Pith. sign in

REVIEW 4 major objections 5 minor 66 references

Hyperspectral Calibration Detection: A Novel Concept For Change Detection With Unsupervised Incremental Safe Pseudo-Labeling Implementation

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read HyperLUCID claims fully unsupervised hyperspectral change detection reaches 93.6–97.9% overall accuracy by iteratively calibrating spectra with safe pseudo-labels.

desk verdict Genuinely new unsupervised calibration-based change detection with solid experiments and released code, but the SOTA claim is undercut by QUEEN-G on two datasets and the final threshold was fitted to the test sets. read the letter →

arxiv 2608.06028 v1 pith:P2HH4EPB submitted 2026-08-06 eess.IV

classification eess.IV
keywords hyperspectralchangedetectionunsupervisedsafepseudo-labelingspectralanglemapperspectrumcalibrationresiduallearninglightweightCNNonboardedgecomputing
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 is trying to establish that change detection between two hyperspectral images of the same scene can be solved with zero ground-truth labels. Its central idea, called hyperspectral calibration detection (HyperCAD), is to learn a calibration function that removes differences in acquisition conditions so that unchanged pixels become indistinguishable across the two dates while changed pixels stay distinguishable. The proposed unsupervised algorithm, HyperLUCID, iteratively selects a small safe set of supposedly unchanged pixels, trains a lightweight network to calibrate the first image onto the second, and repeats while gradually enlarging the safe set. The authors report state-of-the-art accuracy on five real benchmark datasets and argue that the method is fast enough for onboard edge computing, which would matter because labels are often unavailable for newly acquired satellite images.

What carries the argument

The load-bearing machinery is the iteratively augmented safe index set $\Omega_t$ together with the lightweight residual calibration network $f$. At each iteration, $\Omega_t$ is grown by adding the $R_{\mathrm{safe}} = 20\%$ of pixel pairs with the smallest spectral angle mapper (SAM) distance between the current calibrated image and the second image, then taking a union with the previous set to guarantee monotone growth. The network is trained with an L1 masked loss over the safe pairs, which the paper argues is robust to a small number of contaminated changed pixels in the safe set. The residual skip connection lets the network learn only the spectral adjustment rather than the full spectrum, and the group-convolution design keeps the model small. The SAM distance is the key geometric invariant: it is insensitive to multiplicative illumination scaling, which is why low-SAM pixels can be trusted as unchanged at the start.

What would settle it

Construct a synthetic bitemporal hyperspectral pair with a known changed-area ratio above 50%, inject per-pixel multiplicative illumination changes, and run HyperLUCID while measuring the purity of the safe set at each iteration; if the correctly identified unchanged-pixel ratio in $\Omega_t$ falls well below the paper's reported 96–99% and the final overall accuracy drops correspondingly, the central assumption would be falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that a simple, fully unsupervised loop can learn the calibration function that the HyperCAD concept requires. Starting from the 20% of pixel pairs with the smallest spectral angle (SAM), the algorithm trains a small residual group-convolution network to map the first-time spectrum to the second-time spectrum on those safe samples, recalibrates the entire first image, and repeats until the SAM distribution stabilizes. The paper reports that this safe set remains highly pure, with correctly identified unchanged-pixel ratios above 96% on all five datasets, and that the learned calibration function satisfies Property 1 in practice: unchanged pixels become spectrally close after calibration while changed pixels remain distinct. The final change map is then simply a threshold of the SAM values between the calibrated first image and the second image, using one fixed threshold of 0.2. The authors show that this zero-label method achieves 93.6% to 97.9% overall accuracy, frequently outperforming semi-supervised methods, and is one to two orders of magnitude faster than most deep-learning baselines.

Load-bearing premise

The algorithm assumes that the pixels with the smallest spectral angle between the two images are overwhelmingly unchanged locations; if the changed area is large or if lighting changes alter spectral shape rather than just magnitude, this safe seed set can be contaminated and the whole calibration loop can drift.

Editorial extensions

If this is right

  • If HyperLUCID works as claimed, zero-label hyperspectral change detection becomes practical for onboard edge computing, returning change maps in seconds rather than minutes or hours.
  • Because the method outperforms most semi-supervised baselines without any labels, the paper implies that careful radiometric calibration can substitute for supervision in benchmark change-detection scenarios.
  • The unified hyperparameter setting means the same fixed configuration works across five diverse datasets, removing the need for per-dataset tuning during deployment.
  • The lightweight residual architecture is shown to be sufficient; replacing it with a more complex LSTM module degrades accuracy and increases runtime by a factor of five or more.
  • Deployment on a resource-constrained embedded platform (NVIDIA Jetson Orin Nano) preserves detection quality and keeps peak memory far below the 8GB limit, supporting real-time onboard use.

