Pith. sign in

REVIEW 4 major objections 8 minor 69 references

MSP-Net: Manifold-Guided Spectral Prompt Network for Hyperspectral Object Tracking

T0 review · 4 major / 8 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 Hyperspectral tracking reaches new state-of-the-art on HOT2020 and HOT2023 by grouping spectral bands with a learned manifold graph, and refreshing prompts during inference.

desk verdict A novel, well-motivated hyperspectral tracker whose SOTA numbers rest on an uncontrolled benchmark protocol—worth refereeing, but the claims need reining in until baselines are retrained. read the letter →

arxiv 2608.09575 v1 pith:NJSOT7RR submitted 2026-08-10 cs.CV

Juliu Li , Hanlin Qin , Shuowen Yang , Jingjing Li , Yuedong Tan , Shuai Yuan , Huixin Zhou This is my paper · ORCID
classification cs.CV
keywords hyperspectralobjecttrackingspectralmanifoldbandgroupinggraph-drivenroutingpromptmodulationtemporalmemorycross-sensorgeneralizationmixtureofexperts
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

This paper tries to establish that hyperspectral object tracking should not treat the hundreds of spectral bands as a fixed, order-based RGB-like input. It argues that bands sit on a nonlinear manifold, that grouping them by learned graph-routed similarity preserves that structure, and that the spectral condition driving the tracker should be refreshed online as the target changes. On the HOT2020 and HOT2023 benchmarks the resulting network reports state-of-the-art numbers, specifically AUC 0.8072 with DP@20 0.9756 on HOT2020, and AUC 0.8192 with DP@20 0.9666 on the NIR2023 split, at 34.71 FPS. A sympathetic reader would care because sensor-agnostic, adaptive spectral grouping would remove the need to retune a tracker for every hyperspectral camera.

What carries the argument

The load-bearing object is the Graph-Driven Manifold Spectral Routing (GMSR) module, which constructs a band-relation graph in a spectral-manifold space. Each spectral band becomes a node represented by spatially average-pooled features; pairwise squared Euclidean distances between nodes are converted into a similarity adjacency $A_{i,j}=\exp(-\lVert V_i - V_j\rVert_2^2 / (2\zeta(\sigma)^2+\epsilon))$, where $\zeta$ is Softplus and $\sigma$ is a learned bandwidth. The adjacency is symmetrically normalized by the degree matrix and used in one graph-convolution layer with GELU activation, after which a routing matrix assigns bands to groups, using Gumbel-Softmax in training and softmax in inference, and group features are combined with mixture-of-experts weights. This mechanism is what lets non-adjacent but spectrally correlated bands share a group, and it is also what the paper's claim of manifold-guided adaptation rests on. Two secondary mechanisms carry the temporal part: the Decoupled Spectral-Condition Evolution (DSCE) strategy refreshes the prompt condition from intermediate template tokens at layers 3, 6, and 9 during inference with no parameter update, and the Temporal Morphology Memory-Adaptive (TMMA) post-processor gates the scale penalty using a sliding-window memory of confident past boxes.

What would settle it

Take the trained MSP-Net and replace only the GMSR adjacency construction and graph convolution with a plain learned soft-clustering layer that maps the same pooled band features to the same number of groups with the same MoE fusion; if the HOT2020 AUC stays within about 0.01 of 0.8072, the manifold graph is not carrying the reported gain. A second check would be to see whether the learned band groups persist across frames and across sensors; if they reshuffle randomly from frame to frame, there is no stable band manifold being exploited.

Watch

Extended reading notes

Core claim

The central claim is that replacing fixed, sequential, or weight-based band grouping with manifold-aware routing is enough to move hyperspectral tracking across sensors and over long sequences. The paper proposes MSP-Net, in which GMSR treats each band as a graph node, builds the adjacency from Gaussian-kernel distances between spatially pooled band features, convolves over that graph, and routes bands into groups via Gumbel-Softmax at training and softmax at inference. The grouped features and template statistics are fused into a condition vector, and a prompt generator turns it into dynamic prompts injected through cross-attention with cosine-similarity gating; during inference the condition is recomputed at layers 3, 6, and 9 from intermediate template tokens without adding parameters. A temporal post-processing module uses a memory of reliable past boxes to gate the scale penalty. The paper's reported results, including a 4.92% AUC gain over the previous best ProFit on HOT2020 and the leading AUC and DP@20 on NIR2023, are offered as evidence that the full system, and each of its components in ablations, contributes to the improvement.

Load-bearing premise

The load-bearing premise is that pairwise Euclidean distances between spatially averaged band features faithfully represent the true nonlinear structure of the spectral manifold, so the Gaussian-kernel adjacency matrix built from them is a valid description of inter-band relationships; if that premise fails, GMSR is just soft clustering of pooled band statistics and the claimed manifold advantage is unsupported.

Editorial extensions

If this is right

  • If the reported numbers hold, hyperspectral trackers no longer need to be redesigned per sensor: the same grouped spectral representation transfers across HOT2020, HOT2023, HOT2024, and IMEC25 splits.
  • Because DSCE updates conditions at inference time without fine-tuning, long sequences with illumination drift or deformation should keep their prompts aligned with the current target state.
  • Hard Gumbel-Softmax routing at training followed by soft routing at inference gives stable group boundaries without extra parameters, so the benefit is available at 34.71 FPS.
  • TMMA's gating of the scale penalty by historical morphology variance should reduce drift toward similar distractors without suppressing genuine deformation.

