Evolutionary art
Evolutionary art is a form of generative art, are produced at the works of art from the fields of visual arts, music and performing arts with evolutionary algorithms. Evolutionary algorithms are methods to solve optimization problems with principles of natural evolution. By art processes to be construed as an optimization, objects may thus be created, the aesthetic effect to humans. Due to time constraints, the implementation of this class of algorithms is mandatory with the computer could in principle also be calculated by hand. Therefore Evolutionary Art belongs to digital art.
Basics
Basis of evolutionary art is a population of individuals, herein represented by each visual structure as with all evolutionary algorithms. This representation can be done either indirectly by individuals such as in genetic programming each contain a program that creates a visual tree, so that here the biological distinction between genotype and phenotype is maintained. The representation can also be directly, as in the evolution strategy by an individual is considered only as a phenotype, are applied to the evolutionary operations. In this case, an individual has a picture, drawing, moving image or the like includes for the purposes of an image file or video file.
Art process
Almost all applications of evolutionary art that use indirect representations that produce non-figurative visual works. Regardless whether direct or indirect representation, there are few approaches for objective evolutionary art.
In evolutionary art process, a starting population of individuals is first defined. Random programs and thus random visual structures generated - For an indirect representation are - as usual with Genetic Programming. In a direct representation usually non-random visual structures are selected by the artists, such as images from previous evolutionary runs.
The following is a reproduction phase, in which the individuals are increased in accordance with a present reproduction strategy by recombination and mutation operations are applied to the representation structures. The nature of these operations depends on the type of the program or direct visual structures, such as for example, linear and generally require individuals hierarchical structures each adapted recombination and mutation operations on the evolutionary algorithms.
Part of the strategy is the reproduction manner in which individuals are selected for recombination ( selection for the reproduction ). Oriented to the reproductive strategy of genetic algorithms so must first be fitness values for each individual. The frequency of selection for reproduction is a strictly monotonic function of fitness, that is, the higher the fitness, the higher the probability of selection. Oriented to the reproductive strategy of evolution strategies, the selection is equally distributed randomly.
After the reproduction phase results in a population of offspring, each fitness value must be specified for the intended to reflect the aesthetics of the visual structures in any way. An algorithmic definition of these values would require a formal aesthetic model, which is not present or with previous methods for evolutionary art only in its infancy. Therefore, algorithmic methods limited to determining simple properties of the image analysis and instrument models, such as entropiebasierte models. Widespread is the definition of fitness by a human or a group of people (interactive evolution). This is usually the artist who defines the ratings after his subjective aesthetic criteria. Alternative methods of estimating the empirical fitness example, the time considered a viewer a visual tree that is presented to him. In addition, there exist pre-conscious process in which an attempt is made, a correlation between physiologically measurable characteristics of an observer and to derive its aesthetic reviews (e.g., pupil reactions). The most innovative approaches offers here neuroesthetics be identified in brain regions that are involved in aesthetic and reviews are to be produced at the correlations between the activities of these regions and aesthetic Reviews ( analogous methods as neuromarketing ). However, since these approaches require complex and very expensive equipment for medical imaging, their use is limited in evolutionary art so far on small isolated studies.
Do parents and offspring each fitness value, then a selection strategy used is determined by which an individual may exist in the next generation and will possibly reproduce. This selection strategy takes into account either the offspring or the union of parents and offspring. Accesses in addition no stopping criterion, such as reaching a predetermined maximum number of generations, the next iteration of the evolutionary art process is started with a new reproductive phase.
Non- photorealistic rendering
An application of evolutionary art is the non- photorealistic rendering, an area of computer graphics by graphics deliberately not be faithfully represented their physical image. One example is the generation of an artificial painting from a photograph. The British scientist Collomosse and Hall 2005 developed an algorithm that produces paintings from photographs. A painting is conceived as a series of brush strokes, with brush strokes are defined by attributes such as position, direction, color, etc.. A genetic algorithm is used to search the space of all possible in this manner paintings. The fitness function that assigns a quality each candidate solution, the edge image of a candidate compares with a calculated at the beginning Salienzbild. The salience of an image detail indicates how obvious it is to a human viewer. In the algorithm of Collomosse Hall and the salience of image details is composed of three factors: rarity, degree of visibility, and a third factor, the previously learned the taste of users in some areas, to distinguish important from unimportant artifacts for people.
The Salienzberechnung based on the idea that works of art " no mirror " pose (Ernst Gombrich ) of reality, but rather an interpretation of the artist.