Pith. sign in

REVIEW 4 major objections 6 minor 62 references

RMAU-NET: A Residual-Multihead-Attention U-Net Architecture for Landslide Segmentation and Detection from Remote Sensing Images

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A residual multihead-attention U-Net reports top landslide detection and segmentation accuracy on three public remote-sensing benchmarks.

desk verdict A thorough ablation study undermined by test-set selection of the threshold and architecture, and by internal contradictions between the ablation table and the final results table. read the letter →

arxiv 2507.11143 v1 pith:ZAQ2IMXP submitted 2025-07-15 cs.CV cs.LGeess.IV

Lam Pham , Cam Le , Hieu Tang , Khang Truong , Truong Nguyen , Jasmin Lampert , Alexander Schindler , Martin Boyer
show 1 more author
Son Phan
This is my paper · ORCID
classification cs.CVcs.LGeess.IV
keywords landslidedetectionsegmentationremotesensingU-NetmultiheadattentionresiduallearningLandslide4SenseSentinel-2
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 tries to establish that a single deep-learning architecture, RMAU-NET, can automatically detect and segment landslides from free satellite imagery with accuracy high enough to be useful in real landslide observation systems. It reports F1 scores of 98.23 on Landslide4Sense detection and 93.83 on Bijie detection, and mIoU scores of 65.97 on Landslide4Sense segmentation and 76.88 on Nepal segmentation. The authors build the architecture by starting from a U-Net baseline and adding, one at a time, generated spectral bands, a combined Focal and IoU loss, multiple prediction heads at different resolutions, residual-convolutional blocks, a multihead attention layer, and a final threshold. Each addition is shown to improve the scores, which is the evidence that the full design, not any single trick, is what delivers the result.

What carries the argument

The load-bearing object is the RMAU-NET architecture itself. It is a U-Net whose double-convolution blocks are replaced by residual-convolutional layers, with a multihead attention layer inserted after every convolutional layer; the attention layer pools the feature map along each of its width, height, and channel dimensions, applies standard multihead attention to the pooled maps, and multiplies the result back onto the original feature map. Around this backbone the model adds 9 generated bands to the 14 original satellite bands, averages three prediction heads at 64x64, 128x128, and 256x256 resolution, trains with a combined Focal and IoU loss, and applies a 0.95 threshold to the output mask. The argument proceeds by ablation: each component raises F1 and mIoU on Landslide4Sense, from 67.83 and 60.01 for the plain U-Net baseline to 74.63 and 65.97 for the complete RMAU-NET.

What would settle it

Run RMAU-NET on a landslide dataset from a region and time period completely absent from training, or on the official 800-image hidden test set of Landslide4Sense, and check whether the F1 and mIoU values stay near 98.23 and 65.97; a large drop would show the reported gains depend on seeing the test images during development.

Watch

Extended reading notes

Core claim

The central claim is that RMAU-NET, an encoder-decoder with residual skip blocks and a three-dimensional multihead attention mechanism, outperforms previous landslide mapping models on all three benchmark datasets it is tested on. On Landslide4Sense the model reaches 98.23 F1 for image-level detection and 65.97 mIoU for pixel-level segmentation; on Bijie it reaches 93.83 F1 for detection; on Nepal it reaches 76.88 mIoU for segmentation. The authors interpret these numbers as showing the architecture is robust across different sensors, regions, and landslide ratios, and therefore suitable for integration into operational landslide observation systems.

Load-bearing premise

The evaluation assumes that carving an 80:20 split from the Landslide4Sense training set gives a valid held-out test, and that tuning the final threshold on that same split does not inflate the reported scores.

Editorial extensions

If this is right

  • The same architecture handles both image-level detection and pixel-level segmentation, so a single model can be deployed for landslide early-warning and inventory mapping.
  • Generated bands (normalization, NDVI/NDMI/NBR, filters, gradients, edges) contribute beyond the original 14 bands, suggesting spectral and textural indices are worth adding for other remote-sensing tasks.
  • Multiple resolution heads and residual-plus-attention blocks give a concrete recipe for improving U-Net-style models on imbalanced segmentation problems.
  • The reported scores on Landslide4Sense, Bijie, and Nepal position RMAU-NET as a strong candidate baseline for future landslide mapping work.