Reading between the lines

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

  • The manifold here is operational as a Gaussian-kernel soft clustering of spatially pooled band statistics; a natural test is whether a plain learned clustering with the same number of groups and the same MoE fusion reproduces the gains, which would tell us how much graph convolution specifically contributes.
  • The paper's own HOT2024 breakdown shows the advantage concentrates in NIR and weakens in VIS and RedNIR, where MSP-Net trails UBSTrack, suggesting the manifold premise is least reliable when the number of informative bands is small.
  • If the fixed update layers 3, 6, and 9 are the right cadence, one could extend the idea to an adaptive update schedule driven by tracking confidence, which the paper does not test.
  • Because the routing is derived from the search region and applied to the template, the architecture could be dropped into RGB trackers that accept arbitrary channel counts, making unknown-sensor generalization a directly testable downstream claim.
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 / 8 minor

Summary. The paper proposes MSP-Net, a hyperspectral object tracking framework composed of four modules: GMSR (graph-driven manifold spectral routing for adaptive band grouping), DSCPM (dual-stream spectral-conditioned prompt modulation), DSCE (decoupled spectral-condition evolution at inference), and TMMA (temporal morphology memory-adaptive post-processing). The architecture is described in Section III, and experiments are reported on HOT2020, HOT2023, HOT2024, and IMEC25. The central claim is state-of-the-art tracking accuracy, with headline numbers AUC 0.8072 and DP@20 0.9756 on HOT2020 (Table II) and AUC 0.8192 and DP@20 0.9666 on HOT2023 NIR (Table IV). Ablations in Section IV-F evaluate each component, and the paper includes visualizations of grouping, prompt modulation, and failure cases.

Significance. If the empirical claims are sustained, the paper addresses a genuine limitation of existing hyperspectral trackers: fixed band ordering and static fusion that ignore inter-band structure. The core formulation is coherent and built from standard building blocks (Gaussian-kernel graphs, Gumbel-Softmax routing, cross-attention prompts, and temporal post-processing). The evaluation is broad, covering four benchmarks and multiple spectral modalities, and the ablations are designed to isolate each module. The promised code release would aid reproducibility. However, the significance is currently constrained by an uncontrolled comparison protocol, missing statistical validation, and an internally inconsistent ablation narrative, so the state-of-the-art claim is not yet securely established.

major comments (4)
  1. [Section IV-A, Tables II and IV] The training protocol makes the SOTA comparison uncontrolled. The paper states that the model is trained exclusively on the HOT2024 training set and tested on HOT2020, HOT2023, and HOT2024 validation sets, but the baseline numbers in Tables II and IV are taken from prior publications without stating each baseline's training split. If those baselines were trained on the target dataset's own training set or otherwise saw the test distribution, the comparison conflates architecture improvements with training-data effects; if they saw less data, the comparison is still uncontrolled. The load-bearing nature is concrete: the reported gains over ProFit are 4.92% AUC on HOT2020 and 6.52% on NIR2023, while the no-DSCE ablation (Table VIII, Model-4, AUC 0.7980) differs from the full model by only about 1.2% relative, a margin that could be protocol drift or noise. The authors should retrain all baselines under the same protocol, or at minimum provide a per-baseline table of training splits and a sensitivity analysis. Section IV-A also says IMEC25 is trained and evaluated separately, but Section IV-F5 refers to 'joint training on HOT2024 and IMEC25'; this contradiction must be resolved.
  2. [Section IV-F1 and Table VIII] The ablation narrative is internally inconsistent and statistically unsupported. The text says that introducing DSCE (Model-5) 'increases AUC further by 1.17%', but Table VIII shows Model-5 AUC 0.8028 versus Model-4 AUC 0.7980, a relative increase of about 0.60%. In addition, Table XII shows Model-4 without TMMA has AUC 0.8042, which is higher than Model-4 with TMMA (0.7980), so the claim that TMMA consistently improves accuracy is not supported without qualification. All ablations are single runs with no error bars or significance tests, and the headline differences are comparable in size to the observed ablation fluctuations. The authors should report mean and standard deviation over multiple seeds and correct the percentage claims.
  3. [Section III-A, Eq. (2), Tables IX-X] The manifold premise of GMSR is not validated. Eq. (2) constructs a Gaussian-kernel adjacency matrix from pairwise Euclidean distances between spatially pooled band features, and the paper asserts that this 'characterizes the underlying spectral manifold,' but no evidence is given that this distance in pooled-feature space captures nonlinear spectral geometry, that the learned groups are physically coherent, or that the performance gain comes from manifold structure rather than from the added GCN/clustering parameters. The 'GMSR No Manifold' ablation in Table IX removes the graph convolution but retains the learned routing head; a direct comparison against a plain learned-clustering baseline without the manifold vocabulary, plus a coherence metric on the grouped bands (e.g., wavelength contiguity or agreement with material labels), is needed to support the central novelty claim.
  4. [Abstract, Table IV, and Figure 6] The headline claim in the abstract that MSP-Net achieves 'AUC and Precision exceeding 0.80 and 0.96, respectively' is not supported across the HOT2023 modalities. Table IV reports VIS2023 AUC 0.7576 and DP@20 0.9138, and RedNIR2023 AUC 0.6951 and DP@20 0.8339; only NIR2023 meets the thresholds. The abstract and conclusion should qualify the claim to the specific modalities or datasets. Additionally, the Figure 6 legend reports MSP-Net AUC 0.8105 and DP 0.9763, which differ from Table II's 0.8072 and 0.9756 without explanation; the authors should reconcile these values.
minor comments (8)
  1. [Section II title] The section title 'RELATEWORK' should be 'RELATED WORK'.
  2. [Figure 2 caption] The caption reads 'Temporal Pathology Memory Adaptive Post Processing'; this should be 'Temporal Morphology Memory Adaptive Post Processing' to match the method name.
  3. [Table III] The HOT-MoE row contains merged numeric entries without separators (e.g., '0.9650.9670.9670.9940.945 0.937 0.9611.0000.940'), making the table unreadable; the table generation should be fixed.
  4. [Table I] The Trans-DAT row shows '23.230.3GPU', with the FLOPs value and device concatenated; add proper spacing or column separation.
  5. [Eq. (5)] Equation (5) is a gradient-descent training update rule, not part of the forward model; consider moving it to the training setup section or clarifying that it describes parameter optimization.
  6. [Section IV-F5] The sentence 'Despite joint training on HOT2024 and IMEC25' contradicts Section IV-A, which states that IMEC25 is trained and evaluated separately; the wording should be corrected to match the actual protocol.
  7. [Section III-D] The TMMA hyperparameters (lambda = 0.5, K = 8, mu_r = 0.04, mu_a = 0.20, eta_base = 0.99) are set heuristically with no sensitivity analysis; a brief parameter study would strengthen the robustness claims.
  8. [Figure 9] The caption 'Example of cross spectral manifold grouping effect of GMSR model' does not explain how the visualization is produced or what the colors/edges represent; a few sentences of explanation would improve interpretability.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the AUC/DP results are external benchmark predictions; self-citations are related-work mentions and do not carry the argument.

