pith. machine review for the scientific record. sign in

arxiv: 2605.05549 · v1 · submitted 2026-05-07 · 💻 cs.CV

Recognition: unknown

A Novel Graph-Regulated Disentangling Mamba Model with Sparse Tokens for Enhanced Tree Species Classification from MODIS Time Series

Motasem Alkayid , Zhengsen Xu , Saeid Taleghanidoozdoozan , Yimin Zhu , Megan Greenwood , Quinn Ledingham , Zack Dewis , Mabel Heffring
show 2 more authors
Naser El-Sheimy Lincoln Linlin Xu
Authors on Pith no claims yet

Pith reviewed 2026-05-09 16:35 UTC · model grok-4.3

classification 💻 cs.CV
keywords tree species classificationMODIS time seriesMamba modelgraph regulationdisentangled featuressparse tokensremote sensingsatellite imagery
0
0 comments X

The pith

Graph-regulated disentangling Mamba with sparse tokens achieves over 93 percent accuracy classifying tree species from MODIS data.

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

The paper develops a Graph-regulated Disentangled Sparse Mamba model, called GDS-Mamba, to classify tree species using time series data from the MODIS satellite sensor. It addresses key difficulties including subtle differences in species signatures, tightly coupled spatial spectral and temporal information, and the need to model broad topological context across large areas. The approach uses a mini-batch graph method to handle correlations between images, a specialized Mamba structure that separates spatial patterns from spectral and temporal features, and adaptive sparse tokens to focus on the most relevant data points and reduce correlation decay. These changes lead to strong performance on extensive annual data from two Canadian provinces, beating twelve other models. Readers interested in remote sensing and environmental monitoring would value this because better tree species maps can aid in tracking forest health and biodiversity at scale.

Core claim

The authors claim that their novel GDS-Mamba model, through its mini-batch graph-regulated approach for exploring topological correlations, disentangling Mamba architecture for capturing independent spatial, spectral, and temporal information, and adaptive sparse token selection to address correlation decay, enables superior feature extraction and classification of tree species from large-scale MODIS time series data, as demonstrated by accuracies of 93.94 percent in Alberta and 80.19 percent in cross-provincial tests outperforming twelve state-of-the-art models.

What carries the argument

The GDS-Mamba architecture, which integrates graph regulation in mini-batches to model topological effects, disentangling Mamba blocks to separate spatial-spectral-temporal couplings, and adaptive sparse tokens to learn optimal subsets for efficient subtle feature learning.

If this is right

  • The model better captures independent phenology behaviors and spatial patterns in time series data.
  • Graph regulation allows explicit handling of large-scale context information among input images.
  • Sparse tokens improve efficiency and mitigate bottlenecks in standard Mamba models for this task.
  • Superior performance holds in both within-province and cross-province evaluations.
  • Outperformance of twelve other models suggests practical advantages for environmental applications.

Where Pith is reading between the lines

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

  • If the disentangling works as described, similar architectures could apply to other coupled multi-dimensional remote sensing tasks like land cover mapping.
  • Cross-provincial results hint at some generalization, but further tests on global or varied datasets would be needed to confirm broad applicability.
  • The approach may inspire hybrid graph-Mamba designs for other sequence modeling problems with topological structure.

Load-bearing premise

The mini-batch graph-regulated approach, disentangling Mamba blocks, and adaptive sparse tokens genuinely resolve spatial-spectral-temporal coupling and correlation decay without introducing selection bias or overfitting to the specific Canadian MODIS dataset.

What would settle it

A study applying the GDS-Mamba to MODIS data from a different region with distinct tree species or environmental conditions, where it does not achieve higher accuracy than the twelve baseline models, would challenge the central claim.

Figures

Figures reproduced from arXiv: 2605.05549 by Lincoln Linlin Xu, Mabel Heffring, Megan Greenwood, Motasem Alkayid, Naser El-Sheimy, Quinn Ledingham, Saeid Taleghanidoozdoozan, Yimin Zhu, Zack Dewis, Zhengsen Xu.

