Scene Understanding Based on High-Order Potentials and Generative Adversarial Networks

Scene Understanding Based on High-Order Potentials and Generative Adversarial Networks

Scene understanding is to predict a class label at each pixel of an image. In this study, we propose a semantic segmentation framework based on classic generative adversarial nets (GAN) to train a fully convolutional semantic segmentation model along with an adversarial network. To improve the consi...

Full description

Bibliographic Details
Main Authors: Xiaoli Zhao, Guozhong Wang, Jiaqi Zhang, Xiang Zhang
Format: Article
Language:English
Published: Hindawi Limited 2018-01-01
Series:Advances in Multimedia
Online Access:http://dx.doi.org/10.1155/2018/8207201
id doaj-c39e1f67b2034cb3a2d597d0774c6f49
record_format Article
spelling doaj-c39e1f67b2034cb3a2d597d0774c6f492020-11-25T02:51:30ZengHindawi LimitedAdvances in Multimedia1687-56801687-56992018-01-01201810.1155/2018/82072018207201Scene Understanding Based on High-Order Potentials and Generative Adversarial NetworksXiaoli Zhao0Guozhong Wang1Jiaqi Zhang2Xiang Zhang3School of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, ChinaSchool of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, ChinaSchool of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, ChinaSchool of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, ChinaScene understanding is to predict a class label at each pixel of an image. In this study, we propose a semantic segmentation framework based on classic generative adversarial nets (GAN) to train a fully convolutional semantic segmentation model along with an adversarial network. To improve the consistency of the segmented image, the high-order potentials, instead of unary or pairwise potentials, are adopted. We realize the high-order potentials by substituting adversarial network for CRF model, which can continuously improve the consistency and details of the segmented semantic image until it cannot discriminate the segmented result from the ground truth. A number of experiments are conducted on PASCAL VOC 2012 and Cityscapes datasets, and the quantitative and qualitative assessments have shown the effectiveness of our proposed approach.http://dx.doi.org/10.1155/2018/8207201
collection DOAJ
language English
format Article
sources DOAJ
author Xiaoli Zhao
Guozhong Wang
Jiaqi Zhang
Xiang Zhang
spellingShingle Xiaoli Zhao
Guozhong Wang
Jiaqi Zhang
Xiang Zhang
Scene Understanding Based on High-Order Potentials and Generative Adversarial Networks
Advances in Multimedia
author_facet Xiaoli Zhao
Guozhong Wang
Jiaqi Zhang
Xiang Zhang
author_sort Xiaoli Zhao
title Scene Understanding Based on High-Order Potentials and Generative Adversarial Networks
title_short Scene Understanding Based on High-Order Potentials and Generative Adversarial Networks
title_full Scene Understanding Based on High-Order Potentials and Generative Adversarial Networks
title_fullStr Scene Understanding Based on High-Order Potentials and Generative Adversarial Networks
title_full_unstemmed Scene Understanding Based on High-Order Potentials and Generative Adversarial Networks
title_sort scene understanding based on high-order potentials and generative adversarial networks
publisher Hindawi Limited
series Advances in Multimedia
issn 1687-5680
1687-5699
publishDate 2018-01-01
description Scene understanding is to predict a class label at each pixel of an image. In this study, we propose a semantic segmentation framework based on classic generative adversarial nets (GAN) to train a fully convolutional semantic segmentation model along with an adversarial network. To improve the consistency of the segmented image, the high-order potentials, instead of unary or pairwise potentials, are adopted. We realize the high-order potentials by substituting adversarial network for CRF model, which can continuously improve the consistency and details of the segmented semantic image until it cannot discriminate the segmented result from the ground truth. A number of experiments are conducted on PASCAL VOC 2012 and Cityscapes datasets, and the quantitative and qualitative assessments have shown the effectiveness of our proposed approach.
url http://dx.doi.org/10.1155/2018/8207201
work_keys_str_mv AT xiaolizhao sceneunderstandingbasedonhighorderpotentialsandgenerativeadversarialnetworks
AT guozhongwang sceneunderstandingbasedonhighorderpotentialsandgenerativeadversarialnetworks
AT jiaqizhang sceneunderstandingbasedonhighorderpotentialsandgenerativeadversarialnetworks
AT xiangzhang sceneunderstandingbasedonhighorderpotentialsandgenerativeadversarialnetworks
_version_ 1724734235630108672

Similar Items