Image Cartoonizer

Abstract

This report is to make the applications about transferring the real-world images to the cartoon style image based on the CartoonGAN [1]. This is a valuable and useful project on the computer vision and machine learning. CartoonGAN [1] purposed the Generative adversarial networks to train the model of the cartoon style and real-world. And it also purposed two loss function to transfer. Our project mainly focuses on improving the model which is lighter and faster than the model of the CartoonGAN [1] under the condition of not influencing the transferring effect. And do the testing each time finish the coding to check whether the error occurs or not and the effect after the new function. The CartoonGAN [1] purposed the regression testing to ensure system normally run and we use the white box testing to check the applications. This project is to save the time of the artist, and to purpose a platform for people who don’t have the drawing skill but want their own cartoon images. By this project we can study about Computer vision, Generative adversarial networks (GAN), Neural style transfer (NST) and Machine learning (ML).