Conditional Wasserstein GAN with gradient penalty for the generation of synthetic images.
MNIST dataset was used to train the conditional WGAN. Gradient Penalty was also implemented.
Run the cells in sequence in cwgan-gp.ipynb
jupyter notebook. Final cell contains code to create synthetic
image conditioned on a label.
Code in part based on:
- https://keras.io/examples/generative/conditional_gan/
- https://keras.io/examples/generative/wgan_gp/
- https://keras.io/examples/generative/dcgan_overriding_train_step/
- tensorflow
- numpy
- matplotlib
See requirements.txt
file.
Free to use for any purpose