Image genetics

 

Can images be created by a genetic process, comparable to the biological ones ?

Can these images be "beautiful" or at least, original and interesting ?

The Roxame software, and the PAP (Primeval Art Project)

propose an experimental approach.

Photography . Textures

Roxame crosses an image with a set of textures.

The mating operation consists in

- transforming the original image into a map of classes (grays, vivids, blues...),

- replacing the original pixels by the corresponding ones in one texture image.

We get :

Photography . Geometrical forms

We cross an image with a geometrical generation algorithm :

- we segment the original image,

- for each region, we generate a rectangle and/or an ellipse, drawn or painted,

- (to get it more appealing) we apply some gradations and get :

The PAP project

Initiated by Michel Bret and Pierre Berger within the Algorists group

Primeval Art Project (PAP) aims to create a "primeval soup" where

- autonomous generative processes create new images

- select the "best ones" in a Darwinian mood.

Present PAP results

At present, the soup contains works from several artists, members of the group.

The first results use either :

- Roxame algorithms, using several original images provided by members.

- neural network processes developped by Michel Bret.