full rationale

MSP-Net's headline numbers (AUC 0.8072, DP@20 0.9756 on HOT2020; AUC 0.8192, DP@20 0.9666 on NIR2023) are obtained by training on the HOT2024 training set and testing on the HOT2020/HOT2023 validation sets, which are external ground-truth annotations rather than quantities defined by the method's equations. The central derivation therefore does not reduce to its inputs. The only self-citations in the paper are related-work and motivation mentions: [3] appears in a range citation for the general claim that hyperspectral imaging supports reliable identification, [15] is part of a similar range, and [35] is cited when describing that MoE was introduced to hyperspectral tracking. None of these supplies a uniqueness theorem, a fitted parameter, or a premise that the MSP-Net result is constrained by definition. The GMSR 'manifold' construction (Eq. 2) is a Gaussian-kernel adjacency over spatially pooled band features; whether this captures a true physical manifold is a modeling-assumption and external-validity concern, not a circular step, since the reported AUC/DP values are measured against benchmark labels and are not derived from that equation. The DSCE/DSCPM module is a self-conditioning feature-modulation loop (the condition is computed from template and search features and then used to modulate those same features), but this is a standard network mechanism; the benchmark outcome is not definitionally forced by it. The uncontrolled baseline training protocol and the absence of error bars are correctness/reproducibility risks, explicitly outside the circularity definition. Overall: no significant circularity; the low score reflects only minor, non-load-bearing self-citations.

Assumptions & free parameters 10 free parameters · 7 assumptions · 3 invented entities

The central claim rests on several hand-set hyperparameters and domain assumptions about spectral manifolds, but introduces no new physical entities. The paper does not release code or data, so the exact values of learnable parameters such as sigma and the projection matrices are not independently auditable.

free parameters (10)
  • Gaussian kernel bandwidth sigma = learnable, value not reported
    Eq. (2) uses a learnable scalar sigma to control adjacency width; it is optimized on training data and directly shapes which bands are grouped.
  • Manifold node grid size = 4x4, 16 nodes
    Selected by ablation on HOT2020 in Table X; 2x2 and 8x8 both reduce performance, so the central GMSR grouping depends on this hand-chosen resolution.
  • Number of spectral groups G = not specified
    Eq. (6) routes bands into a 'predefined grouping space', but the group count is never reported; it is a free design parameter affecting all downstream condition vectors.
  • Condition update layers Omega = {3,6,9}
    Eq. (15) refreshes spectral conditions only at these Transformer layers; the choice is stated without sensitivity analysis.
  • TMMA displacement tolerance lambda = 0.5
    Eq. (18) uses this hand-set threshold for the spatial jump lock, with no sensitivity study.
  • TMMA memory length K = 8
    Sliding-window morphology memory length in Section III-D; hand-set without sensitivity analysis.
  • TMMA aspect-ratio variance threshold mu_r = 0.04
    Eq. (22) triggers deformation-aware scale modulation; hand-set without sensitivity analysis.
  • TMMA area fluctuation threshold mu_a = 0.20
    Eq. (22) triggers deformation-aware scale modulation; hand-set without sensitivity analysis.
  • TMMA baseline scale penalty eta_base = 0.99
    Eq. (22) sets the baseline background-suppression penalty; hand-set without sensitivity analysis.
  • TMMA reliability threshold tau_s = not reported
    Memory update requires tracking confidence above tau_s in Section III-D, but the value is omitted.
assumptions (7)
  • domain assumption Spectral bands lie on a nonlinear manifold whose intrinsic structure is captured by pairwise Euclidean distances between spatially pooled band features.
    Eq. (2) builds the adjacency matrix from pooled feature distances and calls it the spectral manifold; no independent evidence or metric validates this manifold assumption.
  • domain assumption A routing matrix estimated from the search region can be applied unchanged to the template branch while preserving aligned spectral representations.
    Section III-A states this alignment without proof; template and search regions have different content and scale.
  • domain assumption Intermediate template tokens at layers 3, 6, and 9 provide reliable target state signals for updating prompts during inference.
    DSCE in Eq. (15) refreshes conditions only from template tokens at these layers; there is no analysis of what these tokens encode or why these layers are optimal.
  • ad hoc to paper Hard Gumbel-Softmax routing during training transfers to softmax routing at inference without a harmful distribution shift.
    Eq. (6) switches between hard and soft assignment; the paper reports that soft routing hurts in ablation but still uses it at inference, assuming the discrepancy is benign.
  • domain assumption AUC and DP@20 values from different papers are comparable even when training data and protocols differ.
    Tables II through VII compare numbers drawn from prior papers without controlling training data or implementation details; MSP-Net is trained on HOT2024 and tested elsewhere.
  • standard math Symmetric normalization of the adjacency matrix stabilizes graph convolution gradients.
    Eq. (3) uses standard GCN normalization; this is accepted background material.
  • standard math Gumbel-Softmax provides differentiable discrete routing during training.
    Eq. (6) uses the standard Gumbel-Softmax estimator; this is a standard technique used as background.
invented entities (3)
  • Spectral manifold routing matrix N
    purpose: Partitions C spectral bands into G adaptive groups for feature aggregation.
    Internal latent partition with no external handle; it is evaluated only through downstream tracking scores.
  • 97-dimensional spectral condition vector C_cond
    purpose: Combines group statistics, template appearance, and MoE weights to generate dynamic prompts.
    Internal feature descriptor with no falsifiable prediction outside the network.
  • Temporal morphology memory bank Q_t
    purpose: Stores reliable historical aspect ratios and areas to modulate the scale penalty.
    Internal bookkeeping construct with no physical counterpart and no external verification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MSP-Net: Manifold-Guided Spectral Prompt Network for Hyperspectral Object Tracking." pith.science (2026). https://pith.science/paper/NJSOT7RR

