Nerf.jl

Real-time Neural Radiance Fields in pure Julia.

Requirements

• Julia 1.9 or above.
• AMD or Nvidia GPU (check AMDGPU.jl or CUDA.jl for respective device support).
• At least 4-6 GB of VRAM.

GPU backend

Use Nerf.set_backend!(name) to specify which GPU backend to use, which takes a String with backend name:

• AMD GPU: Nerf.set_backend!("AMDGPU");
• Nvidia: Nerf.set_backend!("CUDA").

And then restart your Julia session if returns true.

Alternatively, if you cloned Nerf.jl repo, you can create LocalPreferences.toml file and specify backend in it with:

[Nerf]
backend = "AMDGPU" # or "CUDA"

Quickstart (NerfGUI.jl)

For standalone application check NerfGUI.jl package, which provides OpenGL application for real-time interaction.

1. Clone git clone https://github.com/JuliaNeuralGraphics/NerfGUI.jl.git.
2. Start Julia repl julia --project=. --threads=auto.
3. Instantiate project with ]up command.
4. Select GPU backend as described above.
5. Launch application:

julia> using NerfGUI

julia> NerfGUI.main()

References

1. Instant Neural Graphics Primitives with a Multiresolution Hash Encoding: https://nvlabs.github.io/instant-ngp/.