Pith. sign in

REVIEW 3 major objections 5 minor 52 references

SpecBPP: A Self-Supervised Learning Approach for Hyperspectral Representation and Soil Organic Carbon Estimation

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read SpecBPP claims that shuffling and unshuffling spectral bands is a powerful self-supervised pretraining task: on EnMAP data, fine-tuned SpecBPP reaches R²=0.9456 for soil organic carbon, beating MAE, I-JEPA, SimCLR, and classical regressors.

desk verdict Spectral jigsaw for HSI is a sensible new application, but the headline R² depends on an unstated spatial-disjointness assumption and a test-set-picked N, so the transfer claim isn't testable yet. read the letter →

arxiv 2507.19781 v1 pith:TXPBNOLK submitted 2025-07-26 eess.IV cs.CVcs.LG

classification eess.IVcs.CVcs.LG
keywords self-supervisedlearninghyperspectralimageryspectralbandpermutationsoilorganiccarbonestimationEnMAPcurriculumrepresentationremotesensing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes SpecBPP, a self-supervised pretraining task in which a model must restore the correct order of shuffled contiguous segments of a hyperspectral spectrum. Trained without labels on EnMAP satellite imagery, then fine-tuned on 1,540 soil samples with laboratory-measured SOC, the model reaches $R^2$ of 0.9456, RMSE of 1.1053%, and RPD of 4.19, exceeding masked autoencoders, JEPA, SimCLR, and traditional regressors. The authors argue that spectral order prediction forces the network to learn global spectral structure, including characteristic absorption features, rather than only local band-to-band smoothness. A curriculum that raises the number of segments from 3 to 8 and biases sampling toward easier permutations makes the task trainable, since direct training with 8 segments nearly fails.

What carries the argument

The central mechanism is spectral permutation prediction: a spectrum with $B$ bands (224 for EnMAP) is cut into $N$ contiguous segments, randomly permuted, and the model must predict the inverse permutation. Because $N!$ grows quickly, the paper uses a factorized prediction head that treats the task as $N$ independent classification problems, and a two-part curriculum that advances from 3 to 8 segments and, within each phase, samples permutations using a temperature schedule over a distance-from-identity measure $\varphi(\pi) = \sum_{i=1}^N |i - \pi(i)|$. This forces the encoder to represent global spectral context rather than local smoothness alone.

What would settle it

Re-run fine-tuning with the 1,540 labeled soil-sample patches excluded from the 196,875 pretraining patches (ideally holding out entire EnMAP scenes) and compare $R^2$; if the 0.9456 collapses, the gain is leakage rather than transferable pretraining.

Watch

Extended reading notes

Core claim

SpecBPP's central claim is that the natural ordering of electromagnetic bands is a usable self-supervised signal: shuffle the segments of a spectrum and train an encoder-plus-head to output the inverse permutation. On the SOC estimation benchmark, pretraining with this task, followed by supervised fine-tuning on limited labels, yields $R^2 = 0.9456$, RMSE $= 1.1053\%$, and RPD $= 4.19$, which the paper reports as state of the art and as outperforming MAE, I-JEPA, and SimCLR pretraining as well as PLSR, RF, SVR, and a from-scratch supervised network. Performance rises with the number of segments up to $N = 7$ and decreases at $N = 8$, suggesting an optimum in task difficulty, and ablations show the curriculum is the largest contributor to the final score.

Load-bearing premise

The paper does not state that the 196,875 unlabeled pretraining patches and the 1,540 labeled soil-sample patches are spatially or spectrally disjoint, so the reported transfer gains assume that no pretraining patch overlaps the fine-tuning or test data.

Editorial extensions

If this is right

  • Unlabeled hyperspectral archives become usable pretraining data for soil mapping, lowering the number of laboratory-measured soil samples required for accurate SOC prediction.
  • Curriculum scheduling is a practical prerequisite for permutation-based pretext tasks: direct training at eight segments nearly fails, while the staged curriculum reaches 84.2% accuracy.
  • There is an optimal segment count (here $N=7$), beyond which downstream performance decreases, so task difficulty should be tuned rather than maximized.
  • Ordering-based self-supervision transfers the jigsaw-puzzle idea from spatial image patches to the spectral dimension, supporting the general view that natural data order is a usable label-free signal.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A decisive check the paper leaves open: holding out entire EnMAP scenes from pretraining would test whether the $R^2$ gain survives spatial and spectral separation; the current split description does not rule out leakage.
  • The same permutation task could be evaluated on other contiguous-spectrum sensors and other soil properties; the paper itself notes cross-sensor generalization is untested, so this is an open extension.
  • Because the model reconstructs segment order, its prediction errors may reveal which absorption features are most diagnostic for SOC, offering an interpretability probe the paper does not develop.
  • The factorized head plus curriculum could be applied to other naturally ordered sequences, such as genomic or time-series data, where reconstruction-based pretraining dominates.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes SpecBPP, a self-supervised pretext task for hyperspectral imagery in which a spectrum is partitioned into N contiguous segments, the segments are shuffled, and an encoder is trained to predict the inverse permutation through a factorized per-position classification head. A curriculum progressively increases both the number of segments and the permutation difficulty. The method is evaluated on soil organic carbon (SOC) estimation from EnMAP data: an encoder pretrained on 196,875 unlabeled 64x64 patches is fine-tuned on 1,540 labeled soil samples, and the paper reports an R2 of 0.9456, RMSE of 1.1053%, and RPD of 4.19 at N=7, outperforming PLSR, RF, SVR, a supervised deep model, MAE, I-JEPA, and SimCLR. Ablations indicate that curriculum learning contributes the largest performance gain. The authors conclude that spectral order prediction is a powerful pretext task for hyperspectral representation learning.

Significance. If the empirical result holds, SpecBPP would be a conceptually simple and potentially useful SSL signal for hyperspectral data, and the curriculum strategy is a sensible response to the factorial growth of the permutation space. The evaluation is not circular: the pretext labels are derived from the input spectra themselves, while the SOC target comes from external laboratory measurements. The paper includes comparisons against several baselines, an ablation study, and a candid limitations section, which are strengths. However, the central transfer claim is currently not testable because the paper does not establish that the pretraining and fine-tuning data are spatially disjoint, and the headline N=7 is selected from a sweep without a stated model-selection rule or error bars. The conceptual contribution is also weakened by the factorized head not enforcing a true permutation. With the data-hygiene and statistical issues resolved, the paper would be a worthwhile contribution to the HSI SSL literature.

major comments (3)
  1. [§4.1 (Experimental Setup), Table 2] The pretraining set (196,875 non-overlapping 64x64 patches from 1,000 EnMAP scenes) and the fine-tuning set (patches for 1,540 soil samples) are described without any statement that the two sets are spatially disjoint, and the 70/15/15 stratified split is not geographically blocked. If any fine-tuning sample lies inside, or adjacent to, a pretraining scene, the encoder can memorize scene-specific spectral statistics before the SOC regression head is trained, so the reported R2=0.9456 would reflect leakage rather than transferable spectral-order learning. Please report scene IDs or coordinates for both sets, ensure a scene-disjoint split (including a buffer around sample locations), and re-evaluate Table 2 under that split; this is the minimal condition for the central transfer claim to be testable.
  2. [§4.2, Table 2] The headline N=7 result is the best of the six values N=3,...,8 reported in Table 2, but no model-selection rule is stated. If N was chosen after inspecting test-set performance, the reported superiority over baselines is optimistically biased. Please specify an a priori validation rule (e.g., choose N on the validation split and report test metrics only for that N) or report all values with confidence intervals and a multiple-comparison correction. In addition, all numbers in Tables 2 and 3 appear to come from single runs; without repeated fine-tuning trials or significance tests, the abstract's claim of 'significantly surpassing' the baselines is not supported.
  3. [§3.2.2, Eqs. (9)-(11)] The permutation prediction head factorizes the task into N independent row-wise classification problems and does not enforce that the predicted inverse mapping is a bijection; two positions can be assigned to the same original segment. The pretext objective is therefore not exactly 'recover the correct order of shuffled segments' but a relaxed per-position classification surrogate. Please report the fraction of predicted outputs that are valid permutations, justify the surrogate's adequacy for the downstream transfer claim, or add an assignment constraint (e.g., Sinkhorn) to enforce permutations. This also affects the interpretation of the 100% and 84.2% accuracy numbers in Table 1.
minor comments (5)
  1. [Eq. (9)] The notation Wp ∈ R^{N×N×d} multiplied by z ∈ R^d is not defined as written; specify the contraction or reshape (e.g., a linear layer from d to N^2 followed by a reshape) so that the architecture is reproducible.
  2. [Eq. (12)] The thresholds α_i are described only as 99%, but the equation suggests a sequence of thresholds; define each α_i and the validation metric used to trigger progression from one curriculum phase to the next.
  3. [Table 1] Clarify how 'accuracy' is computed: per-position placement accuracy, full-sequence exact match, or the fraction of valid permutations; this is especially important for interpreting the N=8 row.
  4. [§5.1] The limitations section is candid about factorial growth and cross-sensor transfer, but it does not mention the data-disjointness or model-selection issues raised above; please add these if they cannot be fully resolved.
  5. [General] The text refers to 'Supplementary Material' for t-SNE visualizations, error analysis, and scatter plots, but no supplement is included with the manuscript; please ensure it is uploaded with the revision.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SpecBPP's pretraining objective is self-supervised, but the SOC evaluation uses external laboratory labels, so reported results are not forced by construction.

