Pith. sign in

REVIEW 4 major objections 5 minor 57 references

ClickTrack: Towards Real-time Interactive Single Object Tracking

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A single click, optionally labeled with a category, can initialize a single-object tracker almost as well as a hand-drawn bounding box.

desk verdict ClickTrack cleanly packages a known point-to-box idea for SOT initialization and backs it with solid GOT-10k evidence, but the LaSOT comparison is unfair and the click simulation is too narrow to support the real-time interactive claim. read the letter →

arxiv 2411.13183 v2 pith:VGTZC4LT submitted 2024-11-20 cs.CV

Kuiran Wang , Xuehui Yu , Wenwen Yu , Guorong Li , Xiangyuan Lan , Qixiang Ye , Jianbin Jiao , Zhenjun Han This is my paper · ORCID
classification cs.CV
keywords singleobjecttrackingclickinteractionpointinitializationGuidedRefinerboundingboxregressionvideosegmentationtext-guidedrefinementreal-timeinteractive
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

Single-object trackers normally demand a precise bounding box around the target in the first frame, and even small annotation errors visibly degrade long-term accuracy. This paper proposes replacing that box with a single click, optionally accompanied by a one-word category when the target is ambiguous, and letting a learned refiner generate the box. The Guided Click Refiner (GCR) turns the click into the four edges of the operator's intended box, is tracker-agnostic, and runs fast enough for real-time interactive use. Combined with a transformer tracker, it reports a 65.0 success score on LaSOT and 63.2 average overlap on GOT-10k, close to the same tracker's performance with hand-drawn boxes. If this holds, pointing at an object becomes a sufficient initialization for robust tracking, and the same refiner also cleans up point-based segmentation with models like SAM.

What carries the argument

The load-bearing object is the Guided Click Refiner (GCR), a point-to-box regressor built from three parts. Guided Convolution (GC) is the core: it takes an RoI feature and a guiding feature — either a CLIP text embedding of the object category or a learnable proposal feature — and uses the guiding feature to generate dynamic projection parameters that modulate the visual feature, so the text can steer which part of the object the box should cover. Prototype Selection (PS) places several anchor boxes around the click, uses GC to adjust them and to predict each adjusted anchor's IoU, and keeps the highest-scoring one as the initial regression region. Iterative Refinement (IR) then cascades several GC-like stages to narrow the box toward the target. The whole pipeline regresses four distances from the click to the box edges, which converts the point into a tracker-ready template.

What would settle it

Run a user study on LaSOT in which operators freely click on the target object and optionally speak a category, feed those genuine clicks through GCR, and compare the success rate with the synthetic-ellipse numbers (62.4 point-only, 65.0 with text); a substantial drop would show that the ellipse simulation does not model real clicking behavior.

Watch

Extended reading notes

Core claim

The paper's central claim is that single-object tracking can be reformulated as $b = \mathcal{T}(\mathcal{S}, \mathcal{R}(p_0))$: given a tracker $\mathcal{T}$, a video $\mathcal{S}$, and an initial click $p_0$, a point-to-box regressor $\mathcal{R}$ produces the initial template box, so no hand-drawn bounding box is required. The regression targets are the distances from the click to the four edges of the target box. To resolve the ambiguity that a bare click carries, $\mathcal{R}$ is the Guided Click Refiner (GCR), which accepts an optional text or learnable guiding feature: a Guided Convolution fuses that feature into the visual RoI representation, a Prototype Selection module picks the most promising anchor region by predicted IoU, and an Iterative Refinement cascade sharpens the box. With this design and CLIP text features for category guidance, GCR plus STARK reaches 65.0 success and 68.3 precision on LaSOT using point-plus-text, and 62.4 success with point alone; on GOT-10k it reaches 63.2 mean average overlap. The same GCR structure, attached to SAM as a prompt generator, lifts COCO mask overlap from roughly 52 to 70 IoU, showing the ambiguity cure transfers beyond tracking.

