Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning
Convolutional neural networks (CNNs) have been widely applied in hyperspectral imagery (HSI) classification. However, their classification performance might be limited by the scarcity of labeled data to be used for training and validation. In this paper, we propose a novel lightweight shuffled group...
Published in: | Remote Sensing |
---|---|
Main Authors: | , , , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
MDPI
2020
|
Subjects: | |
Online Access: | https://hdl.handle.net/10037/22936 https://doi.org/10.3390/rs12111780 |
id |
ftunivtroemsoe:oai:munin.uit.no:10037/22936 |
---|---|
record_format |
openpolar |
spelling |
ftunivtroemsoe:oai:munin.uit.no:10037/22936 2023-05-15T14:26:06+02:00 Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning Liu, Yao Gao, Lianru Xiao, Chenchao Qu, Ying Zheng, Ke Marinoni, Andrea 2020-06-01 https://hdl.handle.net/10037/22936 https://doi.org/10.3390/rs12111780 eng eng MDPI Remote Sensing info:eu-repo/grantAgreement/RCN/SFI/237906/Norway/Centre for Integrated Remote Sensing and Forecasting for Arctic Operations/CIRFA/ Liu, Gao, Xiao, Qu, Zheng, Marinoni. Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning. Remote Sensing. 2020;12(11) FRIDAID 1895743 doi:10.3390/rs12111780 2072-4292 https://hdl.handle.net/10037/22936 openAccess Copyright 2020 The Author(s) VDP::Technology: 500 VDP::Teknologi: 500 VDP::Mathematics and natural science: 400::Physics: 430 VDP::Matematikk og Naturvitenskap: 400::Fysikk: 430 Journal article Tidsskriftartikkel Peer reviewed publishedVersion 2020 ftunivtroemsoe https://doi.org/10.3390/rs12111780 2021-11-10T23:54:29Z Convolutional neural networks (CNNs) have been widely applied in hyperspectral imagery (HSI) classification. However, their classification performance might be limited by the scarcity of labeled data to be used for training and validation. In this paper, we propose a novel lightweight shuffled group convolutional neural network (abbreviated as SG-CNN) to achieve efficient training with a limited training dataset in HSI classification. SG-CNN consists of SG conv units that employ conventional and atrous convolution in different groups, followed by channel shuffle operation and shortcut connection. In this way, SG-CNNs have less trainable parameters, whilst they can still be accurately and efficiently trained with fewer labeled samples. Transfer learning between different HSI datasets is also applied on the SG-CNN to further improve the classification accuracy. To evaluate the effectiveness of SG-CNNs for HSI classification, experiments have been conducted on three public HSI datasets pretrained on HSIs from different sensors. SG-CNNs with different levels of complexity were tested, and their classification results were compared with fine-tuned ShuffleNet2, ResNeXt, and their original counterparts. The experimental results demonstrate that SG-CNNs can achieve competitive classification performance when the amount of labeled data for training is poor, as well as efficiently providing satisfying classification results. Article in Journal/Newspaper Arctic University of Tromsø: Munin Open Research Archive Remote Sensing 12 11 1780 |
institution |
Open Polar |
collection |
University of Tromsø: Munin Open Research Archive |
op_collection_id |
ftunivtroemsoe |
language |
English |
topic |
VDP::Technology: 500 VDP::Teknologi: 500 VDP::Mathematics and natural science: 400::Physics: 430 VDP::Matematikk og Naturvitenskap: 400::Fysikk: 430 |
spellingShingle |
VDP::Technology: 500 VDP::Teknologi: 500 VDP::Mathematics and natural science: 400::Physics: 430 VDP::Matematikk og Naturvitenskap: 400::Fysikk: 430 Liu, Yao Gao, Lianru Xiao, Chenchao Qu, Ying Zheng, Ke Marinoni, Andrea Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning |
topic_facet |
VDP::Technology: 500 VDP::Teknologi: 500 VDP::Mathematics and natural science: 400::Physics: 430 VDP::Matematikk og Naturvitenskap: 400::Fysikk: 430 |
description |
Convolutional neural networks (CNNs) have been widely applied in hyperspectral imagery (HSI) classification. However, their classification performance might be limited by the scarcity of labeled data to be used for training and validation. In this paper, we propose a novel lightweight shuffled group convolutional neural network (abbreviated as SG-CNN) to achieve efficient training with a limited training dataset in HSI classification. SG-CNN consists of SG conv units that employ conventional and atrous convolution in different groups, followed by channel shuffle operation and shortcut connection. In this way, SG-CNNs have less trainable parameters, whilst they can still be accurately and efficiently trained with fewer labeled samples. Transfer learning between different HSI datasets is also applied on the SG-CNN to further improve the classification accuracy. To evaluate the effectiveness of SG-CNNs for HSI classification, experiments have been conducted on three public HSI datasets pretrained on HSIs from different sensors. SG-CNNs with different levels of complexity were tested, and their classification results were compared with fine-tuned ShuffleNet2, ResNeXt, and their original counterparts. The experimental results demonstrate that SG-CNNs can achieve competitive classification performance when the amount of labeled data for training is poor, as well as efficiently providing satisfying classification results. |
format |
Article in Journal/Newspaper |
author |
Liu, Yao Gao, Lianru Xiao, Chenchao Qu, Ying Zheng, Ke Marinoni, Andrea |
author_facet |
Liu, Yao Gao, Lianru Xiao, Chenchao Qu, Ying Zheng, Ke Marinoni, Andrea |
author_sort |
Liu, Yao |
title |
Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning |
title_short |
Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning |
title_full |
Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning |
title_fullStr |
Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning |
title_full_unstemmed |
Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning |
title_sort |
hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning |
publisher |
MDPI |
publishDate |
2020 |
url |
https://hdl.handle.net/10037/22936 https://doi.org/10.3390/rs12111780 |
genre |
Arctic |
genre_facet |
Arctic |
op_relation |
Remote Sensing info:eu-repo/grantAgreement/RCN/SFI/237906/Norway/Centre for Integrated Remote Sensing and Forecasting for Arctic Operations/CIRFA/ Liu, Gao, Xiao, Qu, Zheng, Marinoni. Hyperspectral image classification based on a shuffled group convolutional neural network with transfer learning. Remote Sensing. 2020;12(11) FRIDAID 1895743 doi:10.3390/rs12111780 2072-4292 https://hdl.handle.net/10037/22936 |
op_rights |
openAccess Copyright 2020 The Author(s) |
op_doi |
https://doi.org/10.3390/rs12111780 |
container_title |
Remote Sensing |
container_volume |
12 |
container_issue |
11 |
container_start_page |
1780 |
_version_ |
1766298578999836672 |