Skip to content

Latest commit

 

History

History
14 lines (8 loc) · 887 Bytes

README.md

File metadata and controls

14 lines (8 loc) · 887 Bytes

Digimon-Generator-GAN

This repo contains the code that generates Digimon images using the concept of Generative Adversarial Networks

GANs are very hard to train. So, just clone or download the notebook Generate_new_digimons.ipynb, extract the Digimon.zip file and run all the cells on an external local GPU or on Cloud to see the better results.

This code also adds scalars and test images to the tensorboard for better visualization.

Tensorboard graph when trained on a small dataset(195 images) of digimons:

screenshot from 2018-02-27 17-51-55

Tensorboard graph when trained on a large dataset(1061 images) of digimons:

screenshot from 2018-02-27 19-58-22