Load-bearing premise

The reported stability rests on the assumption that uniformly sampling points inside an ellipse whose semi-axes are one-fourth of the ground-truth box width and height reproduces how real operators actually click; there is no user study checking this, and on video-segmentation benchmarks the category labels were added manually rather than spoken by an operator.

Editorial extensions

If this is right

  • Real-time interactive trackers can be initialized by a point click instead of a laborious box annotation, removing a major usability bottleneck in live systems.
  • Adding a one-word category (e.g., 'license plate' vs 'car') measurably removes ambiguity, raising LaSOT success from 62.4 to 65.0 when combined with text.
  • GCR is tracker-agnostic: it works with STARK, MixFormer, and PrDimp, so the click-to-box module can be bolted onto any existing single-object tracker.
  • The refiner transfers to segmentation: plugging GCR into SAM raises single-point mask IoU on COCO substantially and improves VOS initialization on DAVIS and YT-VOS.
  • GCR runs at 31 FPS and adds only about 0.03 seconds per video sequence, so the new initialization paradigm preserves real-time operation.

Reading between the lines

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

  • The training-time assumption that clicks land uniformly inside an ellipse one-quarter the box width and height is a proxy for real operator behavior; genuine human clicks on boundaries, on occluded parts, or on background clutter may not follow that distribution, so a user study would be the natural next test.
  • Because the guiding feature can be a learnable vector as well as text, GCR could be extended to accept other cheap cues — a voice keyword, a finger gesture, or even a rough sketch — without changing the architecture.
  • The open-vocabulary behavior of the text path is inherited from CLIP, so the practical ceiling of the text guidance is set by the category vocabulary CLIP has seen; a category outside that vocabulary would likely fall back to the learnable-proposal behavior.
  • The same point-plus-optional-category pattern could be applied to other interactive vision tasks that currently require precise boxes or masks, such as interactive detection or multi-object tracking initialization.
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

4 major / 5 minor

Summary. The paper proposes ClickTrack, a new paradigm for initializing single-object trackers in which the operator provides a single click (with optional category text) and a learned point-to-box refiner, GCR, converts the click into an initial bounding box for an off-the-shelf tracker. GCR consists of a Guided Convolution that fuses visual RoI features with a text or learnable guiding feature, a Prototype Selection module that picks an anchor prototype, and an Iterative Refinement module that cascades refinements. Experiments combine GCR with STARK on LaSOT and GOT-10k, reporting 65.0 success on LaSOT with point+text and 63.2 mAO on GOT-10k with point only; ablations analyze the modules, number of stages, anchor settings, and robustness to click position. The paper also extends the GCR head to SAM (GCR-SAM) and reports improved box/mask IoU on COCO and improved VOS initialization on DAVIS 2017 and YT-VOS 2018.

Significance. If the reported results held under realistic interactive conditions, ClickTrack would be a valuable contribution: it replaces precise box drawing with a single click, which is faster and more stable, and the optional text input addresses ambiguity. The GOT-10k comparison, where all methods are trained on the same split, provides some evidence that GCR improves over simpler point-to-box baselines, and the paper ships extensive ablations and a transfer study to SAM. A notable strength is that GCR-SAM is evaluated on DAVIS and YT-VOS without using their training data. However, the significance is tempered by the fact that the robustness evaluation uses the same synthetic click distribution as training, the LaSOT comparison is not controlled for in-domain training, and the GCR-SAM comparison is confounded by the additional text input. The headline claims of real-time interactive stability are therefore not yet fully established.

