Refactored DCGAN Implementation

Refactored the codes from carpedm20

• Make codes re-runnable inside Spyder IDE without needing to restart the kernel
• Make less dependent on command line arguments
• Additional Code Refactor

Original Authors:

• Taehoon Kim / @carpedm20
• @PhoenixDai for MNIST codes

Prerequisites

• Python 2.7 or Python 3.3+
• Tensorflow 0.12.1
• SciPy
• pillow
• tqdm
• (Optional) moviepy (for visualization)
• (Optional) Align&Cropped Images.zip : Large-scale CelebFaces Dataset

Usage

First, download dataset with:

$ python download.py mnist celebA

To train a model with downloaded dataset:

$ python main.py --dataset mnist --train
$ python main.py --dataset celebA --train

To test with an existing model:

$ python main.py --dataset mnist
$ python main.py --dataset celebA

Related works

• BEGAN-tensorflow
• DiscoGAN-pytorch
• simulated-unsupervised-tensorflow