Reading between the lines

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

  • Because the Landslide4Sense test split was cut from the same training set and the 0.95 threshold was chosen on that split, the reported numbers likely overstate performance on truly unseen terrain; a geographically or temporally disjoint evaluation would be a stricter test.
  • The Nepal result (mIoU 76.88 with F1 69.43) suggests the model favors precise boundaries over recall, which may be suitable for inventory maps but could miss smaller landslides in emergency response.
  • The three-dimensional multihead attention design is dataset-agnostic and could be tested on other sparse-target segmentation problems, such as building damage or wildfire scar mapping.
  • A simple extension would be to learn the post-processing threshold per dataset instead of fixing it, which would remove a source of test-set leakage.
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

4 major / 6 minor

Summary. The paper proposes RMAU-NET, a U-Net variant combining residual-convolutional blocks, a multihead attention layer, multi-resolution segmentation heads, 23 input bands, a combined Focal+IoU loss, and threshold-based post-processing, for landslide detection and segmentation from remote sensing images. The authors evaluate the model on three public datasets (Landslide4Sense, Bijie, Nepal) and report F1=98.23 for Landslide4Sense detection, F1=93.83 for Bijie detection, mIoU=65.97 for Landslide4Sense segmentation, and mIoU=76.88 for Nepal segmentation, claiming state-of-the-art performance. The evaluation is developed through an ablation study on an 80:20 split of the Landslide4Sense training set, with the post-processing threshold selected on that same split.

Significance. If the reported numbers were reproducible, RMAU-NET would be a practically valuable contribution to automated landslide mapping, since it addresses both detection and segmentation on publicly available benchmarks and the ablation study systematically isolates the contribution of each component. The use of three public datasets and the clear ablation progression from Table III to Table X are strengths that would help practitioners. However, the central empirical claim is compromised by evaluation on the development split and by internal inconsistencies between tables and the abstract; without code or a corrected evaluation protocol, the reported performance is not verifiable.

major comments (4)
  1. [Section II-C and Table IX/Table XI] The Landslide4Sense evaluation uses a single 80:20 split of the training set, and the same split is used both to select every component (including the 0.95 threshold in Table IX) and to produce the final numbers in Table XI; because the threshold was chosen by maximizing F1/mIoU on this split, the reported F1=98.23 detection and F1=76.90/mIoU=65.97 segmentation numbers are not unbiased estimates of performance on unseen data. A separate validation split, or nested cross-validation, and a final evaluation on the official Landslide4Sense test set are needed before the headline results can be taken as evidence of generalization.
  2. [Table X vs Table XI] The final row of Table X lists the complete RMAU-NET configuration on Landslide4Sense segmentation as F1=74.63, mIoU=65.97, while Table XI reports F1=76.90 for the same task with no described change other than the addition of a detection head; the paper does not state that the model is retrained with a multi-task loss, so the +2.27 F1 increase is unexplained. The abstract also reports mIoU=63.74 for the same dataset, contradicting Table XI's 65.97. These discrepancies mean the central claim in Section V is not reproducible from the experimental record.
  3. [Section V and Table XI] The paper claims state-of-the-art results, but Table XI contains no comparison on Landslide4Sense; only the proposed RMAU-NET is listed for that dataset, and the comparison rows cover only Nepal and Bijie. Without a comparison to published methods on the same Landslide4Sense split, or on the official test set, the state-of-the-art claim is unsupported.
  4. [Section IV-D and Table IX] The threshold of 0.95 is applied to Bijie and Nepal after being selected on the Landslide4Sense 80:20 split, with no sensitivity analysis or validation on those datasets; since threshold is a free parameter, the cross-dataset results in Table XI are not independent evidence of robustness. The authors should report thresholds chosen on a validation set per dataset, or show that the chosen threshold transfers with bounded performance loss.
minor comments (6)
  1. [Section IV-D] The text states the best F1 score is 74.463, but Table X and Table IX report 74.63; the text and tables should be reconciled.
  2. [Section V] The text refers to the proposed architecture as shown in Fig. 8, but Fig. 7 contains the RMAU-NET architecture and Fig. 8 shows segmentation results; the figure references should be corrected.
  3. [Section II-C] The sentence 'remaining the 70:30 ratio for the train and test sets' appears to contain a typo; it should likely read 'retaining the 70:30 ratio'.
  4. [Section II-A] There are typographical errors such as 'RBG' and 'landsline'; a copyedit pass is needed.
  5. [Equation (2)] The loss mixing coefficient alpha is fixed at 0.5 without reporting sensitivity to its value; since alpha is a free parameter, a brief study or robustness check would strengthen the loss-function analysis.
  6. [Experimental settings] No code or complete training details (batch size, learning rate, number of parameters, inference time, and the exact multi-task training protocol for RMAU-NET) are provided, which limits reproducibility.

Circularity Check

1 steps flagged · score 4.0 of 10

Landslide4Sense final mIoU is partially fitted because the 0.95 threshold was selected on the same 20% split later reported as the final test result; no derivation-level circularity otherwise.

