Pith. sign in

REVIEW 4 major objections 6 minor 35 references

Self-Supervised Real-Time Tracking of Military Vehicles in Low-FPS UAV Footage

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The authors show that instance association for low-FPS UAV tracking can be learned from single-frame box annotations alone, with scene context doing the work.

desk verdict Useful dataset and robustness observations, but the evaluation never tests generalization because train and test overlap and no baseline is run. read the letter →

arxiv 2507.05229 v1 pith:JPZJKJ2O submitted 2025-07-07 cs.CV

Markiyan Kostiv , Anatolii Adamovskyi , Yevhen Cherniavskyi , Mykyta Varenyk , Ostap Viniavskyi , Igor Krashenyi , Oles Dobosevych This is my paper · ORCID
classification cs.CV
keywords multi-objecttrackingself-supervisedlearningUAVfootagelow-FPSmilitaryvehicledetectioninstanceassociationscenecontextsurveillance
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that multi-object tracking of military vehicles in low-frame-rate UAV footage can be learned without any tracking annotations: only single-frame bounding-box labels are needed. It adapts a previously proposed self-supervised instance-association framework to this setting and shows that associations are carried by global scene context pooled around each detection, not by fine object appearance. The authors report that this approach stays almost as accurate at 160-pixel input resolution and 32-dimensional embeddings as at full resolution, which would make real-time tracking feasible on limited hardware. They also release a benchmark dataset of 119 annotated UAV videos with 58,897 bounding boxes across three military vehicle classes.

What carries the argument

The load-bearing mechanism is a contrastively trained per-instance embedding that uses ROI-Align to pool features from the full frame-level representation rather than from a tight object crop. This gives each detection a descriptor that includes its spatial position and surrounding scene, which stays informative when the object itself is tiny or heavily compressed. Training uses supervised contrastive loss on positive pairs mined from single-frame annotations plus an object-prior distillation auxiliary task, with a residual convolutional backbone pretrained on a large general-purpose segmentation dataset and fine-tuned on the military-vehicle boxes.

What would settle it

Run the same training procedure on one set of UAV videos and evaluate on a disjoint set from different locations, altitudes, and weather; if HOTA and IDF1 drop far more than the roughly two-point gap the paper reports between full and quarter resolution, then the claimed context-based robustness is in-sample memorization rather than transferable tracking.

Watch

Extended reading notes

Core claim

The central claim is that instance association for low-FPS UAV tracking can be learned directly from single-frame ground-truth bounding boxes, without inter-frame track labels. Training on ground-truth boxes outperforms training on pseudo-labels from a general-purpose segmentation model (HOTA 48.29 vs 47.23, IDF1 56.93 vs 55.45). The resulting model pools frame-level features through ROI-Align, so each instance embedding carries scene context; this is why it can re-identify vehicles that are visually near-identical and only a few pixels across. Reducing input resolution to 160 pixels and the embedding to 32 dimensions lowers HOTA from 48.29 to 46.62 and IDF1 from 56.93 to 54.18, which the authors present as evidence that the approach is suitable for real-time, low-resource deployment. The paper further states that the method struggles with objects that leave the field of view for more than a couple of seconds and with the same object seen from multiple viewpoints simultaneously.

Load-bearing premise

The reported results are only evidence of tracking quality if measuring on the same footage used to train the detector and association model counts as an evaluation of generalization; the paper does not describe a held-out split, an external benchmark, or a comparison with other trackers.

Editorial extensions

If this is right

  • Tracking models can be trained from single-frame detection datasets, removing the need for expensive inter-frame track annotations.
  • Inference can run at 160-pixel resolution and 32-dimensional embeddings with only a small drop in HOTA and IDF1, which supports deployment on low-power edge devices.
  • Ground-truth bounding boxes outperform pseudo-labels from a general-purpose segmentation model for this tracking task.
  • Scene context, not fine object appearance, is the reusable cue for association in low-FPS moving-camera footage.
  • The released benchmark of 119 videos and 58,897 boxes gives the community a common evaluation set for low-FPS military-vehicle tracking.

Reading between the lines

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

  • If scene context carries the association, the method's success may depend on the scene type; a cross-location train/test split would test whether the reported numbers transfer to unfamiliar terrain.
  • Because the evaluation uses detections from a detector trained on the same footage later used for association training, the reported HOTA/IDF1 values are in-sample numbers; a held-out evaluation could be materially lower.
  • The paper's own failure modes (objects leaving view for seconds, multi-camera views) point to adding short-term memory or cross-view consistency as next steps.
  • Ablating context pooling, comparing tight-crop embeddings against full-frame ROI-Align embeddings, would directly test the paper's explanation for why low resolutions remain accurate.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper adapts the MASA self-supervised instance-association framework to low-FPS multi-object tracking of military vehicles in UAV footage. The authors introduce a benchmark dataset of 119 videos with 18,421 frames and 58,897 bounding boxes in three vehicle classes, and train a MASA-based ReID model with a ResNet50 backbone by pre-training on SA-1B and fine-tuning on single-frame ground-truth bounding boxes from the UAV footage. The experiments compare SAM pseudo-labels versus ground-truth annotations, measure tracking performance at reduced input resolutions, and evaluate reduced embedding dimensions, reporting HOTA, MOTA, IDF1, AssA, and IDSW. The paper claims that instance association can be learned from single-frame annotations, that global scene context is crucial for low-FPS association, and that the model remains accurate at 160-pixel resolution and 32-dimensional embeddings.

Significance. If the central claims were properly supported, the paper would make a useful practical contribution: demonstrating that temporal labels are not required for training an association model in a low-FPS, moving-camera military domain could substantially reduce annotation cost, and the resolution/embedding reductions point toward real-time edge deployment. The dataset itself addresses a real gap in publicly available UAV military-vehicle tracking benchmarks, and the authors are explicit about some limitations, such as the absence of long-term memory and multi-view association. However, as presented, the evaluation does not currently test generalization because the detector and MASA fine-tuning are trained on the same footage later used for evaluation, no baseline tracker is included, and all metrics are single numbers without variance estimates. These issues are load-bearing for the manuscript's main claims, so the significance cannot be assessed from the reported numbers alone.

major comments (4)
  1. [IV / Tables I-III] Section IV states that 'All experiments and results are based on the detections obtained from a binary object detection model trained on the same UA V footage dataset later used from MASA training.' No train/test split is described for either the detector or the MASA fine-tuning, so the HOTA/IDF1 values in Tables I-III could be inflated by memorizing the exact vehicles and scenes used during training. The central claim that instance association can be learned from single-frame annotations requires demonstrating transfer to footage and vehicles not seen during fine-tuning; please provide a held-out split, ideally at the video level, and report metrics on it.
  2. [V / Tables I-III] No existing tracker is run as a baseline. Section V compares only training-label conditions, resolutions, and embedding sizes against one another. Without a standard MOT method (e.g., ByteTrack, OC-SORT, DeepSORT) or an unfine-tuned MASA model evaluated on the same detections, the absolute numbers (HOTA around 48, IDF1 around 57) provide no anchor for the claim of 'high association quality' in low-FPS conditions. Please add such comparisons using identical detections.
  3. [V-B] The global-context explanation is asserted rather than tested. Section V-B attributes the robustness to resolution reduction to global scene context pooled by ROI-Align, but this mechanism is never ablated or directly measured. The observation that performance degrades only mildly at 160-pixel resolution could also be explained by appearance invariance or by properties of the detector. Please include an experiment that removes or masks the global-context contribution, or otherwise explicitly varies the context available to the association model, and show how HOTA/IDF1 change.
  4. [Tables I-III / V-C] All tracking metrics are reported as single numbers without error bars, repeated runs, or statistical tests. Many of the comparisons are small (e.g., HOTA 48.29 vs 48.00 vs 46.57), and without an estimate of variance the claim of 'no significant drop' in Section V-C is unsupported. Please report means and standard deviations over multiple splits or seeds, and state how many videos are used in each evaluation condition.
minor comments (6)
  1. [Throughout] The paper frequently writes 'UA V' with a spurious space (e.g., 'UA V footage'); please standardize to 'UAV'.
  2. [V-B] The phrase 'with an area of less than 32 pixels' is ambiguous; specify whether this refers to a side length or a pixel count, and in what units.
  3. [Tables II-III] The resolution labels are inconsistent: Table II uses two numbers (e.g., '1280, 736'), while Table III uses one number ('1280'). Please make the notation uniform and explain what the second number denotes.
  4. [III-B] The dataset is said to be 'available upon request through this form,' but no link or DOI is provided; a permanent, citable URL or repository identifier should be given.
  5. [References] Reference [3] appears to be garbled ('1 C filter'); if this refers to the 1€ filter, please correct the title.
  6. [V-C] The claim that the 160/32 model 'strikes an optimal balance' is not supported by any reported runtime, latency, or memory measurements; either add such numbers or soften the wording.

Circularity Check

1 steps flagged · score 6.0 of 10

Evaluation is in-sample: the detector and the MASA model are trained on the same UAV footage used to compute HOTA/IDF1, with no held-out split, so the central generalization claim is untested.

  1. fitted input called prediction [Section IV (Method), sentence beginning 'We train the MASA model...'; results reported in Section V, Tables I-III]
    "We train the MASA model with a ResNet50 [8] backbone by first pre-training it on a subset of the diverse, general-purpose SA-1B [14] dataset using pseudo-labels generated by Segment Anything Model (SAM) [14], and then fine-tuning it on a smaller UAV footage dataset containing single-frame bounding box annotations of various military vehicle types. All experiments and results are based on the detections obtained from a binary object detection model trained on the same UAV footage dataset later used from MASA training."

    The paper's central demonstration that instance association can be learned from single-frame annotations is measured by HOTA, MOTA, IDF1, and IDSW computed on the same UAV footage used to fine-tune the MASA model and to train the object detector. No train/test split, external benchmark, or baseline tracker is described. The reported tracking metrics can therefore be satisfied by memorizing the vehicles, scenes, and contextual layouts seen during fine-tuning; they are in-sample fits rather than predictions of association quality on unseen footage. The claimed generalization from single-frame annotations is not supported by the protocol, because the evaluation distribution is identical to the training distribution by construction.

full rationale

The paper contains no equation-level circularity: the method inherits MASA's contrastive learning objective from an external source, and no target quantity is defined in terms of its own prediction. It also does not rely on self-citation to justify any core premise. The circularity is instead in the evaluation chain: the central claim is that instance association can be learned from single-frame annotations and remains robust at low resolution and small embedding size. The only evidence is Tables I-III, whose detections come from a detector trained on the same UAV dataset later used for MASA fine-tuning, and whose tracking metrics are computed on that same footage. Because the model is evaluated on its own training distribution, the high HOTA/IDF1 values do not demonstrate transfer to new videos or vehicles, and no comparison to a classical tracker such as ByteTrack or OC-SORT on the same detections anchors the absolute numbers. Thus the central claim reduces to a report of in-sample performance rather than a validated prediction. This is partial circularity of the 'fitted input called prediction' kind, warranting a score of 6 rather than a lower score.

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

The central contribution rests on the new dataset and the evaluation protocol, not on a formal derivation. No new physical or algorithmic entities are introduced. The main unstated assumption is that training and evaluation on the same footage is a valid test of tracking generalization; the paper does not describe a train/test split. The MASA framework itself is imported from prior work, so its correctness is assumed rather than re-derived.

assumptions (4)
  • domain assumption Evaluation on the same UAV dataset used for training is representative of real low-FPS military tracking.
    Section IV states experiments use a detector trained on the same dataset later used for MASA training; no separate test set is mentioned, so reported metrics may only reflect in-distribution performance.
  • ad hoc to paper Global scene context pooled by ROI-Align remains stable across resolutions and is the cause of robustness.
    Section V-B attributes robustness to context, but no ablation removes or perturbs context; alternative explanations such as low appearance diversity are not tested.
  • domain assumption MASA self-supervised contrastive learning correctly transfers from SA-1B pre-training to military UAV footage.
    The method is adopted wholesale from reference [18]; no derivation or independent verification of MASA's assumptions on this domain is provided.
  • domain assumption A binary object detector trained on the same data provides sufficiently accurate detections.
    All tracking metrics depend on detector output, but no detector precision, recall, or independent evaluation is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Self-Supervised Real-Time Tracking of Military Vehicles in Low-FPS UAV Footage." pith.science (2026). https://pith.science/paper/JPZJKJ2O

@misc{pith2026250705229,
  author       = {Pith},
  title        = {Pith review of: Self-Supervised Real-Time Tracking of Military Vehicles in Low-FPS UAV Footage},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JPZJKJ2O}},
  note         = {Machine review of arXiv:2507.05229}
}
read the original abstract

Multi-object tracking (MOT) aims to maintain consistent identities of objects across video frames. Associating objects in low-frame-rate videos captured by moving unmanned aerial vehicles (UAVs) in actual combat scenarios is complex due to rapid changes in object appearance and position within the frame. The task becomes even more challenging due to image degradation caused by cloud video streaming and compression algorithms. We present how instance association learning from single-frame annotations can overcome these challenges. We show that global features of the scene provide crucial context for low-FPS instance association, allowing our solution to be robust to distractors and gaps in detections. We also demonstrate that such a tracking approach maintains high association quality even when reducing the input image resolution and latent representation size for faster inference. Finally, we present a benchmark dataset of annotated military vehicles collected from publicly available data sources. This paper was initially presented at the NATO Science and Technology Organization Symposium (ICMCIS) organized by the Information Systems Technology (IST)Scientific and Technical Committee, IST-209-RSY - the ICMCIS, held in Oeiras, Portugal, 13-14 May 2025.

Figures

Figures reproduced from arXiv: 2507.05229 by the authors.

Figure 1
Figure 1. Track Length Distribution [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Classes distribution [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Examples of annotated videos C. Annotation Process The annotation team consisted of human experts from Ukrainian Catholic University who had prior exposure to Computer Vision. • Tracks were annotated by identifying and labeling indi￾vidual vehicles across frames and assigning unique IDs to maintain track consistency over time [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Comparison of raw images (left), ground truth (middle), and SAM annotations (right). SAM struggles with precise segmentation, often mislabeling [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 31 canonical work pages

  1. [1]

    Simple online and realtime tracking

    Alex Bewley, Zongyuan Ge, Lionel Ott, Fabio Ramos, and Ben Upcroft. Simple online and realtime tracking. In 2016 IEEE International Conference on Image Processing (ICIP) . IEEE, September 2016

    work page 2016

  2. [2]

    End-to-end object detec- tion with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to-end object detec- tion with transformers. In European Conference on Computer Vision (ECCV), pages 213–229. Springer, 2020

    work page 2020

  3. [3]

    1 C filter: a simple speed-based low-pass filter for noisy input in interactive systems

    G ´ery Casiez, Nicolas Roussel, and Daniel V ogel. 1 C filter: a simple speed-based low-pass filter for noisy input in interactive systems. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI ’12, page 2527–2530, New York, NY , USA, 2012. Association for Computing Machinery

    work page 2012

  4. [4]

    A simple framework for contrastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. A simple framework for contrastive learning of visual representations. In Proceedings of the 37th International Conference on Machine Learning , ICML’20. JMLR.org, 2020

    work page 2020

  5. [5]

    A simple framework for contrastive learning of visual representations, 2020

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. A simple framework for contrastive learning of visual representations, 2020

    2020

  6. [6]

    An assessment of multi-object tracking under low framerate conditions

    Gabriele Facciolo, Rafael Grompone von Gio, and Axel Davy. An assessment of multi-object tracking under low framerate conditions. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion Workshops (CVPRW), 2022

    work page 2023

  7. [7]

    Mask r-cnn, 2018

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross Girshick. Mask r-cnn, 2018

    work page 2018

  8. [8]

    Deep residual learning for image recognition, 2015

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition, 2015

    2015

Show all 35 references
  1. [9]

    Occlusion and motion reasoning for long-term tracking

    Yang Hua, Karteek Alahari, and Cordelia Schmid. Occlusion and motion reasoning for long-term tracking. In David Fleet, Tomas Pajdla, Bernt Schiele, and Tinne Tuytelaars, editors, Computer Vision – ECCV 2014 , pages 172–187, Cham, 2014. Springer International Publishing

    2014

  2. [10]

    segment anything

    Ge-Peng Ji, Deng-Ping Fan, Peng Xu, Bowen Zhou, Ming-Ming Cheng, and Luc Van Gool. Sam struggles in concealed scenes — empirical study on “segment anything”. Science China Information Sciences , 66(12), November 2023

    2023

  3. [11]

    R. E. Kalman. A new approach to linear filtering and prediction problems. Journal of Basic Engineering , 82(1):35–45, 03 1960

    1960

  4. [12]

    Oc-sort: Observation- centric sort for robust multi-object tracking, 2022

    Lei Ke, Siyuan Li, Qifeng Chen, and Fisher Yu. Oc-sort: Observation- centric sort for robust multi-object tracking, 2022

    2022

  5. [13]

    Supervised contrastive learning, 2021

    Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. Supervised contrastive learning, 2021

    2021

  6. [14]

    Berg, Wan-Yen Lo, Piotr Doll ´ar, and Ross Girshick

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C. Berg, Wan-Yen Lo, Piotr Doll ´ar, and Ross Girshick. Segment anything, 2023

    2023

  7. [15]

    APPTracker: Appearance and prediction probability tracker for low-fps scenarios

    Rohit Kumar, Martin Dahl, and Eva Peterson. APPTracker: Appearance and prediction probability tracker for low-fps scenarios. In Proceedings of the 2022 International Conference on Computer Vision (ICCV), pages 3451–3460. IEEE, 2022

    2022

  8. [16]

    Collaborative tracking learning for frame-rate-insensitive multi-object tracking, 2023

    Jihoon Lee, Ananya Patel, Claire Williams, and Eric Hann. Collaborative tracking learning for frame-rate-insensitive multi-object tracking, 2023

    2023

  9. [17]

    Huang, and Fisher Yu

    Siyuan Li, Martin Danelljan, Henghui Ding, Thomas E. Huang, and Fisher Yu. Tracking every thing in the wild, 2022

    2022

  10. [18]

    Matching anything by segmenting anything, 2024

    Siyuan Li, Lei Ke, Martin Danelljan, Luigi Piccinelli, Mattia Segu, Luc Van Gool, and Fisher Yu. Matching anything by segmenting anything, 2024

    2024

  11. [19]

    Multi-object tracking meets moving uav

    Shuai Liu, Xin Li, Huchuan Lu, and You He. Multi-object tracking meets moving uav. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 8866–8875, 2022

    2022

  12. [20]

    MeMOTR: Memory-enhanced multiple object tracking with transformers

    Wen Liu, Ming Zhao, and Hao Chen. MeMOTR: Memory-enhanced multiple object tracking with transformers. 2024

    2024

  13. [21]

    Trackformer: Multi-object tracking with transformers, 2022

    Tim Meinhardt, Alexander Kirillov, Laura Leal-Taixe, and Christoph Feichtenhofer. Trackformer: Multi-object tracking with transformers, 2022

    2022

  14. [22]

    Quasi-Dense Similarity Learning for Multiple Object Tracking, 2024

    Bo Pang, Linlu Qiu, Pengwan Li, and Chengke Shen. Quasi-Dense Similarity Learning for Multiple Object Tracking, 2024

    2024

  15. [23]

    Faster r-cnn: Towards real-time object detection with region proposal networks, 2016

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks, 2016

    2016

  16. [24]

    Adapt- ing the segment anything model during usage in novel situations, 2024

    Robin Sch ¨on, Julian Lorenz, Katja Ludwig, and Rainer Lienhart. Adapt- ing the segment anything model during usage in novel situations, 2024

    2024

  17. [25]

    Boosttrack: boosting the similarity measure and detection confidence for improved multiple object tracking

    Vukasin D Stanojevic and Branimir T Todorovic. Boosttrack: boosting the similarity measure and detection confidence for improved multiple object tracking. Machine Vision and Applications , 35(3), 2024

    2024

  18. [26]

    FraMOT: Frame- adaptive multi-object tracking in low framerate conditions, 2022

    Vuk Stojnic, Igor Paunovi ´c, and Bojan Todorovic. FraMOT: Frame- adaptive multi-object tracking in low framerate conditions, 2022

    2022

  19. [27]

    Can sam segment anything? when sam meets camouflaged object detection, 2023

    Lv Tang, Haoke Xiao, and Bo Li. Can sam segment anything? when sam meets camouflaged object detection, 2023

    2023

  20. [28]

    Simple online and realtime tracking with a deep association metric, 2017

    Nicolai Wojke, Alex Bewley, and Dietrich Paulus. Simple online and realtime tracking with a deep association metric, 2017

    2017

  21. [29]

    Segment-anything models achieve zero-shot robustness in autonomous driving, 2024

    Jun Yan, Pengyu Wang, Danni Wang, Weiquan Huang, Daniel Watzenig, and Huilin Yin. Segment-anything models achieve zero-shot robustness in autonomous driving, 2024

    2024

  22. [30]

    Self-supervised multi-object tracking with path consistency, 2023

    Lijun Yang, Difan Wu, and Yan Chen. Self-supervised multi-object tracking with path consistency, 2023

    2023

  23. [31]

    MOTR: End-to-end multiple-object tracking with trans- former

    Fangao Zeng, Bin Dong, Tiancai Wang, Xiangyu Zhang, Yichen Wei, and Jian Sun. MOTR: End-to-end multiple-object tracking with trans- former. 2022

    2022

  24. [32]

    MOTRv3: Refined multiple object tracking with transformers

    Xiao Zhang, Hui Li, Kun Duan, and Jian Sun. MOTRv3: Refined multiple object tracking with transformers. 2024

    2024

  25. [33]

    Bytetrack: Multi- object tracking by associating every detection box, 2022

    Yifu Zhang, Peize Sun, Yi Jiang, Dongdong Yu, Fucheng Weng, Zehuan Yuan, Ping Luo, Wenyu Liu, and Xinggang Wang. Bytetrack: Multi- object tracking by associating every detection box, 2022

    2022

  26. [34]

    Fairmot: On the fairness of detection and re-identification in multiple object tracking, 2020

    Yifu Zhang, Chunyu Wang, Xinggang Wang, