Automated creation of graphic design using convolutional neural network style transfer

摘要

Problem to be solved: to provide an automatic design generation method and system for a graphic design product.The method isStep 800 generates a feature feature layer by inputting silhouette images into a convolution neural networkEnter a style image into a convolution neural networkStep 801 generates a pattern feature layerA reference image is input to the convolution neural network to determine the reference feature layer.For each iteration,Minimizing the combined loss of convolutional neural networksTo obtain an output image containing the abstraction of the style image within the silhouette image, step 802.Diagram