  1. fitted input called prediction [Section II-C (dataset splitting), Section IV-D (Table IX threshold sweep), reported as final in Table XI]
    "We split the training set (3044 images) into two parts, using an 80:20 ratio, for training and testing, respectively. ... The Table IX indicates that we obtained the best F1 score of 74.463 and mIoU score of 65.97 at the threshold value of 0.95."

    The 80:20 Landslide4Sense split functions as both the development split and the final evaluation split. Table IX sweeps post-processing thresholds on this split and selects 0.95 because it maximizes F1/mIoU there. Table XI then reports RMAU-NET's Landslide4Sense segmentation mIoU as 65.97, exactly the value at the selected threshold, as though it were an independent final result. The reported final metric is therefore an optimized value from the evaluation set, not a prediction on unseen data: the threshold parameter is fitted to the same test split on which the headline performance is claimed.

full rationale

The central claim is an empirical benchmark result rather than a formal derivation, so it does not reduce by construction to a fitted constant in the sense of Eq. X = Eq. Y. The main circularity-adjacent problem is the threshold tuning on the Landslide4Sense test split: the final mIoU 65.97 in Table XI is identical to the Table IX value chosen by sweeping thresholds on that split, making that number partly self-confirmatory. The architecture's attention layer is imported from the authors' own IGARSS paper [61], but the current paper independently evaluates it in Table VIII (e.g., U-Net with Multihead Att. 71.45 vs 70.45), so the self-citation is not load-bearing by itself. Separate issues, such as the disagreement between Table X final F1 (74.63) and Table XI F1 (76.90) for the same task and the abstract mIoU (63.74) versus Table XI (65.97), are internal-consistency or reproducibility concerns rather than circularity. Overall the empirical comparisons against published datasets retain independent content, so the score is moderate rather than high.

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

The paper introduces no new physical entities. Its free parameters are engineering choices (loss weight, threshold, filter sizes, split ratios). The main unstated burden is that model selection and threshold selection happen on the same test split used to report the final numbers, so the report treats a fitted evaluation as a prediction.

free parameters (4)
  • Loss mixing coefficient alpha = 0.5
    Set empirically in Equation (2) to combine Focal and IoU losses; no ablation over alpha is shown.
  • Post-processing threshold = 0.95
    Chosen by sweeping thresholds on the Landslide4Sense test split in Table IX, then used in the final RMAU-NET results.
  • Filter kernel size for band generation = 10x10
    Gaussian and median filters with 10x10 kernel are used to create bands 22-23 in Table V; no sensitivity analysis is provided.
  • Train/test split ratios = 80:20 and 70:30
    Landslide4Sense uses an 80:20 split and Bijie a 70:30 split (Section II-C); these choices affect all reported scores and are not cross-validated.
assumptions (4)
  • domain assumption The 80:20 split of Landslide4Sense training images is representative of the hidden test distribution.
    Stated in Section II-C; the model and threshold are selected and evaluated on this split.
  • domain assumption The combination of Focal and IoU loss with alpha=0.5 improves the imbalanced segmentation task.
    Section IV-A; only a small set of loss functions is compared and one combination is chosen without ablating alpha.
  • domain assumption The newly generated bands 15-23 add discriminative signal rather than noise.
    Section IV-B and Table VI; the choice of 23 bands is based on test-set performance, not on an independent theoretical justification.
  • ad hoc to paper The threshold of 0.95 selected on Landslide4Sense transfers to Bijie and Nepal and to deployment.
    Section IV-D; the threshold is tuned on one dataset and applied to other datasets without separate validation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RMAU-NET: A Residual-Multihead-Attention U-Net Architecture for Landslide Segmentation and Detection from Remote Sensing Images." pith.science (2026). https://pith.science/paper/ZAQ2IMXP

@misc{pith2026250711143,
  author       = {Pith},
  title        = {Pith review of: RMAU-NET: A Residual-Multihead-Attention U-Net Architecture for Landslide Segmentation and Detection from Remote Sensing Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZAQ2IMXP}},
  note         = {Machine review of arXiv:2507.11143}
}
read the original abstract

In recent years, landslide disasters have reported frequently due to the extreme weather events of droughts, floods , storms, or the consequence of human activities such as deforestation, excessive exploitation of natural resources. However, automatically observing landslide is challenging due to the extremely large observing area and the rugged topography such as mountain or highland. This motivates us to propose an end-to-end deep-learning-based model which explores the remote sensing images for automatically observing landslide events. By considering remote sensing images as the input data, we can obtain free resource, observe large and rough terrains by time. To explore the remote sensing images, we proposed a novel neural network architecture which is for two tasks of landslide detection and landslide segmentation. We evaluated our proposed model on three different benchmark datasets of LandSlide4Sense, Bijie, and Nepal. By conducting extensive experiments, we achieve F1 scores of 98.23, 93.83 for the landslide detection task on LandSlide4Sense, Bijie datasets; mIoU scores of 63.74, 76.88 on the segmentation tasks regarding LandSlide4Sense, Nepal datasets. These experimental results prove potential to integrate our proposed model into real-life landslide observation systems.

