This repo generates automatically LOD and normal map of high resolution mesh by using differentable rasterization, this repo can generate statically LOD and normal from high resolution mesh input. These contain total two processes.
- Create low resolution mesh of the high from template mesh.
- Create normal map of low resolution mesh by difference between high and low
cuda 11.3 windows
nvdiffrast,
I didn't test the other version of cuda.
The utilized differentiable renderer is nvidia differentiable rasterizer.
High-level functionality was implemented by pytorch. The other dependencies are included in environment.yml, which not indicated here.
git clone --recursive https://github.com/hyeonjang/auto-static-lod
with conda
window
INSTALL.bat
conda create env --file environment.yml
ubuntu
bash INSTALL.sh
conda create env --file environment.yml
will be updated
conda activate lod
python train.py