A single RTX 3090 graphics card is enough to run the code.
- Create an virtual environment using conda:
conda create -n cora python=3.9
conda activate cora
- Install pytorch and torchvision. Here we use
cuda 11.6
as an example. For other cuda version, please find the correspondingwhl
file path at here.
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple \
https://download.pytorch.org/whl/cu116/torch-1.13.1%2Bcu116-cp39-cp39-linux_x86_64.whl \
https://download.pytorch.org/whl/cu116/torchvision-0.14.1%2Bcu116-cp39-cp39-linux_x86_64.whl
- Install pytorch3d. Again, if you want to install pytorch3d in other cuda version, please find the corresponding file path at here.
conda install https://anaconda.org/pytorch3d/pytorch3d/0.7.4/download/linux-64/pytorch3d-0.7.4-py39_cu116_pyt1131.tar.bz2
- Install tinycudann:
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch
- Install other libs:
pip install nerfacc==0.3.1 \
moviepy \
kornia \
pyyaml \
lpips \
tensorboard \
scikit-learn \
PyMCubes==0.1.2 \
trimesh \
pymeshlab \
fake-bpy-module-3.1 \
networks \
scikit-image==0.19.1 \
fvcore \
iopath \
pymeshfix \
pyfacer \
av \
pims \
timm \
face_alignment \
mediapipe \
opencv-python==4.5.2.52 \
opencv-python-headless==4.5.2.52 \
chumpy \
numpy==1.22.4
- Install Blender for UV unwrap:
cd blender
# Download Blender
wget https://mirror.clarkson.edu/blender/release/Blender3.1/blender-3.1.0-linux-x64.tar.xz or download from https://cloud.tsinghua.edu.cn/f/44a204b6c5824133ad92/?dl=1
# extract files
tar -xvf blender-3.1.0-linux-x64.tar.xz