pith. sign in

arxiv: 2604.15828 · v1 · submitted 2026-04-17 · 💻 cs.CV

SSFT: A Lightweight Spectral-Spatial Fusion Transformer for Generic Hyperspectral Classification

Alexander Musiat , Nikolas Ebert , Oliver Wasenm\"uller This is my paper

Pith reviewed 2026-05-10 09:12 UTC · model grok-4.3

classification 💻 cs.CV
keywords hyperspectral classificationspectral-spatial fusionlightweight transformercross-attentionHSI-BenchmarkSpectralEarthparameter efficiency
0
0 comments X

The pith

A compact transformer fuses separate spectral and spatial pathways via cross-attention to lead hyperspectral classification benchmarks while using under 2% of prior model size.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces SSFT, a lightweight transformer that splits representation learning into distinct spectral and spatial streams before merging them with cross-attention. This design targets the challenges of high-dimensional hyperspectral data, limited labels, and domain shifts across applications from earth observation to material recognition. The model ranks first overall on the heterogeneous HSI-Benchmark and remains competitive when transferred to the larger SpectralEarth benchmark. Ablations confirm that both pathways matter and that spatial modeling drives most gains, with the approach holding up without data augmentation. Compact size matters because labeled hyperspectral data stays scarce outside narrow domains, so efficient models open the door to wider use.

Core claim

SSFT factorizes representation learning into spectral and spatial pathways and integrates them via cross-attention to capture complementary wavelength-dependent signatures and structural information, achieving state-of-the-art overall performance on the HSI-Benchmark while using less than 2% of the parameters of the previous leading method and remaining competitive on SpectralEarth transfer.

What carries the argument

Cross-attention fusion between separate spectral and spatial transformer pathways that factorizes feature learning to handle complementary wavelength and structural signals.

If this is right

  • Both spectral and spatial pathways are required, with spatial modeling contributing the larger share of performance.
  • SSFT stays effective without data augmentation on the tested benchmarks.
  • The same compact architecture transfers competitively to a substantially larger hyperspectral collection under its official protocol.
  • The approach supports generic hyperspectral classification across earth observation, fruit assessment, and fine-grained material tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The explicit pathway split may reduce overfitting when labeled samples are few, suggesting similar factorizations could help other high-dimensional imaging tasks.
  • Independent scaling of the spectral versus spatial branches offers a testable route to further efficiency gains.
  • Cross-attention between modality-specific streams provides a template for other multi-channel or multi-sensor classification problems where one modality dominates.

Load-bearing premise

The HSI-Benchmark and SpectralEarth protocols sufficiently represent the range of real-world hyperspectral acquisition conditions and domain shifts.

What would settle it

A new hyperspectral dataset or acquisition regime where models with similar or smaller size outperform SSFT on overall accuracy or where SSFT falls behind prior leaders on the same benchmarks.

Figures

Figures reproduced from arXiv: 2604.15828 by Alexander Musiat, Nikolas Ebert, Oliver Wasenm\"uller.

Figure 1
Figure 1. Figure 1: To demonstrate the general applicability of SSFT, we evaluate on the heterogeneous sub-datasets of the HSI-Benchmark [ [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed SSFT architecture. Given an input hyperspectral data cube, the network factorizes representation learn [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: F1-score vs. Parameters and FLOPs on the Spec [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: t-SNE visualization of feature representations on the Debris validation split. From left to right, the figure shows a representative [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
read the original abstract

Hyperspectral imaging enables fine-grained recognition of materials by capturing rich spectral signatures, but learning robust classifiers is challenging due to high dimensionality, spectral redundancy, limited labeled data, and strong domain shifts. Beyond earth observation, labeled HSI data is often scarce and imbalanced, motivating compact models for generic hyperspectral classification across diverse acquisition regimes. We propose the lightweight Spectral-Spatial Fusion Transformer (SSFT), which factorizes representation learning into spectral and spatial pathways and integrates them via cross-attention to capture complementary wavelength-dependent and structural information. We evaluate our SSFT on the challenging HSI-Benchmark, a heterogeneous multi-dataset benchmark covering earth observation, fruit condition assessment, and fine-grained material recognition. SSFT achieves state-of-the-art overall performance, ranking first while using less than 2% of the parameters of the previous leading method. We further evaluate transfer to the substantially larger SpectralEarth benchmark under the official protocol, where SSFT remains competitive despite its compact size. Ablation studies show that both spectral and spatial pathways are crucial, with spatial modeling contributing most, and that SSFT remains robust without data augmentation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes the Spectral-Spatial Fusion Transformer (SSFT), a lightweight model that factorizes hyperspectral representation learning into separate spectral and spatial pathways integrated via cross-attention. It claims state-of-the-art overall ranking on the heterogeneous HSI-Benchmark (covering earth observation, fruit assessment, and material recognition) while using less than 2% of the parameters of the prior leading method, competitive transfer performance on the larger SpectralEarth benchmark, and ablation results showing both pathways are essential with spatial modeling contributing most and robustness without augmentation.

