Generative Adversarial Networks with Inverse Transformation Unit

TL;DR

This paper proposes a novel GAN architecture incorporating an inverse transformation unit, demonstrating convergence and effectiveness in image sharpening and deblurring tasks, even with non-bijective transformations.

Contribution

Introduces a new GAN structure with an inverse transformation unit, providing theoretical convergence proofs and empirical validation on image sharpening and deblurring.

Findings

Model converges under certain conditions.

Successfully learns to sharpen and deblur images.

Transformation does not need to be bijective.

Abstract

In this paper we introduce a new structure to Generative Adversarial Networks by adding an inverse transformation unit behind the generator. We present two theorems to claim the convergence of the model, and two conjectures to nonideal situations when the transformation is not bijection. A general survey on models with different transformations was done on the MNIST dataset and the Fashion-MNIST dataset, which shows the transformation does not necessarily need to be bijection. Also, with certain transformations that blurs an image, our model successfully learned to sharpen the images and recover blurred images, which was additionally verified by our measurement of sharpness.