Figures

Figures reproduced from arXiv: 2507.11143 by the authors.

Figure 1
Figure 1. Some image samples with 14 band and corresponding mask from [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Some image samples from Nepal Dataset [38] [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. The proposed U-Net baseline architecture for the landslide segmentation. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (3 more)
Figure 6
Figure 6. Figure 6: The proposed multihead-attention layer. TABLE VIII EVALUATE THE NETWORK ARCHITECTURE IMPROVEMENTS (U-NET✝: U-NET BASELINE WITH 23 BAND DATA, COMBINED LOSS FUNCTION, AND MULTIPLE RESOLUTION HEADS). Networks F1 score mIoU U-Net✝ 70.45 62.19 DeepLab-V3 68.30 60.49 MobileN…
Figure 7
Figure 7. Figure 7: The proposed RMAU-NET architecture. TABLE IX EVALUATE THE THRESHOLD VALUES Threshold Values F1 score mIoU None 72.42 63.88 0.4 71.78 63.29 0.5 72.42 63.88 0.6 72.60 64.02 0.75 73.02 64.43 0.85 73.13 64.69 0.9 73.68 65.07 0.95 74.63 65.97 0.99 73.11 64.68 backbone, we f…
Figure 8
Figure 8. Figure 8: Segmentation results obtained from RMAU-NET model on LandSlide4Sense remote sensing images. [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references · 59 canonical work pages

  1. [1]

    Landslides,

    The International Charter Space and Major Disasters, “Landslides,” https://disasterscharter.org/web/guest/ disaster-types/-/article/landslides , Accessed: August 25, 2024

    work page 2024

  2. [2]

    The impact of climate change on landslide hazard and risk,

    Luciano Picarelli, Suzanne Lacasse, and Ken KS Ho, “The impact of climate change on landslide hazard and risk,” Understanding and Reducing Landslide Disaster Risk: Volume 1 Sendai Landslide Partnerships and Kyoto Landslide Commitment 5th , pp. 131–141, 2021

    work page 2021

  3. [3]

    The international emergency disasters database,

    EM-DAT, “The international emergency disasters database,” 2023

    work page 2023

  4. [4]

    Crozier, The Nature of Landslide Hazard Impact, chapter 2, pp

    Thomas Glade and Michael J. Crozier, The Nature of Landslide Hazard Impact, chapter 2, pp. 41–74, John Wiley & Sons, Ltd, 2005

    work page 2005

  5. [5]

    Landslide in india,

    The International Charter Space and Major Disasters, “Landslide in india,” https://disasterscharter. org/web/guest/activations/-/article/ landslide-in-india-activation-900- , Accessed: August 25, 2024. TABLE X APPLY DEEP LEARNING TECHNIQUES TO FURTHER IMRPOVE THE U-N ET BASELINE FOR LANDSLIDE SEGMENTATION ON LANDSLIDE 4SENSE DATASET Network Combined loss ...

    work page 2024

  6. [6]

    Landslide in ethiopia,

    The International Charter Space and Major Disasters, “Landslide in ethiopia,” https://disasterscharter. org/web/guest/activations/-/article/ landslide-in-ethiopia-activation-899- , Accessed: August 25, 2024

    work page 2024

  7. [7]

    Landslide inventory maps: New tools for an old problem,

    Fausto Guzzetti, Alessandro Cesare Mondini, Mauro Cardinali, Feder- ica Fiorucci, Michele Santangelo, and Kang-Tsung Chang, “Landslide inventory maps: New tools for an old problem,” Earth-Science Reviews, vol. 112, no. 1-2, pp. 42–66, 2012

    work page 2012

  8. [8]

    Landslide mapping from aerial photographs using change detection-based markov random field,

    Zhongbin Li, Wenzhong Shi, Ping Lu, Lin Yan, Qunming Wang, and Zelang Miao, “Landslide mapping from aerial photographs using change detection-based markov random field,” Remote Sensing of Environment, vol. 187, pp. 76–90, Dec. 2016

    work page 2016

Show all 62 references
  1. [9]

    Objective landslide detection and surface morphology mapping using high-resolution airborne laser altimetry,

    J McKean and J Roering, “Objective landslide detection and surface morphology mapping using high-resolution airborne laser altimetry,” Geomorphology, vol. 57, no. 3-4, pp. 331–351, 2004

    2004

  2. [10]

    Spatial data for landslide susceptibility, hazard, and vulnerability assessment: An overview,

