Pith. sign in

REVIEW 4 major objections 4 minor 40 references

FusionSORT: Fusion Methods for Online Multi-object Visual Tracking

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper shows that in online multi-object visual tracking, the choice of fusion method for combining motion, appearance, height-IoU, and confidence costs determines whether weak cues improve or degrade association performance.

desk verdict An honest but confounded empirical comparison of fusion operators for SORT-style tracking; the headline ranking overreaches the data. read the letter →

arxiv 2501.00843 v3 pith:SVJ54UMS submitted 2025-01-01 cs.CV

Nathanael L. Baisa This is my paper · ORCID
classification cs.CV
keywords multi-objecttrackingdataassociationfusionmethodsmotionandappearancecuesheight-IoUtrackletconfidenceKalmanfiltergatingSORT
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 argues that the way detection-to-tracklet costs are fused, the fusion method, matters as much as which cues are used in online multi-object visual tracking. It builds a SORT-style tracker and compares four fusion recipes: taking the minimum of per-cue costs, a weighted sum with IoU-based motion, Kalman-filter gating with Mahalanobis motion, and an element-wise (Hadamard) product of costs. On validation splits of MOT17, MOT20, and DanceTrack, the paper finds that the weighted sum based on IoU is the most favorable when weak cues (height-IoU and tracklet confidence) are combined with strong cues, while minimum fusion works for motion plus appearance but degrades once weak cues are added. The intended takeaway is practical guidance for choosing the right fusion method for a given cue set in data association.

What carries the argument

The load-bearing object is the fused cost matrix $C$ whose $(i,j)$ entry is the combined association cost between tracklet $i$ and detection $j$. The paper builds four versions of $C$ from four per-cue cost matrices: appearance cosine distance, IoU distance, height-IoU distance (one-dimensional overlap of the boxes' vertical extent), and confidence distance (absolute difference between the tracklet confidence state and the detection score). The fusion operators are $\min(C_a,C_m,C_h,C_c)$, the weighted sum $\lambda_1 C_m + \lambda_2 C_a + \lambda_3 C_h + \lambda_4 C_c$, KF gating $C = \lambda(C_a + \lambda_h C_h + \lambda_c C_c) + (1-\lambda)C_m$ using Mahalanobis motion subject to gating, and the element-wise product $C_a \odot C_m \odot C_h \odot C_c$. The same Hungarian solver performs the assignment in all cases, so performance differences are attributed to the fusion operator itself.

What would settle it

Re-run the same four fusion methods on the same validation splits while sweeping the fusion weights (for example, varying the appearance weight from 0.05 to 0.5 and the height-IoU and confidence weights from 0 to 0.3) and the matching thresholds; if the weighted-sum advantage over minimum disappears under a different weight setting, the conclusion that the fusion method is the key factor would be shown to be parameter-dependent rather than intrinsic.

Watch

Extended reading notes

Core claim

The paper's central claim is that the fusion method is at least as important as the cues themselves for data association in multi-object visual tracking. Using a tracking-by-detection pipeline with a Kalman filter whose state vector is extended with tracklet confidence and its velocity, the author compares four fusion operators on the validation sets of MOT17, MOT20, and DanceTrack: minimum, weighted sum based on IoU, Kalman-filter gating with Mahalanobis distance, and Hadamard product. The consistent pattern is that minimum and Hadamard fusion benefit from adding appearance to motion but lose performance when height-IoU and confidence cues are added, whereas weighted-sum based on IoU and KF gating can exploit those weak cues. On DanceTrack, KF gating trails the other methods overall, so the author recommends weighted sum based on IoU as the favourable choice when weak cues accompany strong cues.

Load-bearing premise

The reported ranking assumes the hand-set fusion weights, matching rejection thresholds, and detection score thresholds, chosen on the same validation sets used for comparison, are equally fair to all four fusion methods; the paper does not test how sensitive the ranking is to those parameters.