Reading between the lines

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

  • Beyond the paper, the same safe pseudo-label calibration loop could plausibly transfer to multispectral or RGB bitemporal change detection, since SAM and the residual calibration idea do not require hyperspectral band counts.
  • A concrete stress test the paper does not run: increase the changed-area ratio beyond the roughly 20% safe-selection fraction and monitor whether the safe set becomes contaminated and the calibration function drifts, which would identify a practical operating envelope.
  • The learned calibration function itself could be reused as a form of automatic radiometric normalization between different sensors or acquisition dates, potentially benefiting other multitemporal analysis tasks beyond change detection.
  • The paper's sensitivity to the final SAM threshold $\eta_2 = 0.2$ suggests that a scene-adaptive threshold derived from the calibrated SAM distribution, rather than a fixed value, might further improve robustness on unusual scenes.
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 / 5 minor

Summary. The paper introduces HyperLUCID, a fully unsupervised hyperspectral change detection (HCD) algorithm. The method iteratively builds a safe pseudo-label set of unchanged pixels by selecting the 20% lowest spectral-angle-mapper (SAM) pixels between the bitemporal images, monotonically growing this set, and training a lightweight residual calibration network f on the masked L1 loss. After convergence, the final change map is obtained by thresholding the SAM values between the calibrated first image and the second image. The authors claim state-of-the-art accuracy (93.6%–97.9% OA) on five benchmark HCD datasets, outperform most semi-supervised methods, and are 1–2 orders of magnitude faster than deep-learning baselines. The paper also includes ablations, hyperparameter discussions, and an edge-device deployment on NVIDIA Jetson Orin Nano.

Significance. The HyperCAD concept—learning a spectrum calibration function from automatically selected unchanged pixels rather than directly comparing bitemporal spectra—is a novel and potentially useful direction for zero-label HCD. The empirical study is extensive: five real datasets, eight baseline methods, five metrics, ablations (residual connection, SAM vs. Euclidean distance, CNN vs. LSTM), and a raw edge-device profiling appendix. The source code is publicly released, and the reproducibility details (epoch schedule, warm starts, ten Monte Carlo trials) are commendable. If the evaluation protocol's threshold-selection circularity is resolved and the robustness of the low-SAM safe-set assumption is demonstrated, the method would be a strong practical contribution; as presented, however, the unsupervised and state-of-the-art claims are not yet fully supported.

major comments (4)
  1. [§III-E, Fig. 16(b), Algorithm 1 (Line 10)] The final threshold η2 = 0.2 is selected in Fig. 16(b) by evaluating overall accuracy on the five benchmark test sets, and the same η2 is then used in Line 10 of Algorithm 1 to produce the change maps whose OA values appear in Table I. This means the reported accuracies are, in part, the result of fitting a parameter to the test labels, which is incompatible with the claim of a zero-label, user-friendly method. Please either choose η2 without using test labels (e.g., an internal validation split or a label-free stability criterion) or present the benchmark results as an explicit threshold sweep with the OA-versus-η2 curves, clearly stating that the published numbers are the envelope. The same concern applies to the choice of Rsafe = 20%, justified in Section III-E using Figure 15 with test-set OA.
  2. [§II-B, Eq. (4), Property 2, Algorithm 1 (Lines 4–5)] The robustness claim in Property 2 is not sufficient for the central claim. The reformulation in Eq. (6) shows sparsity of the concatenated residual vector, which is an elementwise property and not a sample-wise outlier-rejection guarantee. Because Algorithm 1 grows Ωt monotonically by union (Lines 4–5), any changed pixel that enters the safe set is never removed, and the masked L1 loss (4) trains f_{t+1} to map that changed pixel toward its y_l, further lowering its calibrated SAM and reinforcing its appearance as unchanged. The paper's conclusion explicitly states that the low-SAM assumption 'is currently addressed only by Property 2.' Please add a quantitative contamination experiment: inject controlled fractions of changed pixels into Ω1, measure the resulting recall and the SAM separation between changed and unchanged pixels after calibration, and either strengthen the robustness argument or modify the algorithm to reject sample-wise outliers.
  3. [Section II-B, below Eq. (5); Figure 15] The premise that the 20% lowest-SAM pixels are overwhelmingly unchanged is load-bearing and under-tested. Figure 15 reports R_ciup above 96% on the benchmark datasets, but R_ciup is computed using ground-truth labels that the unsupervised algorithm does not have. The manuscript does not examine the failure regime where this assumption breaks: large changed-area ratios, spectral-shape changes beyond multiplicative illumination, or misregistration. Please add synthetic or real stress tests with varying changed-area ratios and spectral perturbations, and report whether the monotone union of Algorithm 1 still converges to a useful Ω* in those cases.
  4. [Table I; Abstract] The abstract's 'state-of-the-art results' claim is stronger than Table I supports. QUEEN-G achieves a higher OA on Hermiston (0.986 vs. 0.979) and ties on Yancheng (0.969); on Jiangsu, HyperLUCID's recall is 0.633, lower than several deep baselines, which is consistent with the concern in Major Comment 2 that subtle changes may be calibrated away. Please either restrict the claim to 'state-of-the-art among the compared unsupervised methods' or provide a statistical significance test (e.g., a paired McNemar test) and explicitly discuss the cases where the method is not superior.
minor comments (5)
  1. [§III-C, Figures 10–14] The edge-device runs shown in panels (j) are only qualitative; please include the quantitative metrics (OA, κ, F1) for the NVIDIA Orin Nano runs or state clearly that the visual comparison is illustrative.
  2. [§III-C, Table I] The phrase 'around 1 to 2 orders of magnitude faster than most benchmark HCD methods' should not include non-learned baselines such as CVA, which is faster than HyperLUCID on Yancheng (0.044 s vs. 3.576 s); 'deep-learning-based methods' would be accurate.
  3. [Algorithm 1, Line 9] The stopping criterion refers to 'the difference of standard deviation of the SAM values' but does not specify whether this is computed over Ωt, over the whole image, or over the newly selected candidates; please clarify.
  4. [Section II-B, below Eq. (5)] The assertion that 'most pixels in real benchmark HCD datasets are unchanged ones' is an empirical assumption; please report the changed-area ratios for each of the five datasets so the reader can assess the applicability of the 20% safe-sample rule.
  5. [Fig. 16(a)] The text says performance is not sensitive to η1, but the Hermiston curve shows a substantial drop at η1 = 0.005; please add a one-sentence explanation or revise the claim.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported 93.6-97.9% OA is partly a test-set-tuned result: the final SAM threshold η2=0.2 (and R_safe, η1) are selected from OA curves on the same five benchmark GTs used for evaluation; the unsupervised calibration core itself is not circular.

