Image Style Transfer via Multi-Style Geometry Warping

Style transfer of an image has been receiving attention from the scientific community since its inception in 2015. This topic is characterized by accelerated process innovation; it defined techniques that blend content and style, first covering only textural details, subsequently incorporating compositional features. The results such had a significant impact on our understanding inner workings Convolutional Neural Networks. Recent research shown increasing interest geometric deformation images, defining trait for different artists, various art styles, previous methods failed to account for. However, current approaches are limited matching class deformations order obtain adequate outputs. paper solves these limitations combining works framework can perform images using styles multiple artists building architecture uses style one as input. proposed combination other existing frameworks more intriguing artistic result. detects objects classes inside assigns them bounding box, before each detected object found box with similar performing warping basis similarities. Next, algorithm blends back together all warped so they placed position initial image, finally applied between merged chosen image. We manage stylistically pleasing were possible generate reasonable amount time, compared methods.