On Demand Solid Texture Synthesis Using Deep 3D Networks

Jorge Gutierrez (1) , Julien Rabin (2) , Bruno Galerne (3) , Thomas Hurtut (1)

1 Polytechnique Montréal
2 Equipe Image - Laboratoire GREYC - UMR6072
3 MAP5 - UMR 8145 - Mathématiques Appliquées Paris 5

This paper addresses the problem of generating volumetric data for solid texture synthesis. We propose a compact, memory efficient, convolutional neural network (CNN) which is trained from an example image. The features to be reproduced are analyzed from the example using deep CNN filters that are previously learnt from a dataset. After training the generator is capable of synthesizing solid textures of arbitrary sizes controlled by the user. The main advantage of the proposed approach in comparison to previous patch-based methods is that the creation of a new volume of data can be done in interactive rates.

Mots clés: Solid textures; Volumetric data generation; Convolutional neural networks; Texture synthesis

Domaines: Machine Learning [stat.ML]; Traitement des images [eess.IV]; Synthèse d'image et réalité virtuelle [cs.GR]

Jorge Gutierrez, Julien Rabin, Bruno Galerne, Thomas Hurtut. On Demand Solid Texture Synthesis Using Deep 3D Networks. 2018. ⟨hal-01678122v1⟩