It's rare for artists to spring up surprises long after their demise, but some occasions are noteworthy. Think Van Gogh, Rembrandt, Leonardo da Vinci, and Picasso.
Beneath the Old Guitarist (1903-04), Pablo Picasso’s popular Blue Period painting is the figure of a woman in faint markings. She was originally noticed in 1998 when X-rays and infrared light were used by conservators of the Art Institute of Chicago to peep behind the top layer of the paint.
She sits with her right arm tucked onto her lap and her left arm reaching toward the viewer. Strangely enough, she resembles a sketch Picasso sent to a colleague at the same time the painting was made.
The foundation for the neural network was a technique created in 2015 by Leon Gatys at the University of Tübingen, Germany, called neural style transfer. The machine vision technique can indicate the style of a painting and convert it to match another.
For example, it can convert a Paul Cézanne to the style of Michelangelo. This technique has been used on videos as well, so it is possible to watch clips of Ice Age in the style of Van Gogh's Starry Night.
All that could be recognized of the woman in the painting until now was the slightest trace of her outline. Other intricate details like style and color were lost.
Nevertheless, with the use of a neural network trained to differentiate the style of one artist from the other or one period from the other, researchers at University College London have given her a new life, even if it is not true to the original of Picasso.
Researchers George Cann and Anthony Bourached published their paper titled "Raiders of the Lost Art," on the arXiv preprint server. The paper was also featured in a recent article in the MIT Technology Review.
The X-rayed images of the woman’s outlines were plugged into a neural network built to convert images into the style of La Vie (1903), another exceptional piece of Picasso's Blue Period. In as much as the result cannot be a perfect match of Picasso’s painting which was also pained over 10 decades ago, it is at least conforming to the style he used at that period.
In Picasso's The Crouching Beggar, the same technique was also used. It was painted over a landscape thought to have been made by a Spanish painter, Santiago Rusiñol, using Rusiñol's Parc del Laberint d'Horta.
Cann and Bourached said that they hope the results help projects on other lost works of art and also challenge the popular doubt that "algorithm can be creative, or innovate."
"We argue that the use of machine learning as an artistic tool can ultimately broaden creative insight and widen the landscape of inspirational ingenuity by human-AI collaboration," wrote Cann and Bourached.
The Crouching Beggar and Old Guitarist are not the only Picasso works that cover hidden paintings. Earlier pieces are usually painted over by artists, especially when a fresh canvas is hard to come by due to inadequate funds. A portrait of a man resting his hand on his head underneath Picasso's The Blue Room (1901) was found by researchers in 2014.