Editorial extensions

If this is right

  • Trackers that currently fuse many cues with minimum or Hadamard product can expect an association-quality gain by switching to an IoU-based weighted sum, since weak cues then contribute instead of hurting.
  • The two-stage association scheme can keep plain IoU in the low-confidence second stage; fusion is only needed where appearance is available, so practitioners do not pay fusion cost on low-confidence boxes.
  • KF gating's weak-cue gains on crowded scenes and its lower overall results on the diverse-motion benchmark indicate that the choice between IoU-based and Mahalanobis-based motion fusion should depend on the motion pattern of the scene.
  • Extending the Kalman state with confidence and its velocity is a cheap addition that lets a tracker use detection score as a smoothly predicted association cue rather than a per-frame threshold.
  • On uniform-appearance, diverse-motion scenes, minimum fusion with motion plus appearance remains competitive, so the simple minimum rule is still a reasonable default when only strong cues are used.

Reading between the lines

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

  • A natural extension the paper does not run is a weight-sensitivity sweep: the recommendation would be stronger if the weighted-sum advantage persisted across a range of $\lambda$ values, since the current weights are fixed on the same validation sets used for comparison.
  • The conclusion may transfer to other tracking-by-detection systems only insofar as their detectors and appearance embeddings have similar error profiles; testing with a different detector or re-identification model would show whether fusion choice interacts with cue quality.
  • Because height-IoU and confidence costs are nearly free to compute, the weighted-sum result suggests a practical recipe: keep the strong motion term dominant and use weak cues as small tie-breakers rather than as equal partners.
  • One could test the fusion methods on held-out test sets or on other benchmarks with nonlinear motion to see whether the DanceTrack ranking generalizes beyond the three validation sets reported.
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 / 4 minor

Summary. The paper proposes FusionSORT, an online tracking-by-detection framework, and uses it to compare four fusion methods for combining strong cues (motion, appearance) with weak cues (height-IoU, tracklet confidence) in the data-association step. The four methods are minimum fusion, weighted sum based on IoU, Kalman-filter (KF) gating with Mahalanobis distance, and Hadamard product of cost matrices. Experiments are conducted on the validation sets of MOT17, MOT20, and DanceTrack, reporting MOTA, IDF1, and HOTA for different cue combinations. The central claim is that the choice of fusion method matters and that the weighted sum based on IoU is more favorable when weak cues are added to strong cues. The paper also introduces a confidence state into the Kalman-filter state vector and releases source code.

Significance. If the comparison were clean, this study would offer useful practical guidance to the tracking community, since many trackers adopt different fusion rules without a systematic comparison. The paper's strengths include its breadth across three benchmarks, the use of publicly available detectors and re-identification models, the absence of tracklet interpolation, and a supplementary comparison of IoU versus Mahalanobis distance for the second association stage (Table IV). The code is publicly available, which supports reproducibility. However, the empirical comparison does not isolate the fusion operator from other design choices, and the headline conclusion is not consistently supported by the paper's own tables. These issues substantially weaken the central claims in their current form.

