LDOUBLEV / semi-supervised-GAN

There is the semi-supervised implement of 'Improved Techniques for Training GANs '

Semi-supervised-GAN

there is the Tensorflow implement of 'Improved Techniques for Training GANs '

Descriptioins

This project is a Tensorflow implemention of semi-supervised which described in Improved Techniques for Training GANs. We made some changes without changing the original intention. We train our code in mnist dataset, and achieve 0.99 test accuracy only with 200 labeled images. The test acc process is shown followed.

usage

The Tensorflow vision is 1.2 and we use python 2.7 to implement our code.

train

python main.py --model=DCGAN --trainable=True --load_model=False

test

`python main.py --model=DCGAN --trainable=False  --load_model=True '

            `