High-Rankness Regularized Semi-Supervised Deep Metric Learning for Remote Sensing Imagery
Deep metric learning has recently received special attention in the field of remote sensing (RS) scene characterization, owing to its prominent capabilities for modeling distances among RS images based on their semantic information. Most of the existing deep metric learning methods exploit pairwise...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-08-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/12/16/2603 |
id |
doaj-fc69fbc77b744ef0b46baef260058941 |
---|---|
record_format |
Article |
spelling |
doaj-fc69fbc77b744ef0b46baef2600589412020-11-25T02:50:29ZengMDPI AGRemote Sensing2072-42922020-08-01122603260310.3390/rs12162603High-Rankness Regularized Semi-Supervised Deep Metric Learning for Remote Sensing ImageryJian Kang0Ruben Fernandez-Beltran1Zhen Ye2Xiaohua Tong3Pedram Ghamisi4Antonio Plaza5Faculty of Electrical Engineering and Computer Science, Technische Universität Berlin, 10587 Berlin, GermanyInstitute of New Imaging Technologies, University Jaume I, 12071 Castellón de la Plana, SpainCollege of Surveying and Geo-Informatics, Tongji University, Shanghai 200092, ChinaCollege of Surveying and Geo-Informatics, Tongji University, Shanghai 200092, ChinaHelmholtz Zentrum Dresden-Rossendorf, Helmholtz Institute Freiberg for Resource Technology, Exploration Division, Machine Learning Group, 09599 Freiberg, GermanyHyperspectral Computing Laboratory, Department of Technology of Computers and Communications, Escuela Politécnica, University of Extremadura, 10003 Cáceres, SpainDeep metric learning has recently received special attention in the field of remote sensing (RS) scene characterization, owing to its prominent capabilities for modeling distances among RS images based on their semantic information. Most of the existing deep metric learning methods exploit pairwise and triplet losses to learn the feature embeddings with the preservation of semantic-similarity, which requires the construction of image pairs and triplets based on the supervised information (e.g., class labels). However, generating such semantic annotations becomes a completely unaffordable task in large-scale RS archives, which may eventually constrain the availability of sufficient training data for this kind of models. To address this issue, we reformulate the deep metric learning scheme in a semi-supervised manner to effectively characterize RS scenes. Specifically, we aim at learning metric spaces by utilizing the supervised information from a small number of labeled RS images and exploring the potential decision boundaries for massive sets of unlabeled aerial scenes. In order to reach this goal, a joint loss function, composed of a normalized softmax loss with margin and a high-rankness regularization term, is proposed, as well as its corresponding optimization algorithm. The conducted experiments (including different state-of-the-art methods and two benchmark RS archives) validate the effectiveness of the proposed approach for RS image classification, clustering and retrieval tasks. The codes of this paper are publicly availablehttps://www.mdpi.com/2072-4292/12/16/2603deep metric learningremote sensingimage characterizationsemi-supervised learning |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Jian Kang Ruben Fernandez-Beltran Zhen Ye Xiaohua Tong Pedram Ghamisi Antonio Plaza |
spellingShingle |
Jian Kang Ruben Fernandez-Beltran Zhen Ye Xiaohua Tong Pedram Ghamisi Antonio Plaza High-Rankness Regularized Semi-Supervised Deep Metric Learning for Remote Sensing Imagery Remote Sensing deep metric learning remote sensing image characterization semi-supervised learning |
author_facet |
Jian Kang Ruben Fernandez-Beltran Zhen Ye Xiaohua Tong Pedram Ghamisi Antonio Plaza |
author_sort |
Jian Kang |
title |
High-Rankness Regularized Semi-Supervised Deep Metric Learning for Remote Sensing Imagery |
title_short |
High-Rankness Regularized Semi-Supervised Deep Metric Learning for Remote Sensing Imagery |
title_full |
High-Rankness Regularized Semi-Supervised Deep Metric Learning for Remote Sensing Imagery |
title_fullStr |
High-Rankness Regularized Semi-Supervised Deep Metric Learning for Remote Sensing Imagery |
title_full_unstemmed |
High-Rankness Regularized Semi-Supervised Deep Metric Learning for Remote Sensing Imagery |
title_sort |
high-rankness regularized semi-supervised deep metric learning for remote sensing imagery |
publisher |
MDPI AG |
series |
Remote Sensing |
issn |
2072-4292 |
publishDate |
2020-08-01 |
description |
Deep metric learning has recently received special attention in the field of remote sensing (RS) scene characterization, owing to its prominent capabilities for modeling distances among RS images based on their semantic information. Most of the existing deep metric learning methods exploit pairwise and triplet losses to learn the feature embeddings with the preservation of semantic-similarity, which requires the construction of image pairs and triplets based on the supervised information (e.g., class labels). However, generating such semantic annotations becomes a completely unaffordable task in large-scale RS archives, which may eventually constrain the availability of sufficient training data for this kind of models. To address this issue, we reformulate the deep metric learning scheme in a semi-supervised manner to effectively characterize RS scenes. Specifically, we aim at learning metric spaces by utilizing the supervised information from a small number of labeled RS images and exploring the potential decision boundaries for massive sets of unlabeled aerial scenes. In order to reach this goal, a joint loss function, composed of a normalized softmax loss with margin and a high-rankness regularization term, is proposed, as well as its corresponding optimization algorithm. The conducted experiments (including different state-of-the-art methods and two benchmark RS archives) validate the effectiveness of the proposed approach for RS image classification, clustering and retrieval tasks. The codes of this paper are publicly available |
topic |
deep metric learning remote sensing image characterization semi-supervised learning |
url |
https://www.mdpi.com/2072-4292/12/16/2603 |
work_keys_str_mv |
AT jiankang highranknessregularizedsemisuperviseddeepmetriclearningforremotesensingimagery AT rubenfernandezbeltran highranknessregularizedsemisuperviseddeepmetriclearningforremotesensingimagery AT zhenye highranknessregularizedsemisuperviseddeepmetriclearningforremotesensingimagery AT xiaohuatong highranknessregularizedsemisuperviseddeepmetriclearningforremotesensingimagery AT pedramghamisi highranknessregularizedsemisuperviseddeepmetriclearningforremotesensingimagery AT antonioplaza highranknessregularizedsemisuperviseddeepmetriclearningforremotesensingimagery |
_version_ |
1724738313704701952 |