Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

Neural Spatial-Temporal Tensor Representation for Infrared Small Target Detection

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read NeurSTT represents an infrared video's background as a low-rank neural tensor function, so the residual contains only sparse moving targets, and detects them without any labeled training data.

desk verdict A genuinely new unsupervised INR-based ISTD method, but the missing binarization threshold makes the headline IoU/F1 numbers uncheckable. read the letter →

arxiv 2412.17302 v1 pith:TRTZPVJX submitted 2024-12-23 cs.CV

Fengyi Wu , Simin Liu , Haoan Wang , Bingjie Tao , Junhai Luo , Zhenming Peng This is my paper · ORCID
classification cs.CV
keywords infraredsmalltargetdetectionimplicitneuralrepresentationspatial-temporaltensorlow-rankbackgroundtotalvariationunsupervisedlearningsoft-thresholding
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's claim is that small moving targets in infrared video can be detected by representing the background as a neural low-rank tensor function, without any labeled training data. NeurSTT stacks consecutive frames into a 3-D tensor, approximates the slowly changing background with three sinusoidal MLP factor networks over a small learnable core tensor, and treats a soft-thresholded residual as the target. A neural three-dimensional total-variation term enforces smoothness across space and time, which suppresses target-like background clusters. On nine sequences the method reports average IoU 82.82% and F1 90.52%, while using 16.6x fewer parameters and 48.9% less runtime than the 3DSTPM baseline on 256x256 inputs. If correct, this shows that optimization-style low-rank plus sparse decomposition can be made nonlinear and data-efficient enough for practical multi-frame infrared detection.

What carries the argument

The load-bearing object is the neural-represented low-rank background tensor: three sinusoidal MLPs map coordinate vectors along height, width, and time into factor matrices that multiply a learnable core tensor, so the Tucker rank is set by the downsampling rate $r_d = 1/4$. The argument is carried by the interplay of three loss terms: a nuclear-norm loss on each background frame that enforces low rank, a neural 3-D total variation loss computed from the MLP derivatives that enforces local smoothness in space and time, and an $\ell^1$ loss on the soft-thresholded residual that preserves sparse targets. This machinery lets a deep optimizer do the work that classical tensor RPCA solvers did with singular-value thresholding.

What would settle it

Run NeurSTT on a sequence with strong camera pan or vigorous moving clutter, such as waving trees, and measure whether the residual $D - B^{NLR}$ remains sparse: if residual energy concentrates in large moving background structures rather than in the target footprint, the background model is inadequate. A more quantitative version is to sweep $r_d$ from 1/8 to 1 on such a sequence; the method should show a clear performance peak only if the fixed rank assumption matches the sequence's background complexity.

Watch

Extended reading notes

Core claim

The central discovery is that the classical low-rank-plus-sparse decomposition of infrared sequences can be neuralized by writing the background as a continuous tensor function $B^{NLR}_{ijk} = [\mathcal{G}; f_{\theta_h}, f_{\theta_w}, f_{\theta_t}](i,j,k)$ with a small Tucker-style core $\mathcal{G}$ and three shallow SIREN MLPs, then unrolling the optimization: a nuclear-norm loss on each temporal slice, a neural 3-D total variation loss on the MLP derivatives, an $\ell^1$ target loss, and a soft-thresholding update $T = ST_{\lambda/2}(D - B^{NLR})$. This replaces the heavy tensor solvers of prior methods with Adam updates on network parameters while keeping the domain priors explicit. The paper argues that the nonlinear representation captures spatial-temporal correlations that linear Tucker or tensor operations miss, yielding better background suppression and fewer false alarms.

Load-bearing premise

The load-bearing premise is that real infrared video backgrounds are well approximated by the specific low-rank neural function with Tucker ranks set to one quarter of the sequence dimensions, so that after subtracting it the residual is mostly sparse target signal rather than structured background.

Editorial extensions

If this is right

  • Infrared small-target detection can be done with zero labeled training data, removing the labeling bottleneck for video datasets.
  • The same model can process both 256x256 and 720x480 sequences, with parameter counts scaling more gently than 3D-CNN baselines.
  • Adding neural temporal total variation suppresses moving target-like clusters and improves IoU by about 1.28 points over spatial-only TV in ablations.
  • The reported averages across nine sequences include large gains on hard 720x480 scenes where several tensor baselines fail entirely.
  • On 256x256 inputs the method is 16.6x lighter and 48.9% faster than 3DSTPM, making the unsupervised scheme practical as a baseline.

