Skip to content

Latest commit

 

History

History
47 lines (40 loc) · 1.48 KB

installation.md

File metadata and controls

47 lines (40 loc) · 1.48 KB
Raw

Installation

You can either use the dockerfile here or follow the instructions here.

First, recurively clone the git repo and it's linked dependencies:

git clone --recursive git@github.com:shubham-goel/ucmr.git

Next, create a conda environment and install python dependencies

conda create -n ucmr python=3.7 anaconda pytorch=1.1 torchvision=0.3 cudatoolkit=10.0 -c pytorch
conda activate ucmr
pip install absl-py tensorflow tensorboard tensorboardX opencv-python==4.1.0.25 dotmap dominate meshzoo==0.4.3 moviepy visdom chainer

Other Dependencies

The following additional dependencies need to be installed: PyMesh, SoftRas and NMR. For convenience, here's how you can install them:

PyMesh

Note that building pymesh requires CMake1.15

apt install -y libgmp-dev libmpfr-dev libboost-dev libboost-thread-dev
git clone --recursive https://github.com/PyMesh/PyMesh.git
cd PyMesh
./setup.py build
./setup.py install

SoftRas

git clone https://github.com/ShichenLiu/SoftRas.git
cd SoftRas
python setup.py install

NMR

# Install using pip if CUDA9
pip install neural_renderer_pytorch

# Install from source if CUDA10
pip install cupy-cuda100
git clone https://github.com/daniilidis-group/neural_renderer/
cd neural_renderer
python setup.py build
python setup.py install