major comments (4)
  1. [Section 4.2 / Table 7] The training procedure samples random points uniformly inside an ellipse with semi-axes one-fourth of the ground-truth box width and height, and the robustness analysis in Table 7 evaluates the same distribution by re-sampling points at training time. This makes the robustness result an interpolation result under the training distribution rather than evidence about realistic operator clicks, which are known to cluster near meaningful parts or centers and can land on boundaries. Since the paper's value proposition is specifically real-time interactive scenarios, I consider this mismatch load-bearing. Please add at least one of the following: a user study with human clicks, an evaluation on a dataset containing human click annotations, or a sensitivity analysis with alternative click distributions (e.g., center-biased Gaussian, boundary clicks, part-level clicks) that are not used in training.
  2. [Table 1 / Section 4.2] The headline comparison on LaSOT is not apples-to-apples: GCR is trained on the LaSOT training split, while the SAM, OVSAM, and GroundingDINO baselines are used off-the-shelf and are not fine-tuned on LaSOT. The performance gap in Table 1 (e.g., 65.0 vs 60.1 for SAM-B in the Point setting) may therefore be due to in-domain training rather than to the proposed architecture. On GOT-10k all methods are trained on the same split, which supports the comparison there. For a clean LaSOT comparison, please either fine-tune the baselines on LaSOT, report GCR in a zero-shot or cross-dataset configuration, or explicitly state the training protocol of each method.
  3. [Table 8 / Section 4.5] The simulation experiments comparing box and point initialization do not specify how the 'Box' initialization boxes are generated. The reported success of 54.1 for Box is far below the known performance of precise-box STARK on LaSOT (typically above 60), indicating that the boxes are corrupted by an unspecified noise model. Without a definition of this model, the conclusion in Section 4.5 that 'point input provides higher accuracy for single-object tracking initialization' is uninterpretable. Please specify the box noise protocol and, ideally, compare against a precise-box upper bound.
  4. [Section 5.3 / Tables 9-11] The GCR-SAM experiments compare GCR-SAM (which always receives a category text prompt) against SAM baselines that receive only the point prompt. The reported gains could therefore be attributed to the additional text modality rather than to the GCR head, and the claim that the GCR structure itself alleviates SAM's ambiguity is not isolated. Please report a point-only GCR-SAM variant (using the learnable proposal feature) alongside the text-conditioned variants, or otherwise control for the text input in the comparison.
minor comments (5)
  1. [Table 1 / Table 8] Table 1 contains the typos 'Dtector + Point' and 'Detector (top.1)', and Table 8 contains 'Norm-Pricise' instead of 'Norm-Precise'.
  2. [Figure 8] The caption of Figure 8 reads 'The framework of TCR-SAM' but should be 'The framework of GCR-SAM'.
  3. [Section 4.2] The sentence 'Each epoch randomly samples 64,000 images in the training dataset' is ambiguous; please clarify whether 64,000 is the number of images per epoch or per iteration.
  4. [Tables 10-11] In Tables 10 and 11, the 'Upper Bound' row for UNINEXT on YT-VOS presents numbers that run together (e.g., '72.1 72.3 76.2 65.9 74.177.0 76.8 81.0 70.8 79.4'), which should be reformatted for readability.
  5. [General] The paper does not mention whether code or trained models will be released; providing them would support reproducibility of the reported numbers.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ClickTrack is a standard supervised point-to-box pipeline with held-out evaluation, and self-citations are not load-bearing.

full rationale

The paper's derivation chain transforms the classic tracking formulation b = T(S, b0) into b = T(S, R(p0)), where R is a learned point-to-box regressor. There is no equation in which the output quantity is defined in terms of the predicted quantity: GCR is trained with supervised regression against ground-truth bounding-box edge distances (Eqs. 3-5) and evaluated on held-out LaSOT and GOT-10k test splits. The headline numbers (65.0 success on LaSOT, 63.2 mAO on GOT-10k) are genuine empirical outputs of that evaluation, not fitted constants renamed as predictions. The uniform-ellipse point sampling described in Sec. 4.2 is used both during training and in the Table 7 robustness test, which indeed limits how strongly the results transfer to real operator clicks; however, this is an external-validity limitation rather than circularity, because the success/IoU metrics are computed against ground-truth boxes and are independent of how the click positions were sampled. The authors' earlier point-supervision works (Refs. [8] and [11]) are cited as related work and as motivation, but they do not carry the central argument: the Guided Convolution, Prototype Selection, Iterative Refinement, and the ClickTrack integration with STARK are self-contained contributions compared against external baselines such as SAM, GroundingDino, and natural-language trackers. No uniqueness theorem, ansatz, or fitting procedure is imported from a self-citation as the justification for the paper's claims. Accordingly, no circular step can be exhibited from the paper's own equations or citation chain, and the appropriate finding is no significant circularity.