Reading between the lines

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

  • An implication the authors leave implicit: the same neural low-rank-plus-sparse decomposition could be applied to other spatiotemporal inverse problems, such as video background subtraction or moving-object detection in other spectra, wherever a slowly varying background is separated from sparse events.
  • The rank-setting rule $r_d=1/4$ was selected by ablation on the evaluation sequences; a data-driven or adaptive rank selection would be a natural test of whether the method generalizes to backgrounds with higher effective complexity.
  • A testable extension: replacing the SIREN factor MLPs with coordinate-based networks that incorporate temporal motion compensation could extend the method to scenes with global camera motion, which the current static-coordinate representation does not model.
  • The soft-thresholding target update suggests the framework could be extended to jointly estimate target motion or track multiple targets by adding a temporal continuity prior on the target tensor.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes NeurSTT, an unsupervised neural-represented spatial-temporal tensor model for infrared small target detection. The background of a stacked sequence is approximated by a low-rank Tucker tensor function implemented with three SIREN-based MLPs, a neural 3D total variation regularizer is added through derivatives of those networks, and the target tensor is updated by soft-thresholding the residual. All parameters are optimized with Adam using a loss composed of a sum of per-frame nuclear norms, an l1 target term, and the Neur3DTV term. The authors evaluate on nine sequences from two public datasets, compare against twelve single- and multi-frame baselines, and report average IoU 82.82% and F1 90.52%, with 16.6x fewer parameters and 48.9% less runtime than 3DSTPM on 256x256 inputs.

Significance. If the reported numbers are trustworthy, the contribution is useful: it is an unsupervised, label-free method that combines low-rank and sparsity priors with implicit neural representation, and it is systematically compared with a broad set of baselines. The architectural idea of representing the background as a low-rank tensor function and regularizing it through neural derivatives is interesting and reasonably well motivated. The paper also provides ablation studies for the main modules and a promised public code link. However, the central empirical claims are currently weakened by an unspecified binarization threshold, hyperparameter selection on the same test sequences, and an inconsistency between the stated objective and the implemented loss; these issues must be resolved before the quantitative conclusions can be accepted.

major comments (5)
  1. [Section 4.4 and Section 5.2] The adaptive threshold procedure used to binarize the NeurSTT target map is not specified. Section 4.4 states that after solving Eq. (22) the result tensor is post-processed with a 'general tensor-to-sequence operation' and 'an adaptive threshold procedure,' but no formula, criterion, or implementation detail is given, and Section 5.2 only lists thresholds for baseline methods. Because every pixel-level IoU and F1 value in Table 9 is computed on binary maps, the headline average IoU 82.82% and F1 90.52% cannot be reproduced or checked; the paper must state the binarization rule and confirm that it does not use ground-truth information.
  2. [Section 5.3 and Table 9] Hyperparameters L, lambda, phi, kappa, and kmax are selected by varying one parameter at a time on the same six sequences from [42] that are then used in the final evaluation of Table 9. This selection-on-test protocol makes the reported average IoU/F1 on those six sequences partly an artifact of tuning rather than an unbiased estimate of performance; a validation split, cross-validation, or separate development set is needed before claiming superiority over the baselines.
  3. [Eqs. (17), (19), and (24)] The optimization problem stated in Eq. (17) contains the tensor nuclear norm ||BNLR||_*, but the loss actually minimized, Eq. (24), uses LNuc defined in Eq. (19) as the sum of per-frame matrix nuclear norms \sum_k ||B_NLR_k||_*. These are different regularizers for a third-order tensor, so the paper does not optimize the objective it derives; the text should either replace ||BNLR||_* in Eqs. (17) and (18) with the per-frame sum or justify the discrepancy.
  4. [Section 5.3, Config. III, and Table 2] The ablation text for phi says that phi=1e-5 performs best and is selected, but Table 3 shows 5e-5 achieves the highest IoU (81.44) and Table 2 lists the final setting as phi=5e-5. This internal contradiction must be resolved.
  5. [Section 5.3 and Algorithm 1] No multiple runs or random seeds are reported. Since SIREN weights are randomly initialized and optimization is stochastic Adam-based, the reported IoU/F1 values and the hyperparameter choices in Tables 3 and 4 may depend on initialization; providing means and standard deviations over several runs is necessary to support the quantitative claims.
minor comments (5)
  1. [Table 4] The F1 value for NeurSTT (Full) is printed as 6094; this appears to be a typo for 60.94.
  2. [Abstract] The abstract states 'outperforms detection challenges' and 'averaged 19.19% higher in IoU'; the first phrase is ungrammatical and the second should specify that the comparison is with 3DSTPM on the 256x256 sequences and whether the increase is absolute or relative.
  3. [Eq. (13)] The temporal difference operator D_t has an extra closing parenthesis: 'B(i,j,k−1))' should read 'B(i,j,k−1)'.
  4. [Fig. 1] The caption contains the typo 'physic-informed'; it should be 'physics-informed'.
  5. [Section 5.2] The precision formula is written as 'Precision = TP/TP+FP'; parentheses are needed: Precision = TP/(TP+FP).

