API

https://github.com/caseman/noise/blob/bb32991ab97e90882d0e46e578060717c5b90dc5/_noise.h#L29-L61

Atom(config::Config)
Atom([rng=GLOBAL_RNG], config::Config)

Atom struct representing the data structure proposed in Visual Atoms: Pre-training Vision Transformers with Sinusoidal Waves.

Arguments

• config::Config: A configuration object containing the parameter ranges.

Returns

• Atom: A new Atom instance with randomly generated properties.

Atom(rng::AbstractRNG, config::Config)

Arguments

• rng::AbstractRNG: The random number generator to be used.
• config::Config: A configuration object containing the parameter ranges.

Returns

• Atom: A new Atom instance with randomly generated properties.

Config

A struct representing the configuration for generating atom instances and images.

Config(config_path::AbstractString)

Create a new Config instance by reading and parsing the configuration file at the specified config_path.

Arguments

• config_path::AbstractString: Path to the configuration file (TOML format).

Returns

• Config: A new Config instance with properties read from the configuration file.

generate_instances(config::Config; save_root::AbstractString, category_id::Int, num_instances::Int)

Generate instances of Atom images based on the configuration object andsave them to the specified directory. The images will be saved in a subdirectory named after the category ID.

Arguments

• config::Config: The configuration object containing parameter ranges for the Atom generation.
• save_root::AbstractString: The root directory where the images will be saved.
• category_id::Int: The category ID, used for generating the seed and naming the output subdirectory.
• num_instances::Int: The number of instances (images) to generate and save.

Returns

• This function does not return a value; it saves generated images to the specified directory.

pnoise1(x)

Perlin noise – pure Julia implementation. The results of this computation are consistent with the Python package's noise.pnoise1 function.

render!(rng::AbstractRNG, atom::Atom, canvas::AbstractMatrix)

Render an atom onto a canvas, modifying the canvas in-place. The atom's properties and a random number generator are used to generate the image.

Arguments

• rng::AbstractRNG: The random number generator to be used.
• atom::Atom: The atom instance with properties defining its appearance.
• canvas::AbstractMatrix: The canvas (matrix of color values) onto which the atom will be rendered.

Returns

• canvas::AbstractMatrix: The modified canvas with the rendered atom.

save_images(rng::AbstractRNG, atom::Atom; save_dir::AbstractString, num_instances::Int, H::Int, W::Int)

Render and save images of the Atom instance atom onto a canvas with specified dimensions. The images will be saved in the specified directory.

Arguments

• rng::AbstractRNG: The random number generator to be used.
• atom::Atom: The atom instance with properties defining its appearance.
• save_dir::AbstractString: The directory where the images will be saved.
• num_instances::Int: The number of instances (images) to generate and save.
• H::Int: The height of the canvas.
• W::Int: The width of the canvas.

Returns

• This function does not return a value; it saves generated images to the specified directory.

uniform([rng=GLOBAL_RNG], a::T, b::T) where {T<:AbstractFloat}
uniform([rng=GLOBAL_RNG], a::Real, b::Real)

Generate a random number in the interval [a, b), uniformly distributed, using the provided random number generator rng. The arguments a and b must be of the same subtype of AbstractFloat.

Arguments

• rng::AbstractRNG: The random number generator to be used.
• a::T: The lower bound of the interval (inclusive).
• b::T: The upper bound of the interval (exclusive).

Returns

• T: A random number in the interval [a, b).