Figure 1
Figure 1. Figure 1: t-SNE class separability of (A) raw data and (B) GDS-Mamba features view at source ↗
Figure 2
Figure 2. Figure 2: The architecture of the proposed model, with three key contributions. (1) First, the large-scale geographical correlation within view at source ↗
Figure 3
Figure 3. Figure 3: Classification maps achieved by different methods for the Alberta dataset. The zoom in views of four regions (i.e., a, b, c, and d) indicate that our view at source ↗
Figure 4
Figure 4. Figure 4: The spectral-temporal curves of different classes in the Alberta dataset view at source ↗
read the original abstract

Although tree species classification from Moderate Resolution Imaging Spectroradiometer (MODIS) time series data is critical for supporting various environmental applications, it is a challenging task due to several key difficulties: the subtle signature differences among tree species, strong spatial-spectral-temporal information coupling, and the difficulty of modeling large-scale topological context information. To better address these challenges, this paper presents a novel Graph-regulated Disentangled Sparse Mamba model (GDS-Mamba) for enhanced tree species classification, with the following contributions. (1) First, to improve large-scale context modeling, we design a mini-batch graph-regulated approach that explicitly explores topological correlation effects among input images. (2) Second, to disentangle the high-dimensional spatial-spectral-temporal information coupling for improved feature extraction, we propose a novel disentangling Mamba architecture tailored for capturing independent spatial patterns, spectral signatures, and temporal phenology behaviors in MODIS time series. (3) Third, to improve efficiency and subtle feature learning, we design novel sparse token approaches that adaptively learn the optimum subset of tokens to better address the correlation decay problem that bottlenecks standard Mamba models. Extensive experiments using large-scale annual MOD13Q1 data across two Canadian provinces (i.e., Alberta and Saskatchewan) achieved an overall accuracy of 93.94\% in Alberta and 80.19\% in cross-provincial evaluations, outperforming twelve state-of-the-art classification models.

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

3 major / 2 minor

Summary. The paper proposes the Graph-regulated Disentangled Sparse Mamba (GDS-Mamba) model for tree species classification from MODIS time series. It introduces a mini-batch graph-regulated approach to model topological correlations, a disentangling Mamba architecture to separate spatial, spectral, and temporal features, and adaptive sparse tokens to mitigate correlation decay. Experiments on large-scale annual MOD13Q1 data from Alberta and Saskatchewan report 93.94% overall accuracy in Alberta and 80.19% in cross-provincial evaluation, outperforming twelve state-of-the-art models.

Significance. If the performance gains and generalizability hold after rigorous validation, the work could advance remote sensing classification by integrating graph regularization, feature disentanglement, and sparsity into Mamba architectures for handling coupled spatio-temporal data in forestry applications. The cross-provincial evaluation provides a direct test of robustness, though the observed drop limits immediate broader impact.

major comments (3)
  1. Abstract and experimental claims: the reported 93.94% Alberta accuracy versus 80.19% cross-provincial accuracy constitutes a 13.75-point drop. This directly tests the central claim that the mini-batch graph-regulated approach, disentangling Mamba blocks, and adaptive sparse tokens resolve spatial-spectral-temporal coupling and correlation decay in a generalizable way; without ablations isolating each component's contribution to the cross-provincial score, the gains may reflect province-specific tuning rather than the proposed mechanisms.
  2. Experimental section (validation protocol and baselines): the manuscript supplies no details on cross-validation folds, baseline implementations, number of independent runs, or statistical error bars. This absence makes it impossible to verify that the outperformance over twelve SOTA models is robust rather than an artifact of a single split or hyperparameter search on the Canadian MODIS dataset.
  3. Ablation studies: no tables or sections report controlled ablations removing the graph regulation, disentangling blocks, or sparse token selection while measuring impact on cross-provincial accuracy. Such analysis is load-bearing for attributing improvements to the architectural innovations rather than the dataset or training procedure.
minor comments (2)
  1. Notation: the abstract introduces 'GDS-Mamba' and 'disentangling Mamba architecture' without an early equation or diagram defining the disentanglement loss or token selection criterion, which would aid readability.
  2. Figure clarity: if present, the architecture diagram should explicitly label the mini-batch graph edges and sparse token masking to match the textual description of the three contributions.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive review. The comments highlight important aspects of experimental rigor and generalizability that we will address in the revision. We provide point-by-point responses below.

read point-by-point responses

  1. Referee: Abstract and experimental claims: the reported 93.94% Alberta accuracy versus 80.19% cross-provincial accuracy constitutes a 13.75-point drop. This directly tests the central claim that the mini-batch graph-regulated approach, disentangling Mamba blocks, and adaptive sparse tokens resolve spatial-spectral-temporal coupling and correlation decay in a generalizable way; without ablations isolating each component's contribution to the cross-provincial score, the gains may reflect province-specific tuning rather than the proposed mechanisms.