Circularity Check

1 steps flagged · score 6.0 of 10

Partial circularity: hyperparameters are tuned on the same six sequences that dominate the headline nine-sequence average; adaptive-threshold step is underspecified.

  1. fitted input called prediction [Section 5.3, Ablation Studies (Tables 3, 4, and 9)]
    "Notably, we use the average values of two pixel-level metrics, IoU and F1, across six sequences from [42] for this analysis."

    Table 4 chooses rd by maximizing the averaged IoU/F1 on the six [42] sequences, and Table 3 does the same for L, λ, φ, κ, and kmax. The final comparison in Table 9 then reports average IoU/F1 over Seqs. 1-9, where Seqs. 1-6 are exactly those same six [42] sequences. The headline 82.82 IoU / 90.52 F1 is therefore, for six of nine sequences, the tuning objective itself rather than an independent prediction. The reported advantage over 3DSTPM on those sequences is partly a test-set selection effect, not a consequence of the model equations.

full rationale

Modeling-side derivation is not circular: B_NLR (Eq. 11) follows from the tensor-function factorization (Eqs. 8-9), Neur3DTV (Eq. 16) is an explicit derivative-based regularizer, and the target update (Eq. 22) is the standard soft-thresholding step of Eq. (21). These equations do not encode the final IoU/F1 numbers, and the priors borrowed from [36], [37], and [40] are external rather than self-citations. The circularity found is evaluation-side: Section 5.3 tunes L, λ, φ, κ, kmax, and rd on the average IoU/F1 of the six [42] sequences, and Table 9 includes those same six sequences in the reported nine-sequence average (82.82 IoU / 90.52 F1), so the headline performance is partly a fitted quantity. Seqs. 7-9 from [22] were not in the ablations and provide some independent support, so the architecture itself is not predetermined. One missing specification is also noted and weighed: Section 4.4 says only that post-processing uses 'an adaptive threshold procedure,' without giving a formula or fitting rule, and Section 5.2 prescribes thresholds only for other methods. This is a reproducibility gap rather than a demonstrated circular step, but it leaves the binary-metric comparison only partially checkable.

Assumptions & free parameters 9 free parameters · 6 assumptions · 0 invented entities

The ledger lists the tuned hyperparameters and the unproved modeling assumptions behind the background representation and the neural TV regularizer. The method does not introduce new physical entities. The largest burden is the imported factorization theorem and the assumption that coordinate-network derivatives correspond to image gradients.

