Pith. sign in

REVIEW 4 major objections 6 minor 39 references

SMMT: Siamese Motion Mamba with Self-attention for Thermal Infrared Target Tracking

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read SMMT: a Siamese Motion Mamba tracker that reports top accuracy on thermal infrared tracking benchmarks.

desk verdict The claimed temporal motion modeling never appears in the equations, so the paper's main mechanism is unsupported despite a familiar benchmark-style evaluation. read the letter →

arxiv 2505.04088 v3 pith:D3C77GYR submitted 2025-05-07 cs.CV

Shang Zhang , Huanbin Zhang , Dali Feng , Yujie Cui , Ruoyan Xiong , Cen He This is my paper · ORCID
classification cs.CV
keywords thermalinfraredtrackingSiamesenetworkstate-spacemodelMambaself-attentionmotionmodelingparametersharingtarget
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 claims that a Siamese tracker built around a Motion Mamba module — a bidirectional state-space scan with self-attention — plus a parameter-sharing strategy and a motion edge-aware regression loss is the new state of the art in thermal infrared (TIR) object tracking. On VOT-TIR2015 and VOT-TIR2017 it reports the highest accuracy of the compared trackers (0.79 and 0.74) and the best expected average overlap on VOT-TIR2017 (0.345). A sympathetic reader cares because TIR tracking must work under low contrast, occlusion, and motion blur, conditions that the paper argues are addressed by explicit motion modeling rather than appearance matching alone. The paper also reports large gains on LSOTB-TIR and PTB-TIR, with ablations attributing the improvement to the SMM module, the parameter-sharing strategy, and the new loss.

What carries the argument

The central object is the Siamese Motion Mamba (SMM) module, a pair of directional state-space scan branches — a horizontal state-space (HSS) model and a vertical state-space (VSS) model — that scan the feature map row-wise and column-wise, followed by a multi-head self-attention block. It sits inside a SiamRPN++ style Siamese architecture with an adaptive scale-weighting mechanism that favors the 1/4, 1/8, or 1/16 feature scale based on target size. The other two pieces are the Siamese parameter-sharing strategy (SPS), which shares convolutional weights between the two scan directions in layers 3 and 4 of the backbone, and the motion edge-aware regression loss, which combines IoU and L1 box regression with a gradient-edge term and a fine-grained feature-alignment term.

What would settle it

Inspect the SMM module's forward pass or trace Eq. (2): if the module consumes only the template image $I_Z$ and the search-region image $I_X$, with no previous/next frame feature tensor, then no temporal motion modeling takes place. A decisive experiment is to feed the tracker only the current frame; if accuracy is unchanged, the motion-modeling claim is unsupported.

Watch

Extended reading notes

Core claim

Put on its own terms, the paper claims that adding temporal motion modeling to a Siamese region-proposal tracker improves thermal infrared tracking. The Siamese Motion Mamba Tracker (SMMT) processes feature maps at three scales, scans them horizontally and vertically with a bidirectional state-space model, and applies multi-head self-attention at the final layer; a Siamese parameter-sharing strategy makes the two scan directions share convolutional weights in the deeper layers, and a motion edge-aware regression loss supervises both the box regression and the predicted edge/gradient field. The result is a tracker that the paper reports as best among all compared methods on four TIR benchmarks: highest accuracy 0.79 on VOT-TIR2015, 0.74 on VOT-TIR2017, best EAO 0.345 on VOT-TIR2017, and top precision/success on PTB-TIR and LSOTB-TIR.

Load-bearing premise

The load-bearing premise is that the Motion Mamba module actually receives temporal information from previous and next frames; the paper's equations define only template and search-region features, so if that temporal input path is absent the claimed motion modeling would not occur.

