/Structure-Aware-Style-Transfer

Adaptive convolutions for structure-aware style transfer

Primary LanguageJupyter Notebook

Adaptive Convolutions for Structure-Aware Style Transfer

Style transfer between images is an artistic application of CNNs, where the ‘style’ of one image is transferred onto another image while preserving the latter’s content.

Contents

Style Transfer Details

We have two purposes in style transfer

  • Transferring statistical properties
  • Transferring geometric properties

mm

Style Transfer Methods

  • Adaptive Instance Normalization
  • Adaptive Convolutions

AdaIN Style Transfer Network

This method only transfers statistical properties

ada

AdaConv Style Transfer Network

This method transfers both statistical and geometric properties

m