major comments (4)
  1. [§III.C, Tables I-III] The four fusion methods are not compared under controlled conditions, so observed differences cannot be attributed to the fusion operator alone. KF gating (Eq. 19) uses Mahalanobis distance as the motion cost, whereas the other three methods use IoU (Eq. 18, Eq. 20, and the minimum method). The threshold policy also differs: minimum and Hadamard apply the masks of Eqs. (10)-(12), weighted sum applies only the appearance threshold, and KF gating applies no thresholding or 0.5 multiplication. The weight scales differ as well (λ1=1.0, λ2=0.1 for weighted sum versus λ=0.98, λh=λc=0.2 for KF gating). To substantiate the claim that fusion method is key, the paper needs a controlled ablation that keeps the motion metric, threshold policy, and weight scales fixed while changing only the fusion rule, or at least disentangles these effects through a sensitivity analysis.
  2. [§V Conclusion, Tables I and II] The conclusion that 'the weighted sum based on IoU is more favourable when using weak cues along with the strong cues' is contradicted by the reported results. On MOT17 (Table I), the weighted-sum HOTA with all cues is 68.812, which is lower than the 69.379 obtained without weak cues, and KF gating achieves a higher HOTA (69.366) than weighted sum (68.812) when all cues are used. On MOT20 (Table II), KF gating with all cues reaches 58.695 HOTA against 58.368 for weighted sum. Weak cues also degrade KF gating on DanceTrack (Table III: 49.08 with mot+app versus 48.695 with all cues). The per-dataset behavior should be reported honestly, and the global conclusion should be reformulated accordingly or supported by additional evidence.
  3. [§IV.B, Tables I-III] All reported numbers are from single runs without error bars, significance tests, or sensitivity analyses, while some of the key differences are very small (for example, 69.366 versus 68.812 HOTA on MOT17). The fusion weights (λ1...λ4), thresholds (θiou, θemb), and detection score thresholds (τ1, τ2) were selected empirically on the same validation sets used to compare the methods, so the ranking could shift under different parameter choices. The paper should include a sensitivity analysis over these parameters and, where feasible, multiple runs or per-sequence statistics to assess whether the observed differences are robust.
  4. [§III.C.4, Eq. (20)] The Hadamard-product fusion is described as treating all cues with equal importance implicitly, but this is not fully accurate in the implementation because the masks of Eqs. (10)-(12) are applied to the appearance, height-IoU, and confidence costs before multiplication. This means the Hadamard method also includes a thresholding policy that the weighted-sum and KF-gating methods do not share. The paper should clarify how the threshold masks affect the scale and meaning of the Hadamard cost, and whether the observed degradation of Hadamard is due to the multiplicative combination itself or to the interaction with the threshold masks.
minor comments (4)
  1. [§III.B, Eq. (9) and Eq. (17)] The confidence cost is defined twice, in Eq. (9) and Eq. (17); the earlier definition could be removed to avoid redundancy.
  2. [§III.C.3, Eq. (19)] The sentence 'we did not apply any thresholding and 0.5 multiplication as in (10), (11), and (12) in this fusion method' would benefit from a brief explanation of why the 0.5 multiplication and the IoU/appearance threshold masks are omitted for KF gating, since the reader may otherwise suspect an unfair comparison.
  3. [Abstract and Introduction] The claim that 'the choice of a fusion method is key' is vague; it would be more informative to state the conditions (cue combinations, dataset characteristics) under which each fusion method is preferred, given the per-dataset variability reported in the tables.
  4. [§IV.B.1, Table I] In the description of Table I, the statement that weak cues improve KF gating is correct for HOTA and IDF1; however, the corresponding MOTA decreases from 78.035 to 78.003, so the description could note the metric-dependent nature of the improvement.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the paper is an empirical comparison of fusion operators on external benchmarks, with only non-load-bearing self-citations.

full rationale

The paper's core claim is an empirical ranking of four fusion formulas (minimum, weighted sum, KF gating, Hadamard) measured on MOT17, MOT20, and DanceTrack validation sets. The cost matrices and fusion rules are defined directly in Eqs. (10)-(20) from detection and tracklet quantities; no prediction is derived from a fit, and no fitted parameter is renamed as a result. The author's self-citations ([7], [12], [21], [23], [29]) are background or methodological references, and none is invoked as an external uniqueness theorem or as the sole support for the central conclusion. The state-vector derivation in [29] is used to justify a standard constant-velocity Kalman filter extension, not to force the fusion comparison. A legitimate validity concern, distinct from circularity, is that the fusion weights, thresholds, and score cutoffs were set empirically on the same validation sets used for the comparison, and the compared methods differ in motion metric and thresholding policy (Eqs. 18 versus 19), so the ranking is not a clean isolation of the fusion operator. This affects generalizability of the recommendation but does not make the claimed result equivalent to its inputs by construction. Accordingly, the paper is self-contained with respect to circularity, and the score is low.

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

The paper's conclusions rest on standard statistical tools (Kalman filter, Hungarian matching, Mahalanobis gating) plus a set of hand-set hyperparameters. The main free parameters are the fusion weights and thresholds, several of which were tuned per dataset on the validation sets used for the comparison. No new physical or algorithmic entities are introduced; the confidence state is an extension of the existing KF state vector. The key domain assumption is that the pre-trained detector and re-ID models provide reliable cues across all fusion methods.