full rationale

The paper's central evaluation is not circular. SpecBPP pretrains an encoder to predict the inverse permutation of shuffled spectral segments (Eq. 1, 11); the permutation labels are generated from the input spectrum itself, which is the normal structure of a self-supervised pretext task. The downstream claim is transfer to SOC estimation, where the target variable is laboratory-measured SOC, an external quantity not derivable from the pretext labels by construction. Table 2 reports R2, RMSE, MAE, and RPD after fine-tuning on those external labels, so the performance numbers are not equivalent to the pretraining loss. The only in-house citation is HyperKON [3], which appears as a related-work example of contrastive SSL in a list with external methods [2, 10] and is not used to justify any load-bearing premise, uniqueness claim, or design choice. The manuscript also openly lists limitations (Section 5.1), including fixed segmentation and EnMAP-only validation. The skeptical concern about potential train/test leakage between the 196,875 pretraining patches and the 1,540 soil-sample patches is a data-split soundness question, not an instance of circular reasoning under the defined criteria: no equation, parameter, or fitted value in the paper reduces to the evaluation target by construction. Therefore the circularity score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

No physical entities are invented. The central claim rests on method hyperparameters chosen by hand (N, curriculum thresholds, temperature schedule, architecture sizes) and on the unstated assumption that pretraining and fine-tuning patches are disjoint. The model itself is a standard neural encoder plus a permutation head.

free parameters (6)
  • Segment count N selected for downstream SOC = 7 (sweep 3 to 8)
    The headline R2 is obtained at N=7; Table 2 shows performance peaks there and drops at N=8. Model selection is done on the reported sweep, not a clearly held-out validation rule.
  • Validation accuracy thresholds alpha_i for curriculum progression = 0.99 each
    Eq. 12 defines progression at 99% validation accuracy; the values are hand-chosen and are only ablated as a block in 'w/o CL'.
  • Curriculum temperature schedule T_s(t) = not specified
    Eq. 13 uses a temperature schedule that increases with training progress, but the concrete functional form and schedule values are not given.
  • Encoder embedding dimension d = not reported
    The latent dimension is a design choice in Section 3.2.1 but is not reported, leaving the architecture underspecified.
  • Multi-scale convolution kernel sizes = 3, 5, 7
    These scales are chosen by hand in Eq. 4 with no sensitivity analysis.
  • Optimizer, learning rate, and epoch counts = SGD, 1e-3/5e-4, 200/150
    Standard training choices from Section 4.1; they are not ablated but the downstream result depends on them.
assumptions (4)
  • domain assumption EnMAP spectra contain enough ordering information that the permutation task is solvable and that solving it transfers to SOC regression.
    Section 3.4 and Table 1 depend on learnability of permutations; Section 4.2 interprets downstream gains as transfer. Plausible but not demonstrated beyond one dataset.
  • domain assumption Pretraining patches and fine-tuning patches are disjoint and from the same spectral distribution.
    Section 4.1 describes both patch sets without stating disjointness; the transfer comparison is only valid under this assumption.
  • ad hoc to paper A factorized per-position softmax over N positions is a sufficient surrogate for predicting a global permutation.
    Eqs. 9-10 treat each output position as an independent classification and never enforce that rows form a valid permutation matrix; a network could output inconsistent assignments.
  • domain assumption Spectral continuity is a stable property across EnMAP scenes and wavelengths.
    The pretext task is motivated by continuity and ordering of bands [20,40], and the model is expected to learn it from data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SpecBPP: A Self-Supervised Learning Approach for Hyperspectral Representation and Soil Organic Carbon Estimation." pith.science (2026). https://pith.science/paper/TXPBNOLK