    Authors: We acknowledge the 13.75-point drop and its implications for the generalizability claim. In the revised manuscript we will add a new subsection with controlled ablations that isolate the contribution of the graph-regulated module, the disentangling Mamba blocks, and the adaptive sparse tokens specifically on the cross-provincial (Alberta-to-Saskatchewan) accuracy. These results will be presented alongside the original Alberta numbers so readers can directly assess whether the architectural components drive the observed performance rather than province-specific tuning. We will also expand the discussion to interpret the drop in light of domain shift between the two provinces. revision: yes

  2. Referee: Experimental section (validation protocol and baselines): the manuscript supplies no details on cross-validation folds, baseline implementations, number of independent runs, or statistical error bars. This absence makes it impossible to verify that the outperformance over twelve SOTA models is robust rather than an artifact of a single split or hyperparameter search on the Canadian MODIS dataset.

    Authors: We agree that these details are essential for reproducibility and statistical credibility. The revised manuscript will include: (i) the exact cross-validation scheme (number of folds and how provinces were handled), (ii) implementation details and hyperparameter settings for all twelve baseline models, (iii) the number of independent runs performed for each method, and (iv) error bars or standard deviations on all reported accuracy figures. These additions will allow verification that the reported outperformance is robust. revision: yes

  3. Referee: Ablation studies: no tables or sections report controlled ablations removing the graph regulation, disentangling blocks, or sparse token selection while measuring impact on cross-provincial accuracy. Such analysis is load-bearing for attributing improvements to the architectural innovations rather than the dataset or training procedure.

    Authors: We will insert a dedicated ablation study section containing tables that systematically disable each component (mini-batch graph regulation, disentangling Mamba blocks, and adaptive sparse token selection) and report the resulting overall accuracy on both the Alberta test set and the cross-provincial evaluation. This will provide direct quantitative evidence linking the performance gains to the proposed mechanisms across both evaluation settings. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical model proposal with independent evaluation

full rationale

The paper introduces GDS-Mamba via three explicitly motivated architectural components (mini-batch graph regulation, disentangling Mamba blocks, adaptive sparse tokens) to address stated challenges in MODIS time-series classification. These are presented as design choices, not derived quantities. Reported accuracies (93.94% Alberta, 80.19% cross-provincial) are obtained from direct experiments on held-out data and compared against twelve external baselines; no equation or claim reduces these metrics to a parameter fitted inside the same derivation. Cross-provincial evaluation functions as an external generalization test rather than a self-referential loop. Any self-citations to prior Mamba or graph work are standard and non-load-bearing for the empirical results.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim rests on the unverified effectiveness of three newly proposed components whose performance is asserted via empirical results that cannot be inspected from the abstract alone.

free parameters (1)
  • hyperparameters of Mamba blocks, graph layers, and sparse token selection
    Standard deep-learning training parameters whose values are not reported.
axioms (1)
  • domain assumption The disentangling Mamba architecture can independently capture spatial patterns, spectral signatures, and temporal phenology without destructive interference.
    Invoked in the second contribution to justify the architecture design.
invented entities (1)
  • GDS-Mamba model no independent evidence
    purpose: To perform tree species classification from MODIS time series
    Newly proposed architecture combining graph regulation, disentangling Mamba, and sparse tokens.

pith-pipeline@v0.9.0 · 5606 in / 1205 out tokens · 48221 ms · 2026-05-09T16:35:14.368662+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

135 extracted references · 12 canonical work pages · 3 internal anchors

  1. [1]

    , title =

    Author, A. , title =. Journal Name , year =

  2. [2]

    Science of The Total Environment , volume=

    CALIPSO-based aerosol extinction profile estimation from MODIS and MERRA-2 data using a hybrid model of Transformer and CNN , author=. Science of The Total Environment , volume=. 2024 , publisher=

    2024

  3. [3]

    IEEE Transactions on Geoscience and Remote Sensing , year=

    High-Resolution Seamless Mapping of the Leaf Area Index via Multisource Data and the Transformer Deep Learning Model , author=. IEEE Transactions on Geoscience and Remote Sensing , year=

