This code uses a Generative Adversarial Network (GAN) to generate melanoma images using Keras
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
LICENSE
README.md
generative_adversarial_network.py

README.md

GANMole

This code uses a Generative Adversarial Network (GAN) to generate melanoma images using Keras.

To run the code:

$ python generative_adversarial_network

I used 374 images to train the GAN, and have resized the images to 32x32 due to the resource exhaustion I get faced with on the GPU I'm using when going with higher resolution images. You can download the image from, here. I would like to mention that I have renamed the images into sequenced number images, that is 0.jpg, 1.jpg, 2.jpg, ..., 373.jpg. You can use this script to rename the image file names to sequence number names.

Of course, the goal here is not to reach an optimum solution, but rather demonstrate how a GAN can be used to generate melanoma images.

Have fun!