File Exchange

image thumbnail

Generate Pokemon Using GAN - Generative Adversarial Network

version 1.0.1 (173 MB) by Kevin Chng
Explore how to generate fake image using GAN in MATLAB

12 Downloads

Updated 16 Feb 2020

GitHub view license on GitHub

Generate Pokemon Using GAN (Generative Adversarial Network) in MATLAB.

The credit for the code go to moxiegushi (google this name with pokemon GAN). Or you may google Siraj Raval - Generating pokemon with GAN in youtube. They are using Python for the training,however, for latest version of MATLAB R2019b, GAN is officially supported in MATLAB.

With 1xK80 GPU virtual machine in azure, it took around 8 hours to complete 2800 epochs with 128 mini batch size and ADAM optimizer.

Now i got the new pokemon. This is also how you generate fake images when you don't have enough data.

Highlights :
Image Datastore and Image Augmentation
Structure Reinforcement Neural Network (Generator & Discriminator)
Workflow for Reinforcement Learning

Product Focus :
MATLAB
Deep Learning Toolbox
Reinforcement Learning Toolbox

Cite As

Kevin Chng (2020). Generate Pokemon Using GAN - Generative Adversarial Network (https://github.com/KevinChngJY/reinforcementlearning_matlab_pokemon), GitHub. Retrieved .

Comments and Ratings (1)

Kenta

Updates

1.0.1

Update description

MATLAB Release Compatibility
Created with R2019b
Compatible with any release
Platform Compatibility
Windows macOS Linux