[3Xin4] Poster session 1

[3Xin4-35] Origami Image Generation Using Style Transfer

Ukyo Ohashi¹, 〇Hiroaki Kawashima¹ (1.University of Hyogo)

Keywords:Image generation, Style transfer, Origami

This study aims to transfer images of real objects, such as animals, human faces, and objects, into images that appear to have been folded into origami, using images of actual objects and origami as input. While many methods have been proposed for Neural Style Transfer to transform a content image into a specific style, most of them have been applied to paintings and other 2D art forms, and few methods can transform the characteristics of three-dimensional objects, such as their texture and depth, into a style. In this work, we discover among the existing style transfer methods suitable for transferring "origami-like" properties, even when the geometric structures of the input images of actual objects and origami differ significantly, and propose a novel method for its extension.