Editorial extensions

If this is right

  • On LSOTB-TIR, replacing the SiamRPN++ features with the SMM module raises precision from 74.0 to 77.8 and success from 55.4 to 58.6; adding parameter sharing plus the loss (full SMMT) raises them to 85.4 and 62.8.
  • On the VOT-TIR benchmarks, SMMT reports the best accuracy of every compared tracker on both VOT-TIR2015 (0.79) and VOT-TIR2017 (0.74), and the highest EAO on VOT-TIR2017 (0.345).
  • On PTB-TIR, the tracker reaches 83.9% precision and 63.6% success, and the paper reports it as the best method in the occlusion, fast-motion, and motion-blur scenario splits.
  • Because the parameter-sharing strategy reduces redundant computation between scan directions, the paper's design implies motion modeling can be added to Siamese trackers without multiplying the parameter count.

Reading between the lines

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

  • Inference: the claimed temporal motion modeling stands or falls on whether previous and next frame features actually enter the SMM module; Eq. (2) only shows template and search-region features being processed, so the temporal path is the part most worth verifying.
  • Inference: if the gain is real, the same state-space scan plus edge-aware loss could transfer to other low-texture modalities such as depth, radar, or night-vision tracking, where edge detail is similarly scarce.
  • Inference: a natural cheap experiment is to ablate the self-attention block alone and the edge loss alone on a motion-blur-heavy subset; the paper's scenario-level PTB-TIR results suggest the edge loss carries most of the motion-blur improvement.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes SMMT, a Siamese tracker for thermal infrared targets, which combines a ``Motion Mamba'' module with bidirectional horizontal/vertical state-space scans, a Siamese parameter-sharing strategy, and a motion edge-aware regression loss. The method is evaluated on LSOTB-TIR, PTB-TIR, VOT-TIR2015, and VOT-TIR2017, reporting strong results including accuracy scores of 0.79 and 0.74 and EAO scores of 0.376 and 0.345 on the VOT-TIR benchmarks, with component ablations on LSOTB-TIR and a parameter-sensitivity study on PTB-TIR. The central claim is that the proposed motion-modeling architecture and loss improve tracking accuracy and robustness in TIR imagery, especially under motion blur and occlusion.

Significance. If the architecture truly performed temporal motion modeling and the loss functions were as described, the reported gains would be a useful step for TIR tracking, a domain where motion blur, occlusion, and low texture are common. The paper has strengths: it evaluates on four established benchmarks, provides a component-wise ablation in Table 1, and reports scenario-level results on PTB-TIR. However, as written the manuscript does not provide a well-formed definition of the motion-modeling module, leaves several loss terms underspecified, and selects hyperparameters on PTB-TIR before reporting the final PTB-TIR results. These issues are load-bearing because they directly concern the core novelty and the validity of the reported benchmark comparisons. The mechanistic basis for the central claim is therefore not currently established by the text.