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

The GCR architecture is a learned module, not a new physical entity or assumption with independent falsifiable handles. The main assumptions are about the interaction modality and the click distribution, both of which are untested against real users.

free parameters (6)
  • Anchor prototype setting = 4 scales (32^2, 64^2, 128^2, 256^2) x 3 ratios (2:1, 1:1, 1:2)
    Selected via ablation on LaSOT (Table 6) to maximize success rate; this is a hand-tuned design choice that affects the initial regression region.
  • Number of iterative refinement stages = 2
    Chosen via ablation (Table 5), where 2 stages give the best success (65.0) and faster speed than 4 or 6.
  • Middle channel MC of Guided Convolution
    Hyperparameter of the dynamic parameter projection in Eq. 6; its value is not reported in the paper.
  • RoI feature size S
    Spatial size of the RoI feature in GCR; not reported in the paper.
  • Click sampling ellipse semi-axes = one-fourth of ground-truth width and height
    Training-time random points are sampled inside this ellipse (Section 4.2); this defines the simulated click distribution that the robustness results depend on.
  • GCR-SAM training subset from Objects365 = 139,491 images, 284,967 instances
    Subset chosen due to computational limits (Section 5.3); the selection rule is not specified.
assumptions (4)
  • domain assumption Benchmark annotations in LaSOT and GOT-10k accurately reflect the target object and are used without correction.
    All training and evaluation rely on these ground-truth boxes; if they are noisy, the refiner learns that noise.
  • domain assumption A single click plus an optional category word is a sufficient interface for a user to specify the tracking target.
    This is the core interaction premise of ClickTrack; the paper does not conduct a user study to verify that operators can reliably provide such input in real time.
  • domain assumption CLIP text embeddings for category words align with the visual RoI features after the MLP in GC.
    The Guided Convolution relies on this alignment to steer the regression (Section 4.6); the paper provides only qualitative Eigen-CAM visualizations.
  • ad hoc to paper Random points sampled uniformly in an ellipse with semi-axes one-fourth of the box represent real clicks.
    This distribution is used for training and for the robustness test in Table 7; it is an ad hoc modeling choice not validated against human behavior.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ClickTrack: Towards Real-time Interactive Single Object Tracking." pith.science (2026). https://pith.science/paper/VGTZC4LT

@misc{pith2026241113183,
  author       = {Pith},
  title        = {Pith review of: ClickTrack: Towards Real-time Interactive Single Object Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VGTZC4LT}},
  note         = {Machine review of arXiv:2411.13183}
}
read the original abstract

Single object tracking(SOT) relies on precise object bounding box initialization. In this paper, we reconsidered the deficiencies in the current approaches to initializing single object trackers and propose a new paradigm for single object tracking algorithms, ClickTrack, a new paradigm using clicking interaction for real-time scenarios. Moreover, click as an input type inherently lack hierarchical information. To address ambiguity in certain special scenarios, we designed the Guided Click Refiner(GCR), which accepts point and optional textual information as inputs, transforming the point into the bounding box expected by the operator. The bounding box will be used as input of single object trackers. Experiments on LaSOT and GOT-10k benchmarks show that tracker combined with GCR achieves stable performance in real-time interactive scenarios. Furthermore, we explored the integration of GCR into the Segment Anything model(SAM), significantly reducing ambiguity issues when SAM receives point inputs.

