Image manipulation has attracted much research over the years due to the
popularity and commercial importance of the task. In recent years, deep neural
network methods have been proposed for many image manipulation tasks. A major
issue with deep methods is the need to train on large amounts of data from the
same distribution as the target image, whereas collecting datasets encompassing
the entire long-tail of images is impossible. In this paper, we demonstrate
that simply training a conditional adversarial generator on the single target
image is sufficient for performing complex image manipulations. We find that
the key for enabling single image training is extensive augmentation of the
input image and provide a novel augmentation method. Our network learns to map
between a primitive representation of the image (e.g. edges) to the image
itself. At manipulation time, our generator allows for making general image
changes by modifying the primitive input representation and mapping it through
the network. We extensively evaluate our method and find that it provides
remarkable performance.
%0 Generic
%1 vinker2020single
%A Vinker, Yael
%A Horwitz, Eliahu
%A Zabari, Nir
%A Hoshen, Yedid
%D 2020
%K 2020 deep-learning image-processing
%T Deep Single Image Manipulation
%U http://arxiv.org/abs/2007.01289
%X Image manipulation has attracted much research over the years due to the
popularity and commercial importance of the task. In recent years, deep neural
network methods have been proposed for many image manipulation tasks. A major
issue with deep methods is the need to train on large amounts of data from the
same distribution as the target image, whereas collecting datasets encompassing
the entire long-tail of images is impossible. In this paper, we demonstrate
that simply training a conditional adversarial generator on the single target
image is sufficient for performing complex image manipulations. We find that
the key for enabling single image training is extensive augmentation of the
input image and provide a novel augmentation method. Our network learns to map
between a primitive representation of the image (e.g. edges) to the image
itself. At manipulation time, our generator allows for making general image
changes by modifying the primitive input representation and mapping it through
the network. We extensively evaluate our method and find that it provides
remarkable performance.
@misc{vinker2020single,
abstract = {Image manipulation has attracted much research over the years due to the
popularity and commercial importance of the task. In recent years, deep neural
network methods have been proposed for many image manipulation tasks. A major
issue with deep methods is the need to train on large amounts of data from the
same distribution as the target image, whereas collecting datasets encompassing
the entire long-tail of images is impossible. In this paper, we demonstrate
that simply training a conditional adversarial generator on the single target
image is sufficient for performing complex image manipulations. We find that
the key for enabling single image training is extensive augmentation of the
input image and provide a novel augmentation method. Our network learns to map
between a primitive representation of the image (e.g. edges) to the image
itself. At manipulation time, our generator allows for making general image
changes by modifying the primitive input representation and mapping it through
the network. We extensively evaluate our method and find that it provides
remarkable performance.},
added-at = {2020-07-04T20:27:05.000+0200},
author = {Vinker, Yael and Horwitz, Eliahu and Zabari, Nir and Hoshen, Yedid},
biburl = {https://www.bibsonomy.org/bibtex/21a320a9679fddbd651e9223b8125f6d5/analyst},
description = {[2007.01289] Deep Single Image Manipulation},
interhash = {99aa4098723cf288daf1c18861028e1f},
intrahash = {1a320a9679fddbd651e9223b8125f6d5},
keywords = {2020 deep-learning image-processing},
note = {cite arxiv:2007.01289Comment: Project page: http://www.vision.huji.ac.il/deepsim/},
timestamp = {2020-07-04T20:27:05.000+0200},
title = {Deep Single Image Manipulation},
url = {http://arxiv.org/abs/2007.01289},
year = 2020
}