Attention Guided Semisupervised Generative Transfer Learning for Hyperspectral Image Analysis

In geospatial image analysis, domain shifts caused by differences between datasets often undermine the performance of deep learning models due to their limited generalization ability. This issue is particularly pronounced in hyperspectral imagery, given the high dimensionality of the per-pixel refle...

Full description

Saved in:
Bibliographic Details
Main Authors: Anan Yaghmour, Saurabh Prasad, Melba M. Crawford
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10731899/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846170423718313984
author Anan Yaghmour
Saurabh Prasad
Melba M. Crawford
author_facet Anan Yaghmour
Saurabh Prasad
Melba M. Crawford
author_sort Anan Yaghmour
collection DOAJ
description In geospatial image analysis, domain shifts caused by differences between datasets often undermine the performance of deep learning models due to their limited generalization ability. This issue is particularly pronounced in hyperspectral imagery, given the high dimensionality of the per-pixel reflectance vectors and the complexity of the resulting deep learning models. We introduce a semisupervised domain adaptation technique that improves on the adversarial discriminative framework, incorporating a novel multiclass discriminator to address low discriminability and negative transfer issues from which current approaches suffer. Significantly, our method addresses mode collapse by incorporating limited labeled data from the target domain for targeted guidance during adaptation. In addition, we integrate an attention mechanism that focuses on challenging spatial regions for the target mode. We tested our approach on three unique hyperspectral remote sensing datasets to demonstrate its efficacy in diverse conditions (e.g., cloud shadows, atmospheric variability, and terrain). This strategy improves discrimination and reduces negative transfer in domain adaptation for geospatial image analysis.
format Article
id doaj-art-03d06e299d5241e586c8dc2e29e5c47d
institution Kabale University
issn 1939-1404
2151-1535
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-03d06e299d5241e586c8dc2e29e5c47d2024-11-12T00:00:41ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352024-01-0117198841989910.1109/JSTARS.2024.348552810731899Attention Guided Semisupervised Generative Transfer Learning for Hyperspectral Image AnalysisAnan Yaghmour0https://orcid.org/0009-0007-0048-9757Saurabh Prasad1https://orcid.org/0000-0003-3729-9360Melba M. Crawford2https://orcid.org/0000-0003-3459-2094Department of Electrical and Computer Engineering, University of Houston, Houston, TX, USADepartment of Electrical and Computer Engineering, University of Houston, Houston, TX, USALyles School of Civil Engineering and School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USAIn geospatial image analysis, domain shifts caused by differences between datasets often undermine the performance of deep learning models due to their limited generalization ability. This issue is particularly pronounced in hyperspectral imagery, given the high dimensionality of the per-pixel reflectance vectors and the complexity of the resulting deep learning models. We introduce a semisupervised domain adaptation technique that improves on the adversarial discriminative framework, incorporating a novel multiclass discriminator to address low discriminability and negative transfer issues from which current approaches suffer. Significantly, our method addresses mode collapse by incorporating limited labeled data from the target domain for targeted guidance during adaptation. In addition, we integrate an attention mechanism that focuses on challenging spatial regions for the target mode. We tested our approach on three unique hyperspectral remote sensing datasets to demonstrate its efficacy in diverse conditions (e.g., cloud shadows, atmospheric variability, and terrain). This strategy improves discrimination and reduces negative transfer in domain adaptation for geospatial image analysis.https://ieeexplore.ieee.org/document/10731899/Attention GANsdomain adaptationgenerative adversarial learninghyperspectralremote sensingsemantic segmentation
spellingShingle Anan Yaghmour
Saurabh Prasad
Melba M. Crawford
Attention Guided Semisupervised Generative Transfer Learning for Hyperspectral Image Analysis
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Attention GANs
domain adaptation
generative adversarial learning
hyperspectral
remote sensing
semantic segmentation
title Attention Guided Semisupervised Generative Transfer Learning for Hyperspectral Image Analysis
title_full Attention Guided Semisupervised Generative Transfer Learning for Hyperspectral Image Analysis
title_fullStr Attention Guided Semisupervised Generative Transfer Learning for Hyperspectral Image Analysis
title_full_unstemmed Attention Guided Semisupervised Generative Transfer Learning for Hyperspectral Image Analysis
title_short Attention Guided Semisupervised Generative Transfer Learning for Hyperspectral Image Analysis
title_sort attention guided semisupervised generative transfer learning for hyperspectral image analysis
topic Attention GANs
domain adaptation
generative adversarial learning
hyperspectral
remote sensing
semantic segmentation
url https://ieeexplore.ieee.org/document/10731899/
work_keys_str_mv AT ananyaghmour attentionguidedsemisupervisedgenerativetransferlearningforhyperspectralimageanalysis
AT saurabhprasad attentionguidedsemisupervisedgenerativetransferlearningforhyperspectralimageanalysis
AT melbamcrawford attentionguidedsemisupervisedgenerativetransferlearningforhyperspectralimageanalysis