  1. fitted input called prediction [Algorithm 1 lines 1 and 10; Section III-E and Figure 16(b)]
    "Third, as for the threshold SAM value η2 := 0.2 for deciding the final detection map C⋆, some of the best hyperspectral analysis algorithms could achieve SAM values of around 0.2 to 2 degrees [60, Table III]. Thus, η2 := 0.2 is a safe setting to argue that a pixel is unchanged; both a smaller setting of η2 := 0.1 or a larger setting of η2 := 0.3 lead to obvious performance degradation as can be seen from Figure 16(b)."

    Algorithm 1's final change map is C⋆_ℓ = J SAM(x^t_ℓ, y_ℓ) > η2 K (line 10), with η2 fixed at line 1. Section III-E chooses η2 := 0.2 by comparing OA under η2 = 0.1/0.2/0.3 in Figure 16(b), which is plotted on the same five GT-labeled benchmark datasets whose OA is then reported as the headline result. Hence the final decision threshold is fitted to the evaluation ground truth, so the reported 93.6-97.9% OA is not a cleanly predicted quantity: it includes a hyperparameter selected to maximize OA on those exact test sets. R_safe := 20% and η1 := 0.05 are selected in the same benchmark-driven way (Figures 15 and 16(a)), compounding the leakage.

full rationale

The one genuine circular step is the evaluation-set selection of the decision threshold η2 and, to a lesser extent, R_safe and η1. Because the paper reports OA, κ, and F1 on the same benchmark datasets used to choose these hyperparameters, the headline numbers are partially fitted to the test ground truths rather than being independent predictions. This falls under the fitted-input-called-prediction pattern and justifies a score of 6. No other circularity pattern is load-bearing. The low-SAM-pixels-are-unchanged premise (Section II-B) is an assumption, not a tautology: it is empirically checked via R_ciup in Figure 15, and the paper's own conclusion concedes that the assumption is "currently addressed only by Property 2," with drastic seasonal changes and large changed-area ratios left unaddressed. A changed pixel entering the monotonically grown set Ω_t may indeed be calibrated away, but this is a real limitation rather than a circular derivation. Property 2's L1-robustness argument is informal because it imports a sparse-solution property from the L1 literature and assumes Property 1's behavior, yet it is not the source of the reported accuracy and is explicitly flagged as the only safeguard. Self-citation is not load-bearing: the gradually-augmented selection idea is credited to external work [43], and the benchmark comparisons use published table values rather than a self-citation chain. The calibration network itself is trained without ground truth on pseudo-labels, so the unsupervised algorithm has independent content; the circularity is confined to the hyperparameter selection used in reporting final performance.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central algorithm depends on three hand-set thresholds (R_safe, eta_1, eta_2) that are tuned on the benchmark test sets, and on domain assumptions about SAM purity and the existence of an ideal calibration function. No new physical entities are postulated.

free parameters (4)
  • R_safe (safe sample ratio) = 20%
    Selected empirically in Section III-E and Figure 15: 30% introduces more outliers, 10% slows convergence and lowers OA by about 0.5% on Santa Barbara.
  • eta_1 (SAM std stopping threshold) = 0.05
    Chosen from sensitivity analysis in Figure 16(a); values from 0.005 to 0.5 affect OA, with 0.05 giving generally strong results.
  • eta_2 (final SAM change threshold) = 0.2
    Picked by inspecting overall accuracy across benchmark test sets in Figure 16(b); the paper acknowledges sensitivity to eta_2.
  • Epoch schedule 100*0.7^t with warm starts = heuristic
    Training schedule is an empirical choice that affects convergence speed and final accuracy.
assumptions (5)
  • domain assumption Pixels with the lowest SAM values between X and Y are mostly unchanged pixels.
    Underlies the safe pseudo-label selection in Algorithm 1 Line 4; validated empirically in Figure 15 but not guaranteed when changed-area ratio is large.
  • domain assumption An ideal calibration function f exists satisfying Property 1, preserving changed pixels while aligning unchanged ones.
    Defined in Property 1 and Section II-A; the entire HyperCAD concept assumes such a function can be learned from the data.
  • ad hoc to paper Minimizing the L1 masked loss (4) makes training robust to the small fraction of changed pixels that leak into the safe set.
    Property 2 states this as a formal result but the argument relies on sparsity heuristics of L1 norms; no formal guarantee is given for this specific network and data.
  • standard math Monotone increasing finite index sets converge.
    Used in Section II-B to justify the stopping criterion; standard result from Apostol.
  • domain assumption Most pixels in benchmark HCD datasets are unchanged, so the initial 20% safe ratio is reasonable.
    Stated in Section II-B after Equation (5) and used to justify R_safe=20%.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hyperspectral Calibration Detection: A Novel Concept For Change Detection With Unsupervised Incremental Safe Pseudo-Labeling Implementation." pith.science (2026). https://pith.science/paper/P2HH4EPB