  4. [4]

    Remote sensing of environment , volume=

    Development of the GLASS 250-m leaf area index product (version 6) from MODIS data using the bidirectional LSTM deep learning model , author=. Remote sensing of environment , volume=. 2022 , publisher=

    2022

  5. [5]

    IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , volume=

    Spatiotemporal satellite image fusion using deep convolutional neural networks , author=. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , volume=. 2018 , publisher=

    2018

  6. [6]

    Remote sensing , volume=

    A CNN-LSTM model for soil organic carbon content prediction with long time series of MODIS-based phenological variables , author=. Remote sensing , volume=. 2022 , publisher=

    2022

  7. [7]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    A cascaded spectral--spatial CNN model for super-resolution river mapping with MODIS imagery , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2021 , publisher=

    2021

  8. [8]

    Neurocomputing , volume =

    Muhammad Ahmad and Muhammad Hassaan Farooq Butt and Adil Mehmood Khan and Manuel Mazzara and Salvatore Distefano and Muhammad Usama and Swalpa Kumar Roy and Jocelyn Chanussot and Danfeng Hong , title =. Neurocomputing , volume =. 2025 , doi =

    2025

  9. [9]

    Surveying and Land Information Science , volume =

    Craig Rodarmel and Jie Shan , title =. Surveying and Land Information Science , volume =. 2002 , publisher =

    2002

  10. [10]

    Farrell and R.M

    M.D. Farrell and R.M. Mersereau , title =. IEEE Geoscience and Remote Sensing Letters , year =

  11. [11]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Classification of hyperspectral data from urban areas based on extended morphological profiles , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2005 , publisher=

    2005

  12. [12]

    IEEE Transactions on Geoscience and Remote Sensing , year =

    Jing Wang and Chein-I Chang , title =. IEEE Transactions on Geoscience and Remote Sensing , year =

  13. [13]

    Trends in cognitive sciences , volume=

    Independent component analysis: an introduction , author=. Trends in cognitive sciences , volume=. 2002 , publisher=

    2002

  14. [14]

    IEEE transactions on neural networks and learning systems , volume=

    A survey of convolutional neural networks: analysis, applications, and prospects , author=. IEEE transactions on neural networks and learning systems , volume=. 2021 , publisher=

    2021

  15. [15]

    Advances in neural information processing systems , volume=

    Attention is all you need , author=. Advances in neural information processing systems , volume=

  16. [16]

    Spectral–Spatial Residual Network for Hyperspectral Image Classification: A 3-D Deep Learning Framework , year=

    Zhong, Zilong and Li, Jonathan and Luo, Zhiming and Chapman, Michael , journal=. Spectral–Spatial Residual Network for Hyperspectral Image Classification: A 3-D Deep Learning Framework , year=

  17. [17]

    Spatial–Spectral ConvNeXt for Hyperspectral Image Classification , year=

    Zhu, Yimin and Yuan, Kexin and Zhong, Wenlong and Xu, Linlin , journal=. Spatial–Spectral ConvNeXt for Hyperspectral Image Classification , year=

  18. [18]

    Spectral–Spatial Transformer Network for Hyperspectral Image Classification: A Factorized Architecture Search Framework , year=

    Zhong, Zilong and Li, Ying and Ma, Lingfei and Li, Jonathan and Zheng, Wei-Shi , journal=. Spectral–Spatial Transformer Network for Hyperspectral Image Classification: A Factorized Architecture Search Framework , year=

  19. [19]

    Hyperspectral Image Classification Using Groupwise Separable Convolutional Vision Transformer Network , year=

    Zhao, Zhuoyi and Xu, Xiang and Li, Shutao and Plaza, Antonio , journal=. Hyperspectral Image Classification Using Groupwise Separable Convolutional Vision Transformer Network , year=

  20. [20]

    Spectral–Spatial Feature Tokenization Transformer for Hyperspectral Image Classification , year=

    Sun, Le and Zhao, Guangrui and Zheng, Yuhui and Wu, Zebin , journal=. Spectral–Spatial Feature Tokenization Transformer for Hyperspectral Image Classification , year=

  21. [21]

    Classification of Hyperspectral Images via Multitask Generative Adversarial Networks , year=

