JISE


  [ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ] [ 10 ] [ 11 ] [ 12 ] [ 13 ]


Journal of Information Science and Engineering, Vol. 37 No. 3, pp. 575-592


Escher-like Tiling Design from Video Images Using Convolutional Variational Autoencoder


ASUKA HISATOMI1, TOMOFUMI MATSUYAMA1, TAKAHIRO KINOSHITA1,
KAZUNORI MIZUNO2 AND SATOSHI ONO1
1Department of Information Science and Biomedical Engineering
Kagoshima University
Kagoshima, 890-0065 Japan
E-mail: fmc116029; sc114063; sc115015; onog@ibe.kagoshima-u.ac.jp

2Department of Computer Science
Takushoku University
Tokyo, 193-0985 Japan
E-mail: mizuno@cs.takushoku-u.ac.jp


This paper proposes a method that deforms a prominent movie or animation character into a tileable shape. Tiling is the act of covering the plane with one or a very few types of figures without overlaps and/or gaps. Although some previous methods can transform a given shape into a tileable shape, they cannot easily move the character into a suitably tileable pose. The proposed method learns the latent feature space that abstracts the target character’s silhouettes using a convolutional variational autoencoder, and looks for the poses suitable for tiling by optimization in the latent space. Experimental results showed that the proposed method successfully generated tileable figures of the tested character in various poses, some of which were not included in the training dataset.


Keywords: tiling, tessellations, hierarchical optimization, genetic algorithm, convolutional variational autoencoder

  Retrieve PDF document (JISE_202103_05.pdf)