GeneticTextures.jl

GeneticTextures.jl is a Julia package inspired by the paper Artificial Evolution for Computer Graphics by Karl Sims. It allows you to interactively evolve textures using genetic programming techniques, bringing the power of evolutionary algorithms to the world of computer graphics.

Installation

To install GeneticTextures.jl, open your Julia REPL and type the following commands:

using Pkg
Pkg.add("https://github.com/jofrevalles/GeneticTextures.jl.git")

Usage

using GeneticTextures

n = 6
width = 128
height = 128
max_depth = 3

mutation_probs = Dict(
    :rand_expr => 0.06,
    :adjust_scalar => 0.1,
    :adjust_color => 0.1,
    :rand_func => 0.08,
    :add_argument => 0.1,
    :become_argument => 0.08,
    :duplicate_node => 0.03,
)

# Create initial population
original_population, original_image = generate_population(1, primitives_with_arity, max_depth, width, height)
population, images = create_variations(1, original_population, mutation_probs, primitives_with_arity, max_depth, width, height)

# Display original and mutated images
display_images(original_image[1], images)

# Interactively evolve textures
let population = population, images = images
    while true
        best_choice = get_user_choice(n)
        chosen_image = images[best_choice]

        if best_choice === nothing
            break
        end

        population, images = create_variations(best_choice, population, mutation_probs, primitives_with_arity, max_depth, width, height)
        display_images(chosen_image, images)
    end
end

Examples

Here are some examples of textures generated with GeneticTextures.jl:

Benchmarks

We also benchmarked the performance of the different function backends available in GeneticTextures.jl.

Acknowledgments

GeneticTextures.jl makes use of the following third-party packages:

• CoherentNoise (MIT License)
• Colors (MIT License)
• ForwardDiff (MIT License)
• Images (MIT License)
• Plots (MIT License)