@misc{pith2026250719781,
  author       = {Pith},
  title        = {Pith review of: SpecBPP: A Self-Supervised Learning Approach for Hyperspectral Representation and Soil Organic Carbon Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TXPBNOLK}},
  note         = {Machine review of arXiv:2507.19781}
}
abstract

Self-supervised learning has revolutionized representation learning in vision and language, but remains underexplored for hyperspectral imagery (HSI), where the sequential structure of spectral bands offers unique opportunities. In this work, we propose Spectral Band Permutation Prediction (SpecBPP), a novel self-supervised learning framework that leverages the inherent spectral continuity in HSI. Instead of reconstructing masked bands, SpecBPP challenges a model to recover the correct order of shuffled spectral segments, encouraging global spectral understanding. We implement a curriculum-based training strategy that progressively increases permutation difficulty to manage the factorial complexity of the permutation space. Applied to Soil Organic Carbon (SOC) estimation using EnMAP satellite data, our method achieves state-of-the-art results, outperforming both masked autoencoder (MAE) and joint-embedding predictive (JEPA) baselines. Fine-tuned on limited labeled samples, our model yields an $R^2$ of 0.9456, RMSE of 1.1053%, and RPD of 4.19, significantly surpassing traditional and self-supervised benchmarks. Our results demonstrate that spectral order prediction is a powerful pretext task for hyperspectral understanding, opening new avenues for scientific representation learning in remote sensing and beyond.

Figures

Figures reproduced from arXiv: 2507.19781 by the authors.

Figure 1
Figure 1. Overview of the Spectral Band Permutation Prediction (SpecBPP) architecture. The [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Example of spectral band permutation prediction. Each panel shows: (top) the [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 39 canonical work pages

  1. [1]

    Ranking via sinkhorn propagation

    Ryan Prescott Adams and Richard S Zemel. Ranking via sinkhorn propagation. arXiv preprint arXiv:1106.1925, 2011

  2. [2]

    Self-supervised learning from images with a joint-embedding predictive architecture

    Mahmoud Assran, Quentin Duval, Ishan Misra, Piotr Bojanowski, Pascal Vincent, Michael Rabbat, Yann LeCun, and Nicolas Ballas. Self-supervised learning from images with a joint-embedding predictive architecture. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15619–15629, 2023

  3. [3]

    Hyperkon: A self-supervised contrastive network for hyperspectral image analysis

    Daniel La’ah Ayuba, Jean-Yves Guillemaut, Belen Marti-Cardona, and Oscar Mendez. Hyperkon: A self-supervised contrastive network for hyperspectral image analysis. Remote Sensing, 16(18), 2024. ISSN 2072-4292. URL https://www.mdpi.com/ 2072-4292/16/18/3399

  4. [4]

    Curriculum learning

    Yoshua Bengio, Jérôme Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. In Proceedings of the 26th Annual International Conference on Machine Learning, ICML ’09, pages 41–48, 2009. doi: 10.1145/1553374.1553380

  5. [5]

    Using enmap satellite and machine learning for soil organic carbon assessment: A case study in semi-arid region

    Yassine Bouslihim, Abderrazak Rochdi, Nicolas Baghdadi, Carole Sinfort, Amine El- Alem, Anass Chakir, and Said Lahssini. Using enmap satellite and machine learning for soil organic carbon assessment: A case study in semi-arid region. Remote Sensing, 16 (3):387, 2024

  6. [6]

    Transformer- based masked autoencoder with contrastive loss for hyperspectral image classification

    Xianghai Cao, Haifeng Lin, Shuaixu Guo, Tao Xiong, and Licheng Jiao. Transformer- based masked autoencoder with contrastive loss for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 61:1–12, 2023

  7. [7]

