Artificial intelligence is developing rapidly. Recently, we wrote about what AI is capable of today: it recognizes faces, creates new drugs and even writes music. And this, of course, is far from all. This time, the gaze of artificial intelligence fell on the picture of Pablo Picasso's “Old Guitarist”. However, the first thing that appeared on the canvas was the silhouette of a seated woman. In 1998, scientists looked under the top coat of paint in the painting and found there the original Picasso drawing. Today, thanks to artificial intelligence, you and I can enjoy this hidden work of art.
Pictures inside paintings
As much as you and I would not like this, but the artiststhey do not so often surprise us with their works immediately after death, but sometimes there are exceptions. Among them are paintings by such talented artists as Leonardo da Vinci, Rembrandt, Van Gogh and Pablo Picasso. The fact is that these artists often painted on top of ready-made paintings. A large number of cases are known when specialists found out sketches or even whole pictures under the final paint layer.
You can find out what other AI is capable of discovering on our Telegram news channel.
To see the original image,applied by the artist to the canvas, experts used a method called radiography, which uses x-rays and infrared light. This method previously allowed us to determine that under two paintings by Pablo Picasso, “The Poverty Squatting” and “The Old Guitarist”, other works were painted. Also in 2014, researchers discovered a portrait of a man with his hand on his head under Picasso's “Blue Room” (1901). However, the reconstruction of the image, carefully concealed by the artist, was a problem. Art critics could only guess what a particular work of art looked like.
The painting “Old Guitarist” was painted by Picasso in1903 - 1904 Under it, the artist depicted a seated woman. Her left hand reaches for the viewer, and her right hand is pressed to her knees. It is also curious that it corresponds to the sketch that Picasso sent to a colleague at the same time. So far, all that could be found in the lost picture is a faint trace of the silhouette of a woman. Any additional detail - such as color and style - was lost.
How neural networks recreate works of art
However, using a trained neural networkto distinguish the style of one artist from another or (in the case of artists such as Picasso) to distinguish one period from another, researchers from University College London breathed new life into the picture, recreating the original Picasso.
The basis for the neural network was the technique thatIn 2015, Leon Gatis from the University of Tübingen (Germany) developed it, called the transmission of neural style. Machine vision technology can determine the style of a painting and transform it into a style of another painting. Researchers Anthony Burahead and George Canne published an article on arXiv preprint server. During the study, scientists took X-ray images of the outlines of a woman and connected them to a neural network created to convert images into the style of La Vie (1903), another masterpiece of the blue period of Picasso.
By the way, with the participants of our Telegram chat you can discuss not only scientific discoveries, but also works of art.
Although the result cannot accurately recreate the picture,created by Picasso over a century ago, the resulting image corresponds to the style in which the artist painted his paintings in those years. The authors of the study hope that the results of their work will inspire similar projects to search for lost works of art.