Specialists from the Space Research Institute of the Russian Academy of Sciences and employees of the Kirillo-Belozersky Museum-Reserve are considering neural networks as a way to restore the original appearance of the frescoes of the Nativity of the Virgin Cathedral of the Ferapontov Monastery, which is located in the Vologda region.
It is noted that they have already carried out multispectral shooting of the painting of the cult structure. According to the director of the museum, Mikhail Sharomazov, some of the paintings were lost due to the reconstruction of the temple itself.
Parts of the frescoes were damaged when a window was broken through in the 18th century
Neural networks trained on a large sample of frescoes will be able to generate images close to those that were in the cathedral. This work will be entrusted to Stable Diffusion - a neural network capable of creating pictures even based on a description.
The Ferapontov Monastery was founded in 1398. It was an important cultural and religious center of the Moscow Principality. Many of the frescoes in the monastery belong to the icon painter Dionysius, who is considered a successor to the traditions of Andrei Rublev.