High Resolution Using Conditional Generative Adversarial Networks

Abstract

Lately, High-resolution generators by deep learning methods have produced promising impressive results. In this thesis will be shown the implementation steps of the system objective. Which aims to recover a high-resolution from low-resolution image, it’s considered as a classic computer vision issue. Through implementation it’s going to build a generative opposing network called Conditional Generative Adversarial Network from deep learning approach, opposing network that applies the same concept to produce more photorealistic results in this architecture. Not only does it help zooming parts to correctly calculate their lost pixel after losing image resolution and remove the blurred parts, it also gives a multi-size approach that focuses on values and also improves reconstruction coherence in all image sizes.