@misc{pith2026260809575,
  author       = {Pith},
  title        = {Pith review of: MSP-Net: Manifold-Guided Spectral Prompt Network for Hyperspectral Object Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NJSOT7RR}},
  note         = {Machine review of arXiv:2608.09575}
}
read the original abstract

Hyperspectral object tracking leverages abundant spectral information to provide unique advantages for target discrimination in complex scenes. However, existing methods typically treat hyperspectral images as multi-channel extensions of RGB images, performing feature fusion in fixed band order. This approach leads to models dependent on specific sensor configurations while neglecting manifold relationships between bands, making generalization to heterogeneous sensors difficult. Moreover, the discriminative contribution of bands dynamically changes with target attributes and scene variations, further limiting the representational capacity of static fusion strategies. To address this, we propose the Manifold-Guided Spectral Prompt Network (MSP-Net). This network first reconstructs band relationships and forms adaptive spectral grouping through graph-driven manifold routing, then jointly integrates grouped spectral statistics with template appearance to construct target-related dynamic conditional prompts, enhancing target features while suppressing background interference. Furthermore, as tracking progresses, spectral conditions continuously evolve based on intermediate target representations, enabling target prompts to adapt in real-time to appearance and scene changes. Meanwhile, reliable historical states are used to constrain target localization and scale fluctuations, significantly improving temporal stability in cross-sensor tracking. Experiments on HOT2020 and HOT2023 demonstrate that MSP-Net achieves AUC and Precision exceeding 0.80 and 0.96, respectively, exhibiting exceptional robustness under heterogeneous sensors, target deformation, and complex background conditions. The code will be released at https://github.com/GGML668897/MSP-Net.

Figures

Figures reproduced from arXiv: 2608.09575 by the authors.

Figure 1
Figure 1. Comparison between the existing (a) fixed grouping dimensionality [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall framework diagram of MSP-Net, including the Graph Driven Manifold Spectral Routing (GMSR) module, Dual Stream Spectral Conditional [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Overview of the DSCPM module. The spectral conditions generated by DSCE are mapped into dynamic prompts, which modulate the joint [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Overview of the DSCE module. The module integrates the grouped search features, current template tokens, and MoE weights to construct a [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Framework of the TMMA module. The module constructs a temporal [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Precision and success plots of all competing hyperspectral trackers [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Visual comparisons, arranged from top to bottom, cover a set of [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Precision and success plots of all competing hyperspectral trackers [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Example of cross spectral manifold grouping effect of GMSR model. [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 11
Figure 11. Figure 11: MSP Net overlapping tracking results. (a)Student attribute: IV, SV [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]
Figure 12
Figure 12. Figure 12: Failure case analysis of MSP-Net, scenario is football2, pen, and [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

69 extracted references · 47 canonical work pages

  1. [1]

    Hyperspectral object tracking with dimensionality reduction and spa- tial–spectral–temporal modeling,

    Z. Li, J. Wang, J. Zhang, D. Chen, G. Fu, J. Wang, and J. Lu, “Hyperspectral object tracking with dimensionality reduction and spa- tial–spectral–temporal modeling,”ISPRS Journal of Photogrammetry and Remote Sensing, vol. 238, pp. 14–32, 2026

    work page 2026

  2. [2]

    Visual object tracking with discriminative filters and siamese networks: A survey and outlook,

    S. Javed, M. Danelljan, F. S. Khan, M. H. Khan, M. Felsberg, and J. Matas, “Visual object tracking with discriminative filters and siamese networks: A survey and outlook,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 5, pp. 6552–6574, 2023

    work page 2023

  3. [3]

    Hyper- track: A unified network for hyperspectral video object tracking,

    Y . Tan, W. Sun, J. Li, S. Hou, X. Li, Z. Wang, and B. Song, “Hyper- track: A unified network for hyperspectral video object tracking,”IEEE Transactions on Circuits and Systems for Video Technology, vol. 36, no. 1, pp. 1015–1028, 2026

    work page 2026

  4. [4]