    Cees J Van Westen, Enrique Castellanos, and Sekhar L Kuriakose, “Spatial data for landslide susceptibility, hazard, and vulnerability assessment: An overview,” Engineering geology , vol. 102, no. 3-4, pp. 112–131, 2008

    2008

  3. [11]

    Analysis of lidar-derived topo- graphic information for characterizing and differentiating landslide morphology and activity,

    Nancy F. Glenn, David R. Streutker, D. John Chadwick, Glenn D. Thackray, and Stephen J. Dorsch, “Analysis of lidar-derived topo- graphic information for characterizing and differentiating landslide morphology and activity,” Geomorphology, vol. 73, no. 1, pp. 131– 148, 2006

    2006

  4. [12]

    Giant landslides, topography, and erosion,

    Oliver Korup, John J. Clague, Reginald L. Hermanns, Kenneth Hewitt, Alexander L. Strom, and Johannes T. Weidinger, “Giant landslides, topography, and erosion,” Earth and Planetary Science Letters , vol. 261, no. 3, pp. 578–589, 2007

    2007

  5. [13]

    Relation between land cover and landslide susceptibility in val d’aran, pyrenees (spain): Historical aspects, present situation and forward prediction,

    Heping Shu, Marcel H ¨urlimann, Roberto Molowny-Horas, Marta Gonz´alez, Jordi Pinyol, Cl `audia Abanc ´o, and Jinzhu Ma, “Relation between land cover and landslide susceptibility in val d’aran, pyrenees (spain): Historical aspects, present situation and forward prediction,” Sc...

    2019

  6. [14]

    Variations in the susceptibility to landslides, as a consequence of land cover changes: A look to the past, and another towards the future,

    L. Pisano, V . Zumpano, ˇZ. Malek, C.M. Rosskopf, and M. Parise, “Variations in the susceptibility to landslides, as a consequence of land cover changes: A look to the past, and another towards the future,” Science of The Total Environment, vol. 601-602, pp. 1147–1159, 2017

    2017

  7. [15]

    An overview of the applications of earth observation satellite data: impacts and future trends,

    Qiang Zhao, Le Yu, Zhenrong Du, Dailiang Peng, Pengyu Hao, Yongguang Zhang, and Peng Gong, “An overview of the applications of earth observation satellite data: impacts and future trends,” Remote Sensing, vol. 14, no. 8, pp. 1863, 2022

    2022

  8. [16]

    Deep learning for land use and land cover classification based on hyperspectral and multispectral earth observation data: A review,

    Ava Vali, Sara Comai, and Matteo Matteucci, “Deep learning for land use and land cover classification based on hyperspectral and multispectral earth observation data: A review,” Remote Sensing, vol. 12, no. 15, pp. 2495, 2020

    2020

  9. [17]

    Uav & satellite synergies for optical remote sensing applications: A liter- ature review,

    Emilien Alvarez-Vanhard, Thomas Corpetti, and Thomas Houet, “Uav & satellite synergies for optical remote sensing applications: A liter- ature review,” Science of remote sensing , vol. 3, pp. 100019, 2021

    2021

  10. [18]

    Review on remote sensing methods for landslide detec- tion using machine and deep learning,

    Amrita Mohan, Amit Kumar Singh, Basant Kumar, and Ramji Dwivedi, “Review on remote sensing methods for landslide detec- tion using machine and deep learning,” Transactions on Emerging Telecommunications Technologies, vol. 32, no. 7, pp. e3998, 2021

    2021

  11. [19]

    The outcome of the 2022 landslide4sense competi- tion: Advanced landslide detection from multisource satellite imagery,

    Omid Ghorbanzadeh, Yonghao Xu, Hengwei Zhao, Junjue Wang, Yanfei Zhong, Dong Zhao, Qi Zang, Shuang Wang, Fahong Zhang, Yilei Shi, Xiao Xiang Zhu, Lin Bai, Weile Li, Weihang Peng, and Pedram Ghamisi, “The outcome of the 2022 landslide4sense competi- tion: Advanced landslide det...

    2022

  12. [20]

    Ma- chine learning techniques in landslide susceptibility mapping: a survey and a case study,

    Taskin Kavzoglu, Ismail Colkesen, and Emrehan Kutlug Sahin, “Ma- chine learning techniques in landslide susceptibility mapping: a survey and a case study,” Landslides: Theory, practice and modelling , pp. 283–301, 2019

    2019

  13. [21]

    Revamping land cov- erage analysis using aerial satellite image mapping,

    Gowtham Rajmohan, Chandru Vignesh Chinnappan, Alfred Daniel John William, Sivaparthipan Chandrakrishan Balakrishnan, Bala Anand Muthu, and Gunasekaran Manogaran, “Revamping land cov- erage analysis using aerial satellite image mapping,” Transactions on Emerging Telecommunicati...

