-
Notifications
You must be signed in to change notification settings - Fork 199
Windows install tutorial (alpha)
Install Cuda 9.0 (https://developer.nvidia.com/cuda-90-download-archive)
Install Cudnn 7.0.5 for Cuda 9 (https://developer.nvidia.com/cudnn you need to create an account).
See the documentation on how to install cudnn @ http://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html#download-windows
Install Python 3.6.4 (https://www.python.org/downloads/)
Check if correct Python is installed
python --version
Should say 3.6.4
Install virtualenv for windows
pip install virtualenv
pip install virtualenvwrapper-win
Install git for Windows Download and install from https://gitforwindows.org/
Create virtualenv
mkvirtualenv faceswap
workon faceswap
Get faceswap git repo Choose your desired repo. There are several. Default is at: https://github.com/deepfakes/faceswap (still on virtualenv faceswap)
cd faceswap
There might be some errors when when installing the requirements. For example Scikit-image, dlib and pyyaml might throw errors, so we need to install them from different sources i.e. prebuild binaries.
Scikit-image is kindly provided by https://www.lfd.uci.edu/~gohlke/pythonlibs/#scikit-image. In our case, we need the file "scikit_image‑0.13.1‑cp36‑cp36m‑win_amd64.whl".
This information is taken from "https://github.com/deepfakes/faceswap/issues/71"
Download to the current faceswap directory you are in and install with:
pip install scikit_image-0.13.1-cp36-cp36m-win_amd64.whl
Install dlib (older version but precompiled)
pip install dlib==19.7.0
Install Cmake for Windows -> https://cmake.org/download/
Clone dlib sources
git clone https://github.com/davisking/dlib.git
cd dlib
To build with CUDA support and AVX:
mkdir build
cd build
cmake .. -DUSE_AVX_INSTRUCTIONS=1
cmake --build .
This could throw an error You have CUDA installed, but we can't use it unless you put visual studio in 64bit mode. if you have Visual Studio installed.
Your VS should be updated to the latest version. I.e. Update 3 because of c++ compiler issues from previous VS versions.
If that happens type:
cd ..
cmake -G "Visual Studio 14 2015 Win64"
I guess this will use the Visual Studio c++ compiler.
Latest Dlib should now be compile with CUDA and AVX support.
Now we can install the requirements from the txt file:
If you don't have a CUDA card install requirements-python35.txt or requirements-python36.txt depending on your installed python version. If you follow this tutorial you should use requirements-python36.txt
For CUDA card use "requirements-gpu-python36-cuda9.txt"
Go to your faceswap directory
cd ..
pip install -r requirements-gpu-python36-cuda9.txt
This installs all dependencies the author defined in this textfile.
You should now be able to run faceswap.py via this command:
python faceswap.py -h