The aim to this program is to use a Deep Convolutional GAN (Generative Adversarial Network) using Tensorflow.
This program is made possible thanks to another repository: NFTGrabber
-
First, download some NFT images:
- install dependencies:
npm start
- run
npm start [link]
(ex:"npm start https://rarible.com/boredapeyachtclub")
- install dependencies:
-
Then use
avif_converter.py
to convert the images in PNG -
Finally run
gan_2_tf_module.py
to run the gan
In an epoch, the discriminator is trained wuth the database, then the generator is trained using the discriminator
At the end of each epoch, an image is saved to the generated_images folder
For now, the program is made to generate 64x64x3 RGB images. It reshapes the input images with output shape so the inputs can be different.
In the future, the output size can be an argument, or can be chosen to match the input shape.
Romain Dodet
- Twitter🐦: @Romain_Dodet_
Hope it will be helpful :)