Significance. If the empirical claims hold under full verification, SSFT offers a parameter-efficient architecture for hyperspectral classification in data-scarce and domain-shifted settings, which could be valuable for applications beyond standard earth observation. The factorization into spectral-spatial pathways with cross-attention is a plausible design choice for capturing complementary information in high-dimensional HSI data.

major comments (2)
  1. [Abstract] Abstract: The central SOTA ranking claim on HSI-Benchmark lacks any reported error bars, standard deviations across runs, dataset split details, or statistical significance tests, preventing verification that the performance margin over prior methods is robust rather than due to experimental variance.
  2. [Abstract] Abstract: The claim of suitability for 'generic hyperspectral classification across diverse acquisition regimes' relies on HSI-Benchmark representativeness, yet no quantitative analysis (e.g., statistics on spectral band counts, spatial resolutions, sensor types, or domain-shift metrics across the included datasets) is provided to demonstrate coverage of relevant variation.
minor comments (1)
  1. [Ablation studies] Ablation studies: The statement that SSFT 'remains robust without data augmentation' would be strengthened by specifying the exact augmentation types tested and the magnitude of any performance drop.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract. We address the two major comments point by point below and indicate the revisions we will incorporate in the next version of the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central SOTA ranking claim on HSI-Benchmark lacks any reported error bars, standard deviations across runs, dataset split details, or statistical significance tests, preventing verification that the performance margin over prior methods is robust rather than due to experimental variance.

    Authors: We agree that the abstract, being a concise summary, does not include error bars, standard deviations, or statistical tests, which limits immediate verification of robustness. The manuscript describes the evaluation protocol and dataset splits in Section 3 following official per-dataset conventions. To address this concern directly, we will revise the abstract to include a brief qualifier referencing the multi-run evaluation and direct readers to the detailed tables in the experimental section for standard deviations and full results. We will also add a short discussion of result consistency across datasets in the main text. revision: yes

  2. Referee: [Abstract] Abstract: The claim of suitability for 'generic hyperspectral classification across diverse acquisition regimes' relies on HSI-Benchmark representativeness, yet no quantitative analysis (e.g., statistics on spectral band counts, spatial resolutions, sensor types, or domain-shift metrics across the included datasets) is provided to demonstrate coverage of relevant variation.

    Authors: We acknowledge that the abstract's generality claim would be strengthened by quantitative characterization of the benchmark's diversity. The manuscript qualitatively positions HSI-Benchmark as heterogeneous across earth observation, fruit assessment, and material recognition tasks. In the revised manuscript we will add a compact table (or paragraph) in the experimental setup section summarizing key statistics such as spectral band counts, spatial resolutions, and sensor types for each constituent dataset, along with a simple domain-variation metric where feasible. This addition will support the representativeness argument without altering the core claims. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical architecture proposal with benchmark results

full rationale

The paper proposes the SSFT architecture (spectral-spatial factorization with cross-attention) and reports empirical results on HSI-Benchmark and SpectralEarth. No derivation chain, equations, or first-principles claims exist that reduce to fitted parameters or self-citations by construction. Performance rankings and parameter counts are direct experimental outcomes on external datasets, not predictions forced by the model's own inputs. Self-citations, if present, are not load-bearing for any central result. The work is self-contained against the stated benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The work rests on standard transformer assumptions and empirical evaluation; no new physical entities or ad-hoc axioms are introduced beyond typical computer-vision domain assumptions.

free parameters (1)
  • transformer hyperparameters (layers, heads, embedding size)
    Chosen to balance performance and parameter count on the target benchmarks.
axioms (1)
  • domain assumption Cross-attention can effectively integrate complementary spectral and spatial features
    Invoked in the model design description.

pith-pipeline@v0.9.0 · 5501 in / 1129 out tokens · 31652 ms · 2026-05-10T09:12:02.103228+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages

  1. [1]

    Muhammad Ahmad et al. Hyperspectral image classification—traditional to deep models: A survey for future prospects.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 15: 968–999, 2021. 2

    work page 2021

  2. [2]

