Gatys et al. (2015) showed that optimizing pixels to match features in a
convolutional network with respect reference image features is a way to render
images of high visual quality. We show that unrolling this gradient-based
optimization yields a recurrent computation that creates images by
incrementally adding onto a visual "canvas". We propose a recurrent generative
model inspired by this view, and show that it can be trained using adversarial
training to generate very good image samples. We also propose a way to
quantitatively compare adversarial networks by having the generators and
discriminators of these networks compete against each other.
Beschreibung
[1602.05110] Generating images with recurrent adversarial networks
%0 Journal Article
%1 im2016generating
%A Im, Daniel Jiwoong
%A Kim, Chris Dongjoo
%A Jiang, Hui
%A Memisevic, Roland
%D 2016
%K deep-learning generative-models
%T Generating images with recurrent adversarial networks
%U http://arxiv.org/abs/1602.05110
%X Gatys et al. (2015) showed that optimizing pixels to match features in a
convolutional network with respect reference image features is a way to render
images of high visual quality. We show that unrolling this gradient-based
optimization yields a recurrent computation that creates images by
incrementally adding onto a visual "canvas". We propose a recurrent generative
model inspired by this view, and show that it can be trained using adversarial
training to generate very good image samples. We also propose a way to
quantitatively compare adversarial networks by having the generators and
discriminators of these networks compete against each other.
@article{im2016generating,
abstract = {Gatys et al. (2015) showed that optimizing pixels to match features in a
convolutional network with respect reference image features is a way to render
images of high visual quality. We show that unrolling this gradient-based
optimization yields a recurrent computation that creates images by
incrementally adding onto a visual "canvas". We propose a recurrent generative
model inspired by this view, and show that it can be trained using adversarial
training to generate very good image samples. We also propose a way to
quantitatively compare adversarial networks by having the generators and
discriminators of these networks compete against each other.},
added-at = {2019-04-23T13:13:33.000+0200},
author = {Im, Daniel Jiwoong and Kim, Chris Dongjoo and Jiang, Hui and Memisevic, Roland},
biburl = {https://www.bibsonomy.org/bibtex/275982e9b7f1aaa21e00f8ce9c07ba101/kirk86},
description = {[1602.05110] Generating images with recurrent adversarial networks},
interhash = {e43416aaa4961045a0ea27025ad80728},
intrahash = {75982e9b7f1aaa21e00f8ce9c07ba101},
keywords = {deep-learning generative-models},
note = {cite arxiv:1602.05110},
timestamp = {2019-04-23T13:13:33.000+0200},
title = {Generating images with recurrent adversarial networks},
url = {http://arxiv.org/abs/1602.05110},
year = 2016
}