I used a Generative Adversarial Network (GAN) to generate melanoma images, and named the code GANMole. You can find the code on the GANMole GitHub repository.
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 images 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.