    Hang, Renlong and Zhou, Feng and Liu, Qingshan and Ghamisi, Pedram , journal=. Classification of Hyperspectral Images via Multitask Generative Adversarial Networks , year=

  22. [22]

    Generating Long Sequences with Sparse Transformers

    Generating long sequences with sparse transformers , author=. arXiv preprint arXiv:1904.10509 , year=

    work page internal anchor Pith review arXiv 1904

  23. [23]

    Neurocomputing , pages=

    Sparsity in transformers: A systematic literature review , author=. Neurocomputing , pages=. 2024 , publisher=

    2024

  24. [24]

    Transactions of the Association for Computational Linguistics , volume=

    Efficient content-based sparse attention with routing transformers , author=. Transactions of the Association for Computational Linguistics , volume=. 2021 , publisher=

    2021

  25. [25]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Learning a sparse transformer network for effective image deraining , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  26. [26]

    Neurocomputing , volume=

    SparseSwin: Swin transformer with sparse transformer block , author=. Neurocomputing , volume=. 2024 , publisher=

    2024

  27. [27]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Internimage: Exploring large-scale vision foundation models with deformable convolutions , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  28. [28]

    Knowledge-Based Systems , volume=

    DUNet: A deformable network for retinal vessel segmentation , author=. Knowledge-Based Systems , volume=. 2019 , publisher=

    2019

  29. [29]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Vision transformer with deformable attention , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  30. [30]

    Deformable DETR: Deformable Transformers for End-to-End Object Detection

    Deformable detr: Deformable transformers for end-to-end object detection , author=. arXiv preprint arXiv:2010.04159 , year=

    work page internal anchor Pith review arXiv 2010

  31. [31]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Deformable convnets v2: More deformable, better results , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  32. [32]

    Advances in Neural Information Processing Systems , volume=

    Sparse is enough in scaling transformers , author=. Advances in Neural Information Processing Systems , volume=

  33. [33]

    Fabio Giampaolo, Stefano Izzo, Edoardo Prezioso, and Francesco Piccialli

    Adaptively sparse transformers , author=. arXiv preprint arXiv:1909.00015 , year=

    work page arXiv 1909

  34. [34]

    International Conference on Machine Learning , pages=

    Exphormer: Sparse transformers for graphs , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  35. [35]

    Sensors , volume =

    Giannopoulos, Michalis and Aidini, Anastasia and Pentari, Anastasia and Fotiadou, Konstantina and Tsakalides, Panagiotis , title =. Sensors , volume =. 2021 , publisher =

    2021

  36. [36]

    Hyperspectral unmixing overview: Geometrical, statistical, and sparse regression-based approaches , journal =

    Bioucas-Dias, Jos. Hyperspectral unmixing overview: Geometrical, statistical, and sparse regression-based approaches , journal =. 2012 , publisher =

    2012

  37. [37]

    Earth System Science Data Discussions , year =

    Khaldi, Rohaifa and Rodríguez-Ortega, Jesús and Alcaraz-Segura, Domingo and Tabik, Siham , title =. Earth System Science Data Discussions , year =. doi:10.5194/essd-2021-253 , note =

    work page doi:10.5194/essd-2021-253 2021

  38. [38]

    2013 , eprint =

    Graves, Alex , title =. 2013 , eprint =

    2013

  39. [39]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Miao, Lidan and Qi, Hairong , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2007 , publisher =

    2007

  40. [40]

    2025 , eprint=

    Sparse Deformable Mamba for Hyperspectral Image Classification , author=. 2025 , eprint=

    2025

  41. [41]

    Remote sensing of environment , volume=

    Characterizing land cover/land use from multiple years of Landsat and MODIS time series: A novel approach using land surface phenology modeling and random forest classifier , author=. Remote sensing of environment , volume=. 2020 , publisher=

    2020

  42. [42]

    Remote Sensing , volume=

    MSNet: A multi-stream fusion network for remote sensing spatiotemporal fusion based on transformer and convolution , author=. Remote Sensing , volume=. 2021 , publisher=

    2021

  43. [43]

    IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , year=

    Large Area Crops Mapping by Phenological Horizon Attention Transformer (PHAT) Method Using MODIS Time-Series Imagery , author=. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , year=

  44. [44]

