Feeds:
Posts
Comments

Posts Tagged ‘artificial intelligence’

Photo: Remko de Waal/ANP/AFP via Getty Images.
Rembrandt’s restored ‘Night Watch’ at the Rijksmuseum in Amsterdam.

A project to restore a Rembrandt called “Night Watch” has received a lot of attention recently, but at the risk of repeating what you already know, I’d just like to point out that trimming a work of art can seriously affect its greatness.

How many times have building renovations cut paintings to fit or squashed them into too small a space to be properly appreciated. I think, for example, of the many special WPA paintings in US post offices that have been significantly altered over the years. I understand competing needs, but it’s a loss.

What was lost in Rembrandt’s ‘Night Watch,’ the New York Times says, was a sense of movement. The original was “asymmetrical: The large arch that stands behind the crowd was in the middle, and the group’s leaders were on the right. Rembrandt painted them this way to create a sense of movement through the canvas.

“Once the new pieces were restored, so was the balance, [said Rijksmuseum’s director, Taco Dibbits.] ‘You really get the physical feeling that Banninck Cocq and his colleagues really walk towards you.’ “

The main focus of the recent news coverage, however, was on how experts used artificial intelligence (AI) — along with an early copy of the original painting — to reimagine Rembrandt’s intentions.

Nina Siegal reported at the Times, “Rembrandt’s “The Night Watch” has been a national icon in the Netherlands ever since it was painted in 1642, but even that didn’t protect it.

“In 1715, the monumental canvas was cut down on all four sides to fit onto a wall between two doors in Amsterdam’s Town Hall. The snipped pieces were lost. Since the 19th century, the trimmed painting has been housed in the Rijksmuseum, where it is displayed as the museum’s centerpiece, at the focal point of its Gallery of Honor.

“[Now] for the first time in more than three centuries, it will be possible for the public to see the painting ‘nearly as it was intended,’ said the museum’s director, Taco Dibbits. …

“Rather than hiring a painter to reconstruct the missing pieces, the museum’s senior scientist, Robert Erdmann, trained a computer to recreate them pixel by pixel in Rembrandt’s style. A project of this complexity was possible thanks to a relatively new technology known as convolutional neural networks, a class of artificial-intelligence algorithms designed to help computers make sense of images, Erdmann said.”

As amazing as AI is, the work would not have been possible if a less renowned painter hadn’t made an early copy of Rembrandt’s work.

“Indications already existed of how the original ‘Night Watch’ likely looked,” Siegal continues, “thanks to a copy made by Gerrit Lundens, another 17th-century Dutch painter. He made his replica within 12 years of the original, before it was trimmed.

“Lundens’s copy is less than one-fifth the size of Rembrandt’s monumental canvas, but it is thought to be mostly faithful to the original. It was useful as a model for the missing pieces, even if Lundens’s style was nowhere near as detailed as Rembrandt’s. Lundens’s composition is also much looser, with the figures spread out more haphazardly across the canvas, so it could not be used to make a one-to-one reconstruction.

“The Rijksmuseum recently made high-resolution scans of Rembrandt’s ‘Night Watch,’ as part of a multimillion-dollar, multiyear restoration project, initiated in 2019. Those scans provided Erdmann with precise information about the details and colors in Rembrandt’s original, which the algorithms used to recreate the missing sections using Lundens’s copy as a guide. The images were then printed on canvas, attached to metal plates for stability and varnished to look like a painting.” More at the Times, here.

The Guardian also covered the story, quoting the Dibbits as saying, “With the addition especially on the left and the bottom, an empty space is created in the painting where they march towards. When the painting was cut [the lieutenants] were in the centre, but Rembrandt intended them to be off-centre marching towards that empty space, and that is the genius that Rembrandt understands: you create movement, a dynamic of the troops marching towards the left of the painting. …

“I am always hoping that somebody will call up one day to say that they have the missing pieces. I can understand that the bottom part and top might not be saved but on the left hand you have three figures, so it is surprising that they didn’t surface because at the time in 1715 Rembrandt was already much appreciated and an expensive artist.”

Read Full Post »

Photo: Oxia Palus and Lebenson Gallery London.
The Hidden Picture of Beatrice Hastings by Amedeo Modigliani was created by Oxia Palus using AI technology.

Nowadays, art and science work hand in hand. Consider this story about how artificial intelligence was used to reveal an unknown painting by a great master. It starts with the practice of “overpainting.”

Suzanne’s art professor overpainted because he wanted you to sense what was underneath. But at Hyperallergic, Lauren Moya Ford writes, “Artists paint over their finished canvases for many reasons — out of frustration at a failed design, because they lack the funds to buy more material, or even to spite whoever or whatever they’ve depicted.

“The latter was the case in Amedeo Modigliani’s ‘Portrait of a Girl‘ (1917), an oil painting of a sullen, seated brunette now held in the collection of the Tate. X-ray studies of the canvas conducted by the museum in 2018 revealed that the piece was originally a full-length portrait of another woman, a slender blonde with angular, elongated features. A portion of this hidden painting — now on view at Lebenson Gallery in London — was uncovered and reconstructed by two scientists using a combination of stereoscopic imaging, artificial intelligence technology, and 3D printing.

