Pith. sign in

REVIEW 2 major objections 2 minor 21 references

An eight-channel subset selected via floating search matches full multispectral landslide segmentation performance.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-30 22:24 UTC pith:ANTH6KUT

load-bearing objection The paper applies SFFS via a lightweight proxy to pick an 8-channel subset for Landslide4Sense segmentation that claims F1 parity with larger inputs, but offers no evidence the selection transfers to the full model. the 2 major comments →

arxiv 2605.09746 v2 pith:ANTH6KUT submitted 2026-05-10 cs.LG cs.AI

Sequential Feature Selection for Efficient Landslide Segmentation from Multi-Spectral Data

Arsalaan Ahmad , Oktay Karakus , Paul L. Rosin This is my paper
classification cs.LG cs.AI
keywords landslide detectionfeature selectionmultispectral imagerysemantic segmentationSentinel-2SFFSEarth observation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper tests whether most of the 30-plus spectral and topographic channels used in current landslide models are redundant. It applies sequential forward floating selection inside a lightweight proxy network to build and prune feature sets on the Landslide4Sense benchmark. The process yields a stable eight-channel combination drawn from Sentinel-2 bands, ALOS terrain data, and a few engineered indices. This compact set reaches or exceeds the F1 score of the original high-dimensional inputs. The selection sequence itself is then read as evidence for which physical cues the models actually use.

Core claim

Sequential Forward Floating Selection run on a lightweight U-Net++ proxy identifies an eight-channel input set that equals or surpasses the segmentation F1 obtained from any combination of up to thirty channels on the Landslide4Sense benchmark. The same procedure supplies an ordered list of which spectral and topographic features are retained or discarded, thereby exposing the physical signals that drive landslide predictions rather than treating channel choice as a black-box hyperparameter.

What carries the argument

Sequential Forward Floating Selection (SFFS) that iteratively adds then conditionally removes channels while monitoring proxy-model F1.

Load-bearing premise

The ranking found by the lightweight proxy will remain valid when the same eight channels are fed to a full-scale segmentation network.

What would settle it

Train a standard U-Net++ or similar full model on the reported eight channels versus the full thirty-channel stack and measure whether the F1 gap stays within one percent on a held-out Landslide4Sense split.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Model training and inference cost drop because only eight input maps are processed instead of thirty.
  • Physical interpretability increases because the retained channels correspond to measurable surface properties rather than opaque correlations.
  • The Hughes phenomenon is avoided by removing redundant or noisy channels that can degrade performance.
  • Input design for other Earth-observation segmentation tasks can follow the same SFFS protocol instead of using every available band.

Where Pith is reading between the lines

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

  • The method could be rerun on newer Sentinel-2 collections or different landslide inventories to test whether the same eight channels remain optimal.
  • If the eight-channel set generalizes across regions, future satellite missions might prioritize those exact bands for landslide monitoring.
  • The ordering produced by SFFS supplies a quantitative basis for deciding which engineered indices are worth computing in operational pipelines.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The manuscript proposes using Sequential Forward Floating Selection (SFFS) with a lightweight U-Net++ proxy to identify a compact 8-channel subset from Sentinel-2 multispectral bands, ALOS PALSAR terrain data, and 16 engineered spectral/structural indices for landslide segmentation on the Landslide4Sense benchmark. It claims this subset matches or exceeds the F1 performance of models using up to 30 channels while also yielding physical insights into feature reliance, contrasting with single-band ablation or full-input practices.

Significance. If the central transfer claim holds, the work would offer a principled, interaction-aware alternative to ad-hoc high-dimensional inputs in remote-sensing segmentation, directly addressing the Hughes phenomenon and computational overhead. The methodological preference for SFFS over isolated drop tests to capture interactions is a clear strength.

major comments (2)
  1. [Abstract] Abstract: the headline claim that an 8-channel subset 'matches or exceeds the segmentation F1 of configurations using up to 30 channels' is presented without any numerical F1 values, baselines, error bars, or explicit statement that the selected subset was retrained and scored inside the target full-scale architecture rather than only the proxy.
  2. [Methods] Methods (SFFS and proxy description): no ablation, ranking-correlation table, or transfer experiment is described that would confirm the proxy-derived subset ordering is preserved when the identical 8-channel set is trained and evaluated in the full-capacity model; because the proxy has reduced capacity and altered skip connections, this verification is load-bearing for the transferability assertion.
minor comments (2)
  1. [Abstract] Abstract and results sections would be strengthened by reporting concrete F1 scores, standard deviations across folds or seeds, and at least one standard feature-selection baseline (e.g., mutual information or recursive elimination).
  2. [Introduction / Data] Notation for the engineered indices and the precise definition of the 30-channel pool should be tabulated early to improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive critique. The two major comments correctly identify gaps in the current presentation; we will revise the manuscript to address both.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the headline claim that an 8-channel subset 'matches or exceeds the segmentation F1 of configurations using up to 30 channels' is presented without any numerical F1 values, baselines, error bars, or explicit statement that the selected subset was retrained and scored inside the target full-scale architecture rather than only the proxy.