Figures

Figures reproduced from arXiv: 2411.13183 by the authors.

Figure 1
Figure 1. (a) The performance drops when the deviation rate of annotated bounding box increases. (b) The [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Different initialization methods for single object tracker. Building on the considerations outlined above, a Guided Convolution (GC) struc￾ture was initially designed to accommodate both visual and guiding features. The guid￾ing features may be simple category information derived from text generated by a nat￾ural language model, or they can be learnable features. The structure of the guided convolution allows for in… view at source ↗
Figure 3
Figure 3. Tracking ambiguity: when the point is clicked on the overlap region, [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The Guided Click Refiner (GCR) framework, including the Prototype Selection module and Iter [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Visualization of bounding box (blue) generated with the same one point (green) and different text information (red). 4.6. GCR visualization We select representative visualization results, as shown in [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]
Figure 6
Figure 6. Figure 6: Visualization of bounding box generated with di [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: Visualization of RoI features generated by GC structure with di [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]
Figure 8
Figure 8. Figure 8: The framework of TCR-SAM. 5.2. The framework of GCR-SAM In this section, we present the comprehensive framework of GCR-SAM. As de￾picted in [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]
Figure 9
Figure 9. Figure 9: Visualization of GCR-SAM. It is obvious that GCR-SAM e [PITH_FULL_IMAGE:figures/full_fig_p025_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 47 canonical work pages

  1. [1]

    H. Fan, L. Lin, F. Yang, P. Chu, G. Deng, S. Yu, H. Bai, Y . Xu, C. Liao, H. Ling, Lasot: A high-quality benchmark for large-scale single object tracking, in: Pro- ceedings of the IEEE /CVF conference on computer vision and pattern recogni- tion, 2019, pp. 5374–5383

    work page 2019

  2. [2]

    Q. Feng, V . Ablavsky, Q. Bai, G. Li, S. Sclaroff, Real-time visual object tracking with natural language description, in: IEEE Winter Conference on Applications of Computer Vision, W ACV 2020, Snowmass Village, CO, USA, March 1-5, 2020, IEEE, 2020, pp. 689–698

    work page 2020

  3. [3]

    X. Wang, X. Shu, Z. Zhang, B. Jiang, Y . Wang, Y . Tian, F. Wu, Towards more flex- ible and accurate object tracking with natural language: Algorithms and bench- mark, in: Proceedings of the IEEE /CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 13763–13773

    work page 2021

  4. [4]

    Y . Li, J. Yu, Z. Cai, Y . Pan, Cross-modal target retrieval for tracking by natural language, in: IEEE /CVF Conference on Computer Vision and Pattern Recogni- tion Workshops, CVPR Workshops 2022, New Orleans, LA, USA, June 19-20, 2022, IEEE, 2022, pp. 4927–4936

    work page 2022

  5. [5]

    L. Zhou, Z. Zhou, K. Mao, Z. He, Joint visual grounding and tracking with natural language specification, in: Proceedings of the IEEE /CVF Conference on Com- puter Vision and Pattern Recognition, 2023, pp. 23151–23160. 27

    work page 2023

  6. [6]

    Q. Feng, V . Ablavsky, Q. Bai, S. Sclaro ff, Siamese natural language tracker: Tracking by natural language descriptions with siamese trackers, in: E. Zaimis (Ed.), IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021, V ol. 42, Computer Vision Foundation / IEEE, 2021, pp. 5851–5860

    work page 2021

  7. [7]

    D. P. Papadopoulos, J. R. R. Uijlings, F. K. et al., Training object class detectors with click supervision, in: CVPR, 2017

    work page 2017

  8. [8]

