Will artificial intelligence be the ultimate forger? A new array of the Dutch master appeared on the Net, it was created by algorithms and a 3D printer.
346 artworks were scanned, 50 GB of data analysed by Machine Learning algorithms to create “The Next Rembrandt”, a chart that the painter has never painted but which incorporates the style in a purely scientific manner. This is as much the art?
Unlike the psychedelic rantings of the algorithms of machine learning of Google, the portrait generated by algorithms of Dutch researchers is rather convincing. The generated table “weighs” 148 million pixels and took more than 500 hours of calculation. Researchers have implemented Microsoft Azure computing power to perform these calculations.
In the preamble to this project, researchers conducted a detailed statistical study of the works of the artist over a period of 10 years, from 1632 to 1642. This study showed a predominance of portraits of men Caucasian, bearded, aged 30 to 40 years, dressed in dark color and coated a flange and a hat. On average, the majority of the characters painted by Rembrandt are face and look on their right. The study went further still by measuring the size of the eyes, nose, ears, medium on portraits positions but also the thickness of the strokes of the master of the dark light. 2 algorithms have allowed researchers to generate the textures 3D painting, texture that was produced in 3D printing and go up to 30 print layer to find the brush stroke.
The result is rather convincing. Found well du clair obscur characteristic of the Dutch painter and the composition, a simple portrait, is plausible, even though the character seems devoid of any sense.