@misc{pith2026260806028,
  author       = {Pith},
  title        = {Pith review of: Hyperspectral Calibration Detection: A Novel Concept For Change Detection With Unsupervised Incremental Safe Pseudo-Labeling Implementation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P2HH4EPB}},
  note         = {Machine review of arXiv:2608.06028}
}
read the original abstract

Hyperspectral change detection (HCD) has found numerous key applications, such as land cover monitoring. The majority of benchmark HCD algorithms are semi-supervised methods, and some of them can even achieve very low sample labeling rates. However, in some practical scenarios, such as those requiring immediate detection responses for onboard edge computing, we need to achieve the zero-label requirement as ground-truth labeling would not be available onboard for newly acquired images. In this work, we propose a fully unsupervised HCD algorithm, together with a lightweight model, quite suitable for onboard detection missions. Based on an iteratively augmented training set that safely collects some unchanged pixel samples, we learn an iteratively refined spectrum calibration function that eventually compensates the variability of acquisition conditions (often observed in bitemporal images), thereby making the changed pixels easily detectable by analyzing the calibrated spectra. The proposed hyperspectral looping unsupervised calibration and incremental detection (HyperLUCID) algorithm is not only computationally efficient (around 1 to 2 orders of magnitude faster than most benchmark HCD methods), but has also achieved state-of-the-art results (around 93.6% to 97.9% overall accuracy) on several real benchmark HCD datasets. Source codes: https://github.com/IHCLab/HyperLUCID.

Figures

Figures reproduced from arXiv: 2608.06028 by the authors.

Figure 1
Figure 1. Comparison of representative HCD methods on the large-scale Santa [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Graphical illustration of the HyperCAD concept for HCD. Even for an [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 2
Figure 2. The aim of the ideal index set Ω⋆ (and Ωt ) is to identify some safe unchanged samples in a computationally efficient manner, as defined by the four conditions above, whose motivations will be illustrated below. Also, how the four conditions help efficiently obtain the ideal f defined in Property 1 will also be discussed below. To see the motivation of condition (A), note that an ideal f should satisfy Property 1. L… view at source ↗
Figures from the paper (14 more)
Figure 3
Figure 3. Figure 3: A lightweight simple network architecture [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 4
Figure 4. Figure 4: As indicated by the red/green/blue arrows, the unsupervised Hyper [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: False-color images (X, Y ) and the GT change map of the Hermiston dataset. (a) HSI image captured on May 1st, 2004. (b) HSI image captured on May 8th, 2007. (c) GT change map, where changed pixels are shown in white and unchanged pixels in black. (a) (b) (c) [PITH_FUL…
Figure 5
Figure 5. Figure 5: False-color images (X, Y ) and the GT change map of the Yancheng dataset. (a) HSI image captured on May 3rd, 2006. (b) HSI image captured on April 23rd, 2007. (c) GT change map, where changed pixels are shown in white and unchanged pixels in black. that all its paramet…
Figure 7
Figure 7. Figure 7: False-color images (X, Y ) and the GT change map of the Jiangsu dataset. (a) HSI image captured on May 3rd, 2013. (b) HSI image captured on December 31st, 2007. (c) GT change map, where changed pixels are shown in white and unchanged pixels in black. false-color images…
Figure 8
Figure 8. Figure 8: False-color images (X, Y ) and the GT change map of the Bay Area dataset. (a) HSI image captured in 2013. (b) HSI image captured in 2015. (c) GT change map, where changed, unchanged, and unknown pixels are represented in white, black, and grey, respectively. (a) (b) (c…
Figure 9
Figure 9. Figure 9: False-color images (X, Y ) and the GT change map of the Santa Barbara dataset. (a) HSI image captured in 2013. (b) HSI image captured in 2014. (c) GT change map, where changed, unchanged, and unknown pixels are represented in white, black, and grey, respectively. and b…
Figure 10
Figure 10. Figure 10: Change detection results on the Yancheng dataset, where FP and FN are marked in red and green, respectively. (a) HyperNet. (b) S [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]
Figure 11
Figure 11. Figure 11: Change detection results on the Hermiston dataset, where FP and FN are marked in red and green, respectively. (a) HyperNet. (b) S [PITH_FULL_IMAGE:figures/full_fig_p010_11.png]
Figure 12
Figure 12. Figure 12: Change detection results on the Jiangsu dataset, where FP and FN are marked in red and green, respectively. (a) HyperNet. (b) S [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]
Figure 13
Figure 13. Figure 13: Change detection results on the Bay Area dataset, where FP and FN are marked in red and green, respectively. (a) HyperNet. (b) S [PITH_FULL_IMAGE:figures/full_fig_p010_13.png]
Figure 14
Figure 14. Figure 14: Change detection results on the Santa Barbara dataset, where FP and FN are marked in red and green, respectively. (a) HyperNet. (b) S [PITH_FULL_IMAGE:figures/full_fig_p010_14.png]
Figure 16
Figure 16. Figure 16: Demonstration of the local optimality of the unified hyperparameter [PITH_FULL_IMAGE:figures/full_fig_p013_16.png]
Figure 15
Figure 15. Figure 15: For each benchmark testing data, we demonstrate the evolutions [PITH_FULL_IMAGE:figures/full_fig_p013_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 61 canonical work pages

  1. [1]

