drewszurko / tensorflow-WGAN-GP

TensorFlow 2.0 implementation of Improved Training of Wasserstein GANs

TensorFlow 2.0 WGAN-GP

TensorFlow 2.0 implementation of Improved Training of Wasserstein GANs [1]. New/existing TensorFlow features found in this repository include eager execution, AutoGraph, Keras high-level API, and TensorFlow Datasets.

Requirements

• Python 3
• Abseil
• NumPy
• TensorFlow >= 2.0
• TensorFlow Datasets
• tqdm

Datasets

• CIFAR-10
• CelebA
• TF Flowers
• Oxford Flowers 102
• Oxford-IIIT pet

Usage

Install requirements

$ pip install -r requirements.txt

Train model

$ python main.py -dataset celeb_a -batch_size 64 -image_size 64

References

[1] Improved Training of Wasserstein GANs