    Journal of Atmospheric and Solar-Terrestrial Physics , volume=

    Application of Long Short-Term Memory neural network model for the reconstruction of MODIS Land Surface Temperature images , author=. Journal of Atmospheric and Solar-Terrestrial Physics , volume=. 2019 , publisher=

    2019

  45. [45]

    Ieee Access , volume=

    Classification for remote sensing data with improved CNN-SVM method , author=. Ieee Access , volume=. 2019 , publisher=

    2019

  46. [46]

    Remote Sensing , volume=

    Exploiting the classification performance of support vector machines with multi-temporal Moderate-Resolution Imaging Spectroradiometer (MODIS) data in areas of agreement and disagreement of existing land cover products , author=. Remote Sensing , volume=. 2012 , publisher=

    2012

  47. [47]

    Global developments in environmental earth observation from space , pages=

    Land cover classification with Support Vector Machine applied to MODIS imagery , author=. Global developments in environmental earth observation from space , pages=. 2006 , publisher=

    2006

  48. [48]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume=

    Comparison of support vector machine, neural network, and CART algorithms for the land-cover classification using limited training data points , author=. ISPRS Journal of Photogrammetry and Remote Sensing , volume=. 2012 , publisher=

    2012

  49. [49]

    Remote Sensing , volume=

    Feature selection of time series MODIS data for early crop classification using random forest: A case study in Kansas, USA , author=. Remote Sensing , volume=. 2015 , publisher=

    2015

  50. [50]

    IEEE Journal of selected topics in applied earth observations and remote sensing , volume=

    Mapping annual land use and land cover changes using MODIS time series , author=. IEEE Journal of selected topics in applied earth observations and remote sensing , volume=. 2014 , publisher=

    2014

  51. [51]

    Remote Sensing , volume=

    Developing a random forest algorithm for MODIS global burned area classification , author=. Remote Sensing , volume=. 2017 , publisher=

    2017

  52. [52]

    International Journal of Applied Earth Observation and Geoinformation , volume=

    Temporal optimisation of image acquisition for land cover classification with Random Forest and MODIS time-series , author=. International Journal of Applied Earth Observation and Geoinformation , volume=. 2015 , publisher=

    2015

  53. [53]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Spectral--spatial--temporal transformers for hyperspectral image change detection , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2022 , publisher=

    2022

  54. [54]

    3DSS-Mamba: 3D-Spectral-Spatial Mamba for Hyperspectral Image Classification , year=

    He, Yan and Tu, Bing and Liu, Bo and Li, Jun and Plaza, Antonio , journal=. 3DSS-Mamba: 3D-Spectral-Spatial Mamba for Hyperspectral Image Classification , year=

  55. [55]

    Modis collection 5 global land cover: Algorithm re- finements and characterization of new datasets.Remote Sensing of Environment, 114(1): 31 168–182, 2010

    MODIS Collection 5 global land cover: Algorithm refinements and characterization of new datasets , journal =. 2010 , issn =. doi:https://doi.org/10.1016/j.rse.2009.08.016 , url =

    work page doi:10.1016/j.rse.2009.08.016 2010

  56. [56]

    Remote sensing of Environment , volume=

    Cropland distributions from temporal unmixing of MODIS data , author=. Remote sensing of Environment , volume=. 2004 , publisher=

    2004

  57. [57]

    International Journal of remote sensing , volume=

    The pixel: a snare and a delusion , author=. International Journal of remote sensing , volume=. 1997 , publisher=

    1997

  58. [58]

    and Nai-Yu Chen , journal=

    Pi-Fuei Hsieh and Lee, L.C. and Nai-Yu Chen , journal=. Effect of spatial resolution on classification errors of pure and mixed pixels in remote sensing , year=

  59. [59]

    2013 , issn =

    The use of single-date MODIS imagery for estimating large-scale urban impervious surface fraction with spectral mixture analysis and machine learning techniques , journal =. 2013 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2013.09.010 , url =

    work page doi:10.1016/j.isprsjprs.2013.09.010 2013

  60. [60]

    A unified spatial-spectral-temporal fusion model using Landsat and MODIS imagery , year=

    Bin Chen and Bing Xu , booktitle=. A unified spatial-spectral-temporal fusion model using Landsat and MODIS imagery , year=

