blood vessel like fractals

[jakubMitura14]

Hello I am looking for a 3d dichotomic irregular fractal that would look like branching blood vessels and similar to branching planes - to simulate connective tissue layers
The algorithm would need to be parameterized and differentiable.
The basic plan is to learn the parameters of the fractal by comparing l2 similarity between the original and fractal-generated image (additionally for fine-grained texture Fourier will be used.

Would it be possible to use your tool for it ?)

[flpgrz]

For the parallelisation and the learning, using Jax is surely a good idea. fractal-jax employs the just in time (JIT) compilation to speed up things.

However, this library, as it's implemented now, does not operate any learning. The equation that generated the fractals is given, i.e., z = z^2 + c.

If I correctly understood your task, you could substitute the equation that generates the fractal with a parametric function and learn the parameter by optimising a loss.

[jakubMitura14]

Thanks for answer! yes exactly - I do not expect for you to support the training per se I just need differentiable parametrized functions :). Maybe I am not very precise - generally, I am working on some models in Jax and already have an optimization loop, I had never worked with fractals but images of fractals seem quite similar to natural tissue images. Hence my question.