major comments (4)
  1. [§3.2–3.3, Eq. (2)] The claimed temporal motion modeling is not represented in the formal definition. Section 3.2 states that the SMM module extracts motion features from detection features of both the previous and next frame images, but Eq. (2) defines only transformations of the template feature F_base^Z and search feature F_base^X. The operators φ_V and φ_H are described as spatial row/column scans, and φ_SH is a Siamese parameter-sharing strategy rather than a feature transformation. There is no temporal index, no paired-frame input, and no cross-frame state in the equation, and the self-attention block mentioned in Section 3.2 is also absent. Consequently, the architecture as written does not implement temporal motion modeling, and the ablation gains in Table 1 cannot be attributed to that mechanism. The authors should either extend Eq. (2) with the actual temporal inputs and state-propagation equations, or substantially revise the motion-modeling claims.
  2. [§3.4, Eqs. (4)–(7)] The loss functions are not fully specified. Equation (5) refers to gradients of ground-truth and predicted images, ∇I_i and ∇Î_i, but the tracker outputs bounding boxes and does not produce images; the binary mask M_i and the index ``A-th image'' are undefined. Equation (6) uses E_F defined in Eq. (7) as a cross-entropy between distributions p and q, yet p_i and q_i are never defined, and no explanation is given for how they are computed from the tracker's features or output. Equation (7) also introduces W_i, F_i, and F̂_i without defining their provenance or dimensions. These omissions make the loss non-reproducible and prevent verification of the stated contribution of the motion-edge and fine-grained losses. Please provide precise definitions for every variable, including shapes and where each quantity is computed.
  3. [§4.2, Ablation Study of Parameters] The hyperparameters α and β in Eq. (4) are selected by scanning values on PTB-TIR, as described in Section 4.2, and then the final comparison in Section 4.2 reports PTB-TIR precision and success as the main result for that dataset. This is circular for PTB-TIR: the benchmark result is not an independent test of the loss coefficients. To support the reported PTB-TIR numbers, the authors should tune α and β on a held-out validation split and report PTB-TIR only as a test set, or justify the selection procedure and quantify its effect on the other benchmarks.
  4. [Table 2, §4.2 Performance Comparison] The text claims ``the best EAO score of 0.345 on VOT-TIR2017,'' which is consistent with Table 2, but it also states that SMMT ``achieves a score of 0.376 on VOT-TIR2015,'' which is not the best EAO on VOT-TIR2015 because UDCT reports EAO 0.420 on that dataset. The sentence ``SMMT achieves a score of 0.376 on VOT-TIR2015 and the best EAO score of 0.345 on VOT-TIR2017'' should be corrected to avoid implying state-of-the-art performance on both benchmarks, and the ``superior performance'' wording should be qualified accordingly.
minor comments (6)
  1. [§4.1] Reference [24] is cited for the VOT-TIR2017 benchmark, but the bibliography entry is ``The visual object tracking vot2017 challenge results''; please cite the correct VOT-TIR2017 benchmark publication.
  2. [§4.2] The text refers to ``Table 3'' when listing experimental results, but no Table 3 appears in the manuscript; either include the missing table or correct the cross-reference.
  3. [§4.2] Reference [26] is cited as ``CFNet'' in the list of compared trackers, but the bibliography entry for [26] is for CEDNet; the reference and the cited method do not match.
  4. [Figures 4 and 5] Figures 4 and 5 are used both in the parameter ablation (Figs. 4 and 5) and again in the performance comparison section, which also refers to Fig. 4 and Fig. 5; the figure numbering should be made consistent.
  5. [Eq. (5)] The phrase ``M_i is the binary mask for the A-th image'' contains a typo; it should be ``the i-th image.''
  6. [Table 1] The ablation increments in Table 1 are reported without variance or statistical significance, and some differences are small (e.g., Siam-SMM vs Siam-SPS success 58.6 vs 59.0); please report multiple runs or indicate whether the differences are consistent across random seeds.

Circularity Check

1 steps flagged · score 6.0 of 10

PTB-TIR benchmark result is partly circular because the regression-loss coefficients α and β were tuned on PTB-TIR, and the same benchmark is then reported as an independent performance comparison.

  1. fitted input called prediction [Section 4.2 (Ablation Study of Parameters) and Section 4 (Results on PTB-TIR)]
    "we conduct ablation experiments on the IoU loss coefficient α and the L1 loss coefficient β. This experiment aims to explore what values of α and β can make the best precision and success rate of the tracker. ... The results indicate that when α is set to 0.80 and β is set to 0.50, the tracker achieves the highest precision and success rate. Therefore, in this work, α is set to 0.8 and β is set to 0.5. ... As shown in Fig. 6 (a) and (b), our tracker achieves the highest precision and success rate on PTB-TIR, reaching 83.9% and 63.6%, respectively."

