Review on Generative Adversarial Networks

Abstract

Lately, supervised learning is hugely adopted in computer vision. But unsupervised learning has earned less consideration. A branch of CNNs classified as generative adversarial networks (GANs) is made acquainted, it has some architectural restraints, and exhibit that they are a tough contender for unsupervised learning. Training on different datasets of images, it displays conclusive proof that the adversarial pair learns a hierarchy of portrayal from parts to scenes in both the discriminator and generator. Also, the learned features can be used for variety of innovative tasks, indicating their appropriateness as general image representation.