• Jarda Aamir A.
• Priyang Bhatt

Unpaired image to image translation is a most interesting and challenging topic because of graphic problem and loss function. Through this paper, we aim to understand what exactly Cycle GAN are, what the existing applications of such models are and how we improve the loss function. Scenario is we give input is an image and the output is different version of this input image that is changed according to our guideline. The basic goal of this translation is to learn mapping between input image and output image using training set. Initial approach for translate an image from X to Y in the absence of paired data. Model contain two functions G: X -> Y and F: Y -> X where X is source domain and Y is target domain. The first things we should know about this technique is that it uses Generative adversarial network. In this we have two neural network battling each other. Generator tries to create realistic image and discriminator which tries to learn difference between real and fake image. A cycle consistency loss function is introduced to the optimization problem that means if we convert a zebra image to a horse image and then back to a zebra image, we should get the very same input image back.

Generative adversarial network, CycleGAN, L1 loss, Logistic regression, Adam Optimizer

"Improving Efficiency of Loss Function in Cycle-Consistent Adversarial Network", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 6, page no.216-222, June-2019, Available :http://www.jetir.org/papers/JETIR1906465.pdf

"Improving Efficiency of Loss Function in Cycle-Consistent Adversarial Network", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 6, page no. pp216-222, June-2019, Available at : http://www.jetir.org/papers/JETIR1906465.pdf