Content Aware Neural Style Transfer

TL;DR

This paper introduces a content-aware neural style transfer method that improves visual quality by better preserving content and style details, revealing that neural networks do not fully separate style from content.

Contribution

The paper proposes a novel content-aware style transfer algorithm using pre-trained neural networks, enhancing results over previous methods.

Findings

Style pattern and content information are not fully separated by neural networks.

The proposed method yields better visual quality than prior approaches.

Neural networks retain intertwined style and content features.

Abstract

This paper presents a content-aware style transfer algorithm for paintings and photos of similar content using pre-trained neural network, obtaining better results than the previous work. In addition, the numerical experiments show that the style pattern and the content information is not completely separated by neural network.