“Neuroscientist Anthony Bourached and physicist George Cann joined forces in London in January 2019 to found Oxia Palus, a scientific project that uses machine learning to reconstruct what the duo calls ‘NeoMasters,’ or artworks that have been previously hidden from view under the layers of later paintings. Their past efforts have uncovered a Blue Period nude by Picasso, a Madonna by Leonardo da Vinci, and a landscape painting by Santiago Rusiñol that was later painted over by Picasso, the artist’s friend and mentee. To discover these ‘lost’ works, Bourached and Cann apply a neural style transfer algorithm to X-rays of paintings that are suspected to have another artwork hidden below their surfaces. The technology utilizes imagery from the scan, as well as information from the artist’s other works, to reproduce colors, brushstrokes, and other distinguishing features.

“Unlike conservators or other art specialists, Bourached and Cann bring uniquely non-art areas of expertise to the pieces they analyze.

‘George’s inspiration comes from his research on the surface of Mars for the detection of life,’ Bourached explains in a recent email to Hyperallergic. …

“Who was the woman whose likeness has suddenly been unearthed more than 100 years later? She’s thought to be Modigliani’s ex-lover and muse, the English poet, writer, and literary critic Beatrice Hastings. … The two years that the couple shared an apartment in Montparnasse were creatively productive for both: Hastings published prolifically, and is known to have posed for at least 14 of Modigliani’s portraits. But their relationship was also plagued by alcohol addictions, explosive personalities, and violent confrontations. …

“It was perhaps to symbolically scorn his former lover that Modigliani painted over her portrait in 1917, but, thanks to the two London scientists, Hastings has found a way to see the light again. As she wrote in 1937, ‘Civilized woman wants something more than to be the means to a man’s life. She wants to live herself.’ ” More at Hyperallergic, here.

Who gets the last word about what an artist shows to the world? At some point, the work no longer belongs to the artist but to the public. The only way an artist gets final say, I suspect, is to have some acolyte like Jane Austen’s sister Cassandra, who burned all the novelist’s letters after her death. Cassandra thought that whatever her sister wanted done was more important than what posterity might want.

Read Full Post »

Photo: Ismoon
The earliest recognized form of “written” communication may have been small bits of clay, called tokens. This charming one is from the Indus Valley.

Nowadays, one reads almost too much about artificial intelligence, AI. I myself have an ever increasing list of things I’d rather not have AI managing for me. But using it to translate ancient texts is one application that seems to make perfect sense.

Ruth Schuster writes at Haaretz, “Understanding texts written using an unknown system in a tongue that’s been dead for thousands of years is quite the challenge. Reconstructing missing bits of the ancient text is even harder. …

“Filling in missing text starts with being able to read and understand the original text. That requires much donkey work. Now an Israeli team led by Shai Gordin at Ariel University in the West Bank has reinvented the donkey in digital form, harnessing artificial intelligence to help complete fragmented Akkadian cuneiform tablets.

“Their paper, ‘Restoration of Fragmentary Babylonian Texts Using Recurrent Neural Networks,’ was published in the Proceedings of the National Academy of Sciences in September.

“ ‘Neural networks’ … means software inspired by biological nervous systems. The concept dates back more than 70 years. … The base concept is to teach machines to learn, think and make decisions. In this case, the computer decides on the plausible completion of missing text. …

“Gordin and the team feed their machine transliterations of the extant Babylonian texts i.e., what the text would have sounded like.

“Then what? When it comes to missing bits in a papyrus or tablet, humans can intuit that “’…ow is your moth…’ isn’t a query into the well-being of your mothball.

With machines, it’s all about mathematics and probabilities based on knowledge gained so far. …

“It may have been trading that inspired the earliest recognized form of communication: ‘pseudo-writing’ on small bits of clay in Mesopotamia around 7,000 years ago. The clay bits, called tokens, were shaped into simplistic imagery such as a cow or other ancient commodities. …

“Then we start seeing abstract signs; repetitive strokes or depressions are interpreted as numbers (price, perhaps); and possibly also personal names, using the first sounds of different imprints to put together words you can’t draw. …

“Anyway, after pseudo-writing came proto-writing: figurative proto-cuneiform inscribed on tablets, which arose about 5,500 years ago in the city of Uruk. … Within mere centuries, proto-cuneiform evolved to become increasingly schematic and Sumer was apparently where it happened, [Gordin] says. And figurative hieroglyphic script began to appear in ancient Egypt at about the same time, about 5,000 years ago. …

“By the time cuneiform became a thing, writing had passed the stage of ‘Sheep : four : Yerachmiel’ and reached the stage of official records, letters and formulaic recounts of the wondrousness of the ruler. …

“For cuneiform, we have the gargantuan multilingual text at Behistun, Iran. Darius the Great had his exploits described in three different cuneiform scripts. [The] Behistun text was monumental: 15 meters (49 feet) high by 25 meters wide, and 100 meters up a cliff on the road connecting Babylon and Ecbatana, all to describe how Darius vanquished Gaumata and other foes. …

“And over decades, linguists slowly interpreted the languages of Babylon and Assyria, thanks to Darius’ monumental ego. …

“Interpreting a dead language is a mathematical game, Gordin says. … Neural networks are a computerized model that can understand text. How? They turn each symbol or word into a number, he explains. …

“When humans reconstruct missing text, their interpretation may be subjective. To be human is to err with bias, and quantifying the likely accuracy of the completion is impossible. Enter the machine. …

“The machine proved capable of identifying sentence structures – and did better than expected in making semantic identifications on the basis of context-based statistical inference, Gordin says. Its talents were further deduced by designing a completion test, in which the machine-learning model had to answer a multiple-choice question: which word fits in the blank space of a given sentence.”

Not sure how many readers are into that kind of thing, but I do find it intriguing. More at Haaretz, here.

Read Full Post »

%d bloggers like this: