FastTrackTr:Towards Fast Multi-Object Tracking with Transformers

Dingwen Zhang; Di Wu; Feng Yang; Jinwen Yu; Pan Liao; Wenhui Zhao

REVIEW 4 major objections 5 minor 40 references

FastTrackTr:Towards Fast Multi-Object Tracking with Transformers

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 FastTrackTr claims that replacing self-attention with historical cross-attention in the decoder makes transformer multi-object tracking fast enough for real-time deployment without sacrificing accuracy.

desk verdict A concrete, well-ablated architectural tweak makes transformer MOT actually fast, but the missing association/mask thresholds and the bundled speed-accuracy claim need referee attention. read the letter →

arxiv 2411.15811 v4 pith:G5S4ZPEE submitted 2024-11-24 cs.CV cs.AI

Pan Liao , Feng Yang , Di Wu , Jinwen Yu , Wenhui Zhao , Dingwen Zhang This is my paper · ORCID

classification cs.CVcs.AI

keywords multi-objecttrackingtransformerjointdetectionandhistoricalcross-attentionreal-timeinferenceIDembeddingCircleLossDETR

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 is trying to establish that a transformer multi-object tracker can be both accurate and genuinely real-time if history is moved into the decoder's attention rather than added as extra queries. FastTrackTr follows the joint detection and tracking idea: one network produces detections and identity embeddings in a single pass, and a historical cross-attention layer lets each object query read the previous frame's trajectory features without a second decoder. A historical encoder with a masking mechanism filters stale history, and an ID embedding head trained with Circle Loss supplies appearance features for Hungarian matching. The reported result is 166.4 FPS at 640x640 with TensorRT FP16 on an RTX 4090, and 62.4 HOTA on both DanceTrack and MOT17 test sets, above every transformer-based tracker in the comparison.

What carries the argument

The central mechanism is the historical cross-attention layer. In a standard DETR decoder the first sub-layer is self-attention over $N$ object queries; here that layer keeps the queries as $Q_T = \mathrm{Linear}(q_t^d)$ and sets $K_T = V_T = \mathrm{Linear}(\mathrm{concat}(q_t^d, q'_{t-1,f}))$, so the attention map is $A_T \in \mathbb{R}^{2N \times N}$. This lets each current query pull matching features from the previous frame's trajectory query while avoiding a separate set of track queries. The companion historical encoder, through masked multi-head attention, fuses decoder outputs with previous-frame encoder outputs and suppresses low-confidence historical entries; during training the mask comes from ground-truth matches with random flips, and at inference it comes from a confidence threshold.

What would settle it

Run FastTrackTr on DanceTrack and MOT17 while sweeping the inference mask threshold across a wide range, and also corrupt a small fraction of low-confidence detections; if HOTA stays flat across thresholds the mask is robust, and if the 0.3-point HOTA ablation gain disappears or goes negative, the unreported threshold is doing the work.

Watch

Extended reading notes

Core claim

The paper's central claim is that previous DETR-style trackers pay for tracking by appending track queries or extra decoders, which creates variable query counts that are hard to accelerate. FastTrackTr keeps the standard DETR decoder but replaces its self-attention layer with a historical cross-attention layer: the query is the current frame's initial queries, while the key and value are a linear projection of the concatenation of current queries and the previous frame's encoded decoder output. This is the load-bearing step because it cuts the attention cost from quadratic in track-plus-detection queries to a fixed two-query set, and it makes tensor shapes static. Around that core, a historical encoder with a confidence-based mask aggregates temporal features, and a 256-dimensional ID embedding head trained with Circle Loss produces appearance embeddings for association with a Kalman-filter Hungarian matcher and BYTE-style low-confidence matching. The paper claims that this design reaches 62.4 HOTA on DanceTrack and MOT17 while running at 166.4 FPS at 640x640 under TensorRT on an RTX 4090, placing it ahead of the transformer-based trackers in the comparison tables.

Load-bearing premise

The load-bearing premise is that the confidence score used at inference to build the history mask can stand in for the ground-truth matches used during training, even though the paper never reports the confidence threshold or tests how sensitive the result is to it.

Editorial extensions

If this is right

If the speed and accuracy numbers hold, transformer-based multi-object tracking no longer needs a separate set of track queries: the decoder itself propagates identity, so query count stays fixed and TensorRT acceleration becomes routine.
A three-decoder FastTrackTr at 203.7 FPS indicates that latency can be traded off in coarse steps by deleting decoder layers, giving deployers a speed-accuracy dial without changing the tracking paradigm.
Because the ID embedding head is trained with Circle Loss rather than one-hot identity codes, the method does not need a fixed identity vocabulary, so the embedding should transfer to new scenes without retraining the head.
Even on MOT17, where transformer trackers usually struggle because of limited training data, the model surpasses the transformer-based methods compared, suggesting the historical cross-attention is doing load-bearing association work rather than relying only on the ReID head.

Reading between the lines

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

One testable extension is to expose the inference confidence threshold used to build the history mask and sweep it on DanceTrack and MOT17; the paper reports no sensitivity analysis, so the robustness of the 0.3-point HOTA gain from the mask component remains an open question.
The training-versus-inference mask mismatch is a potential point of failure: if a detector mistake produces a low-confidence but still valid track, the inference mask may cut useful history and propagate errors, a regime the paper does not measure.
The historical cross-attention formulation is detector-agnostic, so the same layer could be inserted into any DETR-family detector to build a tracking-by-detection system without retraining an end-to-end tracker; that would be a direct extension beyond what the paper evaluates.
The generalization claim for metric-learned ID embeddings is only demonstrated on people and vehicle benchmarks, so testing on object classes outside those categories would clarify how far the embedding transfers.

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.

Desk Editor's Note

read the letter

Colleague,

Worth a serious look if you work on transformer MOT. The core idea is simple and genuinely different from MOTR-style query feeding or TransTrack's dual decoder: replace the decoder self-attention with a historical cross-attention whose key/value concatenate the current queries with the previous frame's encoded queries. That cuts query count and keeps the architecture lean, which is what lets them hit 166 FPS with TensorRT at 640x640 and 51 FPS in PyTorch FP16. The ablations on DanceTrack val are monotonic and sensible: decoder gives the big jump, encoder adds a bit, mask adds a little. The edge deployment test on a Jetson AGX Orin is a nice touch; most papers stop at a GPU FPS table.

The soft spots are real but mostly addressable. The inference-time mask threshold and the spatial-distance gating threshold for Hungarian assignment are never reported. Training uses GT-derived masks with random flips, so there is a train/inference distribution shift; a sensitivity sweep on that threshold is needed to confirm the mask's 0.3 HOTA gain isn't brittle. The gating threshold directly controls IDS and MOTA and is not ablated; that's the biggest hole in the reported test numbers. Also, the headline bundles speed at 640x640 with accuracy at 800x1333; those are two configurations, not one. And the conclusion itself concedes MOTIP still has higher tracking accuracy, which sits oddly with the claim of surpassing all prior transformer trackers — they mean among the compared ones, but they need to say that explicitly. Eq. 6-7 has a dimension typo (A^T is stated as R^{2N x N} but the multiplication implies N x 2N); it should be fixed but doesn't threaten the method.

None of this is load-bearing. The mechanism is concrete, the ablations support it, and the speed numbers are plausibly measured. The missing thresholds and sensitivity analysis are exactly what a referee should ask for. If they release code and report the thresholds, the central claim will be reproducible.

This paper deserves a serious referee. It is an engineering contribution with a small, clean architectural novelty, and it addresses a real bottleneck. I'd send it to review, and I'd cite it if I were working on real-time transformer trackers.

Referee Report

4 major / 5 minor

Summary. FastTrackTr is a transformer-based joint-detection-and-tracking method built on RT-DETR. The core idea is to replace the decoder self-attention with a historical cross-attention layer that concatenates current queries with encoded queries from the previous frame, and to add a masked Historical Encoder that aggregates temporal information. The method is evaluated on DanceTrack, MOT17, SportsMOT, VisDrone2019, and BDD100K, with additional latency measurements on an RTX 4090 and a Jetson AGX Orin platform. The paper claims real-time inference (up to 166.4 FPS at 640x640 with TensorRT) and competitive accuracy, including HOTA 62.4 on DanceTrack test and 62.4 on MOT17 test, surpassing prior transformer-based trackers on those benchmarks.

Significance. If the reported numbers are reproducible, FastTrackTr is a practically relevant contribution: it keeps the model simple by avoiding separate track decoders, uses a single ID embedding head, and the ablation in Table VII shows a monotonic benefit from each proposed component. The breadth of benchmarks and the edge-device deployment experiment are strengths. However, the central quantitative claims currently rest on undocumented post-processing thresholds and on a comparison protocol that is not stated consistently, so the contribution cannot be fully assessed as written.

major comments (4)

[Section III-C and Table VII] The inference-time mask in the Historical Encoder is generated from an unspecified confidence threshold ('objects below a certain threshold are filtered out'), while the training-time mask is derived from ground-truth matches with random flips. This is a train/inference distribution shift whose magnitude cannot be evaluated from the paper. The mask is ablated only on DanceTrack val (+0.3 HOTA in Table VII), so its effect may be modest, but the threshold also controls which historical queries enter the temporal encoder and can interact with the association stage. Please report the threshold value and provide a sensitivity sweep over thresholds for HOTA, AssA, IDF1, and IDS.
[Section III-D] The association step rejects Hungarian assignments when observations are 'spatially distant' from the Kalman prediction, but no distance threshold, normalization, or units are given. This parameter directly controls identity switches and therefore MOTA/IDF1, and the absence of the value and of a sensitivity analysis makes the reported test-set numbers unreproducible. Please report the threshold (and units) and ablate it on a validation set.
[Section V (Conclusion) vs Section IV-B] The conclusion states that 'our method currently does not surpass the tracking accuracy of MOTIP', which conflicts with the claim in Section IV-B and Tables II-IV that FastTrackTr surpasses all prior transformer-based methods. The two statements can be reconciled only if 'surpass' is scoped to specific benchmarks and resolutions (for example, MOT17 test HOTA 62.4 vs MOTIP 59.2, but DanceTrack val HOTA 56.9 vs MOTIP 59.1). The paper needs to state this scoping explicitly and align the abstract, experimental summary, and conclusion.
[Section IV-B and Table I] The speed number 166.4 FPS at 640x640 with TensorRT corresponds to a validation HOTA of 54.8 in Table I, whereas the highest test HOTA of 62.4 is obtained at 800x1333 with additional augmentations. No single configuration reports latency and test accuracy simultaneously. If the headline claim is that FastTrackTr is simultaneously the fastest and the most accurate transformer tracker, the paper should provide a table with speed and accuracy for identical settings, or clearly separate the two claims.

minor comments (5)

[Equations (6)-(7)] There is a dimension mismatch: with Q^T in R^{N x C} and K^T, V^T in R^{2N x C}, the attention map A^T = Softmax(Q^T (K^T)^T / sqrt(C)) is in R^{N x 2N}, not R^{2N x N}. The subsequent product A^T V^T is defined only in the N x 2N orientation. This is presumably a typo, but it should be corrected.
[Section IV-A] The optimizer is written as 'AdwmW' and should be 'AdamW'.
[Table VI] The entries '56 2' and '54 2' appear to be missing decimal points (should likely be 56.2 and 54.2).
[Figure 1 caption] The phrase 'The Fastest is A bout Times Faster Than MOTIP and MOTR' is garbled and should be rewritten.
[Table II caption] The note 'please pay more attention to the metrics with *' is not followed by a definition of the asterisk.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported tracking accuracy and speed are empirical measurements against external benchmarks and do not reduce to fitted inputs, definitions, or self-citations.

full rationale

FastTrackTr is an architecture-and-benchmark paper rather than a derivation from first principles. The accuracy claims (e.g., 62.4 HOTA on DanceTrack test and 62.4 HOTA on MOT17 test) are measured on public datasets and compared with external methods, while the speed claims are measured latencies under PyTorch and TensorRT. None of the equations (1)-(9) defines a predicted benchmark quantity in terms of the benchmark itself, and no fitted parameter is renamed as a prediction. The only self-citation is [17] (MO-YOLO), used in Section I for the peripheral qualitative point that 'network lightweighting alone yields insufficient speed improvements'; that point is not a load-bearing premise for any reported result, and the same sentence attributes it to the authors' own empirical studies, so it is independent evidence rather than circular support. Undocumented choices such as the inference-time mask confidence threshold (Section III-C) and the spatial-distance assignment rejection (Section III-D) are reproducibility and robustness concerns, not circularity: omitting a threshold does not make an outcome equivalent to its inputs by construction. Accordingly, the circularity burden is not met.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The method's performance rests on a set of hand-chosen hyperparameters (e.g., query count, embedding dimension, masking probabilities) and on domain assumptions about the transferability of RT-DETR features to MOT datasets. No new physical entities are introduced. The most fragile design choice is the inference-time mask confidence threshold, which is not reported. The approach is not an attempt to derive a law from data, so the circularity burden is low.

free parameters (7)

Number of object queries (nq) = 300
Set to 300 for all datasets; fixed query count enables TensorRT acceleration. Not tuned per dataset.
ID embedding dimension = 256
Output dimension of the ID embedding head; chosen without reported tuning.
Tracklet embedding momentum (eta) = 0.9
Exponential moving average update for tracklet embeddings in Eq. 9.
Circle Loss margin (m) = not reported
Controls separation margin in Eq. 3; value not stated in the paper.
Circle Loss scale (gamma) = not reported
Scaling factor in Eq. 4; value not stated.
Inference confidence threshold for mask = not reported
Determines which historical objects are retained; no value provided.
Mask flip probabilities = formulas in Sec. IV-A
Hand-crafted probabilities for random masking during training.

assumptions (4)

domain assumption RT-DETR trained on COCO provides transferable backbone and encoder features for MOT datasets.
The method builds directly on RT-DETR (Ref. [19]) without retraining the backbone on MOT data; no analysis of domain gap is provided.
domain assumption Identity embeddings from DETR query features are sufficiently discriminative for association across frames.
The central design places an ID head on decoder queries; the paper does not prove that these features separate identities well in all scenarios.
domain assumption The Hungarian algorithm with Kalman filtering and a hand-set gating distance is a sound association strategy.
The matching pipeline is taken from JDE and OC-SORT; parameters such as the distance threshold are not reported.
domain assumption Public MOT benchmarks (DanceTrack, MOT17, SportsMOT, VisDrone, BDD100K) are reliable and comparable.
All conclusions rely on standard metrics; the paper does not discuss benchmark evaluation noise.

how reviews work

0 comments

Cite this review

Pith. "Pith review of FastTrackTr:Towards Fast Multi-Object Tracking with Transformers." pith.science (2026). https://pith.science/paper/G5S4ZPEE

@misc{pith2026241115811,
  author       = {Pith},
  title        = {Pith review of: FastTrackTr:Towards Fast Multi-Object Tracking with Transformers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G5S4ZPEE}},
  note         = {Machine review of arXiv:2411.15811}
}

read the original abstract

Transformer-based multi-object tracking (MOT) methods have captured the attention of many researchers in recent years. However, these models often suffer from slow inference speeds due to their structure or other issues. To address this problem, we revisited the Joint Detection and Tracking (JDT) method by looking back at past approaches. By integrating the original JDT approach with some advanced theories, this paper employs an efficient method of information transfer between frames on the DETR, constructing a fast and novel JDT-type MOT framework: FastTrackTr. Thanks to the superiority of this information transfer method, our approach not only reduces the number of queries required during tracking but also avoids the excessive introduction of network structures, ensuring model simplicity. Experimental results indicate that our method has the potential to achieve real-time tracking and exhibits competitive tracking accuracy across multiple datasets.

Figures

Figures reproduced from arXiv: 2411.15811 by the authors.

**Figure 5.** C. Ablation Study We tested the impact of various components of our model and other factors on performance on the val set of DanceTrack. The ablation study in Table VII highlights the progressive TABLE V: Comparison With the SOTA Methods on the VisDrone2019 Test Set Methods MOTA ↑ IDF1 ↑ FP ↓ FN ↓ IDS ↓ MOTR [9] 22.8 41.4 28407 147937 959 TrackFormer [8] 25.0 30.5 25856 141526 4840 MMTrack [5] 36.7 54. 7 23849 1208… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 20 canonical work pages

[1]

Blockchain-empowered distributed multicamera multitarget tracking in edge computing,

S. Wang, H. Sheng, Y . Zhang, D. Yang, J. Shen, and R. Chen, “Blockchain-empowered distributed multicamera multitarget tracking in edge computing,” IEEE Transactions on Industrial Informatics , vol. 20, no. 1, pp. 369–379, 2023

work page 2023
[2]

An efficient edge artificial intelligence multipedestrian tracking method with rank constraint,

H. Yang, J. Wen, X. Wu, L. He, and S. Mumtaz, “An efficient edge artificial intelligence multipedestrian tracking method with rank constraint,” IEEE Transactions on Industrial Informatics , vol. 15, no. 7, pp. 4178–4188, 2019

work page 2019
[3]

A tracking-based burst bubble recognition in flotation using deep graph neural network,

R. Tian, Z. Lv, M. Sun, Y . Tu, Z. Xu, and W. Wang, “A tracking-based burst bubble recognition in flotation using deep graph neural network,” IEEE Transactions on Industrial Informatics , 2025

work page 2025
[4]

Ermot: Evidence reasoning-based robust multiple object tracking method,

X. Xu, X. Wang, F. Wu, Z. Zhang, and L. Chang, “Ermot: Evidence reasoning-based robust multiple object tracking method,” IEEE Trans- actions on Industrial Informatics , 2024

work page 2024
[5]

Rethinking joint detection and embedding for multiobject tracking in multiscenario,

L. Xu and Y . Huang, “Rethinking joint detection and embedding for multiobject tracking in multiscenario,” IEEE Transactions on Industrial Informatics, 2024

work page 2024
[6]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

2017
[7]

Transtrack: Multiple object tracking with transformer,

P. Sun, J. Cao, Y . Jiang, R. Zhang, E. Xie, Z. Yuan, C. Wang, and P. Luo, “Transtrack: Multiple object tracking with transformer,” arXiv preprint arXiv:2012.15460, 2020

arXiv 2012
[8]

Track- former: Multi-object tracking with transformers,

T. Meinhardt, A. Kirillov, L. Leal-Taixe, and C. Feichtenhofer, “Track- former: Multi-object tracking with transformers,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 8844–8854

2022

Show all 40 references

[9]

Motr: End-to-end multiple-object tracking with transformer,

F. Zeng, B. Dong, Y . Zhang, T. Wang, X. Zhang, and Y . Wei, “Motr: End-to-end multiple-object tracking with transformer,” in European Conference on Computer Vision . Springer, 2022, pp. 659–675

2022
[10]

Bridging the Gap Between End-to-end and Non-End-to-end Multi-Object Tracking

F. Yan, W. Luo, Y . Zhong, Y . Gan, and L. Ma, “Bridging the Gap Between End-to-end and Non-End-to-end Multi-Object Tracking.”
[11]

Memotr: Long-term memory-augmented trans- former for multi-object tracking,

R. Gao and L. Wang, “Memotr: Long-term memory-augmented trans- former for multi-object tracking,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 9901–9910

2023
[12]

Motrv3: Release-fetch supervision for end-to-end multi-object tracking,

E. Yu, T. Wang, Z. Li, Y . Zhang, X. Zhang, and W. Tao, “Motrv3: Release-fetch supervision for end-to-end multi-object tracking,” arXiv preprint arXiv:2305.14298, 2023

2023 arXiv
[13]

Multiple object tracking as id prediction,

R. Gao, Y . Zhang, and L. Wang, “Multiple object tracking as id prediction,” arXiv preprint arXiv:2403.16848 , 2024

2024 arXiv
[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,” Feb. 2023

2023
[15]

Hybrid-sort: Weak cues matter for online multi-object tracking,

M. Yang, G. Han, B. Yan, W. Zhang, J. Qi, H. Lu, and D. Wang, “Hybrid-sort: Weak cues matter for online multi-object tracking,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 7, 2024, pp. 6504–6512

2024
[16]

Observation- centric sort: Rethinking sort for robust multi-object tracking,

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

2023
[17]

Mo-yolo: End-to-end multiple-object tracking method with yolo and motr,

L. Pan, Y . Feng, W. Di, L. Bo, and Z. Xingle, “Mo-yolo: End-to-end multiple-object tracking method with yolo and motr,” arXiv preprint arXiv:2310.17170, 2023

2023 arXiv
[18]

Deformable detr: Deformable transformers for end-to-end object detection,

X. Zhu, W. Su, L. Lu, B. Li, X. Wang, and J. Dai, “Deformable detr: Deformable transformers for end-to-end object detection,”arXiv preprint arXiv:2010.04159, 2020. 10

2010 arXiv
[19]

Detrs beat yolos on real-time object detection,

Y . Zhao, W. Lv, S. Xu, J. Wei, G. Wang, Q. Dang, Y . Liu, and J. Chen, “Detrs beat yolos on real-time object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 16 965–16 974

2024
[20]

Transcenter: Transformers with dense representations for multiple- object tracking,

Y . Xu, Y . Ban, G. Delorme, C. Gan, D. Rus, and X. Alameda-Pineda, “Transcenter: Transformers with dense representations for multiple- object tracking,” IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 6, pp. 7820–7835, 2022

2022
[21]

Memot: Multi-object tracking with memory,

J. Cai, M. Xu, W. Li, Y . Xiong, W. Xia, Z. Tu, and S. Soatto, “Memot: Multi-object tracking with memory,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 8090–8100

2022
[22]

Fairmot: On the fairness of detection and re-identification in multiple object tracking,

Y . Zhang, C. Wang, X. Wang, W. Zeng, and W. Liu, “Fairmot: On the fairness of detection and re-identification in multiple object tracking,” International journal of computer vision, vol. 129, pp. 3069–3087, 2021

2021
[23]

Towards real-time multi-object tracking,

Z. Wang, L. Zheng, Y . Liu, Y . Li, and S. Wang, “Towards real-time multi-object tracking,” in European conference on computer vision . Springer, 2020, pp. 107–122

2020
[24]

DanceTrack: Multi-Object Tracking in Uniform Appearance and Di- verse Motion,

P. Sun, J. Cao, Y . Jiang, Z. Yuan, S. Bai, K. Kitani, and P. Luo, “DanceTrack: Multi-Object Tracking in Uniform Appearance and Di- verse Motion,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA: IEEE, Jun. 2022, pp. 20 961–20 970

2022
[25]

Sportsmot: A large multi-object tracking dataset in multiple sports scenes,

Y . Cui, C. Zeng, X. Zhao, Y . Yang, G. Wu, and L. Wang, “Sportsmot: A large multi-object tracking dataset in multiple sports scenes,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 9921–9931

2023
[26]

MOT16: A Benchmark for Multi-Object Tracking,

A. Milan, L. Leal-Taixe, I. Reid, S. Roth, and K. Schindler, “MOT16: A Benchmark for Multi-Object Tracking,” May 2016

2016
[27]

Detection and tracking meet drones challenge,

P. Zhu, L. Wen, D. Du, X. Bian, H. Fan, Q. Hu, and H. Ling, “Detection and tracking meet drones challenge,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 11, pp. 7380–7399, 2021

2021
[28]

Bdd100k: A diverse driving dataset for heterogeneous multitask learning,

F. Yu, H. Chen, X. Wang, W. Xian, Y . Chen, F. Liu, V . Madhavan, and T. Darrell, “Bdd100k: A diverse driving dataset for heterogeneous multitask learning,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 2636–2645

2020
[29]

Motrv2: Bootstrapping end-to-end multi-object tracking by pretrained object detectors,

Y . Zhang, T. Wang, and X. Zhang, “Motrv2: Bootstrapping end-to-end multi-object tracking by pretrained object detectors,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 22 056–22 065

2023
[30]

Putr: A pure transformer for decou- pled and online multi-object tracking,

C. Liu, H. Li, Z. Wang, and R. Xu, “Putr: A pure transformer for decou- pled and online multi-object tracking,” arXiv preprint arXiv:2405.14119, 2024

2024 arXiv
[31]

ByteTrack: Multi-object Tracking by Associating Every Detection Box,

Y . Zhang, P. Sun, Y . Jiang, D. Yu, F. Weng, Z. Yuan, P. Luo, W. Liu, and X. Wang, “ByteTrack: Multi-object Tracking by Associating Every Detection Box,” in Computer Vision – ECCV 2022 , ser. Lecture Notes in Computer Science, S. Avidan, G. Brostow, M. Ciss ´e, G. M. Farinell...

