Skip to content

Generating Art using Generative Adversarial Network (GAN)

Notifications You must be signed in to change notification settings

FingerLeakers/GAN_Canvas_Maker

 
 

Repository files navigation

GAN_Canvas_Maker

Generating Art using Generative Adversarial Network (GAN)


Python code to generate art with Generative Adversarial Network:

https://towardsdatascience.com/generating-modern-arts-using-generative-adversarial-network-gan-on-spell-39f67f83c7b4


The code is made to run it on spell:

https://web.spell.ml/


But it can also be launched locally, but obviously you need a good graphic card.

To launch the GAN locally you can use this command:

python art_gan.py -m cubism_data.npy


The code I started from had a couple of indentation errors, and some unrecognized characters, which I fixed.


Requirements:

-tensorflow
-PIL
-os
-numpy
-spell


The repo also contains a small dataset (taken from wikiart) which contains 1000 images.

The complete dataset (25.4 Gb) can be downloaded from here:

https://github.com/cs-chan/ArtGAN/tree/master/WikiArt%20Dataset



Personally I had several problems starting the script from the command prompt,
so I opted to start the script directly from the Spell ide,
in any case I also encountered several problems given by the compatibility of the libraries versions,
if you need see the working setup to run the script from Spell
you can look at the screenshots i inserted with my working setup.


Alt text


About

Generating Art using Generative Adversarial Network (GAN)

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%