A differentiable rasterizer
- Render primitive at random pose and save the results.
julia --threads=auto --project=. examples/primitive.jl
Antialiased | Render | UV | Triangle ID |
---|---|---|---|
![]() |
![]() |
![]() |
![]() |
- Learn vertex positions and colors of a randomly initialized cube.
julia --threads=auto --project=. examples/cube.jl
cube.mp4
- Julia 1.8+ is needed.
- Supports only instance mode for now.
- Batch matrix multiplication is done using
NNlib.⊠
operator, for AMDGPU we'd need to add that. rocBLAS should have the necessary stuff.