Skip to content
TensorFlow implementation of Deep Convolutional Generative Adversarial Networks
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
assets
LICENSE
README.md
main.py
model.py
requirements.txt
utils.py

README.md

TensorFlow DCGAN

Usage

Train a model.
$ python main.py --dataset='celeb_a' --height=64 --width=64 output_dir=''

Options.

  --batch_size: Batch size of images to train.
    (default: '64')
    (an integer)
  --beta: Training beta.
    (default: '0.5')
    (a number)
  --cache: Optional: [None, memory, disk]. If specified, data will be cached
    for faster training. 
    memory: slower, disposable. 
    disk: faster, requires space.
    (default: '')
  --[no]crop: Center crop image.
    (default: 'false')
  --dataset: Dataset to train [cifar10, celeb_a, tf_flowers]
    (default: 'cifar10')
  --epochs: Number of epochs to train.
    (default: '100')
    (an integer)
  --height: Image input height.
    (default: '64')
    (an integer)
  --width: Image input width.
    (default: '64')
    (an integer)
  --learning_rate: Learning rate.
    (default: '0.0002')
    (a number)
  --output_dir: Directory to save output.
    (default: '')

Results

CelebA
Generated image of CelebA dataset Generated gif of CelebA dataset

CelebA cropped
Generated image of CelebA dataset Generated gif of CelebA dataset

Flowers ~3,600 images
Generated image of Flower dataset

You can’t perform that action at this time.