free parameters (9)
  • lambda1 (motion weight in weighted sum) = 1.0 for all datasets
    Set by hand in Eq. (18); controls balance between motion and other costs.
  • lambda2 (appearance weight in weighted sum) = 0.1 for MOT17/MOT20, 0.2 for DanceTrack
    Tuned per dataset on the validation sets used for comparison; affects the relative ranking of fusion methods.
  • lambda3 (height-IoU weight in weighted sum) = 0.1
    Set by hand; controls contribution of the weak height cue.
  • lambda4 (confidence weight in weighted sum) = 0.1
    Set by hand; controls contribution of the weak confidence cue.
  • KF gating weights (lambda, lambda_h, lambda_c) = 0.98, 0.2, 0.2
    Set by hand in Eq. (19); influence the KF-gating fusion results.
  • Appearance and IoU thresholds (theta_emb, theta_iou) = 0.25, 0.5
    Set by hand in Eqs. (10)-(12); used to discard low-quality matches and affect all fusion methods.
  • Detection score thresholds (tau1, tau2) = 0.6, 0.1
    Set empirically in Section IV.A.3; separate high- and low-confidence detections.
  • KF noise factors (sigma_p, sigma_v, sigma_m) = 0.05, 0.00625, 0.05
    Taken from prior work for 30 fps; appear in Qk and Rk and affect motion cost.
  • EMA momentum alpha = 0.9
    Set by hand for appearance embedding update in Eq. (7).
assumptions (6)
  • domain assumption Kalman filter with constant-velocity model is a reliable motion predictor for tracklets.
    Used throughout Section III; standard in SORT-style trackers but a modeling choice.
  • standard math Hungarian algorithm optimally solves the bipartite matching for data association.
    Invoked in Section III and Section IV.A.3 as the assignment solver.
  • standard math Mahalanobis distance with a 2-degree-of-freedom chi-square 0.95 quantile provides valid KF gating.
    Used in Section III.C.3 for the KF gating fusion method.
  • domain assumption Pre-trained YOLOX-X detector and FastReID SBS/ResNeSt50 appearance model provide reliable detection and appearance cues across all fusion methods.
    Stated in Section IV.A.3; the comparison's fairness depends on these models being equally suitable for all fusion methods.
  • domain assumption Setting width, height, and confidence derivatives to zero before KF prediction improves re-activation after occlusion.
    Applied in Section III for height, width, and confidence preservation; effects are sequence-dependent, as the author notes.
  • domain assumption Validation-set performance is a faithful proxy for test-set performance on MOT17, MOT20, and DanceTrack.
    All conclusions are drawn from validation sets, as described in Section IV.A.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FusionSORT: Fusion Methods for Online Multi-object Visual Tracking." pith.science (2026). https://pith.science/paper/SVJ54UMS

@misc{pith2026250100843,
  author       = {Pith},
  title        = {Pith review of: FusionSORT: Fusion Methods for Online Multi-object Visual Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SVJ54UMS}},
  note         = {Machine review of arXiv:2501.00843}
}
read the original abstract

In this work, we investigate four different fusion methods for associating detections to tracklets in multi-object visual tracking. In addition to considering strong cues such as motion and appearance information, we also consider weak cues such as height intersection-over-union (height-IoU) and tracklet confidence information in the data association using different fusion methods. These fusion methods include minimum, weighted sum based on IoU, Kalman filter (KF) gating, and hadamard product of costs due to the different cues. We conduct extensive evaluations on validation sets of MOT17, MOT20 and DanceTrack datasets, and find out that the choice of a fusion method is key for data association in multi-object visual tracking. We hope that this investigative work helps the computer vision research community to use the right fusion method for data association in multi-object visual tracking.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 28 canonical work pages

  1. [1]

    Simple online and realtime tracking,

    A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, “Simple online and realtime tracking,” in 2016 IEEE International Conference on Image Processing (ICIP) , 2016, pp. 3464–3468

    2016

  2. [2]

    Simple online and rea ltime tracking with a deep association metric,

    N. Wojke, A. Bewley, and D. Paulus, “Simple online and rea ltime tracking with a deep association metric,” in 2017 IEEE International Conference on Image Processing (ICIP) . IEEE, 2017, pp. 3645–3649

    work page 2017

  3. [3]

    Strong- sort: Make deepsort great again,

    Y . Du, Z. Zhao, Y . Song, Y . Zhao, F. Su, T. Gong, and H. Meng, “Strong- sort: Make deepsort great again,” IEEE Transactions on Multimedia , 2023

    2023

  4. [4]

