Fable.jl is an open research project with the goal of creating a new, general purpose rendering engine based on Iterated Function Systems (IFS). The principle is simple: an IFS can draw any object with the right set of equations, so if we can find that set of equations, we can render dynamic scenes.
This package uses KernelAbstractions.jl and is built with parallelism and performance in mind. Here is a quick list of various research directions for the project:
- Performance: Right now, Fable is using a naive chaos game kernel for all visualizations. It would be interesting to explore higher-performance methods for this. In addition, object rendering is currently done layer-by-layer and it would be interesting to be able to render multiple objects with the same IFS. Relevant issue.
- Generalizations: Right now, all fractals are made by first starting with a set of equations and then drawing those equations, but it would be interesting to explore the other direction. If the kernels are performant enough, we can do multiple IFS iterations and use some form of optimal control to dynamically learn equations from an input image. This would mean we could essentially turn any image into a fractal, similar to fractal compression. Relevant Issue.
- Synergy with other rendering methods: There are still advantages to using raytracing, raymarching, or rasterization, but performing clear analyses between the methods and using them together to create a general-purpose rendering library would be an interesting direction.
As it stands, the documentation will be a bit sparse until we hit v0.0.1, which should be in a few months, but if you have any questions, comments, or concerns, please create an issue or pull request!