    Authors: We agree. The abstract currently states the claim qualitatively. The full manuscript reports the relevant F1 scores (with standard deviations across folds) in the results section and states that the final 8-channel model was trained and evaluated in the target architecture. We will move the key numerical values and an explicit statement of the evaluation protocol into the abstract. revision: yes

  2. Referee: [Methods] Methods (SFFS and proxy description): no ablation, ranking-correlation table, or transfer experiment is described that would confirm the proxy-derived subset ordering is preserved when the identical 8-channel set is trained and evaluated in the full-capacity model; because the proxy has reduced capacity and altered skip connections, this verification is load-bearing for the transferability assertion.

    Authors: The observation is accurate: the manuscript does not contain a dedicated transfer experiment or correlation table between proxy and full-model rankings. We will add (i) a direct comparison of the 8-channel subset performance when trained in the full-capacity model versus the proxy, (ii) a table showing rank correlation of feature importance between the two architectures, and (iii) a brief discussion of any discrepancies. These additions will be placed in a new subsection of the methods/results. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical SFFS selection is independent of inputs

full rationale

The paper's core claim—an 8-channel subset matching or exceeding F1 of up to 30-channel models—is obtained by running Sequential Forward Floating Selection on the Landslide4Sense benchmark using a lightweight U-Net++ proxy. This is a standard algorithmic search over feature subsets evaluated by direct model training and scoring; the selected subset and its performance are not defined in terms of themselves, nor derived from any fitted parameter that is then renamed as a prediction. No equations appear in the provided text, no self-citations are invoked to justify uniqueness or an ansatz, and the result is benchmarked against external configurations rather than reducing to prior author work. The derivation chain is therefore self-contained against the empirical data.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; full text would be needed to audit model hyperparameters, data assumptions, or proxy-model choices.

pith-pipeline@v0.9.1-grok · 5765 in / 968 out tokens · 21978 ms · 2026-06-30T22:24:21.971261+00:00 · methodology

0 comments
read the original abstract

Landslide detection from satellite imagery has advanced through deep learning, yet most models rely on large, highly correlated spectral-topographic inputs whose contributions remain poorly understood. The question of which channels are actually necessary has received surprisingly little attention. This matters: redundant or correlated inputs obscure physical interpretability, inflate computational overhead, and can actively degrade model performance through the Hughes Phenomenon. We present a systematic, explainable channel-selection framework for the Landslide4Sense benchmark, combining Sentinel-2 multispectral and ALOS PALSAR terrain data with 16 engineered spectral and structural indices. Rather than relying on conventional single-band drop tests, which evaluate channels in isolation and miss interaction effects, we apply Sequential Forward Floating Selection (SFFS) to iteratively build and prune a candidate feature pool using a lightweight U-Net++ proxy model. Beyond identifying a compact 8-channel subset that matches or exceeds the segmentation F1 of configurations using up to 30 channels, we use the selection process itself to interrogate which spectral and topographic features landslide models genuinely rely on, and what this reveals about the physical cues driving their predictions. We argue that SFFS represents a principled feature selection approach to input design in Earth observation, in contrast to the prevailing practice of appending every available band and hoping the model learns what to ignore.

Figures

Figures reproduced from arXiv: 2605.09746 by Arsalaan Ahmad, Oktay Karakus, Paul L. Rosin.

Figure 1
Figure 1. Figure 1: Illustration of the Sequential Forward Floating Selec [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Channel selection results. (a) SFFS-selected subset [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Review history (2 revisions) →

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

21 extracted references · 6 canonical work pages · 3 internal anchors

  1. [1]

    Sentinel-2: ESA’s optical high-resolution mission for GMES operational services,

    M. Drusch, U. Del Bello, S. Carlier, O. Colin, V . Fernandez, F. Gascon, B. Hoersch, C. Isola, P. Laberinti, P. Martimort, A. Meygret, F. Spoto, O. Sy, F. Marchese, and P. Bargellini, “Sentinel-2: ESA’s optical high-resolution mission for GMES operational services,”Remote Sensing of Environment, vol. 120, pp. 25–36, 2012, the Sentinel Missions - New Oppor...