    Partial least squares for discrimination.Journal of Chemometrics, 17(3):166– 173, 2003

    Matthew Barker and William Rayens. Partial least squares for discrimination.Journal of Chemometrics, 17(3):166– 173, 2003. 6

    work page 2003

  3. [3]

    Albrecht, Julien Mairal, Jocelyn Chanussot, Yi Wang, and Xiao Xiang Zhu

    Nassim Ait Ali Braham, Conrad M. Albrecht, Julien Mairal, Jocelyn Chanussot, Yi Wang, and Xiao Xiang Zhu. Spec- tralearth: Training hyperspectral foundation models at scale. IEEE Journal of Selected Topics in Applied Earth Observa- tions and Remote Sensing, 18:16780–16797, 2025. 2, 5, 6, 7

    work page 2025

  4. [4]

    Spectralnet: Ex- ploring spatial-spectral waveletcnn for hyperspectral image classification.arXiv preprint, 2021

    Tanmay Chakraborty and Utkarsh Trehan. Spectralnet: Ex- ploring spatial-spectral waveletcnn for hyperspectral image classification.arXiv preprint, 2021. 2, 6

    work page 2021

  5. [5]

    A lightweight spectral–spatial feature extraction and fusion network for hyperspectral image classification.Remote Sensing, 12(9): 1395, 2020

    Linlin Chen, Zhihui Wei, and Yang Xu. A lightweight spectral–spatial feature extraction and fusion network for hyperspectral image classification.Remote Sensing, 12(9): 1395, 2020. 3

    work page 2020

  6. [6]

    Greenhyperspectra: A multi-source hyper- spectral dataset for global vegetation trait prediction.arXiv preprint arXiv:2507.06806, 2025

    Eya Cherif et al. Greenhyperspectra: A multi-source hyper- spectral dataset for global vegetation trait prediction.arXiv preprint arXiv:2507.06806, 2025. 2

    work page arXiv 2025

  7. [7]

    Cam- bridge University Press, Cambridge, 2004

    Nello Cristianini and John Shawe-Taylor.Support Vector Machines and Other Kernel-Based Learning Methods. Cam- bridge University Press, Cambridge, 2004. 6

    work page 2004

  8. [8]

    Christian Debes, Andreas Merentitis, Roel Heremans, Jerome Hahn, Nikolaos Frangiadakis, Thomas van Kasteren, Wenzhi Liao, Rudy Bellens, Siddharth Gautama, Wilfried Philips, Satya Prasad, Qian Du, and Fabio Pacifici. Hyper- spectral and lidar data fusion: Outcome of the 2013 grss data fusion contest.IEEE Journal of Selected Topics in Applied Earth Observat...

    work page 2013

  9. [9]

    An image is worth 16x16 words: Trans- formers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, et al. An image is worth 16x16 words: Trans- formers for image recognition at scale. InInternational Con- ference on Learning Representations (ICLR), 2021. 4

    work page 2021

  10. [10]

    Plg- vit: Vision transformer with parallel local and global self- attention.Sensors, 2023

    Nikolas Ebert, Didier Stricker, and Oliver Wasenm¨uller. Plg- vit: Vision transformer with parallel local and global self- attention.Sensors, 2023. 4

    work page 2023

  11. [11]

    En- hancing robustness and generalization in microbiological few-shot detection through synthetic data generation and contrastive learning.Computers in Biology and Medicine,

    Nikolas Ebert, Didier Stricker, and Oliver Wasenm ¨uller. En- hancing robustness and generalization in microbiological few-shot detection through synthetic data generation and contrastive learning.Computers in Biology and Medicine,

    work page

  12. [12]

    3d residual spatial–spectral convolution network for hyperspec- tral remote sensing image classification.Neural Computing and Applications, 35(6):4479–4497, 2023

    H ¨useyin Firat, Ayseg ¨ul Yilmaz, and Ertugrul Bilgili. 3d residual spatial–spectral convolution network for hyperspec- tral remote sensing image classification.Neural Computing and Applications, 35(6):4479–4497, 2023. 2

    work page 2023

  13. [13]

    Hy- perspectral benchmark: Bridging the gap between hsi ap- plications through comprehensive dataset and pretraining

    Hannah Frank, Leon Amadeus Varga, and Andreas Zell. Hy- perspectral benchmark: Bridging the gap between hsi ap- plications through comprehensive dataset and pretraining. arXiv preprint arXiv:2309.11122, 2023. 1, 2, 5, 6, 7

    work page arXiv 2023

  14. [14]

