New Architecture of Generator and Discriminator

Features

• Improvements in the generator architecture to maintain low-level characteristics leading to greater training stability;
• Discriminator improvements to implement spectral normalization to improve training stability;

Improvings on v1

New features:

• Implement post processing to image and video generator;

---

Old Features:

• Train your own GAN with the images of your choice;

• Generate as many images as you want after completing the training;

• Produce videos through interpolation of the generated images;

• If you want to upscale the generated images or video, use the --upscale argument followed by the width value;

Initial Release

Features:

• Train your own GAN with the images of your choice;

• Generate as many images as you want after completing the training;

• Produce videos through interpolation of the generated images;

• If you want to upscale the generated images or video, use the --upscale argument followed by the width value;