free parameters (9)
  • Temporal index L = 80
    Number of frames stacked into the spatial-temporal tensor; chosen by ablation Config I to maximize averaged IoU/F1 on the six sequences from [42] used later for evaluation.
  • Sparsity weight lambda = 0.20
    Balances background low-rankness and target sparsity in Eq. (17); selected in Config II from values 0.05 to 0.25 on the evaluation sequences.
  • Neur3DTV weight phi = 5e-5
    Weights the neural total variation loss; Table 3 shows 5e-5 gives the best IoU, although the text says phi=1e-5 was selected. Tables 2 and 4 use 5e-5.
  • Temporal derivative weight kappa = 100
    Controls temporal emphasis in Neur3DTV; selected in Config IV from values 1, 10, 100, 1000 on evaluation data.
  • Max iterations kmax = 2000
    Stopping criterion for Adam; selected in Config V from 1000 to 3000 on evaluation data.
  • Downsampling rate rd = 1/4
    Sets Tucker ranks as n_i/rd in Algorithm 1; selected in Table 4 by maximizing IoU/F1 on the six evaluation sequences. This is the main determinant of model capacity.
  • MLP channel size ci = ci=ni
    Network width equal to input dimensions per coordinate; fixed by design in Section 4.5, no sensitivity study.
  • SIREN initialization std = 5
    Uniform initialization scale for sine MLPs, Section 4.5; chosen by hand following [40].
  • Learning rate and weight decay = 0.0005, 0.01
    Adam settings reported in Section 5.1; no sensitivity study.
assumptions (6)
  • domain assumption Observed tensor is exactly background plus target (D=B+T).
    Eq. (3) in Section 4.1; no sensor noise term appears in the model. Used to justify soft-thresholding residual extraction.
  • domain assumption Infrared backgrounds are low-rank and targets are sparse.
    Eq. (4)-(5) in Section 4.1; standard ISTD prior, but the paper provides no measurement of actual rank or sparsity for the nine test scenes.
  • standard math Low F-rank tensor functions admit factorization G x1 f_x(.) x2 f_y(.) x3 f_z(.).
    Eq. (8)-(9) in Section 4.2, imported from Luo et al. [36] without proof or numerical verification in this paper.
  • domain assumption SIREN sine activation and its cosine derivative capture complex structure and remain differentiable.
    Used in Section 4.2 and 4.5, based on [40]; needed for both background approximation and Neur3DTV derivative computation.
  • ad hoc to paper Derivatives of factor MLPs with respect to coordinate inputs act as spatial and temporal total variation of the reconstructed background.
    Section 4.3, Eq. (16) and Appendix A; no derivation connects the Jacobian of coordinate MLPs to the discrete image gradient of B_NLR.
  • ad hoc to paper Sum of per-frame nuclear norms is a valid low-rank background regularizer.
    Eq. (19) replaces the tensor nuclear norm in Eq. (17) with a sum of matrix nuclear norms; the equivalence is asserted through the learning loss rather than proven.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural Spatial-Temporal Tensor Representation for Infrared Small Target Detection." pith.science (2026). https://pith.science/paper/TRTZPVJX

@misc{pith2026241217302,
  author       = {Pith},
  title        = {Pith review of: Neural Spatial-Temporal Tensor Representation for Infrared Small Target Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TRTZPVJX}},
  note         = {Machine review of arXiv:2412.17302}
}
abstract

Optimization-based approaches dominate infrared small target detection as they leverage infrared imagery's intrinsic low-rankness and sparsity. While effective for single-frame images, they struggle with dynamic changes in multi-frame scenarios as traditional spatial-temporal representations often fail to adapt. To address these challenges, we introduce a Neural-represented Spatial-Temporal Tensor (NeurSTT) model. This framework employs nonlinear networks to enhance spatial-temporal feature correlations in background approximation, thereby supporting target detection in an unsupervised manner. Specifically, we employ neural layers to approximate sequential backgrounds within a low-rank informed deep scheme. A neural three-dimensional total variation is developed to refine background smoothness while reducing static target-like clusters in sequences. Traditional sparsity constraints are incorporated into the loss functions to preserve potential targets. By replacing complex solvers with a deep updating strategy, NeurSTT simplifies the optimization process in a domain-awareness way. Visual and numerical results across various datasets demonstrate that our method outperforms detection challenges. Notably, it has 16.6$\times$ fewer parameters and averaged 19.19\% higher in $IoU$ compared to the suboptimal method on $256 \times 256$ sequences.

Figures

Figures reproduced from arXiv: 2412.17302 by the authors.