    2021

  14. [22]

    Landslide susceptibility mapping using frequency ratio, analytic hierarchy process, logistic regression, and artificial neural network methods at the inje area, korea,

    Soyoung Park, Chuluong Choi, Byungwoo Kim, and Jinsoo Kim, “Landslide susceptibility mapping using frequency ratio, analytic hierarchy process, logistic regression, and artificial neural network methods at the inje area, korea,” Environmental earth sciences , vol. 68, pp. 1443...

    2013

  15. [23]

    Object-oriented landslide mapping using zy-3 satellite imagery, random forest and mathematical morphology, for the three-gorges reservoir, china,

    Tao Chen, John C Trinder, and Ruiqing Niu, “Object-oriented landslide mapping using zy-3 satellite imagery, random forest and mathematical morphology, for the three-gorges reservoir, china,” Remote sensing , vol. 9, no. 4, pp. 333, 2017

    2017

  16. [24]

    Reg- sa–unet++: A lightweight landslide detection network based on single- temporal images captured postlandslide,

    Chaoyang Niu, Ouyang Gao, Wanjie Lu, Wei Liu, and Tao Lai, “Reg- sa–unet++: A lightweight landslide detection network based on single- temporal images captured postlandslide,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , vol. 15, pp. 9746–...

    2022

  17. [25]

    A landslide extraction method of channel attention mechanism u-net network based on sentinel-2a remote sensing images,

    Hesheng Chen, Yi He, Lifeng Zhang, Sheng Yao, Wang Yang, Yumin Fang, Yaoxiang Liu, and Binghai Gao, “A landslide extraction method of channel attention mechanism u-net network based on sentinel-2a remote sensing images,” International Journal of Digital Earth , vol. 16, no. 1,...

    2023

  18. [26]

    Drs-unet: A deep semantic segmentation network for the recognition of active landslides from insar imagery in the three rivers region of the qinghai–tibet plateau,

    Ximing Chen, Xin Yao, Zhenkai Zhou, Yang Liu, Chuangchuang Yao, and Kaiyu Ren, “Drs-unet: A deep semantic segmentation network for the recognition of active landslides from insar imagery in the three rivers region of the qinghai–tibet plateau,” Remote Sensing , vol. 14, no. 8,...

    2022

  19. [27]

    Landslide susceptibility prediction based on image semantic segmentation,

    Bowen Du, Zirong Zhao, Xiao Hu, Guanghui Wu, Liangzhe Han, Leilei Sun, and Qiang Gao, “Landslide susceptibility prediction based on image semantic segmentation,” Computers & Geosciences , vol. 155, pp. 104860, 2021

    2021

  20. [28]

    Landslide recognition from multi- feature remote sensing data based on improved transformers,

    Renxiang Huang and Tao Chen, “Landslide recognition from multi- feature remote sensing data based on improved transformers,” Remote Sensing, vol. 15, no. 13, pp. 3340, 2023

    2023

  21. [29]

    An improved segmentation method for automatic mapping of cone karst from remote sensing data based on deeplab v3+ model,

    Han Fu, Bihong Fu, and Pilong Shi, “An improved segmentation method for automatic mapping of cone karst from remote sensing data based on deeplab v3+ model,” Remote Sensing , vol. 13, no. 3, pp. 441, 2021

    2021

  22. [30]

    Evaluation of different machine learning methods and deep-learning convolutional neural networks for landslide detection,

    Omid Ghorbanzadeh, Thomas Blaschke, Khalil Gholamnia, Sansar Raj Meena, Dirk Tiede, and Jagannath Aryal, “Evaluation of different machine learning methods and deep-learning convolutional neural networks for landslide detection,” Remote Sensing , vol. 11, no. 2, pp. 196, 2019

    2019

  23. [31]

    Relict landslide detection using deep- learning architectures for image segmentation in rainforest areas: a new framework,

    Guilherme Pereira Bento Garcia, LP Soares, Mateus Espadoto, and Carlos Henrique Grohmann, “Relict landslide detection using deep- learning architectures for image segmentation in rainforest areas: a new framework,” International Journal of Remote Sensing , vol. 44, no. 7, pp. ...

    2023

  24. [32]

    Deep learning for landslide recognition in satellite architecture,

    Trong-An Bui, Pei-Jun Lee, Kai-Yew Lum, Clarissa Loh, and Kyo Tan, “Deep learning for landslide recognition in satellite architecture,” IEEE Access, vol. 8, pp. 143665–143678, 2020

    2020

  25. [33]

    Landslide detection using multi-scale image segmentation and different machine learning models in the higher himalayas,

    Sepideh Tavakkoli Piralilou, Hejar Shahabi, Ben Jarihani, Omid Ghor- banzadeh, Thomas Blaschke, Khalil Gholamnia, Sansar Raj Meena, and Jagannath Aryal, “Landslide detection using multi-scale image segmentation and different machine learning models in the higher himalayas,” Re...