    P. Chen, X. Yu, X. Han, N. Hassan, K. Wang, J. Li, J. Zhao, H. Shi, Z. Han, Q. Ye, Point-to-box network for accurate object detection via single point supervision, in: European Conference on Computer Vision, Springer, 2022, pp. 51–67

    work page 2022

Show all 57 references
  1. [9]

    Kirillov, E

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, et al., Segment anything, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 4015– 4026

    2023

  2. [10]

    M. Liao, Z. Guo, , Y . W. et al., Attentionshift: Iteratively estimated part-based attention map for pointly supervised instance segmentation, in: CVPR, 2023

    2023

  3. [11]

    X. Yu, P. Chen, D. Wu, N. Hassan, G. Li, J. Yan, H. Shi, Q. Ye, Z. Han, Ob- ject localization under single coarse point supervision, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 4868–4877

    2022

  4. [12]

    Huang, X

    L. Huang, X. Zhao, K. Huang, Got-10k: A large high-diversity benchmark for generic object tracking in the wild, IEEE transactions on pattern analysis and machine intelligence 43 (5) (2019) 1562–1577

    2019

  5. [13]

    B. Yan, H. Peng, J. Fu, D. Wang, H. Lu, Learning spatio-temporal transformer for visual tracking, in: Proceedings of the IEEE /CVF International Conference on Computer Vision, 2021, pp. 10448–10457. 28

    2021

  6. [14]

    Khoreva, R

    A. Khoreva, R. Benenson, E. Ilg, T. Brox, B. Schiele, Lucid data dreaming for video object segmentation, International Journal of Computer Vision 127 (9) (2019) 1175–1197

    2019

  7. [15]

    B. Yan, Y . Jiang, J. Wu, D. Wang, P. Luo, Z. Yuan, H. Lu, Universal instance per- ception as object discovery and retrieval, in: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, 2023, pp. 15325–15336

    2023

  8. [16]

    Zhang, Z

    Y . Zhang, Z. Wu, H. Peng, S. Lin, A transductive approach for video object seg- mentation, in: Proceedings of the IEEE /CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 56949–6958

    2020

  9. [17]

    Robinson, F

    A. Robinson, F. J. Lawin, M. Danelljan, F. S. Khan, M. Felsberg, Learning fast and robust target models for video object segmentation, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 7406–7415

    2020

  10. [18]

    Z. Li, R. Tao, E. Gavves, C. G. M. Snoek, A. W. M. Smeulders, Tracking by natural language specification, in: 2017 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, Honolulu, HI, USA, July 21-26, 2017, IEEE Computer Society, 2017, pp. 7350–7358

    2017

  11. [19]

    Z. Yang, T. Kumar, T. Chen, J. Su, J. Luo, Grounding-tracking-integration, IEEE Transactions on Circuits and Systems for Video Technology 31 (9) (2020) 3433– 3443

    2020

  12. [20]

    Bertinetto, J

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

    2016

  13. [21]

    B. Li, J. Yan, W. Wu, Z. Zhu, X. Hu, High performance visual tracking with siamese region proposal network, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 8971–8980

    2018

  14. [22]

    B. Li, W. Wu, Q. Wang, F. Zhang, J. Xing, J. Yan, Siamrpn ++: Evolution of siamese visual tracking with very deep networks, in: Proceedings of the 29 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 4282–4291

    2019

  15. [23]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural infor- mation processing systems 30 (2017)

    2017

  16. [24]

    Y . Li, X. Liu, D. Yuan, J. Wang, P. Wu, J. Liu, A transformer-based visual object tracker via learning immediate appearance change, Pattern Recognition (2024) 110705

    2024

  17. [25]

    L. Gao, L. Chen, P. Liu, Y . Jiang, Y . Li, J. Ning, Transformer-based visual ob- ject tracking via fine–coarse concatenated attention and cross concatenated mlp, Pattern Recognition 146 (2024) 109964

    2024

  18. [26]

    Q. Song, C. Wang, Z. Jiang, Y . Wang, Y . Tai, C. Wang, J. Li, F. Huang, Y . Wu, Rethinking counting and localization in crowds: A purely point-based framework, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 3365–3374

    2021

  19. [27]

    L. Chen, T. Yang, X. Zhang, W. Zhang, J. Sun, Points as queries: Weakly semi- supervised object detection by points, in: Proceedings of the IEEE /CVF Confer- ence on Computer Vision and Pattern Recognition, 2021, pp. 8823–8832

    2021

  20. [28]

    Tabernik, J

    D. Tabernik, J. Muhovi ˇc, D. Skoˇcaj, Dense center-direction regression for object counting and localization with point supervision, Pattern Recognition 153 (2024) 110540

    2024

  21. [29]

    J. Liew, Y . Wei, W. Xiong, S.-H. Ong, J. Feng, Regional interactive image seg- mentation networks, in: 2017 IEEE international conference on computer vision (ICCV), IEEE, 2017, pp. 2746–2754

    2017

  22. [30]

    Maninis, S

    K.-K. Maninis, S. Caelles, J. Pont-Tuset, L. Van Gool, Deep extreme cut: From extreme points to object segmentation, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 616–625. 30

    2018

  23. [31]

    Z. Li, Q. Chen, V . Koltun, Interactive image segmentation with latent diversity, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recog- nition, 2018, pp. 577–585

    2018

  24. [32]

    Benenson, S

    R. Benenson, S. Popov, V . Ferrari, Large-scale interactive object segmentation with human annotators, in: Proceedings of the IEEE /CVF Conference on Com- puter Vision and Pattern Recognition, 2019, pp. 11700–11709

    2019

  25. [33]

    J. Lu, D. Batra, D. Parikh, S. Lee, Vilbert: Pretraining task-agnostic visiolinguis- tic representations for vision-and-language tasks, Advances in neural information processing systems 32 (2019)

    2019

  26. [34]

    L. Zhou, H. Palangi, L. Zhang, H. Hu, J. Corso, J. Gao, Unified vision-language pre-training for image captioning and vqa, in: Proceedings of the AAAI confer- ence on artificial intelligence, V ol. 34, 2020, pp. 13041–13049

    2020

  27. [35]

    Z. Fan, Z. Wei, S. Wang, R. Wang, Z. Li, H. Shan, X. Huang, Tcic: Theme concepts learning cross language and vision for image captioning, in: IJCAI, 2021, pp. 657–663

    2021

  28. [36]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al., Learning transferable visual models from natural language supervision, in: International Conference on Machine Learning, PMLR, 2021, pp. 8748–8763

    2021

  29. [37]

    Kim, Y .-W

    Y .-E. Kim, Y .-W. Lee, S.-W. Lee, Lc-msm: Language-conditioned masked seg- mentation model for unsupervised domain adaptation, Pattern Recognition 148 (2024) 110201

    2024

  30. [38]

    H. Song, J. Bang, Prompt-guided detr with roi-pruned masked attention for open- vocabulary object detection, Pattern Recognition (2024) 110648

    2024

  31. [39]

    L. H. Li, P. Zhang, H. Zhang, J. Yang, C. Li, Y . Zhong, L. Wang, L. Yuan, L. Zhang, J.-N. Hwang, et al., Grounded language-image pre-training, in: Pro- ceedings of the IEEE /CVF Conference on Computer Vision and Pattern Recog- nitio, 2022, pp. 10965–10975. 31

    2022

  32. [40]

    H. Luo, L. Ji, M. Zhong, Y . Chen, W. Lei, N. Duan, T. Li, Clip4clip: An empirical study of clip for end to end video clip retrieval and captioning, Neurocomputing 508 (2022) 293–304

    2022

  33. [41]

    S. Ren, K. He, R. Girshick, J. Sun, Faster r-cnn: Towards real-time object detec- tion with region proposal networks, Advances in neural information processing systems 28 (2015)

