross-Country Drive in 2021, when gallery visits revealed how much art remains hidden due to damage and recovery backlogs. If someone recovers as a hobby, he understood both the problem and the potential for a technological solution.
To demonstrate his method, Kachkine opted for a challenging test case: an oil painting from the 15th century that requires repairs in 5,612 individual regions. An AI model identified damage patterns and generated 57,314 different colors to match the original work. The entire restoration process reportedly lasted 3.5 hours-hevanger 66 times faster than traditional manufacturing methods.
Alex Kachkine, who developed the AI-beneficial film technology.
Credit: MIT
In particular, with the help of generative AI models such as stable diffusion or the “full application” of generative opponents (GANs) for the digital restoration step with the help of generative AI models. According to natural paper, these models cause “spatial distortion” that would prevent the correct coordination between the recovered image and the damaged original.