    Mask-enhanced contrastive learning for hyperspectral image classification

    Xianghai Cao, Jiayu Yu, Ruijie Xu, Jiaxuan Wei, and Licheng Jiao. Mask-enhanced contrastive learning for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 2024

  8. [8]

    The enmap spaceborne imaging spectroscopy mission: Initial scientific results two years after launch

    Sabine Chabrillat, Saskia Foerster, Karl Segl, Alison Beamish, Maximilian Brell, Saeid Asadzadeh, Robert Milewski, Kathrin J Ward, Arlena Brosinsky, Katrin Koch, et al. The enmap spaceborne imaging spectroscopy mission: Initial scientific results two years after launch. Remote Sensing of Environment, 315:114379, 2024

Show all 52 references
  1. [9]

    Laird, Maurice J

    Cheng-Wen Chang, David A. Laird, Maurice J. Mausbach, and Charles R. Hurburgh. Near-infrared reflectance spectroscopy–principal components regression analyses of soil properties. Soil Science Society of America Journal , 65(2):480–490, 2001. doi: 10.2136/sssaj2001.652480x

  2. [10]

    A simple framework for contrastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. A simple framework for contrastive learning of visual representations. In Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, pages 1597...

  3. [11]

    Exploring simple siamese representation learning

    Xinlei Chen and Kaiming He. Exploring simple siamese representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 15750–15758, 2021. doi: 10.1109/CVPR46437.2021.01549. 12 DANIEL ET AL: SPECBPP

  4. [12]

    Xception: Deep learning with depthwise separable convolutions

    François Chollet. Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 1251–1258, 2017. doi: 10.1109/CVPR.2017.195

  5. [13]

    Soil moisture, organic carbon, and nitrogen content prediction with hyperspectral data using regression models

    Dristi Datta, Manoranjan Paul, Manzur Murshed, Shyh Wei Teng, and Leigh Schmidtke. Soil moisture, organic carbon, and nitrogen content prediction with hyperspectral data using regression models. Sensors, 22(20):7998, 2022

  6. [14]

    Bayesian optimization over permutation spaces

    Aryan Deshwal, Syrine Belakaria, Janardhan Rao Doppa, and Dae Hyun Kim. Bayesian optimization over permutation spaces. In Proceedings of the AAAI conference on artificial intelligence, pages 6515–6523, 2022

  7. [15]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language...

  8. [16]

    Carl Doersch, Abhinav Gupta, and Alexei A. Efros. Unsupervised visual representation learning by context prediction. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), pages 1422–1430, 2015. doi: 10.1109/ICCV .2015.167

  9. [17]

    Self-supervised video representation learning with odd-one-out networks

    Basura Fernando, Hakan Bilen, Efstratios Gavves, and Stephen Gould. Self-supervised video representation learning with odd-one-out networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3636–3645, 2017

  10. [18]

    Dual attention network for scene segmentation

    Jun Fu, Jing Liu, Haijie Tian, Yong Li, Yongjun Bao, Zhiwei Fang, and Hanqing Lu. Dual attention network for scene segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3146–3154, 2019

  11. [19]

    Gholizadeh, M

    A. Gholizadeh, M. Saberioon, L. Bo ˇr˚ uvka, A. Wayayok, and M.A.M. Soom. Leaf chlorophyll and nitrogen dynamics and their relationship to lowland rice yield for site- specific paddy management. Information Processing in Agriculture , 4(4):259–268, 2017

  12. [20]

    Alexander F. H. Goetz. Three decades of hyperspectral remote sensing of the earth: A personal view. Remote Sensing of Environment, 113:S5–S16, 2009. doi: 10.1016/j.rse. 2007.12.014

  13. [21]

    Richemond, Elena Buchatskaya, Carl Doersch, Bernardo Ávila Pires, Zhaohan Daniel Guo, Moham- mad Gheshlaghi Azar, Bilal Piot, Koray Kavukcuoglu, Rémi Munos, and Michal Valko

    Jean-Bastien Grill, Florian Strub, Florent Altché, Corentin Tallec, Pierre H. Richemond, Elena Buchatskaya, Carl Doersch, Bernardo Ávila Pires, Zhaohan Daniel Guo, Moham- mad Gheshlaghi Azar, Bilal Piot, Koray Kavukcuoglu, Rémi Munos, and Michal Valko. Bootstrap your own laten...

