Who is the author of the painting? Ask Panther, a national software that automatically learns to identify the painter of a work of art and situate it in an artistic current. Developed at the University of Aveiro (UAveiro), Panther promises to be a tireless feline when it comes to discovering the true authors of works or hunting down forgers.
Caravaggio, Klimt, Miró or Picasso, but also the lesser known O’Keeffe, Seurat or Schiele, there are already dozens of artists that Panther can identify through the analysis of the digitized image of the painting, potentially recognizing the authenticity of the work and, as a true human expert, in which aesthetic trend it belongs.
Created at the Institute of Electronics and Informatics Engineering of Aveiro (IEETA), one of UAveiro’s research units, the software bears the signature (true and authenticated!) of researchers Jorge Silva, Diogo Pratas, Rui Antunes, Sérgio Matos, and Armando Pinho.
“Panther contains an algorithm that measures probabilistic-algorithmic information from artistic paintings and uses it to describe how each author typically composes and distributes elements across the canvas, and therefore how their work is perceived,” explain doctoral students Jorge Silva and Rui Antunes. The software, they add, “also allows you to identify patterns and hidden relationships present in art paintings and perform a classification, namely author and style identification.”
By analyzing more than 4,200 scanned images of paintings – a number that may grow to include more and more artists and their works – Panther creates what the research team calls a fingerprint of the author. In essence, the software manages to approximate the unique and untransferable way in which each author distributes the elements of his or her painting on the canvas.
This information, the paper’s authors point out, “is very descriptive and together with compression modeling and neural network software can be used for author and style classification purposes with great accuracy.”
The authors believe that in the future these kinds of tools “will become less dependent on the presence of an expert to do authentication of works.” However, the researchers warn, “the presence of an expert human being cannot be ruled out in the near future because there are counterfeiters capable of making forgeries of works at a very detailed level that sometimes deceive experts in the field themselves.”
Jorge Silva reminds us that Panther only uses a digital scan of the image: “A painting is a four-dimensional and physical object in which certain characteristics, such as the thickness of the paint used and its age, are difficult to be observed only by photography”, he admits.
The work of the IEETA team is accompanied by the website http://panther.web.ua.pt/ that shows, among other information, a complete catalog of the fingerprints by each author, as well as several examples of paintings by each of them. This catalog allows us to understand how the average local complexity (the fingerprints) is an adequate means of explaining the content of the artwork, as well as being valuable for attribution and validation of artwork authorship. On the website you can also see the phylogenetic tree of the relationship between the complexity of the work of the various authors, which allows you to see the influences and techniques shared by the painters at various times in their artistic lives.