    UIU-Net: U-Net in U-Net for infrared small object detection,

    X. Wu, D. Hong, and J. Chanussot, “UIU-Net: U-Net in U-Net for infrared small object detection,”IEEE Trans. Image Process., vol. 32, pp. 364–376, Dec. 2022

  2. [2]

    Nonlocal patch tensor sparse representation for hyperspectral image super-resolution,

    Y . Xu, Z. Wu, J. Chanussot, and Z. Wei, “Nonlocal patch tensor sparse representation for hyperspectral image super-resolution,”IEEE Trans. Image Process., vol. 28, pp. 3034–3047, Jan. 2019

  3. [3]

    An augmented linear mixing model to address spectral variability for hyperspectral unmixing,

    D. Hong, N. Yokoya, J. Chanussot, and X. X. Zhu, “An augmented linear mixing model to address spectral variability for hyperspectral unmixing,” IEEE Trans. Image Process., vol. 28, pp. 1923–1938, Nov. 2019

  4. [4]

    A generalized non-convex surrogated framework for anomaly detection on blurred hyperspectral images,

    Y . Wang, W. Li, Y . Gui, H. Xie, and L. Zhang, “A generalized non-convex surrogated framework for anomaly detection on blurred hyperspectral images,”IEEE Trans. Image Process., 2025

  5. [5]

    Flex-DLD: Deep low- rank decomposition model with flexible priors for hyperspectral image denoising and restoration,

    Y . Chen, H. Zhang, Y . Wang, Y . Yang, and J. Wu, “Flex-DLD: Deep low- rank decomposition model with flexible priors for hyperspectral image denoising and restoration,”IEEE Trans. Image Process., vol. 33, pp. 1211–1226, Feb. 2024

  6. [6]

    A supervised segmentation network for hyperspectral image classification,

    H. Sun, X. Zheng, and X. Lu, “A supervised segmentation network for hyperspectral image classification,”IEEE Trans. Image Process., vol. 30, pp. 2810–2825, Feb. 2021

  7. [7]

    Identifiability of the simplex volume minimization criterion for blind hyperspectral unmixing: The no-pure-pixel case,

    C.-H. Lin, W.-K. Ma, W.-C. Li, C.-Y . Chi, and A. Ambikapathi, “Identifiability of the simplex volume minimization criterion for blind hyperspectral unmixing: The no-pure-pixel case,”IEEE Trans. Geosci. Remote Sens., vol. 53, no. 10, pp. 5530–5546, May 2015

  8. [8]

    Nonnegative blind source separation for ill-conditioned mixtures via John ellipsoid,

    C.-H. Lin and J. M. Bioucas-Dias, “Nonnegative blind source separation for ill-conditioned mixtures via John ellipsoid,”IEEE Trans. Neural Netw. Learn. Syst., vol. 32, no. 5, pp. 2209–2223, Jul. 2020. 3https://docs.nvidia.com/deeplearning/tensorrt/latest/performance/ benchmarking.html 15

Show all 66 references
  1. [9]

    PCA-based land-use change detection and analysis using multitemporal and multisensor satellite data,

    J. Deng, K. Wang, Y . Deng, and G. Qi, “PCA-based land-use change detection and analysis using multitemporal and multisensor satellite data,”Int. J. Remote Sens., vol. 29, no. 16, pp. 4823–4838, 2008

  2. [10]

    Multivariate alteration detection (MAD) and MAF postprocessing in multispectral, bitemporal image data: New approaches to change detection studies,

    A. A. Nielsen, K. Conradsen, and J. J. Simpson, “Multivariate alteration detection (MAD) and MAF postprocessing in multispectral, bitemporal image data: New approaches to change detection studies,”Remote Sens. Environ., vol. 64, no. 1, pp. 1–19, 1998

  3. [11]

    Multiple support vector machines for land cover change detection: An application for mapping urban extensions,

    H. Nemmour and Y . Chibani, “Multiple support vector machines for land cover change detection: An application for mapping urban extensions,” ISPRS J. Photogramm. Remote Sens., vol. 61, no. 2, pp. 125–133, 2006

  4. [12]

    Extreme learning machine: theory and applications,

    G.-B. Huang, Q.-Y . Zhu, and C.-K. Siew, “Extreme learning machine: theory and applications,”Neurocomputing, vol. 70, no. 1-3, pp. 489–501, 2006

  5. [13]

    Automatic analysis of the difference image for unsupervised change detection,

    L. Bruzzone and D. F. Prieto, “Automatic analysis of the difference image for unsupervised change detection,”IEEE Trans. Geosci. Remote Sens., vol. 38, no. 3, pp. 1171–1182, 2000

  6. [14]

    SAAN: Similarity-aware attention flow network for change detection with VHR remote sensing images,

    H. Guo, X. Su, C. Wu, B. Du, and L. Zhang, “SAAN: Similarity-aware attention flow network for change detection with VHR remote sensing images,”IEEE Trans. Image Process., vol. 33, pp. 2599–2613, Mar. 2024

  7. [15]

    Iterative robust graph for unsupervised change detection of heterogeneous remote sensing images,

    Y . Sun, L. Lei, D. Guan, and G. Kuang, “Iterative robust graph for unsupervised change detection of heterogeneous remote sensing images,”IEEE Trans. Image Process., vol. 30, pp. 6277–6291, Jul. 2021

  8. [16]

    A context-sensitive technique robust to registration noise for change detection in VHR multispectral images,

    S. Marchesi, F. Bovolo, and L. Bruzzone, “A context-sensitive technique robust to registration noise for change detection in VHR multispectral images,”IEEE Trans. Image Process., vol. 19, pp. 1877–1889, Mar. 2010

  9. [17]

    Binary change guided hyperspec- tral multiclass change detection,

    M. Hu, C. Wu, B. Du, and L. Zhang, “Binary change guided hyperspec- tral multiclass change detection,”IEEE Trans. Image Process., vol. 32, pp. 791–806, Jan. 2023

  10. [18]

    Deep learning for change detection in remote sensing images: Comprehensive review and meta-analysis,

    L. Khelifi and M. Mignotte, “Deep learning for change detection in remote sensing images: Comprehensive review and meta-analysis,”IEEE Access, vol. 8, pp. 126 385–126 400, 2020

  11. [19]

    Backpropagation applied to handwritten zip code recognition,

    Y . LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, and L. D. Jackel, “Backpropagation applied to handwritten zip code recognition,”Neural Comput., vol. 1, pp. 541–551, 1989

  12. [20]

    Multispectral change detection with bilinear convolutional neural networks,

    Y . Lin, S. Li, L. Fang, and P. Ghamisi, “Multispectral change detection with bilinear convolutional neural networks,”IEEE Geosci. Remote Sens. Lett., vol. 17, no. 10, pp. 1757–1761, 2019

  13. [21]

    SwinSUNet: Pure transformer network for remote sensing image change detection,

    C. Zhang, L. Wang, S. Cheng, and Y . Li, “SwinSUNet: Pure transformer network for remote sensing image change detection,”IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 1–13, 2022

  14. [22]

    TransUNetCD: A hybrid trans- former network for change detection in optical remote-sensing images,

    Q. Li, R. Zhong, X. Du, and Y . Du, “TransUNetCD: A hybrid trans- former network for change detection in optical remote-sensing images,” IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 1–19, 2022

  15. [23]

    A non-local sparse unmixing based hyperspectral change detection with unsupervised deep clustering,

    T. Gao, M. Gong, X. Jiang, Y . Zhao, H. Liu, and Y . Pu, “A non-local sparse unmixing based hyperspectral change detection with unsupervised deep clustering,”Knowl.-Based Syst., vol. 317, p. 113408, 2025

  16. [24]

    Maximum volume inscribed ellipsoid: A new simplex-structured matrix factoriza- tion framework via facet enumeration and convex optimization,

    C.-H. Lin, R. Wu, W.-K. Ma, C.-Y . Chi, and Y . Wang, “Maximum volume inscribed ellipsoid: A new simplex-structured matrix factoriza- tion framework via facet enumeration and convex optimization,”SIAM Journal on Imaging Sciences, vol. 11, no. 2, pp. 1651–1679, 2018

  17. [25]

    HyperNet: Self-supervised hyperspectral spatial–spectral feature understanding network for hyperspectral change detection,

    M. Hu, C. Wu, and L. Zhang, “HyperNet: Self-supervised hyperspectral spatial–spectral feature understanding network for hyperspectral change detection,”IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 1–17, 2022

  18. [26]

    Hyperspectral change detection using semi- supervised graph neural network and convex deep learning,

    T.-H. Lin and C.-H. Lin, “Hyperspectral change detection using semi- supervised graph neural network and convex deep learning,”IEEE Trans. Geosci. Remote Sens., vol. 61, pp. 1–18, Jun. 2023

  19. [27]

    Quantum Information- Empowered Graph Neural Network for Hyperspectral Change Detec- tion,

    C.-H. Lin, T.-H. Lin, and J. Chanussot, “Quantum Information- Empowered Graph Neural Network for Hyperspectral Change Detec- tion,”IEEE Trans. Geosci. Remote Sens., vol. 62, pp. 1–15, Nov. 2024

  20. [28]

    A theoretical framework for unsupervised change detection based on change vector analysis in the polar domain,

    F. Bovolo and L. Bruzzone, “A theoretical framework for unsupervised change detection based on change vector analysis in the polar domain,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 1, pp. 218–236, 2006

  21. [29]

    Unsupervised change detection in satellite images using principal component analysis andk-means clustering,

    T. Celik, “Unsupervised change detection in satellite images using principal component analysis andk-means clustering,”IEEE Geosci. Remote Sens. Lett., vol. 6, no. 4, pp. 772–776, 2009

  22. [30]

    GETNET: A general end-to-end 2-D CNN framework for hyperspectral image change detection,

    Q. Wang, Z. Yuan, Q. Du, and X. Li, “GETNET: A general end-to-end 2-D CNN framework for hyperspectral image change detection,”IEEE Trans. Geosci. Remote Sens., vol. 57, no. 1, pp. 3–13, 2018

  23. [31]

    CSANet: Cross-temporal interaction symmetric attention network for hyperspectral image change detection,

    R. Song, W. Ni, W. Cheng, and X. Wang, “CSANet: Cross-temporal interaction symmetric attention network for hyperspectral image change detection,”IEEE Geosci. Remote Sens. Lett., vol. 19, pp. 1–5, 2022

  24. [32]

    A multilevel encoder– decoder attention network for change detection in hyperspectral images,

    J. Qu, S. Hou, W. Dong, Y . Li, and W. Xie, “A multilevel encoder– decoder attention network for change detection in hyperspectral images,” IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 1–13, 2021

  25. [33]

    SSA-SiamNet: Spectral–spatial-wise attention-based Siamese network for hyperspectral image change detection,

