Adversarial Learning for Semi-Supervised Semantic Segmentation
W. Hung, Y. Tsai, Y. Liou, Y. Lin, and M. Yang. (2018)cite arxiv:1802.07934Comment: Accepted in BMVC 2018. Code and models available at https://github.com/hfslyc/AdvSemiSeg.
Abstract
We propose a method for semi-supervised semantic segmentation using an
adversarial network. While most existing discriminators are trained to classify
input images as real or fake on the image level, we design a discriminator in a
fully convolutional manner to differentiate the predicted probability maps from
the ground truth segmentation distribution with the consideration of the
spatial resolution. We show that the proposed discriminator can be used to
improve semantic segmentation accuracy by coupling the adversarial loss with
the standard cross entropy loss of the proposed model. In addition, the fully
convolutional discriminator enables semi-supervised learning through
discovering the trustworthy regions in predicted results of unlabeled images,
thereby providing additional supervisory signals. In contrast to existing
methods that utilize weakly-labeled images, our method leverages unlabeled
images to enhance the segmentation model. Experimental results on the PASCAL
VOC 2012 and Cityscapes datasets demonstrate the effectiveness of the proposed
algorithm.
%0 Generic
%1 hung2018adversarial
%A Hung, Wei-Chih
%A Tsai, Yi-Hsuan
%A Liou, Yan-Ting
%A Lin, Yen-Yu
%A Yang, Ming-Hsuan
%D 2018
%K 2018 GAN arxiv computer-vision paper segmentation
%T Adversarial Learning for Semi-Supervised Semantic Segmentation
%U http://arxiv.org/abs/1802.07934
%X We propose a method for semi-supervised semantic segmentation using an
adversarial network. While most existing discriminators are trained to classify
input images as real or fake on the image level, we design a discriminator in a
fully convolutional manner to differentiate the predicted probability maps from
the ground truth segmentation distribution with the consideration of the
spatial resolution. We show that the proposed discriminator can be used to
improve semantic segmentation accuracy by coupling the adversarial loss with
the standard cross entropy loss of the proposed model. In addition, the fully
convolutional discriminator enables semi-supervised learning through
discovering the trustworthy regions in predicted results of unlabeled images,
thereby providing additional supervisory signals. In contrast to existing
methods that utilize weakly-labeled images, our method leverages unlabeled
images to enhance the segmentation model. Experimental results on the PASCAL
VOC 2012 and Cityscapes datasets demonstrate the effectiveness of the proposed
algorithm.
@misc{hung2018adversarial,
abstract = {We propose a method for semi-supervised semantic segmentation using an
adversarial network. While most existing discriminators are trained to classify
input images as real or fake on the image level, we design a discriminator in a
fully convolutional manner to differentiate the predicted probability maps from
the ground truth segmentation distribution with the consideration of the
spatial resolution. We show that the proposed discriminator can be used to
improve semantic segmentation accuracy by coupling the adversarial loss with
the standard cross entropy loss of the proposed model. In addition, the fully
convolutional discriminator enables semi-supervised learning through
discovering the trustworthy regions in predicted results of unlabeled images,
thereby providing additional supervisory signals. In contrast to existing
methods that utilize weakly-labeled images, our method leverages unlabeled
images to enhance the segmentation model. Experimental results on the PASCAL
VOC 2012 and Cityscapes datasets demonstrate the effectiveness of the proposed
algorithm.},
added-at = {2018-09-11T19:10:15.000+0200},
author = {Hung, Wei-Chih and Tsai, Yi-Hsuan and Liou, Yan-Ting and Lin, Yen-Yu and Yang, Ming-Hsuan},
biburl = {https://www.bibsonomy.org/bibtex/22d16f344adf449d2eb65346205decf8d/analyst},
description = {1802.07934.pdf},
interhash = {ea7b30b3709ec90d7d90602252acbd28},
intrahash = {2d16f344adf449d2eb65346205decf8d},
keywords = {2018 GAN arxiv computer-vision paper segmentation},
note = {cite arxiv:1802.07934Comment: Accepted in BMVC 2018. Code and models available at https://github.com/hfslyc/AdvSemiSeg},
timestamp = {2018-09-11T19:10:15.000+0200},
title = {Adversarial Learning for Semi-Supervised Semantic Segmentation},
url = {http://arxiv.org/abs/1802.07934},
year = 2018
}