    Agricultural plant hyperspectral imaging dataset.Computer Optics, 47 (3):442–450, 2023

    Andrey Viktorovich Gaidel, Vladimir Vladimirovich Podlip- nov, Nikolay Aleksandrovich Ivliev, Rustam Alexandrovich Paringer, Pavel Alexandrovich Ishkin, Sergey Vladimirovich Mashkov, and Roman Viktorovich Skidanov. Agricultural plant hyperspectral imaging dataset.Computer Optics, 47 (3):442–450, 2023. 2

    work page 2023

  15. [15]

    An exten- sive review of hyperspectral image classification and predic- tion: techniques and challenges.Multimedia Tools and Ap- plications, 83(34):80941–81038, 2024

    Tejasree Ganji and Loganathan Agilandeeswari. An exten- sive review of hyperspectral image classification and predic- tion: techniques and challenges.Multimedia Tools and Ap- plications, 83(34):80941–81038, 2024. 2

    work page 2024

  16. [16]

    Pedram Ghamisi, Emmanuel Maggiori, Shutao Li, and et al. New frontiers in spectral-spatial hyperspectral image clas- sification: The latest advances based on mathematical mor- phology, markov random fields, segmentation, sparse repre- sentation, and deep learning.IEEE Geoscience and Remote Sensing Magazine, 6(3):10–43, 2018. 6

    work page 2018

  17. [17]

    Gra ˜na, M

    M. Gra ˜na, M. A. Veganzons, and B. Ayerdi. Hyperspec- tral remote sensing scenes. Grupo de Inteligencia Computa- cional (GIC), University of the Basque Country (UPV/EHU),

    work page

  18. [18]

    Accessed: 2026-02-20

    Last edited: 12 Jul 2021. Accessed: 2026-02-20. 2

    work page 2021

  19. [19]

    Hyperspectral image classification with attention-aided cnns.IEEE Transactions on Geoscience and Remote Sensing, 59(3):2281–2293, 2020

    Renlong Hang et al. Hyperspectral image classification with attention-aided cnns.IEEE Transactions on Geoscience and Remote Sensing, 59(3):2281–2293, 2020. 2

    work page 2020

  20. [20]

    Ji He et al. Hsi-bert: Hyperspectral image classification us- ing the bidirectional encoder representation from transform- ers.IEEE Transactions on Geoscience and Remote Sensing, 58(1):165–178, 2019. 2

    work page 2019

  21. [21]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InProceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016. 4, 6

    work page 2016

  22. [22]

    Multi-scale 3d deep convolutional neural network for hyperspectral image classi- fication

    Mingyi He, Bo Li, and Huahui Chen. Multi-scale 3d deep convolutional neural network for hyperspectral image classi- fication. In2017 IEEE International Conference on Image Processing (ICIP), pages 3904–3908. IEEE, 2017. 2

    work page 2017

  23. [23]

    Spectralformer: Rethinking hyper- spectral image classification with transformers.IEEE Trans- actions on Geoscience and Remote Sensing, 60:1–15, 2021

    Danfeng Hong and et al. Spectralformer: Rethinking hyper- spectral image classification with transformers.IEEE Trans- actions on Geoscience and Remote Sensing, 60:1–15, 2021. 2, 3, 6

    work page 2021

  24. [24]

    A survey: Deep learning for hyperspectral im- age classification with few labeled samples.Neurocomput- ing, 448:179–204, 2021

    Sen Jia et al. A survey: Deep learning for hyperspectral im- age classification with few labeled samples.Neurocomput- ing, 448:179–204, 2021. 3

    work page 2021

  25. [25]

    A hyperspectral data 3d convolutional neural network classification model for diagnosis of gray mold disease in strawberry leaves.Frontiers in Plant Sci- ence, 13:837020, 2022

    Dae-Hyun Jung, Seul-Ki Lee, Youngjo Kim, Won Seok Lee, and Minyoung Kim. A hyperspectral data 3d convolutional neural network classification model for diagnosis of gray mold disease in strawberry leaves.Frontiers in Plant Sci- ence, 13:837020, 2022. 2

    work page 2022

  26. [26]

    Khonina et al

    Svetlana N. Khonina et al. Synergy between artificial intel- ligence and hyperspectral imaging: A review.Technologies, 12(9):163, 2024. 3

    work page 2024

  27. [27]

    Deep learning for hyperspectral image classi- fication: A survey.Computer Science Review, 53:100658,