    The two loss weights α and β are selected by scanning values on PTB-TIR to maximize precision and success rate on that benchmark (Figs. 4–5). The final 'Results on PTB-TIR' section then reports the same benchmark as evidence of superiority. Thus the PTB-TIR numbers are not an independent prediction of the proposed architecture; they are a re-reporting of the tuning objective. The claimed 'highest precision and success rate on PTB-TIR' is therefore at least partly forced by the parameter search rather than derived from the model design. The VOT-TIR2015, VOT-TIR2017, and LSOTB-TIR results are not affected by this particular fit.

full rationale

The only concrete circular step is the PTB-TIR coefficient search: α and β are chosen by maximizing PTB-TIR precision and success rate, and the same PTB-TIR benchmark is then presented as an evaluation result. This makes the PTB-TIR comparison partly selected rather than predicted. The other three benchmarks, LSOTB-TIR, VOT-TIR2015, and VOT-TIR2017, are not used for this parameter fit and provide independent comparative evidence. No load-bearing self-citation chain appears in the paper: the authors' earlier related works are cited only as background context, and the Mamba/self-attention sources are external. The inconsistency between the claim that the SMM module extracts features from previous and next frame images and Eq. (2), which defines transformations only on template and search features, is a mechanistic gap and correctness risk rather than a circular reduction; it does not show an output equaling an input by definition. Overall, the PTB-TIR tuning warrants a partial circularity finding, while the central architectural comparison on the remaining benchmarks retains independent content.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several unstated numerical choices (alpha, beta, and many unspecified loss weights) and on architectural assumptions that are not validated or fully described.

free parameters (6)
  • alpha (IoU loss weight) = 0.8
    Tuned on PTB-TIR benchmark (Fig. 4) to maximize precision and success rate.
  • beta (L1 loss weight) = 0.5
    Tuned on PTB-TIR benchmark (Fig. 5).
  • lambda_reg, lambda_mot, lambda_fine = not specified
    Weights in Eq. (3) for regression, motion edge, and fine-grained losses; values are not given.
  • u, v (edge loss weights) = not specified
    Weights in Eq. (5); values are not given.
  • lambda_1, lambda_2, gamma (fine-grained loss weights) = not specified
    Weights in Eq. (6); values are not given.
  • T_small, T_large (scale thresholds) = not specified
    Thresholds for adaptive scale selection in Section 3.2; values are not given.
assumptions (4)
  • domain assumption Mamba-style selective state-space models are effective for sequence modeling
    The paper relies on the effectiveness of Mamba from [16] without justifying its application to 2D image features.
  • domain assumption ResNet-50 features are suitable for TIR tracking
    The backbone is assumed to provide good features for thermal infrared, though TIR has low resolution and sparse details.
  • domain assumption Gradient difference between ground truth and predicted images is a valid edge supervision signal
    The motion edge loss (Eq. 5) assumes predicted images exist and their gradients approximate target boundaries.
  • standard math Cross-entropy between distributions p and q measures global feature alignment
    Used in Eq. (7) but p and q are not defined in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SMMT: Siamese Motion Mamba with Self-attention for Thermal Infrared Target Tracking." pith.science (2026). https://pith.science/paper/D3C77GYR

@misc{pith2026250504088,
  author       = {Pith},
  title        = {Pith review of: SMMT: Siamese Motion Mamba with Self-attention for Thermal Infrared Target Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D3C77GYR}},
  note         = {Machine review of arXiv:2505.04088}
}
read the original abstract

Thermal infrared (TIR) object tracking often suffers from challenges such as target occlusion, motion blur, and background clutter, which significantly degrade the performance of trackers. To address these issues, this paper pro-poses a novel Siamese Motion Mamba Tracker (SMMT), which integrates a bidirectional state-space model and a self-attention mechanism. Specifically, we introduce the Motion Mamba module into the Siamese architecture to ex-tract motion features and recover overlooked edge details using bidirectional modeling and self-attention. We propose a Siamese parameter-sharing strate-gy that allows certain convolutional layers to share weights. This approach reduces computational redundancy while preserving strong feature represen-tation. In addition, we design a motion edge-aware regression loss to improve tracking accuracy, especially for motion-blurred targets. Extensive experi-ments are conducted on four TIR tracking benchmarks, including LSOTB-TIR, PTB-TIR, VOT-TIR2015, and VOT-TIR 2017. The results show that SMMT achieves superior performance in TIR target tracking.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 37 canonical work pages