    2019

  26. [34]

    Sar-based landslide classification pretraining leads to better segmentation,

    Vanessa B ¨ohm, Wei Ji Leong, Ragini Bal Mahesh, Ioannis Prapas, Edoardo Nemni, Freddie Kalaitzis, Siddha Ganju, and Raul Ramos- Pollan, “Sar-based landslide classification pretraining leads to better segmentation,” arXiv preprint arXiv:2211.09927 , 2022

    2022 arXiv

  27. [35]

    Landslide identification from post-earthquake high-resolution remote sensing images based on resunet–bfa,

    Zhenyu Zhao, Shucheng Tan, Yiquan Yang, and Qinghua Zhang, “Landslide identification from post-earthquake high-resolution remote sensing images based on resunet–bfa,” Remote Sensing , vol. 17, no. 6, pp. 995, 2025

    2025

  28. [36]

    Enhanced u-net++ for improved semantic segmentation in landslide detection,

    Meng Tang, Yuelin He, Muhammed Aslam, Edore Akpokodje, and Syeda Fizzah Jilani, “Enhanced u-net++ for improved semantic segmentation in landslide detection,” Sensors, vol. 25, no. 9, pp. 2670, 2025

    2025

  29. [37]

    Landslide detection from an open satellite imagery and digital elevation model dataset using attention boosted convolutional neural networks,

    Shunping Ji, Dawen Yu, Chaoyong Shen, Weile Li, and Qiang Xu, “Landslide detection from an open satellite imagery and digital elevation model dataset using attention boosted convolutional neural networks,” Landslides, vol. 17, pp. 1337–1352, 2020

    2020

  30. [38]

    Convolutional neural networks applied to semantic segmentation of landslide scars,

    L. Bragagnolo, L.R. Rezende, R.V . da Silva, and J.M.V . Grzybowski, “Convolutional neural networks applied to semantic segmentation of landslide scars,” CATENA, vol. 201, pp. 105189, 2021

    2021

  31. [39]

    A practical trial of landslide detection from single-temporal landsat8 images using contour-based proposals and random forest: A case study of national nepal,

    Fang Chen, Bo Yu, and Bin Li, “A practical trial of landslide detection from single-temporal landsat8 images using contour-based proposals and random forest: A case study of national nepal,” Landslides, vol. 15, pp. 453–464, 2018

    2018

  32. [40]

    Landslide detection based on contour-based deep learning framework in case of national scale of nepal in 2015,

    Bo Yu, Fang Chen, and Chong Xu, “Landslide detection based on contour-based deep learning framework in case of national scale of nepal in 2015,” Computers & Geosciences , vol. 135, pp. 104388, 2020

    2015

  33. [41]

    Landslide detection from open satellite imagery using distant domain transfer learning,

    Shengwu Qin, Xu Guo, Jingbo Sun, Shuangshuang Qiao, Lingshuai Zhang, Jingyu Yao, Qiushi Cheng, and Yanqing Zhang, “Landslide detection from open satellite imagery using distant domain transfer learning,” Remote sensing, vol. 13, no. 17, pp. 3383, 2021

    2021

  34. [42]

    Landslide detection for remote sensing images using a multi-label classification network based on bijie landslide dataset,

    Yongxin Li, Zhihui Xin, Guisheng Liao, Penghui Huang, and Mengt- ing Yuan, “Landslide detection for remote sensing images using a multi-label classification network based on bijie landslide dataset,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sens...

    2024

  35. [43]

    Landslide4sense: Reference benchmark data and deep learning models for landslide detection,

    Omid Ghorbanzadeh, Yonghao Xu, Pedram Ghamisi, Michael Kopp, and David Kreil, “Landslide4sense: Reference benchmark data and deep learning models for landslide detection,” IEEE Transactions on Geoscience and Remote Sensing , vol. 60, pp. 1–17, 2022

    2022

  36. [44]

    Adam: A method for stochastic optimization,

    P. K. Diederik and B. Jimmy, “Adam: A method for stochastic optimization,” CoRR, vol. abs/1412.6980, 2015

    2015 arXiv

  37. [45]

    Cutmix: Regularization strategy to train strong classifiers with localizable features,

    Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Chun, Junsuk Choe, and Youngjoon Yoo, “Cutmix: Regularization strategy to train strong classifiers with localizable features,” in Proceedings of the IEEE/CVF international conference on computer vision , 2019, pp. 6023–6032

