Fusing Transformers in a Tuning Fork Structure for Hyperspectral Image Classification Across Disjoint Samples

The 3-D swin transformer (3DST) and spatial–spectral transformer (SST) each excel in capturing distinct aspects of image information: the 3DST with hierarchical attention and window-based processing, and the SST with self-attention mechanisms for long-range dependencies. However, applying...

Full description

Saved in:
Bibliographic Details
Main Authors: Muhammad Ahmad, Muhammad Usama, Manuel Mazzara, Salvatore Distefano, Hamad Ahmed Altuwaijri, Silvia Liberata Ullo
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/10685113/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846163869374873600
author Muhammad Ahmad
Muhammad Usama
Manuel Mazzara
Salvatore Distefano
Hamad Ahmed Altuwaijri
Silvia Liberata Ullo
author_facet Muhammad Ahmad
Muhammad Usama
Manuel Mazzara
Salvatore Distefano
Hamad Ahmed Altuwaijri
Silvia Liberata Ullo
author_sort Muhammad Ahmad
collection DOAJ
description The 3-D swin transformer (3DST) and spatial–spectral transformer (SST) each excel in capturing distinct aspects of image information: the 3DST with hierarchical attention and window-based processing, and the SST with self-attention mechanisms for long-range dependencies. However, applying them independently reveals the following limitations: the 3DST struggles with spectral information, while the SST lacks in capturing fine spatial details. In this article, we propose a novel tuning fork fusion approach to overcome these shortcomings, integrating the 3DST and SST to enhance the hyperspectral image (HSI) classification (HSIC). Our method integrates the hierarchical attention mechanism from the 3DST with the long-range dependence modeling of the SST. This combination refines spatial and spectral information representation and merges insights from both transformers at a fine-grained level. By emphasizing the fusion of attention mechanisms from both architectures, our approach significantly enhances the model's ability to capture complex spatial–spectral relationships, resulting in improved HSIC accuracy. In addition, we highlight the importance of disjoint training, validation, and test samples to enhance model generalization. Experimentation on benchmark HSI datasets demonstrates the superiority of our fusion approach over other state-of-the-art methods and standalone transformers. The source code has been developed from scratch and will be made public upon acceptance.
format Article
id doaj-art-db3482e359d246c78013a2c7bebf6d7a
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-db3482e359d246c78013a2c7bebf6d7a2024-11-19T00:00:55ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352024-01-0117181671818110.1109/JSTARS.2024.346583110685113Fusing Transformers in a Tuning Fork Structure for Hyperspectral Image Classification Across Disjoint SamplesMuhammad Ahmad0https://orcid.org/0000-0002-3320-2261Muhammad Usama1https://orcid.org/0000-0001-5015-8605Manuel Mazzara2https://orcid.org/0000-0002-3860-4948Salvatore Distefano3Hamad Ahmed Altuwaijri4https://orcid.org/0000-0002-2604-5974Silvia Liberata Ullo5https://orcid.org/0000-0001-6294-0581Department of Computer Science, National University of Computer and Emerging Sciences Islamabad, Chiniot-Faisalabad Campus, Chiniot, PakistanDepartment of Computer Science, National University of Computer and Emerging Sciences Islamabad, Chiniot-Faisalabad Campus, Chiniot, PakistanInstitute of Software Development and Engineering, Innopolis University, Innopolis, RussiaDipartimento di Matematica e Informatica—MIFT, University of Messina, Messina, ItalyDepartment of Geography, College of Humanities and Social Sciences, King Saud University, Riyadh, Saudi ArabiaDepartment of Engineering, University of Sannio, Benevento, ItalyThe 3-D swin transformer (3DST) and spatial–spectral transformer (SST) each excel in capturing distinct aspects of image information: the 3DST with hierarchical attention and window-based processing, and the SST with self-attention mechanisms for long-range dependencies. However, applying them independently reveals the following limitations: the 3DST struggles with spectral information, while the SST lacks in capturing fine spatial details. In this article, we propose a novel tuning fork fusion approach to overcome these shortcomings, integrating the 3DST and SST to enhance the hyperspectral image (HSI) classification (HSIC). Our method integrates the hierarchical attention mechanism from the 3DST with the long-range dependence modeling of the SST. This combination refines spatial and spectral information representation and merges insights from both transformers at a fine-grained level. By emphasizing the fusion of attention mechanisms from both architectures, our approach significantly enhances the model's ability to capture complex spatial–spectral relationships, resulting in improved HSIC accuracy. In addition, we highlight the importance of disjoint training, validation, and test samples to enhance model generalization. Experimentation on benchmark HSI datasets demonstrates the superiority of our fusion approach over other state-of-the-art methods and standalone transformers. The source code has been developed from scratch and will be made public upon acceptance.https://ieeexplore.ieee.org/document/10685113/3-D Swin Transformer (3DST)feature fusionhyperspectral image classification (HSIC)spatial–spectral featuresspatial–spectral transformer (SST)
spellingShingle Muhammad Ahmad
Muhammad Usama
Manuel Mazzara
Salvatore Distefano
Hamad Ahmed Altuwaijri
Silvia Liberata Ullo
Fusing Transformers in a Tuning Fork Structure for Hyperspectral Image Classification Across Disjoint Samples
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
3-D Swin Transformer (3DST)
feature fusion
hyperspectral image classification (HSIC)
spatial–spectral features
spatial–spectral transformer (SST)
title Fusing Transformers in a Tuning Fork Structure for Hyperspectral Image Classification Across Disjoint Samples
title_full Fusing Transformers in a Tuning Fork Structure for Hyperspectral Image Classification Across Disjoint Samples
title_fullStr Fusing Transformers in a Tuning Fork Structure for Hyperspectral Image Classification Across Disjoint Samples
title_full_unstemmed Fusing Transformers in a Tuning Fork Structure for Hyperspectral Image Classification Across Disjoint Samples
title_short Fusing Transformers in a Tuning Fork Structure for Hyperspectral Image Classification Across Disjoint Samples
title_sort fusing transformers in a tuning fork structure for hyperspectral image classification across disjoint samples
topic 3-D Swin Transformer (3DST)
feature fusion
hyperspectral image classification (HSIC)
spatial–spectral features
spatial–spectral transformer (SST)
url https://ieeexplore.ieee.org/document/10685113/
work_keys_str_mv AT muhammadahmad fusingtransformersinatuningforkstructureforhyperspectralimageclassificationacrossdisjointsamples
AT muhammadusama fusingtransformersinatuningforkstructureforhyperspectralimageclassificationacrossdisjointsamples
AT manuelmazzara fusingtransformersinatuningforkstructureforhyperspectralimageclassificationacrossdisjointsamples
AT salvatoredistefano fusingtransformersinatuningforkstructureforhyperspectralimageclassificationacrossdisjointsamples
AT hamadahmedaltuwaijri fusingtransformersinatuningforkstructureforhyperspectralimageclassificationacrossdisjointsamples
AT silvialiberataullo fusingtransformersinatuningforkstructureforhyperspectralimageclassificationacrossdisjointsamples