2022
[32]

Tracking objects as points,

X. Zhou, V . Koltun, and P. Kr ¨ahenb¨uhl, “Tracking objects as points,” in European conference on computer vision. Springer, 2020, pp. 474–490

2020
[33]

Circle loss: A unified perspective of pair similarity optimization,

Y . Sun, C. Cheng, Y . Zhang, C. Zhang, L. Zheng, Z. Wang, and Y . Wei, “Circle loss: A unified perspective of pair similarity optimization,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 6398–6407

2020
[34]

Yolox: Exceeding yolo series in 2021,

Z. Ge, “Yolox: Exceeding yolo series in 2021,” arXiv preprint arXiv:2107.08430, 2021

2021 arXiv
[35]

Diffmot: A real-time diffusion-based multiple object tracker with non-linear prediction,

W. Lv, Y . Huang, N. Zhang, R.-S. Lin, M. Han, and D. Zeng, “Diffmot: A real-time diffusion-based multiple object tracker with non-linear prediction,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 19 321–19 330

2024
[36]

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

E. Ristani, F. Solera, R. Zou, R. Cucchiara, and C. Tomasi, “Performance measures and a data set for multi-target, multi-camera tracking,” in European conference on computer vision . Springer, 2016, pp. 17–35

2016
[37]

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

J. Luiten, A. Osep, P. Dendorfer, P. Torr, A. Geiger, L. Leal-Taix ´e, and B. Leibe, “Hota: A higher order metric for evaluating multi-object tracking,” International journal of computer vision , vol. 129, pp. 548– 578, 2021

2021
[38]

Qdtrack: Quasi-dense similarity learning for appearance-only multiple object tracking,

T. Fischer, T. E. Huang, J. Pang, L. Qiu, H. Chen, T. Darrell, and F. Yu, “Qdtrack: Quasi-dense similarity learning for appearance-only multiple object tracking,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

2023
[39]

Crowdhuman: A benchmark for detecting human in a crowd,

S. Shao, Z. Zhao, B. Li, T. Xiao, G. Yu, X. Zhang, and J. Sun, “Crowdhuman: A benchmark for detecting human in a crowd,” arXiv preprint arXiv:1805.00123, 2018

2018 arXiv
[40]

Facenet: A unified embed- ding for face recognition and clustering,

F. Schroff, D. Kalenichenko, and J. Philbin, “Facenet: A unified embed- ding for face recognition and clustering,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2015, pp. 815– 823

2015

Pith tools

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

[1] [1]

Blockchain-empowered distributed multicamera multitarget tracking in edge computing,

S. Wang, H. Sheng, Y . Zhang, D. Yang, J. Shen, and R. Chen, “Blockchain-empowered distributed multicamera multitarget tracking in edge computing,” IEEE Transactions on Industrial Informatics , vol. 20, no. 1, pp. 369–379, 2023

work page 2023

[2] [2]

An efficient edge artificial intelligence multipedestrian tracking method with rank constraint,

H. Yang, J. Wen, X. Wu, L. He, and S. Mumtaz, “An efficient edge artificial intelligence multipedestrian tracking method with rank constraint,” IEEE Transactions on Industrial Informatics , vol. 15, no. 7, pp. 4178–4188, 2019

work page 2019

[3] [3]

A tracking-based burst bubble recognition in flotation using deep graph neural network,

R. Tian, Z. Lv, M. Sun, Y . Tu, Z. Xu, and W. Wang, “A tracking-based burst bubble recognition in flotation using deep graph neural network,” IEEE Transactions on Industrial Informatics , 2025

work page 2025

[4] [4]

Ermot: Evidence reasoning-based robust multiple object tracking method,

X. Xu, X. Wang, F. Wu, Z. Zhang, and L. Chang, “Ermot: Evidence reasoning-based robust multiple object tracking method,” IEEE Trans- actions on Industrial Informatics , 2024

work page 2024

[5] [5]

Rethinking joint detection and embedding for multiobject tracking in multiscenario,

L. Xu and Y . Huang, “Rethinking joint detection and embedding for multiobject tracking in multiscenario,” IEEE Transactions on Industrial Informatics, 2024

work page 2024

[6] [6]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

2017

[7] [7]

Transtrack: Multiple object tracking with transformer,

P. Sun, J. Cao, Y . Jiang, R. Zhang, E. Xie, Z. Yuan, C. Wang, and P. Luo, “Transtrack: Multiple object tracking with transformer,” arXiv preprint arXiv:2012.15460, 2020

arXiv 2012

[8] [8]

Track- former: Multi-object tracking with transformers,

T. Meinhardt, A. Kirillov, L. Leal-Taixe, and C. Feichtenhofer, “Track- former: Multi-object tracking with transformers,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 8844–8854

2022