    Sp- kan: Sparse-sine perception kolmogorov–arnold networks for infrared small target detection,

    S. Yuan, Y . Liu, X. Zhang, X. Yan, H. Qin, and N. Akhtar, “Sp- kan: Sparse-sine perception kolmogorov–arnold networks for infrared small target detection,”ISPRS Journal of Photogrammetry and Remote Sensing, vol. 234, pp. 1–19, 2026

    work page 2026

  5. [5]

    Online object tracking: A benchmark,

    Y . Wu, J. Lim, and M.-H. Yang, “Online object tracking: A benchmark,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2013

    work page 2013

  6. [6]

    Cass: A context-aware memory framework with spectral-spatial modulation for uav multispectral video tracking,

    Y . Chen, Q. Yuan, B. Zhong, X. Hu, J. Zhou, X. Zhao, X. Su, Y . Tang, and L. Zhang, “Cass: A context-aware memory framework with spectral-spatial modulation for uav multispectral video tracking,” IEEE Transactions on Circuits and Systems for Video Technology, 2026

    work page 2026

  7. [7]

    High-speed tracking with kernelized correlation filters,

    J. F. Henriques, R. Caseiro, P. Martins, and J. Batista, “High-speed tracking with kernelized correlation filters,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no. 3, pp. 583–596, 2015

    2015

  8. [8]

    Learning background- aware correlation filters for visual tracking,

    H. Kiani Galoogahi, A. Fagg, and S. Lucey, “Learning background- aware correlation filters for visual tracking,” inProceedings of the IEEE International Conference on Computer Vision (ICCV), Oct 2017

    work page 2017

Show all 69 references
  1. [9]

    Fully-convolutional siamese networks for object tracking,

    L. Bertinetto, J. Valmadre, J. F. Henriques, A. Vedaldi, and P. H. Torr, “Fully-convolutional siamese networks for object tracking,” inEuropean conference on computer vision. Springer, 2016, pp. 850–865

    2016

  2. [10]

    Joint feature learning and relation modeling for tracking: A one-stream framework,

    B. Ye, H. Chang, B. Ma, S. Shan, and X. Chen, “Joint feature learning and relation modeling for tracking: A one-stream framework,” inEuropean conference on computer vision. Springer, 2022, pp. 341– 357

    2022

  3. [11]

    Mixformer: End-to-end tracking with iterative mixed attention,

    Y . Cui, C. Jiang, L. Wang, and G. Wu, “Mixformer: End-to-end tracking with iterative mixed attention,” inProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, 2022, pp. 13 608– 13 618. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13

    2022

  4. [12]

    Learning spatio-temporal transformer for visual tracking,

    B. Yan, H. Peng, J. Fu, D. Wang, and H. Lu, “Learning spatio-temporal transformer for visual tracking,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 448–10 457

    2021

  5. [13]

    Material based object tracking in hyperspectral videos,

    F. Xiong, J. Zhou, and Y . Qian, “Material based object tracking in hyperspectral videos,”IEEE Transactions on Image Processing, vol. 29, pp. 3719–3733, 2020

    2020

  6. [14]

    Material- guided multiview fusion network for hyperspectral object tracking,

    Z. Li, F. Xiong, J. Zhou, J. Lu, Z. Zhao, and Y . Qian, “Material- guided multiview fusion network for hyperspectral object tracking,” IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–15, 2024

    2024

  7. [15]

    Variational bayesian inference-inspired network for compressive spectral imaging,

    S. Yang, H. Qin, S. Yuan, C. Tian, J. Li, and X. Yuan, “Variational bayesian inference-inspired network for compressive spectral imaging,” IEEE Transactions on Geoscience and Remote Sensing, 2025

    2025

  8. [16]

    Causal hyperprompter: A framework for unbiased hyperspectral camouflaged object tracking,

    H. Wang, W. Li, X.-G. Xia, and Q. Du, “Causal hyperprompter: A framework for unbiased hyperspectral camouflaged object tracking,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    2025

  9. [17]

    Hhtrack: Hyperspectral object tracking based on hybrid attention,

    Y . Tan, W. Sun, J. Yuan, W. Du, Z. Wang, N. Mao, and B. Song, “Hhtrack: Hyperspectral object tracking based on hybrid attention,” in 2023 13th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing (WHISPERS). IEEE, 2023, pp. 1–5

    2023

  10. [18]

    End-to-end unmixing with material prompts for hyperspectral object tracking,

    X. Han, M. A. Islam, L. Wang, Z. Long, G. Fu, W. Cai, K. K. Paliwal, and J. Zhou, “End-to-end unmixing with material prompts for hyperspectral object tracking,” 2026

    2026

  11. [19]

    Learning a deep ensemble network with band importance for hyperspectral object tracking,

    Z. Li, F. Xiong, J. Zhou, J. Lu, and Y . Qian, “Learning a deep ensemble network with band importance for hyperspectral object tracking,”IEEE Transactions on Image Processing, vol. 32, pp. 2901–2914, 2023

    2023

  12. [20]

    Siambag: Band attention grouping- based siamese object tracking network for hyperspectral videos,

    W. Li, Z. Hou, J. Zhou, and R. Tao, “Siambag: Band attention grouping- based siamese object tracking network for hyperspectral videos,”IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–12, 2023

    2023

  13. [21]

    Hyperspectral tracker with constrained object adaptive learning and trajectory construc- tion,

    Y . Wang, M. Ma, G. Zhang, Y . Liu, T. Gao, and S. Mei, “Hyperspectral tracker with constrained object adaptive learning and trajectory construc- tion,”IEEE Transactions on Circuits and Systems for Video Technology, 2025

