In 2016, researchers from St. Petersburg State Electrotechnical University LETI had the idea of creating a robot artist. The goal of the project was to teach the machine to paint in such a way that it could use paints and brushes to convey the style characteristics of an artist or artistic movement. And it was important that in creating a picture, the result of the robot's work would have minimal differences from the original.
Development of algorithms for receiving different colours was a nontrivial task, as regularities describing mixture of paints with high accuracy were not yet investigated. Considering the principles of working with a palette, which are used by real painters, scientists decided to mix not more than four pure colors simultaneously. This allows you to get the most vivid and saturated shades, and at the same time simplifies the task from a mathematical point of view. They then did a series of experiments in which the robot mixed pure colors in different ratios, yielding more than 600 samples. Only five pure colours were used - black, white, red, blue and yellow. The resulting shades and the proportions of colours that correspond to them have been used to create a mathematical model which reproduces the regularity of mixing of colours. Based on this model, an algorithm has been developed that allows the robot artist to automatically create any colour.
The machine has already successfully painted several pictures, some of which have been generated by a neural network. A paper based on the results has been submitted to a prestigious robotics science journal and is currently under review.
"In the long term, our robot could become a mechanical assistant to a professional artist, a machine for high-precision restoration of paintings and creation of reproductions. If we talk about the fundamental importance of our work, the development undoubtedly has great potential as a tool for testing new machine learning algorithms in the field of creativity," explains Artur Karimov.
The development of the artist robot was supported by a grant from the Russian Science Foundation (No. 22-79-00171).