Neural Painters: A learned differentiable constraint for generating
brushstroke paintings
R. Nakano. (2019)cite arxiv:1904.08410Comment: Added more references and acknowledgments.
Abstract
We explore neural painters, a generative model for brushstrokes learned from
a real non-differentiable and non-deterministic painting program. We show that
when training an agent to "paint" images using brushstrokes, using a
differentiable neural painter leads to much faster convergence. We propose a
method for encouraging this agent to follow human-like strokes when
reconstructing digits. We also explore the use of a neural painter as a
differentiable image parameterization. By directly optimizing brushstrokes to
activate neurons in a pre-trained convolutional network, we can directly
visualize ImageNet categories and generate "ideal" paintings of each class.
Finally, we present a new concept called intrinsic style transfer. By
minimizing only the content loss from neural style transfer, we allow the
artistic medium, in this case, brushstrokes, to naturally dictate the resulting
style.
Description
Neural Painters: A learned differentiable constraint for generating brushstroke paintings
%0 Generic
%1 nakano2019neural
%A Nakano, Reiichiro
%D 2019
%K cv dl
%T Neural Painters: A learned differentiable constraint for generating
brushstroke paintings
%U http://arxiv.org/abs/1904.08410
%X We explore neural painters, a generative model for brushstrokes learned from
a real non-differentiable and non-deterministic painting program. We show that
when training an agent to "paint" images using brushstrokes, using a
differentiable neural painter leads to much faster convergence. We propose a
method for encouraging this agent to follow human-like strokes when
reconstructing digits. We also explore the use of a neural painter as a
differentiable image parameterization. By directly optimizing brushstrokes to
activate neurons in a pre-trained convolutional network, we can directly
visualize ImageNet categories and generate "ideal" paintings of each class.
Finally, we present a new concept called intrinsic style transfer. By
minimizing only the content loss from neural style transfer, we allow the
artistic medium, in this case, brushstrokes, to naturally dictate the resulting
style.
@misc{nakano2019neural,
abstract = {We explore neural painters, a generative model for brushstrokes learned from
a real non-differentiable and non-deterministic painting program. We show that
when training an agent to "paint" images using brushstrokes, using a
differentiable neural painter leads to much faster convergence. We propose a
method for encouraging this agent to follow human-like strokes when
reconstructing digits. We also explore the use of a neural painter as a
differentiable image parameterization. By directly optimizing brushstrokes to
activate neurons in a pre-trained convolutional network, we can directly
visualize ImageNet categories and generate "ideal" paintings of each class.
Finally, we present a new concept called intrinsic style transfer. By
minimizing only the content loss from neural style transfer, we allow the
artistic medium, in this case, brushstrokes, to naturally dictate the resulting
style.},
added-at = {2019-05-20T22:33:08.000+0200},
author = {Nakano, Reiichiro},
biburl = {https://www.bibsonomy.org/bibtex/236621ab8dcf8cf01aed0ebe0cf742a53/bechr7},
description = {Neural Painters: A learned differentiable constraint for generating brushstroke paintings},
interhash = {77c6670a112b8244fea73154430a9a14},
intrahash = {36621ab8dcf8cf01aed0ebe0cf742a53},
keywords = {cv dl},
note = {cite arxiv:1904.08410Comment: Added more references and acknowledgments},
timestamp = {2019-05-20T22:33:08.000+0200},
title = {Neural Painters: A learned differentiable constraint for generating
brushstroke paintings},
url = {http://arxiv.org/abs/1904.08410},
year = 2019
}