Instance Normalization: The Missing Ingredient for Fast Stylization

TL;DR

This paper demonstrates that replacing batch normalization with instance normalization in fast stylization models significantly improves image quality, enabling real-time high-performance stylization.

Contribution

The key contribution is identifying instance normalization as a crucial modification that enhances the quality of fast stylization methods.

Findings

Qualitative improvement in stylized images

Effective for real-time image generation

Simple architectural change yields significant benefits

Abstract

It this paper we revisit the fast stylization method introduced in Ulyanov et. al. (2016). We show how a small change in the stylization architecture results in a significant qualitative improvement in the generated images. The change is limited to swapping batch normalization with instance normalization, and to apply the latter both at training and testing times. The resulting method can be used to train high-performance architectures for real-time image generation. The code will is made available on github at https://github.com/DmitryUlyanov/texture_nets. Full paper can be found at arXiv:1701.02096.