Mario Klingemann: Cameraless Photography with Neural Networks

Video of talk from @mario-klingemann at The Photographers Gallery, London, on the subject of making Art with neural networks:

Coinciding with his current project on our Media Wall, Mario Klingemann will be in conversation with Daniel Rourke discussing his recent work which employs machine learning to create visual imagery in a process he calls “Neurography”. In recent months, attention has been drawn to the use of adversarial neural networks to produce entirely fictitious photographic images and their potential use and mis-use in a post-truth age. Reflecting on the ways in which machines are being trained at an increasing rate to learn new skills - from creating to interpreting images - Mario will unpack the creative possibilities presented by neural networks. 

Some of Mario’s work is currently being shown at the gallery - you can find out more here

Look at walls splashed with a number of stains, or stones of various mixed colours. If you have to invent some scene, you can see there resemblances to a number of landscapes, adorned with mountains, rivers, rocks, trees, great plains, valleys and hills, in various ways. Also you can see various battles, and lively postures of strange figures, expressions on faces, costumes and an infinite number of things, which you can reduce to good integrated form. This happens on such walls and varicoloured stones, (which act) like the sound of bells, in whose peeling you can find every name and word that you can imagine. Do not despise my opinion, when I remind you that it should not hard for you to stop sometimes and look into the stains of walls, or the ashes of a fire, or clouds, or mud or like places, in which, if you consider them well, you may find really marvelous ideas. The mind of the painter is stimulated to new discoveries, the composition of battles of animals and men, various compositions of landscapes and monstrous things, such as devils and similar things, which may bring you honor, because by indistinct things the mind is stimulated to new inventions.

Leonardo da Vinci

CycleGAN feedback loop

Visual experiment by @mario-klingemann resembles a combination of DeepDream and cellular automata, generating abstract patterns with facial features:

Two CycleGAN models trained on face transformations are passing their outputs back and forth creating an infinite loop of ever-changing Turing patterns. 
This is an experiment where 2 CycleGAN models are pitted against each other and transform their reciprocal output images in a feedback loop. The models have originally been trained on transforming faces which is why you see a lot of eyes and noses being generated. The process is seeded with an initial image but it quickly descends into a semi-stable state that reminds of complex cellular automata or Turing patterns.