FishGAN
Keywords - machine learning, Generative Adversarial Networks
This is a fun mini-project to learn about Generative adversarial networks(GANs). A DC-GAN has been trained on fish images from the dataset the QUT FISH dataset containing around 4k fish images(64 x 64)of various species. The code used for DCGAN and WGAN is taken from their official repositories. This project also discusses mode collapse in GANs and some ways to address it.
More information can be found here:
Images generated by GAN with each epoch