Windows Installation Guide - YuvalNirkin/fsgan GitHub Wiki

Dependencies

Install all of the following:

Latest NVIDIA driver
CUDA Toolkit
Python distribution like Anaconda
Git client

FFmpeg

Download and extract the FFmpeg binaries
Download and extract the OpenH264 library to the "bin" directory under the FFmpeg root directory.
Add the FFmpeg "bin" directory to Windows path:

Example

Python packages

conda install pytorch torchvision cudatoolkit=10.1 -c pytorch -y
conda install -c conda-forge yacs -y
pip install opencv-python ffmpeg-python

v1 only:

conda install -c 1adrianb face_alignment

Training only:

pip install tensorflow tensorboardX

FSGAN

Suggested directory structure

mkdir -p C:\data
mkdir -p C:\data\dev\projects
mkdir -p C:\data\datasets
mkdir -p C:\data\output

Clone repositories

cd C:\data\dev\projects   # Or any other path
git clone https://github.com/YuvalNirkin/face_detection_dsfd
git clone https://github.com/YuvalNirkin/fsgan

Add the projects directory to Python path:

Example