    L. Wang, L. Wang, Q. Wang, and P. M. Atkinson, “SSA-SiamNet: Spectral–spatial-wise attention-based Siamese network for hyperspectral image change detection,”IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 1–18, 2022

  26. [34]

    Three-order tucker decomposition and reconstruction detector for unsupervised hyperspectral change de- tection,

    Z. Hou, W. Li, R. Tao, and Q. Du, “Three-order tucker decomposition and reconstruction detector for unsupervised hyperspectral change de- tection,”IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 14, pp. 6194–6205, 2021

  27. [35]

    A patch tensor-based change detection method for hyperspectral images,

    Z. Hou, W. Li, and Q. Du, “A patch tensor-based change detection method for hyperspectral images,” inIEEE IGARSS, 2021, pp. 4328– 4331

  28. [36]

    S3Net: Superpixel- guided self-supervised learning network for multitemporal image change detection,

    T. Zhan, M. Gong, X. Jiang, and E. Zhang, “S3Net: Superpixel- guided self-supervised learning network for multitemporal image change detection,”IEEE Geosci. Remote Sens. Lett., vol. 20, pp. 1–5, 2023

  29. [37]

    Hypergraph self-supervised learning-based joint spectral-spatial-temporal feature representation for hyperspectral image change detection,

    P. Jian, Y . Ou, and K. Chen, “Hypergraph self-supervised learning-based joint spectral-spatial-temporal feature representation for hyperspectral image change detection,”IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 18, pp. 741–756, 2025

  30. [38]

    ADMM-ADAM: A new inverse imaging framework blending the advantages of convex optimization and deep learning,

    C.-H. Lin, Y .-C. Lin, and P.-W. Tang, “ADMM-ADAM: A new inverse imaging framework blending the advantages of convex optimization and deep learning,”IEEE Trans. Geosci. Remote Sens.ing, vol. 60, pp. 1–16, Sep. 2021

  31. [39]

    Underdetermined blind source separation via weighted simplex shrinkage regularization and quantum deep image prior,

    C.-H. Lin and S.-S. Young, “Underdetermined blind source separation via weighted simplex shrinkage regularization and quantum deep image prior,”IEEE Transactions on Image Processing, no. 99, pp. 1–1, 2026

  32. [40]

    HyperKING: Quantum-classical generative adversarial networks for hyperspectral image restoration,

    ——, “HyperKING: Quantum-classical generative adversarial networks for hyperspectral image restoration,”IEEE Trans. Geosci. Remote Sens., vol. 63, pp. 1–19, Apr. 2025

  33. [41]

    Change representation and extraction in stripes: Rethinking unsupervised hyperspectral image change detection with an untrained network,

    B. Yang, Y . Mao, L. Liu, L. Fang, and X. Liu, “Change representation and extraction in stripes: Rethinking unsupervised hyperspectral image change detection with an untrained network,”IEEE Transactions on Image Processing, vol. 33, pp. 5098–5113, 2024

  34. [42]

    Unsupervised change detection in multitemporal multispectral satellite images: A convex relaxation approach,

    W.-C. Zheng, C.-H. Lin, K.-H. Tseng, C.-Y . Huang, T.-H. Lin, C.-H. Wang, and C.-Y . Chi, “Unsupervised change detection in multitemporal multispectral satellite images: A convex relaxation approach,” inIEEE IGARSS, Yokohama, Japan, Jul. 2019, pp. 1546–1549

  35. [43]

    Active sample selection and correction propagation on a gradually-augmented graph,

    H. Su, Z. Yin, T. Kanade, and S. Huh, “Active sample selection and correction propagation on a gradually-augmented graph,” inIEEE CVPR, 2015, pp. 1975–1983

  36. [44]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” inProc. Int. Conf. Learn. Represent., San Diego, CA, USA, May 2015

  37. [45]

    Discrimination among semi-arid landscape endmembers using the spectral angle mapper (SAM) algorithm,

    R. H. Yuhas, A. F. H. Goetz, and J. W. Boardman, “Discrimination among semi-arid landscape endmembers using the spectral angle mapper (SAM) algorithm,” inProc. 3rd Annual JPL Airborne Geoscience Workshop. Vol. 1: AVIRIS Workshop, Pasadena, CA, Jun. 1992

  38. [46]

    Hyperspectral anomaly detection via global and local joint modeling of background,

    Z. Wu, W. Zhu, J. Chanussot, Y . Xu, and S. Osher, “Hyperspectral anomaly detection via global and local joint modeling of background,” IEEE Trans. Signal Process., vol. 67, no. 14, pp. 3858–3869, 2019

  39. [47]

    T. M. Apostol,Mathematical Analysis. 2nd Edition, Pearson Edu. Taiwan Ltd., 2007

  40. [48]

    Chi, W.-C

    C.-Y . Chi, W.-C. Li, and C.-H. Lin,Convex Optimization for Signal Processing and Communications: From Fundamentals to Applications. Boca Raton, FL, USA: CRC Press, 2017

  41. [49]

    Band selection- based dimensionality reduction for change detection in multi-temporal hyperspectral images,

    S. Liu, Q. Du, X. Tong, A. Samat, H. Pan, and X. Ma, “Band selection- based dimensionality reduction for change detection in multi-temporal hyperspectral images,”Remote Sens., vol. 9, no. 10, p. 1008, 2017

  42. [50]

    Unsupervised change detection in multispectral remote sensing images via spectral- spatial band expansion,

