ghanashyamvtatti/PokeGAN

A tf.keras implementation of DCGAN to generate images of new Pokemon

PokeGAN - A tf.keras implementation of Deep Convolutional Generative Adversarial Networks

I created this repository to explore tf.keras and get comfortable with GANs.

Dataset

819 images of Pokemon from Kaggle Pokemon Images Dataset Used utils.py to convert them to jpg resized them to 64x64

© 2010 Pokémon. © 1995-2010 Nintendo/Creatures Inc./GAME FREAK inc.

Results

Epoch 37

Epoch 45

Setup

1. Download the dataset from here
2. Convert images to jpg from png (utils.py could help)
3. Install the dependencies using pip install -r requirements.txt
4. Change the dataset_path in pokeGAN.py if necessary
5. Run the program python pokeGAN.py

References

• Radford, Alec, Luke Metz, and Soumith Chintala. "Unsupervised representation learning with deep convolutional generative adversarial networks." arXiv preprint arXiv:1511.06434 (2015).
• Goodfellow, Ian. "NIPS 2016 tutorial: Generative adversarial networks." arXiv preprint arXiv:1701.00160 (2016).

Credits

• Pokemon Images Dataset
• How to Train a GAN? Tips and tricks to make GANs work
• Anime-Face-GAN-Keras
• DCGAN and WGAN implementation on Keras for Bird Generation