Control Method Of perceptual Elements Based On Universal Neural Style Transfer
Xiya Guo, Shaobin Li, Jun ping Yin · 2019
Universal style transfer has been implemented on the VGG network. Our work proposes control methods for universal style transfer. In general, the proposed methods consists of a texture segmentation method, and a method to control the weights of different layers in style transfer. Based on the results of texture segmentation, spatial control and background control can be realized. Increasing these control dimensions results in a more diverse result.