  14. [22]

    Cross-domain contrastive learning for hyperspectral image classification

    Peiyan Guan and Edmund Y Lam. Cross-domain contrastive learning for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 60:1–13, 2022. DANIEL ET AL: SPECBPP 13

  15. [23]

    On the power of curriculum learning in training deep networks

    Guy Hacohen and Daphna Weinshall. On the power of curriculum learning in training deep networks. In Proceedings of the 36th International Conference on Machine Learning, Proceedings of Machine Learning Research, pages 2535–2544, 2019

  16. [24]

    Momentum contrast for unsupervised visual representation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 9729–9738,

  17. [25]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked autoencoders are scalable vision learners. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 16000–16009,

  18. [26]

    Spectralformer: Rethinking hyperspectral image classification with transformers

    Danfeng Hong, Zhu Han, Jing Yao, Lianru Gao, Bing Zhang, Antonio Plaza, and Jocelyn Chanussot. Spectralformer: Rethinking hyperspectral image classification with transformers. IEEE Transactions on Geoscience and Remote Sensing, 60:1–15, 2022. doi: 10.1109/TGRS.2021.3130716

  19. [27]

    Hyper- spectral imagery classification based on contrastive learning

    Sikang Hou, Hongye Shi, Xianghai Cao, Xiaohua Zhang, and Licheng Jiao. Hyper- spectral imagery classification based on contrastive learning. IEEE Transactions on Geoscience and Remote Sensing, 60:1–13, 2021

  20. [28]

    Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, and Hartwig Adam

    Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, and Hartwig Adam. Mobilenets: Efficient convolu- tional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861, 2017

  21. [29]

    Contrastive learning based on transformer for hyperspectral image classification

    Xiang Hu, Teng Li, Tong Zhou, Yu Liu, and Yuanxi Peng. Contrastive learning based on transformer for hyperspectral image classification. Applied Sciences, 11(18):8670, 2021

  22. [30]

    Spectral–spatial masked transformer with supervised and contrastive learning for hyperspectral image classification

    Lingbo Huang, Yushi Chen, and Xin He. Spectral–spatial masked transformer with supervised and contrastive learning for hyperspectral image classification. IEEE Trans- actions on Geoscience and Remote Sensing, 61:1–18, 2023

  23. [31]

    Instructional mask autoencoder: A scalable learner for hyperspectral image classification

    Weili Kong, Baisen Liu, Xiaojun Bi, Jiaming Pei, and Zheng Chen. Instructional mask autoencoder: A scalable learner for hyperspectral image classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 17:1348–1362, 2023

  24. [32]

    Soil carbon sequestration impacts on global climate change and food security

    Rattan Lal. Soil carbon sequestration impacts on global climate change and food security. Science, 304(5677):1623–1627, 2004. doi: 10.1126/science.1097396

  25. [33]

    Demae: Diffusion enhanced masked autoencoder for hyperspectral image classification with few labeled samples

    Ziyu Li, Zhaohui Xue, Mingming Jia, Xiangyu Nie, Hao Wu, Mengxue Zhang, and Hongjun Su. Demae: Diffusion enhanced masked autoencoder for hyperspectral image classification with few labeled samples. IEEE Transactions on Geoscience and Remote Sensing, 2024

  26. [34]

    Self-supervised feature learning based on spectral masking for hyperspectral image classification

    Weiwei Liu, Kai Liu, Weiwei Sun, Gang Yang, Kai Ren, Xiangchao Meng, and Jiangtao Peng. Self-supervised feature learning based on spectral masking for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 61:1–15, 2023. 14 DANIEL ET AL: SPECBPP

  27. [35]

    Lawrence Zitnick, and Martial Hebert

    Ishan Misra, C. Lawrence Zitnick, and Martial Hebert. Shuffle and learn: Unsupervised learning using temporal order verification. In Computer Vision – ECCV 2016, pages 527–544, 2016. doi: 10.1007/978-3-319-46448-0_32

  28. [36]

    Unsupervised learning of visual representations by solving jigsaw puzzles

    Mehdi Noroozi and Paolo Favaro. Unsupervised learning of visual representations by solving jigsaw puzzles. In Computer Vision – ECCV 2016, pages 69–84, 2016. doi: 10.1007/978-3-319-46466-4_5