  61. [61]

    and Olthof, Ian , year =

    Latifovic, Rasim and Pouliot, D.A. and Olthof, Ian , year =. Circa 2010 Land Cover of Canada: Local Optimization Methodology and Product Development , volume =. Remote Sensing , doi =

    2010

  62. [62]

    Advances in Neural Information Processing Systems , volume=

    Earthformer: Exploring space-time transformers for earth system forecasting , author=. Advances in Neural Information Processing Systems , volume=

  63. [63]

    Scientific Reports , volume=

    Transformer-based land use and land cover classification with explainability using satellite imagery , author=. Scientific Reports , volume=. 2024 , publisher=

    2024

  64. [64]

    Forest Ecology and Management , volume=

    The use of remote sensing and GIS in watershed level analyses of non-point source pollution problems , author=. Forest Ecology and Management , volume=. 2000 , publisher=

    2000

  65. [65]

    2024 , issn =

    A local enhanced mamba network for hyperspectral image classification , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.jag.2024.104092 , url =

    work page doi:10.1016/j.jag.2024.104092 2024

  66. [66]

    Ahmad, S

    A Comprehensive Survey for Hyperspectral Image Classification: The Evolution from Conventional to Transformers and Mamba Models , author=. arXiv preprint arXiv:2404.14955 , year=

    work page arXiv

  67. [67]

    International journal of remote sensing , volume=

    Using long short-term memory recurrent neural network in land cover classification on Landsat and Cropland data layer time series , author=. International journal of remote sensing , volume=. 2019 , publisher=

    2019

  68. [68]

    ISPRS International Journal of Geo-Information , volume=

    Multi-temporal land cover classification with sequential recurrent encoders , author=. ISPRS International Journal of Geo-Information , volume=. 2018 , publisher=

    2018

  69. [69]

    Sustainability , volume=

    Short-term forecasting of land use change using recurrent neural network models , author=. Sustainability , volume=. 2019 , publisher=

    2019

  70. [70]

    Deep Residual Learning for Image Recognition , year=

    He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian , booktitle=. Deep Residual Learning for Image Recognition , year=

  71. [71]

    The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences , volume=

    Land Use Land Cover Mapping Using Uas Imagery: Scene Classification and Semantic Segmentation , author=. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences , volume=. 2022 , publisher=

    2022

  72. [72]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    Extended vision transformer (ExViT) for land use and land cover classification: A multimodal deep learning framework , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2023 , publisher=

    2023

  73. [73]

    Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

    Vision transformer for multispectral satellite imagery: Advancing landcover classification , author=. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

  74. [74]

    IEEE Transactions on Geoscience and Remote Sensing , year=

    HyperMamba: A Spectral-Spatial Adaptive Mamba for Hyperspectral Image Classification , author=. IEEE Transactions on Geoscience and Remote Sensing , year=

  75. [75]

    IEEE Transactions on Geoscience and Remote Sensing , year=

    MambaHSI: Spatial-spectral mamba for hyperspectral image classification , author=. IEEE Transactions on Geoscience and Remote Sensing , year=

  76. [76]

    arXiv preprint , volume =

    Kaiming He and Xiangyu Zhang and Shaoqing Ren and Jian Sun , title =. arXiv preprint , volume =. 2015 , url =

    2015

  77. [77]

    ICCV, 2021.https://arxiv.or g/abs/2103.14030 35 Supplementary Material S1

    Swin Transformer: Hierarchical Vision Transformer using Shifted Windows , author =. arXiv preprint arXiv:2103.14030 , year =. doi:10.48550/arXiv.2103.14030 , url =

    work page doi:10.48550/arxiv.2103.14030

  78. [78]

    A convnet for the 2020s.arXiv preprint arXiv:2201.03545, 2022

    A ConvNet for the 2020s , author =. arXiv preprint arXiv:2201.03545 , year =. doi:10.48550/arXiv.2201.03545 , url =

    work page doi:10.48550/arxiv.2201.03545

  79. [79]

    Proceedings of the International Conference on Learning Representations (ICLR) , year =

    An Image is Worth 16×16 Words: Transformers for Image Recognition at Scale , author =. Proceedings of the International Conference on Learning Representations (ICLR) , year =

  80. [80]

    Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling , journal =

    Junyoung Chung and. Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling , journal =. 2014 , url =

    2014

Showing first 80 references.