    ByteTrack: Multi-object tracking by associat ing every detection box,

    Y . Zhang, P . Sun, Y . Jiang, D. Y u, F. Weng, Z. Y uan, P . Luo, W . Liu, and X. Wang, “ByteTrack: Multi-object tracking by associat ing every detection box,” 2022

    work page 2022

  5. [5]

    BoT-SORT: Ro bust associations multi-pedestrian tracking,

    N. Aharon, R. Orfaig, and B.-Z. Bobrovsky, “BoT-SORT: Ro bust associations multi-pedestrian tracking,” 2022. [Online] . Available: https://arxiv.org/abs/2206.14651

    arXiv 2022

  6. [6]

    Simple cues lead to a strong multi-object tracker,

    J. Seidenschwarz, G. Braso, V . C. Serrano, I. Elezi, and L . Leal-Taixe, “Simple cues lead to a strong multi-object tracker,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , March 2023

    work page 2023

  7. [7]

    Occlusion-robust online multi-object vis ual tracking using a GM-PHD filter with CNN-based re- identification,

    N. L. Baisa, “Occlusion-robust online multi-object vis ual tracking using a GM-PHD filter with CNN-based re- identification,” Journal of Visual Communication and Image Representation, vol. 80, p. 103279, 2021. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1047320321001814

    work page 2021

  8. [8]

    An introduction to the kalman filt er,

    G. Welch and G. Bishop, “An introduction to the kalman filt er,” 2006

    work page 2006

Show all 40 references
  1. [9]

    A detailed study of the association task in t racking-by- detection-based multi-person tracking,

    D. Stadler, “A detailed study of the association task in t racking-by- detection-based multi-person tracking,” in Proceedings of the 2022 Joint W orkshop of Fraunhofer IOSB and Institute for Anthropomati cs, Vision and Fusion Laboratory. Ed.: J. Beyerer . KIT Scientific Publish...

    2022

  2. [10]

    T he mahalanobis distance,

    R. De Maesschalck, D. Jouan-Rimbaud, and D. Massart, “T he mahalanobis distance,” Chemometrics and Intelligent Laboratory Systems, vol. 50, no. 1, pp. 1–18, 2000. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0169743999000477

    2000

  3. [11]

    FastReID: A pytorch toolbox for general instance re-ident ification,

    L. He, X. Liao, W. Liu, X. Liu, P . Cheng, and T. Mei, “FastReID: A pytorch toolbox for general instance re-ident ification,” in Proceedings of the 31st ACM International Conference on Multimedia, ser. MM ’23. New Y ork, NY , USA: Association for Computing Machinery, 2023, p. 9...

    2023

  4. [12]

    Local-aware global attention network for person re-identification based on body and hand images,

    N. L. Baisa, “Local-aware global attention network for person re-identification based on body and hand images,” Journal of Visual Communication and Image Represen- tation, vol. 103, p. 104207, 2024. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1047320...

    2024

  5. [13]

    Ob servation- centric sort: Rethinking sort for robust multi-object trac king,

    J. Cao, J. Pang, X. Weng, R. Khirodkar, and K. Kitani, “Ob servation- centric sort: Rethinking sort for robust multi-object trac king,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and P attern Recognition, 2023, pp. 9686–9696

    2023

  6. [14]

    Deep-OC -SORT: Multi-pedestrian tracking by adaptive re-identification,

    G. Maggiolino, A. Ahmad, J. Cao, and K. Kitani, “Deep-OC -SORT: Multi-pedestrian tracking by adaptive re-identification, ” arXiv preprint arXiv:2302.11813, 2023

    2023 arXiv

  7. [15]

    Towards rea l-time multi-object tracking,

    Z. Wang, L. Zheng, Y . Liu, Y . Li, and S. Wang, “Towards rea l-time multi-object tracking,” in Computer Vision – ECCV 2020 , A. V edaldi, H. Bischof, T. Brox, and J.-M. Frahm, Eds. Cham: Springer Int erna- tional Publishing, 2020, pp. 107–122

    2020

  8. [16]

    FairMOT: On the fairness of detection and re-identification in multiple obj ect tracking,