  1. [1]

    F., Vedaldi, A., Torr, P

    Bertinetto, L., Valmadre, J., Henriques, J. F., Vedaldi, A., Torr, P. H.: Fully -convolutional siamese networks for object tracking. In: ECCV Workshops, 14, 850-865 (2016)

    work page 2016

  2. [2]

    In: CVPR, 8971-8980 (2018)

    Li, B., Yan, J., Wu, W., Zhu, Z., Hu, X.: High performance visual tracking with siamese region proposal network. In: CVPR, 8971-8980 (2018)

    work page 2018

  3. [3]

    In: ICCV, 4282-4291 (2019)

    Li, B., Wu, W., Wang, Q., Zhang, F., Xing, J., Yan, J.: Siamrpn++: Evolution of siamese visual tracking with very deep networks. In: ICCV, 4282-4291 (2019)

    work page 2019

  4. [4]

    In: CVPR, 9686-9696 (2023)

    Cao, J.; Pang, J.; Weng, X.; Khirodkar, R.; and Kitani, K.: Observation-centric sort: Re- thinking sort for robust multi-object tracking. In: CVPR, 9686-9696 (2023)

    work page 2023

  5. [5]

    : FOLT: Fast Multiple Object Tracking from UAV-captured Videos Based on Optical Flow

    Yao, M.; Wang, J.; Peng, J.; Chi, M.; and Liu, C. : FOLT: Fast Multiple Object Tracking from UAV-captured Videos Based on Optical Flow. In: ICME, 3375–3383 (2023)

    work page 2023

  6. [6]

    In: IROS, 3086–3092 (2021)

    Cao, Z.; Fu, C.; Ye, J.; Li, B.; Li, Y.: SiamAPN++: Siamese attentional aggregation network for real-time UAV tracking. In: IROS, 3086–3092 (2021)

    work page 2021

  7. [7]

    In: Trans

    Chen, Z., Zhong, B., Li, G., Zhang, S., Ji, R., Tang, Z., et al.: SiamBAN: Targetaware track- ing with siamese box adaptive network. In: Trans. Pattern Anal. Machine Intell, 5158–5173 (2022)

    work page 2022

  8. [8]

    In: CVPR, 6728-6737 (2020)

    Yu Y, Xiong Y, et al.: Deformable Siamese attention network for visual object tracking. In: CVPR, 6728-6737 (2020)

    work page 2020

Show all 39 references
  1. [9]

    Zhao, X.; Gong, D.; and Medioni, G. 2012. : Tracking using motion patterns for very crowded scenes. In ECCV, 315-328 (2012)

    2012

  2. [10]

    In: Infrared Physics & Technology, 103825 (2021)

    Yao T, Hu J, Zhang B, et al.: Scale and appearance variation enhanced siamese network for thermal infrared target tracking. In: Infrared Physics & Technology, 103825 (2021)

    2021

  3. [11]

    In: IEEE Conf

    Shuai, B.; Berneshawi, A.; Li, X.; Modolo, D.; and Tighe, J.: SiamMOT: Siamese Multi - Object Tracking. In: IEEE Conf. Comput. Vis. Pattern Recog, 12372–12382 (2021)

    2021

  4. [12]

    In: ICPRAI

    Xiong, R., Zhang, S., Zhang, Y., & Zhang, M.: SPECTER: A Tracker with Super-Resolution Siamese Network Reconstruction and Multi -Scale Feature Extraction for Thermal Infrared Pedestrian Tracking. In: ICPRAI. 7-16 (2024) SMMT: Siamese Motion Mamba Tracker 15

    2024

  5. [13]

    In: ICIC, 145-156 (2024)

