/Face_Conversion_Conditonal_GAN

It uses Conditional GAN(Generative adversarial networks) to convert a front face image into a more primitive representation of the face.

Primary LanguageJupyter Notebook

Face Conversion using Conditional GAN

Demo

alt text

installation

sudo pip3 install tensorflow
sudo pip3 install keras
sudo pip3 install Pillow
scipy and numpy will automatically get installed

Contents

open jupyter notebook " Experiment-4.ipynb" to view training code
open jupyter notebook " tester.ipynb" to view testing code.

for floyd hub

run terminal in this directory
floyd login
floyd init project
floyd run --mode jupyter --env tensorflow --gpu
now run cells one after another

requirements

keras -2.0.2
tensorflow -1.0.1
PIL or Pillow
python 3
numpy
scipy

thank you to

https://github.com/legokichi/keras-pix2pix

https://github.com/ray0809/pix2pix

https://github.com/wiseodd/generative-models

https://github.com/r0nn13/conditional-dcgan-keras

https://github.com/yiyuezhuo/GAN_experiment

you tube

https://www.youtube.com/user/hvasslabs

https://www.youtube.com/watch?v=u7kQ5lNfUfg

blogs to read

https://blog.openai.com/generative-models/

http://blog.aylien.com/introduction-generative-adversarial-networks-code-tensorflow/

https://affinelayer.com/pix2pix/

https://bamos.github.io/2016/08/09/deep-completion/

http://www.foldl.me/2015/conditional-gans-face-generation/

http://wiseodd.github.io/techblog/2016/12/24/conditional-gan-tensorflow/

online demo

https://affinelayer.com/pixsrv/

future use

https://github.com/richliao/SketchToFace