    2019

  38. [46]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift,

    Sergey I. and Christian S., “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in Proc. ICML, 2015, pp. 448–456

    2015

  39. [47]

    Rectifier nonlinearities improve neural network acoustic models,

    Andrew L Maas, Awni Y Hannun, Andrew Y Ng, et al., “Rectifier nonlinearities improve neural network acoustic models,” in Proc. ICML, 2013, vol. 30, p. 3

    2013

  40. [48]

    Focal loss for dense object detection,

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Doll´ar, “Focal loss for dense object detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 42, no. 2, pp. 318–327, 2020

    2020

  41. [49]

    A comprehensive survey of loss functions in machine learning,

    Qi Wang, Yue Ma, Kun Zhao, and Yingjie Tian, “A comprehensive survey of loss functions in machine learning,” Annals of Data Science, pp. 1–26, 2020

    2020

  42. [50]

    Generalized intersection over union: A metric and a loss for bounding box regression,

    Hamid Rezatofighi, Nathan Tsoi, JunYoung Gwak, Amir Sadeghian, Ian Reid, and Silvio Savarese, “Generalized intersection over union: A metric and a loss for bounding box regression,” in Proc. CVPR, 2019, pp. 658–666

    2019

  43. [51]

    Tversky loss function for image segmentation using 3d fully convolu- tional deep networks,

    Seyed Sadegh Mohseni Salehi, Deniz Erdogmus, and Ali Gholipour, “Tversky loss function for image segmentation using 3d fully convolu- tional deep networks,” in International workshop on machine learning in medical imaging . Springer, 2017, pp. 379–387

    2017

  44. [52]

    The lov ´asz hinge: A novel convex surrogate for submodular losses,

    Jiaqian Yu and Matthew B Blaschko, “The lov ´asz hinge: A novel convex surrogate for submodular losses,” IEEE transactions on pattern analysis and machine intelligence , vol. 42, no. 3, pp. 735–748, 2018

    2018

  45. [53]

    Boundary loss for remote sensing imagery semantic segmentation,

    Alexey Bokhovkin and Evgeny Burnaev, “Boundary loss for remote sensing imagery semantic segmentation,” in International Symposium on Neural Networks , 2019, pp. 388–401

    2019

  46. [54]

    A compre- hensive study on center loss for deep face recognition,

    Yandong Wen, Kaipeng Zhang, Zhifeng Li, and Yu Qiao, “A compre- hensive study on center loss for deep face recognition,” International Journal of Computer Vision , vol. 127, pp. 668–683, 2019

    2019

  47. [55]

    A simple method to improve the quality of ndvi time-series data by integrating spatiotemporal information with the savitzky-golay filter,

    Ruyin Cao, Yang Chen, Miaogen Shen, Jin Chen, Ji Zhou, Cong Wang, and Wei Yang, “A simple method to improve the quality of ndvi time-series data by integrating spatiotemporal information with the savitzky-golay filter,” Remote Sensing of Environment , vol. 217, pp. 244–257, 2018

    2018

  48. [56]

    Rethinking atrous convolution for semantic image segmentation,

    Liang-Chieh Chen, “Rethinking atrous convolution for semantic image segmentation,” arXiv preprint arXiv:1706.05587 , 2017

    2017 arXiv

  49. [57]

    Mobilenetv3 for image classification,

    Siying Qian, Chenran Ning, and Yuepeng Hu, “Mobilenetv3 for image classification,” in 2021 IEEE 2nd International Conference on Big Data, Artificial Intelligence and Internet of Things Engineering (ICBAIE), 2021, pp. 490–497

    2021

  50. [58]

    Efficientnet,

    Brett Koonce and Brett Koonce, “Efficientnet,” Convolutional neural networks with swift for Tensorflow: image recognition and dataset categorization, pp. 109–123, 2021

    2021

  51. [59]

    Squeeze-and-excitation networks,

    Jie Hu, Li Shen, and Gang Sun, “Squeeze-and-excitation networks,” in Proc. CVPR, June 2018

    2018

  52. [60]

    Cbam: Convolutional block attention module,

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon, “Cbam: Convolutional block attention module,” in Proc. ECCV, 2018, pp. 3–19

    2018

  53. [61]

    Landslide detection and segmentation using remote sensing images and deep neural networks,

    Cam Le, Lam Pham, Jasmin Lampert, Matthias Schl ¨ogl, and Alexan- der Schindler, “Landslide detection and segmentation using remote sensing images and deep neural networks,” in IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium, 2024, pp. 9582–9586

    2024

  54. [62]

    At- tention is all you need,

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin, “At- tention is all you need,” Advances in neural information processing systems, vol. 30, 2017

    2017

Pith tools

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