    Vinod Kumar, Ravi Shankar Singh, Medara Rambabu, and Yaman Dua. Deep learning for hyperspectral image classi- fication: A survey.Computer Science Review, 53:100658,

    work page

  28. [28]

    Hierarchical multi-scale convolutional neural networks for hyperspectral image clas- sification.Sensors, 19(7):1714, 2019

    Simin Li, Xueyu Zhu, and Jie Bao. Hierarchical multi-scale convolutional neural networks for hyperspectral image clas- sification.Sensors, 19(7):1714, 2019. 3

    work page 2019

  29. [29]

    Data aug- mentation for hyperspectral image classification with deep CNN.IEEE Geoscience and Remote Sensing Letters, 16(4): 593–597, 2018

    Wei Li, Guodong Wu, Fulin Zhang, and Qian Du. Data aug- mentation for hyperspectral image classification with deep CNN.IEEE Geoscience and Remote Sensing Letters, 16(4): 593–597, 2018. 3

    work page 2018

  30. [30]

    Spectral–spatial classification of hyperspectral imagery with 3d convolutional neural network.Remote Sensing, 9(1):67, 2017

    Ying Li, Haokui Zhang, and Qiang Shen. Spectral–spatial classification of hyperspectral imagery with 3d convolutional neural network.Remote Sensing, 9(1):67, 2017. 2

    work page 2017

  31. [31]

    A spectral– spatial fusion transformer network for hyperspectral image classification.IEEE Transactions on Geoscience and Remote Sensing, 61:1–16, 2023

    Diling Liao, Cuiping Shi, and Liguo Wang. A spectral– spatial fusion transformer network for hyperspectral image classification.IEEE Transactions on Geoscience and Remote Sensing, 61:1–16, 2023. 3

    work page 2023

  32. [32]

    Deeply-supervised pseudo learning with small class- imbalanced samples for hyperspectral image classification

    Weiran Luo, Zhaoxiang Huang, Wei Li, Yanning Zhang, et al. Deeply-supervised pseudo learning with small class- imbalanced samples for hyperspectral image classification. International Journal of Applied Earth Observation and Geoinformation, 112:102949, 2022. 3

    work page 2022

  33. [33]

    Hyperspectral data augmentation.arXiv preprint arXiv:1903.05580, 2019

    Jakub Nalepa, Michal Myller, and Michal Kawu- lok. Hyperspectral data augmentation.arXiv preprint arXiv:1903.05580, 2019. 3

    work page arXiv 1903

  34. [34]

    Genformer–generated im- ages are all you need to improve robustness of transformers on small datasets

    Sven Oehri, Nikolas Ebert, Ahmed Abdullah, Didier Stricker, and Oliver Wasenm¨uller. Genformer–generated im- ages are all you need to improve robustness of transformers on small datasets. InInternational Conference on Pattern Recognition (ICPR). Springer, 2024. 3

    work page 2024

  35. [35]

    Samuel Ortega, Laura Quintana-Quintana, Raquel Leon, Hi- mar Fabelo, Mar ´ıa de la Luz Plaza, Rafael Camacho, and Gustavo M. Callico. Histological hyperspectral glioblastoma dataset (histologyhsi-gb).Scientific Data, 11(1):681, 2024. 2

    work page 2024

  36. [36]

    Haut, Javier Plaza, and Antonio Plaza

    Mercedes Eugenia Paoletti, Juan M. Haut, Javier Plaza, and Antonio Plaza. Deep learning classifiers for hyperspectral imaging: A review.ISPRS Journal of Photogrammetry and Remote Sensing, 158:279–317, 2019. 1, 2, 4, 6

    work page 2019

  37. [37]

    Paoletti, Juan M

    Mercedes E. Paoletti, Juan M. Haut, Ricardo Fernandez- Beltran, Javier Plaza, and Antonio Plaza. Rotation equiv- ariant convolutional neural networks for hyperspectral image classification.IEEE Access, 8:179575–179591, 2020. 2

    work page 2020

  38. [38]

    One-dimensional convolutional neural networks for hyperspectral analysis of nitrogen in plant leaves.Applied Sciences, 11(24):11853, 2021

    Razieh Pourdarbani et al. One-dimensional convolutional neural networks for hyperspectral analysis of nitrogen in plant leaves.Applied Sciences, 11(24):11853, 2021. 2

    work page 2021

  39. [39]

    Hyperspectral band se- lection using attention-based convolutional neural networks