    2015

  34. [42]

    Z. Tian, C. Shen, H. Chen, T. He, Fcos: Fully convolutional one-stage object detection, in: Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 9627–9636

    2019

  35. [43]

    K. He, G. Gkioxari, P. Doll ´ar, R. Girshick, Mask r-cnn, in: Proceedings of the IEEE international conference on computer vision, 2017, pp. 2961–2969

    2017

  36. [44]

    Y . Chen, X. Dai, M. Liu, D. Chen, L. Yuan, Z. Liu, Dynamic convolution: At- tention over convolution kernels, in: 2020 IEEE /CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020, Seattle, W A, USA, June 13-19, 2020, Computer Vision Foundation / IEEE, 2020, pp....

    2020

  37. [45]

    Contributors, Mmtracking: Openmmlab video perception toolbox and bench- mark (2020)

    M. Contributors, Mmtracking: Openmmlab video perception toolbox and bench- mark (2020)

    2020

  38. [46]

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

    2016

  39. [47]

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, L. Fei-Fei, Imagenet: A large-scale hierarchical image database, in: 2009 IEEE conference on computer vision and pattern recognition, Ieee, 2009, pp. 248–255

    2009

  40. [48]

    S. Liu, Z. Zeng, T. Ren, F. Li, H. Zhang, J. Yang, C. Li, J. Yang, H. Su, J. Zhu, et al., Grounding dino: Marrying dino with grounded pre-training for open-set object detection, arXiv preprint arXiv:2303.05499 (2023). 32

    2023 arXiv

  41. [49]

    H. Yuan, X. Li, C. Zhou, Y . Li, K. Chen, C. C. Loy, Open-vocabulary sam: Seg- ment and recognize twenty-thousand classes interactively, in: ECCV , 2024

    2024

  42. [50]

    Jocher, A

    G. Jocher, A. Stoken, J. Borovec, A. Chaurasia, L. Changyu, A. Hogan, J. Hajek, L. Diaconu, Y . Kwon, Y . Defretin, et al., ultralytics/yolov5: v5. 0-yolov5-p6 1280 models, aws, supervise. ly and youtube integrations, Zenodo (2021)

    2021

  43. [51]

    Y . Cui, C. Jiang, L. Wang, G. Wu, Mixformer: End-to-end tracking with itera- tive mixed attention, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 13608–13618

    2022

  44. [52]

    Danelljan, L

    M. Danelljan, L. V . Gool, R. Timofte, Probabilistic regression for visual tracking, in: Proceedings of the IEEE /CVF conference on computer vision and pattern recognition, 2020, pp. 7183–7192

    2020

  45. [53]

    M. B. Muhammad, M. Yeasin, Eigen-cam: Class activation map using principal components, in: 2020 international joint conference on neural networks (IJCNN), IEEE, 2020, pp. 1–7

    2020

  46. [54]

    Q. Wang, L. Zhang, L. Bertinetto, W. Hu, P. H. Torr, Fast online object tracking and segmentation: A unifying approach, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2019, pp. 1328–1338

    2019

  47. [55]

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, C. L. Zitnick, Microsoft coco: Common objects in context, in: Computer Vision– ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13, Springer, 2014...

    2014

  48. [56]

    Pont-Tuset, F

    J. Pont-Tuset, F. Perazzi, S. Caelles, P. Arbel ´aez, A. Sorkine-Hornung, L. Van Gool, The 2017 davis challenge on video object segmentation, arXiv preprint arXiv:1704.00675 (2017)

    2017 arXiv

  49. [57]

    N. Xu, L. Yang, Y . Fan, D. Yue, Y . Liang, J. Yang, T. Huang, Youtube- vos: A large-scale video object segmentation benchmark, arXiv preprint arXiv:1809.03327 (2018). 33 Kuiran Wang received the B.E. degree in computer science and technology from Central South University, Chi...

    2018 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.