    Xiong, R., Zhang, S., Zou, Y., & Zhang, Y.: SRCFT: A Correlation Filter Tracker with Siamese Super-Resolution Network and Sample Reliability Awareness for Thermal Infrared Target Tracking. In: ICIC, 145-156 (2024)

    2024

  6. [14]

    Bytetrack: Multiobject tracking by associating every detection box

    Zhang, Y.; Sun, P.; Jiang, Y.; Yu, D.; Weng, F.; Yuan, Z.; Luo, P.; Liu, W.; and Wang, X. Bytetrack: Multiobject tracking by associating every detection box. In: ECCV, 1–21. (2022)

    2022

  7. [15]

    : Chained-tracker: Chaining paired attentive regression results for end -to-end joint multiple-object detection and tracking

    Peng, J.; Wang, C.; Wan, F.; Wu, Y.; Wang, Y.; Tai, Y.; Wang, C.; Li, J.; Huang, F.; and Fu, Y. : Chained-tracker: Chaining paired attentive regression results for end -to-end joint multiple-object detection and tracking. In ECCV, 145161 (2020)

    2020

  8. [16]

    In: arXiv preprint arXiv, 2312.00752 (2023)

    Gu A, Dao T.: Mamba: Linear-time sequence modeling with selective state spaces. In: arXiv preprint arXiv, 2312.00752 (2023)

    2023 arXiv

  9. [17]

    : MM-Tracker: Motion Mamba with Margin Loss for UAV - platform Multiple Object Tracking

    Yao M, Peng J, He Q, et al. : MM-Tracker: Motion Mamba with Margin Loss for UAV - platform Multiple Object Tracking. In: arXiv preprint arXiv, 2407.10485 (2024)

    2024 arXiv

  10. [18]

    In: arXiv preprint arXiv, 2407.08083 (2024)

    Hatamizadeh A, Kautz J.: Mambavision: A hybrid mamba-transformer vision backbone. In: arXiv preprint arXiv, 2407.08083 (2024)

    2024 arXiv

  11. [19]

    : Parameter sharing and multi -granularity feature learning for cross-modality person re-identification

    Chan S, Du F, Tang T, et al. : Parameter sharing and multi -granularity feature learning for cross-modality person re-identification. In: Complex & Intelligent Systems, 949-962 (2024)

    2024

  12. [20]

    In: IJCAI 1092–1099 (2018)

    Ye M, Wang Z, Lan X, Yuen PC .: Visible thermal person re -identification via dual-con- strained top-ranking. In: IJCAI 1092–1099 (2018)

    2018

  13. [21]

    IEEE Trans

    Liu, Q., He, Z., Li, X., Zheng, Y.: PTB-TIR: A thermal infrared pedestrian tracking bench- mark. IEEE Trans. Multimedia, 22, 666-675 (2019)

    2019

  14. [22]

    In: ACM MM, 3847-3856 (2020)

    Liu, Q., Li, X., He, Z., Li, C., Li, J., Zhou, Z., Zheng, F.: LSOTB -TIR: A large-scale high- diversity thermal infrared object tracking benchmark. In: ACM MM, 3847-3856 (2020)

    2020

  15. [23]

    In: ICCV Workshops, 76-88 (2015)

    Felsberg, M., Berg, A., Hager, G., Ahlberg, J., et al.: The thermal infrared visual object tracking VOT-TIR2015 challenge results. In: ICCV Workshops, 76-88 (2015)

    2015

  16. [24]

    In: ICCV Workshops, 1949–1972 (2017)

    Kristan, M., Leonardis, A., Matas, J., Felsberg, M., Pflugfelder, R., et al.: The visual object tracking vot2017 challenge results. In: ICCV Workshops, 1949–1972 (2017)

    2017

  17. [25]

    In: ICCV, 1369-1378 (2019)

    Li, X., Ma, C., Wu, B., et al.: Target-aware deep tracking. In: ICCV, 1369-1378 (2019)

    2019

  18. [26]