    2025

  14. [22]

    Domain adapter for visual object tracking based on hyperspectral video,

    L. Gao, Y . Zhang, L. Chen, Y . Jiang, G. He, W. Xie, and Y . Li, “Domain adapter for visual object tracking based on hyperspectral video,”Pattern Recognition, p. 112296, 2025

    2025

  15. [23]

    Transformer- based band regrouping with feature refinement for hyperspectral ob- ject tracking,

    H. Wang, W. Li, X.-G. Xia, Q. Du, J. Tian, and Q. Shen, “Transformer- based band regrouping with feature refinement for hyperspectral ob- ject tracking,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–14, 2024

    2024

  16. [24]

    Tracking in aerial hyperspectral videos using deep kernelized correlation filters,

    B. Uzkent, A. Rangnekar, and M. J. Hoffman, “Tracking in aerial hyperspectral videos using deep kernelized correlation filters,”IEEE Transactions on Geoscience and Remote Sensing, vol. 57, no. 1, pp. 449–461, 2019

    2019

  17. [25]

    Object tracking in hyperspectral videos with convolutional features and kernelized correla- tion filter,

    K. Qian, J. Zhou, F. Xiong, H. Zhou, and J. Du, “Object tracking in hyperspectral videos with convolutional features and kernelized correla- tion filter,” inInternational conference on smart multimedia. Springer, 2018, pp. 308–319

    2018

  18. [26]

    Bae-net: A band attention aware ensemble network for hyperspectral object tracking,

    Z. Li, F. Xiong, J. Zhou, J. Wang, J. Lu, and Y . Qian, “Bae-net: A band attention aware ensemble network for hyperspectral object tracking,” in 2020 IEEE international Conference on image processing (ICIP). IEEE, 2020, pp. 2106–2110

    2020

  19. [27]

    A deep temporal- spectral–spatial anchor-free siamese tracking network for hyperspectral video object tracking,

    Z. Liu, Y . Zhong, G. Ma, X. Wang, and L. Zhang, “A deep temporal- spectral–spatial anchor-free siamese tracking network for hyperspectral video object tracking,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–16, 2024

    2024

  20. [28]

    Siamhyper: Learning a hyperspectral object tracker from an rgb-based tracker,

    Z. Liu, X. Wang, Y . Zhong, M. Shu, and C. Sun, “Siamhyper: Learning a hyperspectral object tracker from an rgb-based tracker,”IEEE Trans- actions on Image Processing, vol. 31, pp. 7116–7129, 2022

    2022

  21. [29]

    Spectral feature gating adapter for hyperspectral object tracking,

    W. Zhao, Y . Zhang, S. Gao, and G. He, “Spectral feature gating adapter for hyperspectral object tracking,”IEEE Access, vol. 14, pp. 35 427– 35 439, 2026

    2026

  22. [30]

    Outrageously large neural networks: The sparsely-gated mixture-of-experts layer,

    N. Shazeer, A. Mirhoseini, K. Maziarz, A. Davis, Q. Le, G. Hinton, and J. Dean, “Outrageously large neural networks: The sparsely-gated mixture-of-experts layer,” 2017

    2017

  23. [31]

    Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity,

    W. Fedus, B. Zoph, and N. Shazeer, “Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity,”Journal of Machine Learning Research, vol. 23, no. 120, pp. 1–39, 2022

    2022

  24. [32]

    Scaling vision with sparse mixture of experts,

    C. Riquelme, J. Puigcerver, B. Mustafa, M. Neumann, R. Jenatton, A. Susano Pinto, D. Keysers, and N. Houlsby, “Scaling vision with sparse mixture of experts,” inAdvances in Neural Information Process- ing Systems, M. Ranzato, A. Beygelzimer, Y . Dauphin, P. Liang, and J. W. Va...

    2021

  25. [33]

    Deepseekmoe: Towards ultimate expert specialization in mixture-of-experts language models,

    D. Dai, C. Deng, C. Zhao, R. X. Xu, H. Gao, D. Chen, J. Li, W. Zeng, X. Yu, Y . Wu, Z. Xie, Y . K. Li, P. Huang, F. Luo, C. Ruan, Z. Sui, and W. Liang, “Deepseekmoe: Towards ultimate expert specialization in mixture-of-experts language models,” 2024

    2024

  26. [34]

    Mixture of cluster-conditional lora experts for vision- language instruction tuning,

    Y . Gou, Z. Liu, K. Chen, L. Hong, H. Xu, Z. Li, D.-Y . Yeung, J. T. Kwok, and Y . Zhang, “Mixture of cluster-conditional lora experts for vision- language instruction tuning,”IEEE Transactions on Image Processing, vol. 35, pp. 3881–3892, 2026

    2026

  27. [35]

    Hotmoe: Exploring sparse mixture-of-experts for hyperspectral object tracking,

    W. Sun, Y . Tan, J. Li, S. Hou, X. Li, Y . Shao, Z. Wang, and B. Song, “Hotmoe: Exploring sparse mixture-of-experts for hyperspectral object tracking,”IEEE Transactions on Multimedia, vol. 27, pp. 4072–4083, 2025

    2025

  28. [36]

    Fsmt: Frequency- spatial mixture-of-experts hyperspectral object tracking,

    L. Gao, Z. Liu, Y . Zhang, Y . Jiang, W. Xie, and Y . Li, “Fsmt: Frequency- spatial mixture-of-experts hyperspectral object tracking,”IEEE Transac- tions on Circuits and Systems for Video Technology, 2026

    2026

  29. [37]