Figure 1
Figure 1. Comparison of the existing spatial-temporal tensor schemes for ISTD and our NeurSTT with [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Schematic of a tensor function representation method in Eq. ( [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Overall procedure of NeurSTT model for ISTD. The background spatial-temporal tensor is repre [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Example of different configuration performance on different epochs on Seq. 3, where Config. II is in gray, Config. III is in skyblue, and Config. IV (NeurSTT) is in purple. Notably, the proposed model can effectively reduce false alarms and preserve targets as the epoc…
Figure 5
Figure 5. Figure 5: Example of different nuclear norm loss LNuc performance via different configurations on Seq. 3. Our NeurSTT model is the fastest at recovering the background at early epochs while being relatively stable compared with the model without Neur3DTV. the background, we exam…
Figure 6
Figure 6. Figure 6: Visual comparison of detection results from di [PITH_FULL_IMAGE:figures/full_fig_p022_6.png]
Figure 7
Figure 7. Figure 7: Visual comparison of detection results from di [PITH_FULL_IMAGE:figures/full_fig_p023_7.png]
Figure 8
Figure 8. Figure 8: Visual comparison of detection results from di [PITH_FULL_IMAGE:figures/full_fig_p024_8.png]
Figure 9
Figure 9. Figure 9: 3-D ROC curves and corresponding 2-D ROC curves (ROC [PITH_FULL_IMAGE:figures/full_fig_p025_9.png]
Figure 10
Figure 10. Figure 10: 3-D ROC curves and corresponding 2-D ROC curves (ROC [PITH_FULL_IMAGE:figures/full_fig_p026_10.png]
Figure 11
Figure 11. Figure 11: 3-D ROC curves and corresponding 2-D ROC curves (ROC [PITH_FULL_IMAGE:figures/full_fig_p026_11.png]
Figure 12
Figure 12. Figure 12: (a) Visualized examples of different noise interferences on Seq. 1, along with corresponding results and ground truth labels; (b) and (c) show the corresponding averaged F1 and IoU evaluations on six sequences from [42]. they use pre-trained weights on a GPU without c…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Continuous Representation Methods, Theories, and Applications: An Overview and Perspectives

    cs.CV 2025-05 conditional novelty 2.0 of 10

    A survey organizing continuous representation methods into parametric models, structural modeling, theory, and applications, with a curated open-source reference repository.

Reference graph

Works this paper leans on

43 extracted references · 42 canonical work pages · cited by 1 Pith paper

  1. [42]

    B. Hui, Z. Song, H. Fan, P. Zhong, W. Hu, X. Zhang, J. Lin, H. Su, W. Jin, Y . Zhang, et al., A dataset for infrared image dim-small aircraft target detection and tracking underground/air background, Sci. Data Bank 5 (12) (2019) 4

    work page 2019

  2. [1]

    R. Kou, C. Wang, Z. Peng, Z. Zhao, Y . Chen, J. Han, F. Huang, Y . Yu, Q. Fu, Infrared small target segmentation networks: A survey, Pattern Recognit. 143 (2023) 109788

    work page 2023

  3. [2]

    M. Zhao, W. Li, L. Li, J. Hu, P. Ma, R. Tao, Single-frame infrared small-target detection: A survey, IEEE Geosci. Remote Sens. Mag. 10 (2) (2022) 87–119

    work page 2022

  4. [3]

    Zhang, Y

    M. Zhang, Y . Wang, J. Guo, Y . Li, X. Gao, J. Zhang, Irsam: Advancing segment anything model for infrared small target detection, in: Proc. Eur. Conf. Comput. Vis. (ECCV), 2024

    work page 2024

  5. [4]

    C. Gao, D. Meng, Y . Yang, Y . Wang, X. Zhou, A. G. Hauptmann, Infrared patch- image model for small target detection in a single image, IEEE Trans. Image Process. 22 (12) (2013) 4996–5009. 31

    work page 2013

  6. [5]

    Y . Dai, Y . Wu, Reweighted infrared patch-tensor model with both nonlocal and local priors for single-frame small target detection, IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens. 10 (8) (2017) 3752–3767

    work page 2017

  7. [6]

    F. Wu, T. Zhang, L. Li, Y . Huang, Z. Peng, Rpcanet: Deep unfolding rpca based infrared small target detection, in: Proc. IEEE Winter Conf. Appl. Comput. Vis. (W ACV), 2024, pp. 4809–4818

    work page 2024

  8. [7]

    C. Kwan, B. Budavari, Enhancing small moving target detection performance in low-quality and long-range infrared videos using optical flow techniques, Remote Sens. 12 (24) (2020) 4024

    work page 2020

Show all 43 references
  1. [8]

    T. Liu, J. Yang, B. Li, C. Xiao, Y . Sun, Y . Wang, W. An, Nonconvex tensor low-rank approximation for infrared small target detection, IEEE Trans. Geosci. Remote Sens. 60 (2021) 1–18

    2021

  2. [9]