  29. [37]

    Paoletti, Juan M

    Mercedes E. 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. doi: 10.1016/j.isprsjprs.2019.09.006

  30. [38]

    Integrating soil spectral library and prisma data to estimate soil organic carbon in crop lands.IEEE Geoscience and Remote Sensing Letters, 2024

    Bonthu Sandeep Reddy and HR Shwetha. Integrating soil spectral library and prisma data to estimate soil organic carbon in crop lands.IEEE Geoscience and Remote Sensing Letters, 2024

  31. [39]

    Guided curriculum model adap- tation and uncertainty-aware evaluation for semantic nighttime image segmentation

    Christos Sakaridis, Dengxin Dai, and Luc Van Gool. Guided curriculum model adap- tation and uncertainty-aware evaluation for semantic nighttime image segmentation. In Proceedings of the IEEE/CVF international conference on computer vision, pages 7374–7383, 2019

  32. [40]

    Shaw and Heidi-Marie K

    Gary A. Shaw and Heidi-Marie K. Burke. Spectral imaging for remote sensing. Lincoln Laboratory Journal, 14(1):3–28, 2003

  33. [41]

    Cur- riculum learning: A survey

    Petru Soviany, Radu Tudor Ardei, Radu Tudor Ionescu, and Marius Leordeanu. Cur- riculum learning: A survey. International Journal of Computer Vision, 130:1526–1565,

  34. [42]

    Viscarra Rossel, Abdul M

    Bo Stenberg, Raphael A. Viscarra Rossel, Abdul M. Mouazen, and Johanna Wetterlind. Visible and near infrared spectroscopy in soil science. Advances in Agronomy, 107: 163–215, 2010. doi: 10.1016/S0065-2113(10)07005-7

  35. [43]

    Prediction of soil organic carbon at the european scale by visible and near infrared reflectance spectroscopy

    Antoine Stevens, Marco Nocita, Gergely Tóth, Luca Montanarella, and Bas van Wesemael. Prediction of soil organic carbon at the european scale by visible and near infrared reflectance spectroscopy. PLOS ONE , 8(6):e66409, 2013. doi: 10.1371/journal.pone.0066409

  36. [44]

    doi: 10.1007/s11263-022-01611-x

  37. [45]

    Spectral–spatial feature to- kenization transformer for hyperspectral image classification

    Lei Sun, Guang Zhao, Yingqian Zheng, and Zebin Wu. Spectral–spatial feature to- kenization transformer for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 60:1–14, 2022

  38. [46]

    Hsimae: A unified masked autoencoder with large-scale pre-training for hyperspectral image classification

    Yue Wang, Ming Wen, Hailiang Zhang, Jinyu Sun, Qiong Yang, Zhimin Zhang, and Hongmei Lu. Hsimae: A unified masked autoencoder with large-scale pre-training for hyperspectral image classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,...

  39. [47]

    The enmap imaging spectroscopy mission towards operations

    Tobias Storch, Hans-Peter Honold, Sabine Chabrillat, Martin Habermeyer, Paul Tucker, Maximilian Brell, Andreas Ohndorf, Katrin Wirth, Matthias Betz, Michael Kuchler, et al. The enmap imaging spectroscopy mission towards operations. Remote Sensing of Environment, 294:113632, 2023

  40. [48]

    Xlnet: Generalized autoregressive pretraining for language understanding

    Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. Xlnet: Generalized autoregressive pretraining for language understanding. Advances in neural information processing systems, 32, 2019

  41. [49]

    Cross-domain few-shot contrastive learning for hyperspectral images classification

    Suhua Zhang, Zhikui Chen, Dan Wang, and Z Jane Wang. Cross-domain few-shot contrastive learning for hyperspectral images classification. IEEE Geoscience and Remote Sensing Letters, 19:1–5, 2022

  42. [50]

    Cbam: Convolu- tional block attention module

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. Cbam: Convolu- tional block attention module. In Proceedings of the European Conference on Computer Vision (ECCV), pages 3–19, 2018

  43. [2020]

    doi: 10.1109/CVPR42600.2020.00975

  44. [2022]

    doi: 10.1109/CVPR52688.2022.01553

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.