    Mamba: Linear-time sequence modeling with selective state spaces,

    A. Gu and T. Dao, “Mamba: Linear-time sequence modeling with selective state spaces,” 2024

    2024

  30. [38]

    The power of scale for parameter-efficient prompt tuning,

    B. Lester, R. Al-Rfou, and N. Constant, “The power of scale for parameter-efficient prompt tuning,” inProceedings of the 2021 Con- ference on Empirical Methods in Natural Language Processing. As- sociation for Computational Linguistics, 2021, pp. 3045–3059

    2021

  31. [39]

    Visual prompt tuning,

    M. Jia, L. Tang, B.-C. Chen, C. Cardie, S. Belongie, B. Hariharan, and S.-N. Lim, “Visual prompt tuning,” inComputer Vision–ECCV 2022. Springer, 2022, pp. 709–727

    2022

  32. [40]

    Adapt- Former: Adapting vision transformers for scalable visual recognition,

    S. Chen, C. Ge, Z. Tong, J. Wang, Y . Song, J. Wang, and P. Luo, “Adapt- Former: Adapting vision transformers for scalable visual recognition,” inAdvances in Neural Information Processing Systems, vol. 35, 2022, pp. 16 664–16 678

    2022

  33. [41]

    Explicit visual prompting for low-level structure segmentations,

    W. Liu, X. Shen, C.-M. Pun, and X. Cun, “Explicit visual prompting for low-level structure segmentations,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 19 434–19 445

    2023

  34. [42]

    Domain prompt learning with quaternion networks,

    Q. Cao, Z. Xu, Y . Chen, C. Ma, and X. Yang, “Domain prompt learning with quaternion networks,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 26 637–26 646

    2024

  35. [43]

    Prompting for multi-modal tracking,

    J. Yang, Z. Li, F. Zheng, A. Leonardis, and J. Song, “Prompting for multi-modal tracking,” inProceedings of the 30th ACM International Conference on Multimedia, 2022, pp. 3492–3500

    2022

  36. [44]

    Visual prompt multi- modal tracking,

    J. Zhu, S. Lai, X. Chen, D. Wang, and H. Lu, “Visual prompt multi- modal tracking,” inProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, 2023, pp. 9516–9526

    2023

  37. [45]

    Bi-directional adapter for multi- modal tracking,

    B. Cao, J. Guo, P. Zhu, and Q. Hu, “Bi-directional adapter for multi- modal tracking,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 2, 2024, pp. 927–935

    2024

  38. [46]

    Hyperspectral object tracking with spectral information prompt,

    G. He, L. Gao, L. Chen, Y . Jiang, W. Xie, and Y . Li, “Hyperspectral object tracking with spectral information prompt,”IEEE Transactions on Circuits and Systems for Video Technology, 2025

    2025

  39. [47]

    PHTrack: Prompting for hyperspectral video tracking,

    Y . Chen, Y . Tang, X. Su, J. Li, Y . Xiao, J. He, and Q. Yuan, “PHTrack: Prompting for hyperspectral video tracking,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–18, 2024, art. no. 5533918

    2024

  40. [48]

    ProFiT: A prompt-guided frequency-aware filtering and template-enhanced interaction framework for hyperspectral video track- ing,

    Y . Chen, Q. Yuan, Y . Tang, X. Wang, Y . Xiao, J. He, Z. Lihe, and X. Jin, “ProFiT: A prompt-guided frequency-aware filtering and template-enhanced interaction framework for hyperspectral video track- ing,”ISPRS Journal of Photogrammetry and Remote Sensing, vol. 226, pp. 164–...

    2025

  41. [49]

    Nonlinear dimensionality reduction by locally linear embedding,

    S. T. Roweis and L. K. Saul, “Nonlinear dimensionality reduction by locally linear embedding,”science, vol. 290, no. 5500, pp. 2323–2326, 2000

    2000

  42. [50]

    Manifold-learning- based feature extraction for classification of hyperspectral data: A review of advances in manifold learning,

    D. Lunga, S. Prasad, M. M. Crawford, and O. Ersoy, “Manifold-learning- based feature extraction for classification of hyperspectral data: A review of advances in manifold learning,”IEEE Signal Processing Magazine, vol. 31, no. 1, pp. 55–66, 2013

    2013

  43. [51]

    Learning a robust local manifold representation for hyperspectral dimensionality reduction,

    D. Hong, N. Yokoya, and X. X. Zhu, “Learning a robust local manifold representation for hyperspectral dimensionality reduction,”IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 10, no. 6, pp. 2960–2975, 2017

    2017

  44. [52]

    Local-manifold- learning-based graph construction for semisupervised hyperspectral image classification,

    L. Ma, M. M. Crawford, X. Yang, and Y . Guo, “Local-manifold- learning-based graph construction for semisupervised hyperspectral image classification,”IEEE Transactions on Geoscience and Remote Sensing, vol. 53, no. 5, pp. 2832–2844, 2014

    2014

  45. [53]

    Graph-regularized fast and robust principal compo- nent analysis for hyperspectral band selection,

    W. Sun and Q. Du, “Graph-regularized fast and robust principal compo- nent analysis for hyperspectral band selection,”IEEE Transactions on Geoscience and Remote Sensing, vol. 56, no. 6, pp. 3185–3195, 2018

    2018

  46. [54]

    Laplacian-regularized low- rank subspace clustering for hyperspectral image band selection,

    H. Zhai, H. Zhang, L. Zhang, and P. Li, “Laplacian-regularized low- rank subspace clustering for hyperspectral image band selection,”IEEE Transactions on Geoscience and Remote Sensing, vol. 57, no. 3, pp. 1723–1740, 2019. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 14