    S. Chen, L. Ji, J. Zhu, M. Ye, X. Yao, Sstnet: Sliced spatio-temporal network with cross-slice convlstm for moving infrared dim-small target detection, IEEE Trans. Geosci. Remote Sens. (2024)

    2024

  3. [10]

    Y . Liu, J. Liu, Z. Long, C. Zhu, Y . Liu, J. Liu, Z. Long, C. Zhu, Deep networks for tensor approximation, Tensor Computation for Data Analysis (2022) 265–284

    2022

  4. [11]

    T. Liu, Q. Yin, J. Yang, Y . Wang, W. An, Combining deep denoiser and low-rank priors for infrared small target detection, Pattern Recognit. 135 (2023) 109184

    2023

  5. [12]

    Miao, X.-L

    Y .-C. Miao, X.-L. Zhao, X. Fu, J.-L. Wang, Y .-B. Zheng, Hyperspectral denoising using unsupervised disentangled spatiospectral deep priors., IEEE Trans. Geosci. Remote. Sens. 60 (2022) 1–16

    2022

  6. [13]

    Zhang, P

    Z. Zhang, P. Gao, S. Ji, X. Wang, P. Zhang, Infrared small target detection com- bining deep spatial-temporal prior with traditional priors, IEEE Trans. Geosci. Remote Sens. (2023)

    2023

  7. [14]

    Molaei, A

    A. Molaei, A. Aminimehr, A. Tavakoli, A. Kazerouni, B. Azad, R. Azad, D. Mer- hof, Implicit neural representation in medical imaging: A comparative survey, in: Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2023, pp. 2381–2391. 32

    2023

  8. [15]

    D. P. Kingma, J. Ba, Adam: A method for stochastic optimization, arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  9. [16]

    X. Bai, F. Zhou, Analysis of new top-hat transformation and the application for infrared dim small target detection, Pattern Recognit. 43 (6) (2010) 2145–2156

    2010

  10. [17]

    Y . Wei, X. You, H. Li, Multiscale patch-based contrast measure for small infrared target detection, Pattern Recognit. 58 (2016) 216–226

    2016

  11. [18]

    Zhang, Z

    L. Zhang, Z. Peng, Infrared small target detection based on partial sum of the tensor nuclear norm, Remote Sens. 11 (4) (2019) 382

    2019

  12. [19]

    X. Kong, C. Yang, S. Cao, C. Li, Z. Peng, Infrared small target detection via nonconvex tensor fibered rank approximation, IEEE Trans. Geosci. Remote Sens. 60 (2021) 1–21

    2021

  13. [20]

    B. Li, C. Xiao, L. Wang, Y . Wang, Z. Lin, M. Li, W. An, Y . Guo, Dense nested attention network for infrared small target detection, IEEE Trans. Image Process. 32 (2022) 1745–1758

    2022

  14. [21]

    Zhang, L

    T. Zhang, L. Li, S. Cao, T. Pu, Z. Peng, Attention-guided pyramid context net- works for detecting infrared small target under complex background, IEEE Trans. Aerosp. Electron. Syst. 59 (4) (2023) 4250–4261

    2023

  15. [22]

    H. Sun, J. Bai, F. Yang, X. Bai, Receptive-field and direction induced attention network for infrared dim small target detection with a large-scale dataset irdst, IEEE Trans. Geosci. Remote Sens. 61 (2023) 1–13

    2023

  16. [23]

    Y . Dai, Y . Wu, F. Zhou, K. Barnard, Attentional local contrast networks for in- frared small target detection, IEEE Trans. Geosci. Remote Sens. 59 (11) (2021) 9813–9824

    2021

  17. [24]

    Zhang, R

    M. Zhang, R. Zhang, Y . Yang, H. Bai, J. Zhang, J. Guo, Isnet: Shape matters for infrared small target detection, in: Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 877–886. 33

    2022

  18. [25]

    P. Du, A. Hamdulla, Infrared moving small-target detection using spatial– temporal local di fference measure, IEEE Geosci. Remote Sens. Lett. 17 (10) (2019) 1817–1821

    2019

  19. [26]

    Y . Luo, X. Li, J. Wang, S. Chen, Clustering and tracking-guided infrared spatial- temporal small target detection, IEEE Trans. Geosci. Remote Sens. (2024)

    2024

  20. [27]

