Everything related with my graduation project "An Analysis of Techniques for Building Generative Adversarial Networks"

The document can be found at: https://repositorio.ufsc.br/handle/123456789/223064

• Overleaf - all files used in the overleaf project to produce the final document.
• drawio - files for the diagrams used in the document, can be opened and edited at draw.io.
• images - images referenced in the README files in this repository

All code except for the ones contained in the utils directory is written in Jupyter notebooks.

• utils - python module for custom functions frequently used in other parts of the code.
• Classifiers - implementation of the MNIST, Fashion MNIST, and CIFAR-10 classifiers. These are used to calculate the Classifier Score (CS) and Fréchet Classifier Distance (FCD) as described in the document in section 4.4 EVALUATING GANS.
• GAN - Code related to the simple GAN implementation described in sections 4.2 THE GAN ARCHITECTURE and 5.1 SIMPLE GAN of the document.
• DCGAN - Code related to the Deep Convolutional GAN (DCGAN) implementation described in subsection 4.3.1 DCGAN and section 5.2 DCGAN of the document.
• CGAN - Code related to the Conditional GAN (CGAN) implementation described in subsection 4.3.2 Conditional GAN and section 5.3 CGAN of the document.
• WGAN - Code related to the Wasserstein GAN (WGAN) implementation described in subsection 4.3.3 Wasserstein GAN and section 5.4 WGAN of the document.
• WGAN-GP - Code related to the WGAN with Gradient Penalty (WGAN-GP) implementation described in subsection 4.3.4 WGAN with Gradient Penalty and section 5.5 WGAN-GP of the document.
• Other - Code for creating the different visualizations shown in the document.