In this repository you will find the Colab Notebook for generating synthetic images for skin cancer lesions. The implemented network is only a PoC (Proof of Concept) to test the feasibility of the task.
Fig.1: GAN framework implemented
The Generative Adversarial Network (GAN) implemented is shown in detail in Fig.1. In particular, with G has been indicated the generator network, while with D the discriminator network. The generator receives in input a random normal noise vector (latent space) with shape [100] and performs an upsampling to produce an output synthetic RGB image of size [64,64,3].
The dataset used is here
Some qualitative results are shown below: |
 |
 |
 |
 |
 |
 |
 |
|---|