    H.-K. Liu, L. Zhang, H. Huang, Small target detection in infrared videos based on spatio-temporal tensor model, IEEE Trans. Geosci. Remote Sens. 58 (12) (2020) 8689–8700

    2020

  21. [28]

    G. Wang, B. Tao, X. Kong, Z. Peng, Infrared small target detection using nonover- lapping patch spatial–temporal tensor factorization with capped nuclear norm reg- ularization, IEEE Trans. Geosci. Remote Sens. 60 (2021) 1–17

    2021

  22. [29]

    F. Wu, H. Yu, A. Liu, J. Luo, Z. Peng, Infrared small target detection using spatiotemporal 4-d tensor train and ring unfolding, IEEE Trans. Geosci. Remote Sens. 61 (2023) 1–22

    2023

  23. [30]

    P. Yan, R. Hou, X. Duan, C. Yue, X. Wang, X. Cao, Stdmanet: Spatio-temporal differential multiscale attention network for small moving infrared target detec- tion, IEEE Trans. Geosci. Remote Sens. 61 (2023) 1–16

    2023

  24. [31]

    R. Li, W. An, C. Xiao, B. Li, Y . Wang, M. Li, Y . Guo, Direction-coded temporal u- shape module for multiframe infrared small target detection, IEEE Trans. Neural Networks Learn. Syst. (2023)

    2023

  25. [32]

    H. Deng, Y . Zhang, Y . Li, K. Cheng, Z. Chen, Bemst: Multi-frame infrared small- dim target detection using probabilistic estimation of sequential backgrounds, IEEE Trans. Geosci. Remote Sens. (2024)

    2024

  26. [33]

    Heide, M

    F. Heide, M. Steinberger, Y .-T. Tsai, M. Rouf, D. Paj˛ ak, D. Reddy, O. Gallo, J. Liu, W. Heidrich, K. Egiazarian, et al., Flexisp: A flexible camera image pro- cessing framework, ACM Trans. Graph. 33 (6) (2014) 1–13. 34

    2014

  27. [34]

    Zhang, W

    K. Zhang, W. Zuo, L. Zhang, Ffdnet: Toward a fast and flexible solution for cnn- based image denoising, IEEE Trans. Image Process. 27 (9) (2018) 4608–4622

    2018

  28. [35]

    Luo, X.-L

    Y .-S. Luo, X.-L. Zhao, T.-X. Jiang, Y . Chang, M. K. Ng, C. Li, Self-supervised nonlinear transform-based tensor nuclear norm for multi-dimensional image re- covery, IEEE Trans. Image Process. 31 (2022) 3793–3808

    2022

  29. [36]

    Y . Luo, X. Zhao, Z. Li, M. K. Ng, D. Meng, Low-rank tensor function representa- tion for multi-dimensional data recovery, IEEE Trans. Pattern Anal. Mach. Intell. (2023)

    2023

  30. [37]

    Y . Luo, X. Zhao, K. Ye, D. Meng, Neurtv: Total variation on the neural domain, arXiv preprint arXiv:2405.17241 (2024)

    2024 arXiv

  31. [38]

    T. G. Kolda, B. W. Bader, Tensor decompositions and applications, SIAM review 51 (3) (2009) 455–500

    2009

  32. [39]

    T. Liu, J. Yang, B. Li, Y . Wang, W. An, Infrared small target detection via noncon- vex tensor tucker decomposition with factor prior, IEEE Trans. Geosci. Remote Sens. 61 (2023) 1–17

    2023

  33. [40]

    Sitzmann, J

    V . Sitzmann, J. Martel, A. Bergman, D. Lindell, G. Wetzstein, Implicit neural rep- resentations with periodic activation functions, in: Proc. Adv. Neural Inf. Process. Syst. (NIPS), V ol. 33, 2020, pp. 7462–7473

    2020

  34. [41]

    S. Ma, D. Goldfarb, L. Chen, Fixed point and bregman iterative methods for matrix rank minimization, Math. Program. 128 (1) (2011) 321–353

    2011

  35. [43]

    Chang, An e ffective evaluation tool for hyperspectral target detection: 3d re- ceiver operating characteristic curve analysis, IEEE Trans

    C.-I. Chang, An e ffective evaluation tool for hyperspectral target detection: 3d re- ceiver operating characteristic curve analysis, IEEE Trans. Geosci. Remote Sens. 59 (6) (2020) 5131–5153. 35

    2020

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.