Different Techniques in Neural Style Transfer-A Review

Abstract

AbstractNeural Style Transfer is a very good example to show the capability of Convolutional Neural Network (CNN) to generated high-quality images. In style transfer algorithms two input images are present one is a content image and the other is a style image and it generates output in such a way that the output is a mix of both content and style image. There are different methods and approaches that can be used to implement the style transfer. This paper will focus briefly on neural networks and their architecture used in the approaches and briefly describe the important loss functions and the implementation of neural style transfer. This review paper will explain the four major approaches or techniques that have been used to implement neural style transfer, these approaches will also tell how they have improved on the original method put forward by Gatys et al. 2015 to produce good quality results and then find the advantages and disadvantages by comparison of each method that is explained. This paper aims to provide knowledge of how different approaches work.KeywordsNeural style transferStyle imageNeural networksStyle transfer algorithms