    Y . Zhang, C. Wang, X. Wang, W. Zeng, and W. Liu, “FairMOT: On the fairness of detection and re-identification in multiple obj ect tracking,” Int. J. Comput. Vision , vol. 129, no. 11, p. 3069–3087, Nov. 2021. [Online]. Available: https://doi.org/10.1007/s11263-0 21-01513-4

    2021 doi

  9. [17]

    An improved association pip eline for multi- person tracking,

    D. Stadler and J. Beyerer, “An improved association pip eline for multi- person tracking,” in 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition W orkshops (CVPRW) , 2023, pp. 3170–3179

    2023

  10. [18]

    Hybrid-SORT: Weak cues matter for online multi-object tra cking,

    M. Y ang, G. Han, B. Y an, W. Zhang, J. Qi, H. Lu, and D. Wang, “Hybrid-SORT: Weak cues matter for online multi-object tra cking,” in Proceedings of the AAAI Conference on Artificial Intelligen ce, vol. 38, no. 7, 2024, pp. 6504–6512

    2024

  11. [19]

    POI: Multip le object tracking with high performance detection and appearance fe ature,

    F. Y u, W. Li, Q. Li, Y . Liu, X. Shi, and J. Y an, “POI: Multip le object tracking with high performance detection and appearance fe ature,” in Computer Vision – ECCV 2016 W orkshops , G. Hua and H. J´ egou, Eds. Cham: Springer International Publishing, 2016, pp. 36–42

    2016

  12. [20]

    Real-time m ultiple people tracking with deeply learned candidate selection an d person re- identification,

    C. Long, A. Haizhou, Z. Zijie, and S. Chong, “Real-time m ultiple people tracking with deeply learned candidate selection an d person re- identification,” in ICME, 2018

    2018

  13. [21]

    Online multi-object visual tracking usin g a GM-PHD filter with deep appearance learning,

    N. L. Baisa, “Online multi-object visual tracking usin g a GM-PHD filter with deep appearance learning,” in 2019 22th International Conference on Information Fusion (FUSION) , 2019, pp. 1–8

    2019

  14. [22]

    GM-P HD filter based online multiple human tracking using deep discrimina tive corre- lation matching,

    Z. Fu, F. Angelini, S. M. Naqvi, and J. A. Chambers, “GM-P HD filter based online multiple human tracking using deep discrimina tive corre- lation matching,” in 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2018, pp. 4299–4303

    2018

  15. [23]

    Development of a N- type GM-PHD filter for multiple target, multiple type visual tracking,

    N. L. Baisa and A. Wallace, “Development of a N- type GM-PHD filter for multiple target, multiple type visual tracking,” Journal of Visual Communication and Image Representation, vol. 59, pp. 257–271, 2019. [Online]. Available: https://www.sciencedirect.com/science/article/pii/...

    2019

  16. [24]

    Fo cus on details: Online multi-object tracking with diverse fine-gr ained represen- tation,

    H. Ren, S. Han, H. Ding, Z. Zhang, H. Wang, and F. Wang, “Fo cus on details: Online multi-object tracking with diverse fine-gr ained represen- tation,” in 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2023, pp. 11 289–11 298

    2023

  17. [25]

    UTM: A unified multip le object tracking model with identity-aware feature enhance ment,

    S. Y ou, H. Y ao, B.-K. Bao, and C. Xu, “UTM: A unified multip le object tracking model with identity-aware feature enhance ment,” in Proceedings of the IEEE/CVF Conference on Computer Vision a nd Pattern Recognition (CVPR) , June 2023, pp. 21 876–21 886

    2023

  18. [26]

    FeatureSOR T: Essential features for effective tracking,

    H. Hashempoor, R. Koikara, and Y . D. Hwang, “FeatureSOR T: Essential features for effective tracking,” 2024. [Online ]. Available: https://arxiv.org/abs/2407.04249

    2024

  19. [27]

    The hungarian method for the as- signment problem,

    H. W. Kuhn, “The hungarian method for the as- signment problem,” Naval Research Logistics Quarterly , vol. 2, no. 1-2, pp. 83–97, 1955. [Online]. Available: https://onlinelibrary.wiley.com/doi/abs/10.1002/nav.3800020109

    1955 doi

  20. [28]

    Tracki ng without bells and whistles,

    P . Bergmann, T. Meinhardt, and L. Leal-Taix´ e, “Tracki ng without bells and whistles,” in 2019 IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 941–951