    Pablo Ribalta Lorenzo and et al. Hyperspectral band se- lection using attention-based convolutional neural networks. IEEE Access, 8:42384–42403, 2020. 6

    work page 2020

  40. [40]

    Hybridsn: Exploring 3-d–2-d cnn feature hierarchy for hyperspectral image classification

    Swalpa Kumar Roy and et al. Hybridsn: Exploring 3-d–2-d cnn feature hierarchy for hyperspectral image classification. IEEE Geoscience and Remote Sensing Letters, 17(2):277– 281, 2019. 6

    work page 2019

  41. [41]

    Rumelhart, Geoffrey E

    David E. Rumelhart, Geoffrey E. Hinton, and Ronald J. Williams. Learning representations by back-propagating er- rors.Nature, 323(6088):533–536, 1986. 6

    work page 1986

  42. [42]

    Deep learning meets hyperspectral image anal- ysis: A multidisciplinary review.Journal of Imaging, 7(6): 109, 2021

    Alberto Signoroni, Marco Savardi, Alice Baronio, and Ser- gio Benini. Deep learning meets hyperspectral image anal- ysis: A multidisciplinary review.Journal of Imaging, 7(6): 109, 2021. 3

    work page 2021

  43. [43]

    Massformer: Memory-augmented spectral- spatial transformer for hyperspectral image classification

    Le Sun et al. Massformer: Memory-augmented spectral- spatial transformer for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 62: 1–15, 2024. 3

    work page 2024

  44. [44]

    Visualizing data using t-sne.Journal of Machine Learning Research, 9 (86):2579–2605, 2008

    Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t-sne.Journal of Machine Learning Research, 9 (86):2579–2605, 2008. 7

    work page 2008

  45. [45]

    L. A. Varga, J. Makowski, and A. Zell. Measuring the ripeness of fruit with hyperspectral imaging and deep learn- ing. In2021 International Joint Conference on Neural Net- works (IJCNN), pages 1–8. IEEE, 2021. 1, 6

    work page 2021

  46. [46]

    Self-supervised pretraining for hyperspectral classification of fruit ripeness

    Leon Amadeus Varga, Hannah Frank, and Andreas Zell. Self-supervised pretraining for hyperspectral classification of fruit ripeness. In6th International Conference on Opti- cal Characterization of Materials (OCM), 2023. Conference paper. 6

    work page 2023

  47. [47]

    L. A. Varga, M. Messmer, N. Benbarka, and et al. Wavelength-aware 2d convolutions for hyperspectral imag- ing. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 3777–3786. IEEE, 2023. 1, 6

    work page 2023

  48. [48]

    Xiaofei Yang, Yongliang Ye, Xin Li, Rynson W. H. Lau, and Xiaofei Zhang. Hyperspectral image classification with deep learning models.IEEE Transactions on Geoscience and Re- mote Sensing, 56(9):5408–5423, 2018. 2

    work page 2018

  49. [49]

    X. Yang, W. Cao, Y . Lu, and et al. Hyperspectral image transformer classification networks.IEEE Transactions on Geoscience and Remote Sensing, 60:1–15, 2022. 2, 6

    work page 2022

  50. [50]

    Spectral data classification by one- dimensional convolutional neural networks

    Fanguo Zeng et al. Spectral data classification by one- dimensional convolutional neural networks. In2021 IEEE International Performance, Computing, and Communica- tions Conference (IPCCC). IEEE, 2021. 2

    work page 2021

  51. [51]

    Hyperspectral image classification us- ing mixed convolutions and covariance pooling.IEEE Trans- actions on Geoscience and Remote Sensing, 59(1):522–534,

    Jianwei Zheng et al. Hyperspectral image classification us- ing mixed convolutions and covariance pooling.IEEE Trans- actions on Geoscience and Remote Sensing, 59(1):522–534,

    work page

  52. [52]

    Hyperspectral image classification with imbalanced data based on a new data augmentation method.Applied Sciences, 12(8):3943, 2022

    Xiaoyong Zheng et al. Hyperspectral image classification with imbalanced data based on a new data augmentation method.Applied Sciences, 12(8):3943, 2022. 3

    work page 2022

  53. [53]

    An enhanced spectral fusion 3d cnn model for hyperspectral image classification.Remote Sensing, 14 (21):5334, 2022

    Junbo Zhou, Shan Zeng, Zuyin Xiao, Jinbo Zhou, Hao Li, and Zhen Kang. An enhanced spectral fusion 3d cnn model for hyperspectral image classification.Remote Sensing, 14 (21):5334, 2022. 2

    work page 2022