Creators : @alessiapacca

Ubuntu 18.04 Kernel 5.4.0
Python 3.7
Cuda 10.2
Cuda Drivers 440.33.01

Use Conda environment:

Install packages by running

conda install pyqt
sudo apt install ffmpeg
sudo apt-get install libportaudio2

Install the dependencies by running

pip install -r requirements.txt

Install also pytorch and torchvision with

conda install pytorch==1.1.0 mkl==2018 torchvision==0.2.1 cudatoolkit=10.0 -c pytorch

Clone the library for face alignment:

git clone https://github.com/1adrianb/face-alignment
cd face-alignment
pip install -r requirements.txt
python setup.py install

All paths are hard-coded inside of the scripts, make sure you change them before running.

The script in scripts/preprocess_driving_video.py is responsible to detect and crop the driving videos in a preprocessing step (if needed). Just run it (change the paths accordingly)

python scripts/preprocess_driving_video.py

The script in scripts/preprocess_images.py is responsible to crop the images to squares in a preprocessing step (if needed). Just run it (change the paths accordingly)

python scripts/preprocess_images.py

Before testing the model, you need to download the weights:
https://drive.google.com/file/d/1wjUDZP7rXRAF6hQ4TisZ1YvCRwn-ur5W/view?usp=sharing
After downloading put the in the folder checkpoints of the repo. Just remember that the code requires that the object in the first frame of the video and in the source image have the same pose.

python scripts/exhibition_script.py

Added the code for enhancing the generated videos inside the script.

python scripts/exhibition_script_enhance.py

In order to be able to use this however, some packages are needed.
Install them with conda:

conda install tensorflow-gpu=1.13.1 
conda install keras-base
conda install keras 

If any problem about tensorflow is outputted, just install these ones too:

pip install dask==1.2.0

Audio generation is heavily based on Corentin Jemine's repository https://github.com/CorentinJ/Real-Time-Voice-Cloning

python /scripts/deep_vocoder.py

Copy the files and put them in the main main folder.

To run the server:

cd exhibiton-code
export FLASK_APP=server.py
flask run

Running on the local server https://127.0.0.1:5000/.