Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints

Shaun Schreiber; Jaco Geldenhuys; Hendrik De Villiers

doi:10.1109/RoboMech.2016.7813173

Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints

Shaun Schreiber, Jaco Geldenhuys, Hendrik De Villiers

Post Harvest Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › Academic › peer-review

5 Citations (Scopus)

Abstract

Procedural texture generation enables the creation of more rich and detailed virtual environments without the help of an artist. However, finding a flexible generative model of real world textures remains an open problem. We present a novel Convolutional Neural Network based texture model consisting of two summary statistics (the Gramian and Translation Gramian matrices), as well as spectral constraints. We investigate the Fourier Transform or Window Fourier Transform in applying spectral constraints, and find that the Window Fourier Transform improved the quality of the generated textures. We demonstrate the efficacy of our system by comparing generated output with that of related state of the art systems.

Original language	English
Title of host publication	2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016
Publisher	Institute of Electrical and Electronics Engineers Inc.
Number of pages	6
ISBN (Print)	9781509033355
DOIs	https://doi.org/10.1109/RoboMech.2016.7813173
Publication status	Published - 2017
Event	2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016 - Stellenbosch, South Africa Duration: 30 Nov 2016 → 2 Dec 2016

Conference

Conference	2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016
Country/Territory	South Africa
City	Stellenbosch
Period	30/11/16 → 2/12/16

Access to Document

10.1109/RoboMech.2016.7813173

Cite this

Schreiber, S., Geldenhuys, J., & De Villiers, H. (2017). Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints. In 2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016 [7813173] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/RoboMech.2016.7813173

@inproceedings{4fee145e7629426dbb7f57ef192e5533,

title = "Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints",

abstract = "Procedural texture generation enables the creation of more rich and detailed virtual environments without the help of an artist. However, finding a flexible generative model of real world textures remains an open problem. We present a novel Convolutional Neural Network based texture model consisting of two summary statistics (the Gramian and Translation Gramian matrices), as well as spectral constraints. We investigate the Fourier Transform or Window Fourier Transform in applying spectral constraints, and find that the Window Fourier Transform improved the quality of the generated textures. We demonstrate the efficacy of our system by comparing generated output with that of related state of the art systems.",

author = "Shaun Schreiber and Jaco Geldenhuys and {De Villiers}, Hendrik",

year = "2017",

doi = "10.1109/RoboMech.2016.7813173",

language = "English",

isbn = "9781509033355",

booktitle = "2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

note = "2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016 ; Conference date: 30-11-2016 Through 02-12-2016",

}

Schreiber, S, Geldenhuys, J & De Villiers, H 2017, Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints. in 2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016., 7813173, Institute of Electrical and Electronics Engineers Inc., 2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016, Stellenbosch, South Africa, 30/11/16. https://doi.org/10.1109/RoboMech.2016.7813173

Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints. / Schreiber, Shaun; Geldenhuys, Jaco; De Villiers, Hendrik.
2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016. Institute of Electrical and Electronics Engineers Inc., 2017. 7813173.

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › Academic › peer-review

TY - GEN

T1 - Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints

AU - Schreiber, Shaun

AU - Geldenhuys, Jaco

AU - De Villiers, Hendrik

PY - 2017

Y1 - 2017

N2 - Procedural texture generation enables the creation of more rich and detailed virtual environments without the help of an artist. However, finding a flexible generative model of real world textures remains an open problem. We present a novel Convolutional Neural Network based texture model consisting of two summary statistics (the Gramian and Translation Gramian matrices), as well as spectral constraints. We investigate the Fourier Transform or Window Fourier Transform in applying spectral constraints, and find that the Window Fourier Transform improved the quality of the generated textures. We demonstrate the efficacy of our system by comparing generated output with that of related state of the art systems.

AB - Procedural texture generation enables the creation of more rich and detailed virtual environments without the help of an artist. However, finding a flexible generative model of real world textures remains an open problem. We present a novel Convolutional Neural Network based texture model consisting of two summary statistics (the Gramian and Translation Gramian matrices), as well as spectral constraints. We investigate the Fourier Transform or Window Fourier Transform in applying spectral constraints, and find that the Window Fourier Transform improved the quality of the generated textures. We demonstrate the efficacy of our system by comparing generated output with that of related state of the art systems.

U2 - 10.1109/RoboMech.2016.7813173

DO - 10.1109/RoboMech.2016.7813173

M3 - Conference paper

AN - SCOPUS:85012034940

SN - 9781509033355

BT - 2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016

Y2 - 30 November 2016 through 2 December 2016

ER -

Schreiber S, Geldenhuys J, De Villiers H. Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints. In 2016 Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference, PRASA-RobMech 2016. Institute of Electrical and Electronics Engineers Inc. 2017. 7813173 doi: 10.1109/RoboMech.2016.7813173

Texture synthesis using convolutional neural networks with long-range consistency and spectral constraints

Abstract

Conference

Access to Document

Fingerprint

Cite this