    2012

  2. [2]

    U-Net: Convolutional net- works for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional net- works for biomedical image segmentation,” inMedical Image Comput- ing and Computer-Assisted Intervention – MICCAI 2015, N. Navab, J. Hornegger, W. M. Wells, and A. F. Frangi, Eds. Cham: Springer International Publishing, 2015, pp. 234–241

    2015

  3. [3]

    A standardized catalogue of spectral indices to advance the use of remote sensing in Earth system research,

    D. Montero, C. Aybar, M. D. Mahecha, F. Martinuzzi, M. Söchting, and S. Wieneke, “A standardized catalogue of spectral indices to advance the use of remote sensing in Earth system research,”Scientific Data, vol. 10, no. 197, 2023

    2023

  4. [4]

    RMAU-NET: A residual- multihead-attention U-Net architecture for landslide segmentation and detection from remote sensing images,

    L. Pham, C. Le, H. Tang, K. Truong, T. Nguyen, J. Lampert, A. Schindler, M. Boyer, and S. Phan, “RMAU-NET: A residual- multihead-attention U-Net architecture for landslide segmentation and detection from remote sensing images,” 2025. [Online]. Available: https://arxiv.org/abs/2507.11143

    work page arXiv 2025

  5. [5]

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

    C. Le, L. Pham, J. Lampert, M. Schlögl, and A. Schindler, “Landslide detection and segmentation using remote sensing images and deep neural network,” 2023. [Online]. Available: https://arxiv.org/abs/2312.16717

    work page arXiv 2023

  6. [6]

    On the mean accuracy of statistical pattern recognizers,

    G. Hughes, “On the mean accuracy of statistical pattern recognizers,” IEEE Transactions on Information Theory, vol. 14, no. 1, pp. 55–63, 1968

    1968

  7. [7]

    Feature selection for classification of hyperspectral data by SVM,

    M. Pal and G. M. Foody, “Feature selection for classification of hyperspectral data by SVM,”IEEE Transactions on Geoscience and Remote Sensing, vol. 48, no. 5, pp. 2297–2307, 2010

    2010

  8. [8]

    An introduction to variable and feature selection,

    I. Guyon and A. Elisseeff, “An introduction to variable and feature selection,”Journal of machine learning research, vol. 3, no. Mar, pp. 1157–1182, 2003

    2003

  9. [9]

    Landslide4Sense: Reference benchmark data and deep learning models for landslide detection,

    O. Ghorbanzadeh, Y . Xu, P. Ghamisi, M. Kopp, and D. Kreil, “Landslide4Sense: Reference benchmark data and deep learning models for landslide detection,”IEEE Transactions on Geoscience and Remote Sensing, vol. 60, p. 1–17, 2022. [Online]. Available: http://dx.doi.org/10.1109/TGRS.2022.3215209

    work page doi:10.1109/tgrs.2022.3215209 2022

  10. [10]

    Floating search methods in feature-selection,

    P. Pudil, J. Novovicova, and J. Kittler, “Floating search methods in feature-selection,”Pattern Recognition Letters, vol. 15, no. 11, pp. 1119– 1125, 1994

    1994

  11. [11]

    UNet++: A Nested U-Net Architecture for Medical Image Segmentation

    Z. Zhou, M. M. R. Siddiquee, N. Tajbakhsh, and J. Liang, “UNet++: A nested U-Net architecture for medical image segmentation,” 2018. [Online]. Available: https://arxiv.org/abs/1807.10165

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  12. [12]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778

    2016

  13. [13]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,”

  14. [14]

    Adam: A Method for Stochastic Optimization

    [Online]. Available: https://arxiv.org/abs/1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv

  15. [15]

    Focal Loss for Dense Object Detection

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” 2018. [Online]. Available: https: //arxiv.org/abs/1708.02002

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  16. [16]

    Generalised Dice overlap as a deep learning loss function for highly un- balanced segmentations,

    C. H. Sudre, W. Li, T. Vercauteren, S. Ourselin, and M. Jorge Cardoso, “Generalised Dice overlap as a deep learning loss function for highly un- balanced segmentations,” inInternational Workshop on Deep Learning in Medical Image Analysis. Springer, 2017, pp. 240–248

    2017

  17. [17]

    Random forests,

    L. Breiman, “Random forests,”Machine learning, vol. 45, no. 1, pp. 5–32, 2001

    2001

  18. [18]

    PyTorch: An imperative style, high-performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antigaet al., “PyTorch: An imperative style, high-performance deep learning library,”Advances in neural information processing systems, vol. 32, 2019

    2019

  19. [19]

    Evaluation of Sentinel-2 red-edge bands for empirical estimation of green lai and chlorophyll content,

    J. Delegido, J. Verrelst, L. Alonso, and J. Moreno, “Evaluation of Sentinel-2 red-edge bands for empirical estimation of green lai and chlorophyll content,”Sensors, vol. 11, no. 7, pp. 7063–7081, 2011

    2011

  20. [20]

    Deep learning in remote sensing applications: A meta-analysis and review,

    L. Ma, Y . Liu, X. Zhang, Y . Ye, G. Yin, and B. A. Johnson, “Deep learning in remote sensing applications: A meta-analysis and review,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 152, pp. 166–177, 2019. [Online]. Available: https://www.sciencedirect.com/ science/article/pii/S0924271619301108

    2019

  21. [21]

    Dimensionality reduc- tion and classification of hyperspectral remote sensing image feature extraction,

    H. Li, J. Cui, X. Zhang, Y . Han, and L. Cao, “Dimensionality reduc- tion and classification of hyperspectral remote sensing image feature extraction,”Remote Sensing, vol. 14, no. 18, p. 4579, 2022

    2022