3S-Net: Arbitrary Semantic-Aware Style Transfer With Controllable ROI Choice

A cluster of stunning style transfer approaches evolve to include single style transfer and multi-style transfer these years. However, few studies consider the style consistency between identical semantic objects in style images and the content image respectively. Especially for multi-style transfer, the merged style is obtained through a simple linear combination with given weights. So, the textures of each source style are mixed, which makes the result lack aesthetic value. To overcome this problem, we propose a 3S-Net to achieve semantic-aware style transfer mainly by Two-Step Semantic Instance Normalization (2SSIN) and Semantic Style Swap. 3S-Net can automatically accomplish the procedure of matching semantic objects between style images and content image. Besides, our method is adaptive and flexible. It succeeds in single style transfer and multi-style transfer from a semantic perspective, even ROI (Region of Interest) style transfer with user’s preference.