Style Transfer of Black and White Silhouette Images using CycleGAN and a Randomly Generated Dataset

TL;DR

This paper introduces a novel approach using CycleGAN trained on randomly generated data to transfer artistic styles to black and white silhouette images, achieving better results than previous neural style transfer methods.

Contribution

It proposes a new training method with randomly generated datasets for style transfer on silhouettes, improving over existing neural style transfer techniques.

Findings

Better style transfer quality than previous methods

Effective use of CycleGAN with synthetic data

Identified areas for further improvement like artifact removal

Abstract

CycleGAN can be used to transfer an artistic style to an image. It does not require pairs of source and stylized images to train a model. Taking this advantage, we propose using randomly generated data to train a machine learning model that can transfer traditional art style to a black and white silhouette image. The result is noticeably better than the previous neural style transfer methods. However, there are some areas for improvement, such as removing artifacts and spikes from the transformed image.