Art, especially painting, is fascinating. Many artists have always tried to reproduce the paintings of the great masters. If some are successful, an informed eye will almost always detect a small difference. For a machine, the difference would be even more glaring. However, a team of MIT researchers has developed an artificial intelligence capable of understanding the techniques used by painters.
Researchers at the Massachusetts Institute of Technology (MIT) computer science and artificial intelligence research laboratory (CSAIL) have developed an artificial intelligence (AI) algorithm, called Timecraft , capable of deducing the way in which a painting was made. and the different brush strokes used to create the painting.
AI was trained with 200 timelapse of paintings, to learn the different painting techniques used by artists. The goal of such an innovation? Better understand how a work is created. By determining the technique for creating the paintings , we can determine how the great masters produced their most famous masterpieces, and why not, learn to imitate them better.
This algorithm would allow researchers to better understand how a work is designed. Thus, they have already been able to figure out that painters tend to first work on the main lines of their works, before then looking into each detail (they therefore work from the largest to the finest). Similarly, the researchers were able to discover that one or two colors at a time are used for a single area of the work.
The result is still conclusive, since 50% of those questioned have already been “deceived” by the provenance of the work , confusing that made by the AI with the original work. Timecraft source code will be available soon. Do you want to reproduce the works of great painters?