    S. Liu, Q. Du, X. Tong, A. Samat, and L. Bruzzone, “Unsupervised change detection in multispectral remote sensing images via spectral- spatial band expansion,”IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 12, no. 9, pp. 3578–3587, 2019

  43. [51]

    Change detection for hyperspectral images via convolutional sparse analysis and temporal spectral unmixing,

    Q. Guo, J. Zhang, C. Zhong, and Y . Zhang, “Change detection for hyperspectral images via convolutional sparse analysis and temporal spectral unmixing,”IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 14, pp. 4417–4426, 2021

  44. [52]

    Adaptive center-focused hybrid attention network for change detection in hyperspectral images,

    F. Jiang, S. Zhang, M. Zhang, M. Gong, Y . Zhou, W. Zhao, and Z. Guan, “Adaptive center-focused hybrid attention network for change detection in hyperspectral images,”IEEE Transactions on Geoscience and Remote Sensing, 2025

  45. [53]

    Accuracy and inaccuracy assessments in land- cover classification,

    R. Nishii and S. Tanaka, “Accuracy and inaccuracy assessments in land- cover classification,”IEEE Trans. Geosci. Remote Sens., vol. 37, no. 1, pp. 491–498, 2002

  46. [54]

    Coefficient kappa: Some uses, misuses, and alternatives,

    R. L. Brennan and D. J. Prediger, “Coefficient kappa: Some uses, misuses, and alternatives,”Educ. Psychol. Meas., vol. 41, no. 3, pp. 687–699, 1981

  47. [55]

    Confusion matrix-based perfor- mance evaluation metrics,

    S. Sathyanarayanan and B. R. Tantri, “Confusion matrix-based perfor- mance evaluation metrics,”Afr. J. Biomed. Res., pp. 4023–4031, Nov. 2024. 16

  48. [56]

    On loss functions for deep neural networks in classification,

    K. Janocha and W. M. Czarnecki, “On loss functions for deep neural networks in classification,”arXiv preprint arXiv:1702.05659, 2017

  49. [57]

    SLIC superpixels compared to state-of-the-art superpixel methods,

    R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, and S. S ¨usstrunk, “SLIC superpixels compared to state-of-the-art superpixel methods,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 34, pp. 2274–2282, 2012

  50. [58]

    Survey over image thresholding techniques and quantitative performance evaluation,

    M. Sezgin and B. l. Sankur, “Survey over image thresholding techniques and quantitative performance evaluation,”J. Electron. Imaging, vol. 13, no. 1, pp. 146–168, 2004

  51. [59]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,”Neural Comput., vol. 9, no. 8, pp. 1735–1780, 1997

  52. [60]

    A fast hyperplane- based minimum-volume enclosing simplex algorithm for blind hyper- spectral unmixing,

    C.-H. Lin, C.-Y . Chi, Y .-H. Wang, and T.-H. Chan, “A fast hyperplane- based minimum-volume enclosing simplex algorithm for blind hyper- spectral unmixing,”IEEE Trans. Signal Process., vol. 64, no. 8, pp. 1946–1961, Dec. 2015

  53. [61]

    Unsupervised spatial-spectral hyperspectral image reconstruction and clustering with diffusion geometry,

    K. Cui, R. Li, S. L. Polk, J. M. Murphy, R. J. Plemmons, and R. H. Chan, “Unsupervised spatial-spectral hyperspectral image reconstruction and clustering with diffusion geometry,”arXiv preprint arXiv:2204.13497, 2022

  54. [62]

    AI-powered video monitoring: Assessing the NVIDIA Jetson Orin devices for edge com- puting applications,

    F. P. Scalcon, R. Tahal, M. Ahrabi, Y . Huangfu, R. Ahmed, B. Nahid- Mobarakeh, S. Shirani, C. Vidal, and A. Emadi, “AI-powered video monitoring: Assessing the NVIDIA Jetson Orin devices for edge com- puting applications,” inProc. IEEE Transp. Electrific. Conf. Expo (ITEC), Ch...

  55. [63]

    Regularization parameter selection in minimum volume hyperspectral unmixing,

    L. Zhuang, C.-H. Lin, M. A. Figueiredo, and J. M. Bioucas-Dias, “Regularization parameter selection in minimum volume hyperspectral unmixing,”IEEE Trans. Geosci. Remote Sens., vol. 57, no. 12, pp. 9858– 9877, 2019

  56. [64]

    Demystifying tensorRT: Characterizing neural network inference engine on nvidia edge devices,

    O. Shafi, C. Rai, R. Sen, and G. Ananthanarayanan, “Demystifying tensorRT: Characterizing neural network inference engine on nvidia edge devices,” in2021 IEEE International Symposium on Workload Characterization (IISWC), 2021, pp. 226–237

  57. [65]

    Under- standing the runtime overheads of deep learning inference on edge devices,

    X. Ma, G. Li, L. Liu, H. Liu, L. Liu, and X. Feng, “Under- standing the runtime overheads of deep learning inference on edge devices,” in2021 IEEE Intl Conf on Parallel & Distributed Pro- cessing with Applications, Big Data & Cloud Computing, Sustain- able Computing & Communic...

  58. [66]

    Detecting vehicles on the edge: Knowledge distillation to improve performance in heteroge- neous road traffic,

    M. Bharadhwaj, G. Ramadurai, and B. Ravindran, “Detecting vehicles on the edge: Knowledge distillation to improve performance in heteroge- neous road traffic,” in2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2022, pp. 3191–3197. Chia-Hs...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.