    2019

  47. [55]

    Unsupervised manifold-preserving and weakly redundant band selection method for hyperspectral imagery,

    C. Sui, C. Li, J. Feng, and X. Mei, “Unsupervised manifold-preserving and weakly redundant band selection method for hyperspectral imagery,” IEEE Transactions on Geoscience and Remote Sensing, vol. 58, no. 2, pp. 1156–1170, 2020

    2020

  48. [56]

    Salient band selection for hyperspectral image classification via manifold ranking,

    Q. Wang, J. Lin, and Y . Yuan, “Salient band selection for hyperspectral image classification via manifold ranking,”IEEE Transactions on Neural Networks and Learning Systems, vol. 27, no. 6, pp. 1279–1289, 2016

    2016

  49. [57]

    High-dimensional noise to low-dimensional manifolds: A manifold-space diffusion framework for degraded hyperspectral image classification,

    B. Yang, N. Chen, X. Yue, Y . Luo, Y . Fan, H. Zhang, H. Ma, J. Yue, and S. Mao, “High-dimensional noise to low-dimensional manifolds: A manifold-space diffusion framework for degraded hyperspectral image classification,” 2026

    2026

  50. [58]

    Histograms of oriented mosaic gradients for snapshot spectral image description,

    L. Chen, Y . Zhao, J. C.-W. Chan, and S. G. Kong, “Histograms of oriented mosaic gradients for snapshot spectral image description,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 183, pp. 79–93, 2022

    2022

  51. [59]

    Ssttrack: A unified hyperspectral video tracking framework via modeling spectral-spatial-temporal conditions,

    Y . Chen, Q. Yuan, Y . Tang, Y . Xiao, J. He, T. Han, Z. Liu, and L. Zhang, “Ssttrack: A unified hyperspectral video tracking framework via modeling spectral-spatial-temporal conditions,”Information Fusion, vol. 114, p. 102658, 2025

    2025

  52. [60]

    Spirit: Spectral awareness interaction network with dynamic template for hyperspectral object tracking,

    Y . Chen, Q. Yuan, Y . Tang, Y . Xiao, J. He, and L. Zhang, “Spirit: Spectral awareness interaction network with dynamic template for hyperspectral object tracking,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–16, 2023

    2023

  53. [61]

    Domain adaptation- aware transformer for hyperspectral object tracking,

    Y . Wu, L. Jiao, X. Liu, F. Liu, S. Yang, and L. Li, “Domain adaptation- aware transformer for hyperspectral object tracking,”IEEE Transactions on Circuits and Systems for Video Technology, vol. 34, no. 9, pp. 8041– 8052, 2024

    2024

  54. [62]

    Tipsst: Transformer-based hyperspectral object tracking with illumi- nation prompts and spatial-spectral-temporal context,

    X. Teng, W. Jiang, P. Xiang, D. Zhao, S. Qi, J. Song, and H. Zhou, “Tipsst: Transformer-based hyperspectral object tracking with illumi- nation prompts and spatial-spectral-temporal context,”Applied Soft Computing, p. 114807, 2026

    2026

  55. [63]

    Sense: Hyperspectral video object tracker via fusing material and motion cues,

    Y . Chen, Q. Yuan, Y . Tang, Y . Xiao, J. He, and Z. Liu, “Sense: Hyperspectral video object tracker via fusing material and motion cues,” Information Fusion, vol. 109, p. 102395, 2024

    2024

  56. [64]

    Multi-domain universal representation learning for hyperspectral object tracking,

    Z. Li, F. Xiong, J. Lu, J. Wang, D. Chen, J. Zhou, and Y . Qian, “Multi-domain universal representation learning for hyperspectral object tracking,”Pattern Recognition, vol. 162, p. 111389, 2025

    2025

  57. [65]

    Hyperspectral mamba for hyperspectral object tracking,

    L. Gao, Y . Zhang, Y . Jiang, W. Xie, and Y . Li, “Hyperspectral mamba for hyperspectral object tracking,”arXiv preprint arXiv:2509.08265, 2025

    2025 arXiv

  58. [66]

    A unified spatial-spectral-temporal network for hyperspectral object tracking,

    Z. Li, J. Wang, J. Zhang, D. Zhao, G. Fu, J. Wang, and J. Lu, “A unified spatial-spectral-temporal network for hyperspectral object tracking,”Pattern Recognition, p. 113005, 2025

    2025

  59. [67]

    Ubstrack: Unified band selection and multi-model ensemble for hyperspectral object tracking,

    M. A. Islam, J. Zhou, W. Xing, Y . Gao, and K. K. Paliwal, “Ubstrack: Unified band selection and multi-model ensemble for hyperspectral object tracking,”IEEE Transactions on Geoscience and Remote Sensing, 2025

    2025

  60. [68]

    Siamtu: Hyperspectral video tracking against appearance changes using improved siamegat with adaptive template updating,

    J. Chen, K. Qian, W. Song, Y . Qiu, and S. Wang, “Siamtu: Hyperspectral video tracking against appearance changes using improved siamegat with adaptive template updating,”Infrared Physics & Technology, vol. 147, p. 105811, 2025

    2025

  61. [69]

    Sptrack: Spectral similarity prompt learning for hyperspectral object tracking,

    G. Guo, Z. Li, W. An, Y . Wang, X. He, Y . Luo, Q. Ling, M. Li, and Z. Lin, “Sptrack: Spectral similarity prompt learning for hyperspectral object tracking,”Remote Sensing, vol. 16, no. 16, p. 2975, 2024

    2024

Pith tools

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