    In: Pattern Recognition, 2302.06052 (2023)

    Zhang, G., Li, Z., Tang, C., Li, J., Hu, X.: CEDNet: A Cascade Encoder -Decoder Network for Dense Prediction. In: Pattern Recognition, 2302.06052 (2023)

    2023 arXiv

  19. [27]

    IEEE Trans

    Liu, Q., Li, X., He, Z., Fan, N., Yuan, D., Wang, H.: Learning deep multi -level similarity for thermal infrared object tracking. IEEE Trans. Multimedia, 23, 2114-2126 (2020)

    2020

  20. [28]

    H.: Vital: Visual tracking via adversarial learning

    Song, Y., Ma, C., Wu, X., Gong, L., Bao, L., Zuo, W., Yang, M. H.: Vital: Visual tracking via adversarial learning. In: CVPR, 8990-8999 (2018)

    2018

  21. [29]

    H.: Learning spatial -temporal regularized correlation filters for visual tracking

    Li, F., Tian, C., Zuo, W., Zhang, L., Yang, M. H.: Learning spatial -temporal regularized correlation filters for visual tracking. In: CVPR, 4904-4913 (2018)

    2018

  22. [30]

    In: CVPR, 4293-4302 (2016)

    Nam, H., Han, B.: Learning multi-domain convolutional neural networks for visual tracking. In: CVPR, 4293-4302 (2016)

    2016

  23. [31]

    Plants, 12, 2806 (2023)

    Gao, X., Tang, Z., Deng, Y., Hu, S., Zhao, H., Zhou, G.: HSSNet: A end-to-end network for detecting tiny targets of apple leaf diseases in complex backgrounds. Plants, 12, 2806 (2023)

    2023

  24. [32]

    S., Felsberg, M.: Atom: Accurate tracking by overlap maximization

    Danelljan, M., Bhat, G., Khan, F. S., Felsberg, M.: Atom: Accurate tracking by overlap maximization. In: ICCV, 4660-4669 (2019)

    2019

  25. [33]

    IEEE Trans

    Zhang, L., Gonzalez-Garcia, A., Van De Weijer, J., et al.: Synthetic data generation for end- to-end thermal infrared tracking. IEEE Trans. Image Process., 28, 1837-1850 (2018)

    2018

  26. [34]

    In: CVPR, 6638-6646 (2017)

    Danelljan, M., Bhat, G., Shahbaz Khan, F., Felsberg, M.: Eco: Efficient convolution opera- tors for tracking. In: CVPR, 6638-6646 (2017)

    2017

  27. [35]

    H.: Staple: Complementary learners for real-time tracking

    Bertinetto, L., Valmadre, J., Golodetz, S., Miksik, O., Torr, P. H.: Staple: Complementary learners for real-time tracking. In: CVPR, 1401-1409 (2016)

    2016

  28. [36]

    In: CVPR, 4844-4853 (2018) 16 XXXX et al

    Wang, N., Zhou, W., Tian, Q., Hong, R., Wang, M., Li, H.: Multi-cue correlation filters for robust visual tracking. In: CVPR, 4844-4853 (2018) 16 XXXX et al

    2018

  29. [37]

    Danelljan, G

    M. Danelljan, G. Häger, F. S. Khan, and M. Felsberg.: Accurate scale estimation for robust visual tracking. In: Proc. Brit. Mach. Vis. Conf, 1–5 (2014)

    2014

  30. [38]

    : Learning support correlation filters for visual tracking

    Zuo, W.; Wu, X.; Lin, L.; Zhang, L.; Yang, M.-H. : Learning support correlation filters for visual tracking. In: IEEE Trans. Pattern Anal, 1158–1172 (2018)

    2018

  31. [39]

    In: ICCV, 3074-3082 (2015)

    Ma C, Huang J B, Yang X, et al.: Hierarchical convolutional features for visual tracking. In: ICCV, 3074-3082 (2015)

    2015

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.