Neural Art – Pollock pattern

Vincent Brinkmann / Cologne — Jun 24, 2017

Neural Art – Pollock pattern

uses Deep Convolutional Generative Adversarial Networks (DCGAN), which is developed by Facebook AI Research. Neural Art is an Artificial Intelligence which generates art images from a Google search of Jackson Pollock acktion paintings.

Can a neural network generate art?

Action Painting or gestural abstraction is an American art movmend, and is closely associated with abstract expressionism, and was significant in the 1940s to 1960s. Jackson Pollock (1912-1956) is one of the most important representatives in this field.

Jackson Pollock paint no longer the paint with the brush, but used wooden sticks, trowels, and dropped them from cans. Similar to Tachism and the Informel on the European continent, Action Painting sought access to unconscious painting.
The predecessors were in the literature of the Dadaismus and, above all, the French surrealism with its Écriture automatique. This attempt to eliminate consciousness by means of an automated procedure – based on Freud’s theory of the unconscious. Action painting renounces the planning and organization of an image content and the figurative representation. The process is the goal. Randomes plays a significant role during the process. The unconscious should pretend the direction. The whole body is used. It is a physical dive with color.

Neural networks are a network combination of neurons that learn through a process of recurrence. In the process of recurrence, information is abstracted and downsampled. This allows a neural network to process information more easily. Random values play an importend role in the initialization of the learning results.

The intuitive act of action paitings is the counter concept to neural networks and the most elementary difference. Neural networks can not handled intuition.

The Neural Art has methods of abstraction and random methods to become an unique Neural Art Paintings and draw with a certain lerned experience.
So how big is the difference really?

The aim of this project is to create new paintings from a database of »Action Paintings« using the described neural networks and to investigate the question what is Neural art?


Vincent Brinkmann is a new media artist, he studied »Digital Media« at the University of Art Bremen and »ExMedia« at the Academy of Media Arts Cologne. He is interested in the connection between virtual and physical, to bring the digital in physical space and to create a connection between code and material. In his works, technology takes an important role in how we act in a world with more and more technology. Notions like, »sculpture,« »space« and »material« a common themes of Vincent’s Works.

Find more contributions in the archive