    2019

  21. [29]

    Derivation of a constant velocity motion m odel for visual tracking,

    N. L. Baisa, “Derivation of a constant velocity motion m odel for visual tracking,” 2020. [Online]. Available: https://arxiv.org /abs/2005.00844

    2020 arXiv

  22. [30]

    Bag of tricks and a strong baseline for deep person re-identification,

    H. Luo, Y . Gu, X. Liao, S. Lai, and W. Jiang, “Bag of tricks and a strong baseline for deep person re-identification,” in 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition W ork shops (CVPRW), 2019, pp. 1487–1495

    2019

  23. [31]

    ResNeSt: Split -attention networks,

    H. Zhang, C. Wu, Z. Zhang, Y . Zhu, H. Lin, Z. Zhang, Y . Sun, T. He, J. Mueller, R. Manmatha, M. Li, and A. Smola, “ResNeSt: Split -attention networks,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition W orkshops (CVPRW), 2022, pp. 2735–2745

    2022

  24. [32]

    MOT16: A benchmark for multi-object tracking,

    A. Milan, L. Leal-Taixe, I. Reid, S. Roth, and K. Schindl er, “MOT16: A benchmark for multi-object tracking,” 2016. [Online]. Av ailable: https://arxiv.org/abs/1603.00831

    2016 arXiv

  25. [33]

    MOT20: A benchma rk for multi object tracking in crowded scenes,

    P . Dendorfer, H. Rezatofighi, A. Milan, J. Shi, D. Cremer s, I. Reid, S. Roth, K. Schindler, and L. Leal-Taix´ e, “MOT20: A benchma rk for multi object tracking in crowded scenes,” 2020. [Online]. A vailable: https://arxiv.org/abs/2003.09003

    2020 arXiv

  26. [34]

    DanceTrack: Multi-object tracking in uniform appearance and diverse motion,

    P . Sun, J. Cao, Y . Jiang, Z. Y uan, S. Bai, K. Kitani, and P . Luo, “DanceTrack: Multi-object tracking in uniform appearance and diverse motion,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2022, pp. 20 961–20 970

    2022

  27. [35]

    Tracking objects as points,

    X. Zhou, V . Koltun, and P . Kr¨ ahenb¨ uhl, “Tracking objects as points,” in Computer Vision – ECCV 2020: 16th European Conference, Glas gow, UK, August 23–28, 2020, Proceedings, Part IV . Berlin, Heidelberg: Springer-V erlag, 2020, p. 474–490

    2020

  28. [36]

    Evaluating multipl e object tracking performance: the CLEAR MOT metrics,

    K. Bernardin and R. Stiefelhagen, “Evaluating multipl e object tracking performance: the CLEAR MOT metrics,” J. Image Video Process. , vol. 2008, Jan. 2008. [Online]. Available: https://doi.org/10.1155/2008/246309

    2008 doi

  29. [37]

    Performance measures and a data set for multi-target, multi-camera trac king,

    E. Ristani, F. Solera, R. Zou, R. Cucchiara, and C. Tomas i, “Performance measures and a data set for multi-target, multi-camera trac king,” in Computer Vision – ECCV 2016 W orkshops , G. Hua and H. J´ egou, Eds. Cham: Springer International Publishing, 2016, pp. 17–35

    2016

  30. [38]

    HOTA: A higher order metric for evaluating multi- object tracking,

    J. Luiten, A. Osep, P . Dendorfer, P . Torr, A. Geiger, L. L eal-Taix´ e, and B. Leibe, “HOTA: A higher order metric for evaluating multi- object tracking,” Int. J. Comput. Vision , vol. 129, no. 2, p. 548–578, Feb

  31. [39]

    YOLOX: Exceeding YOLO series in 2021,

    Z. Ge, S. Liu, F. Wang, Z. Li, and J. Sun, “YOLOX: Exceeding YOLO series in 2021,” 2021. [Online]. Available: https://arxiv.org/abs/2107.08430

    2021 arXiv

  32. [2021]

    Available: https://doi.org/10.1007/s11 263-020-01375-2

    [Online]. Available: https://doi.org/10.1007/s11 263-020-01375-2

    doi

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.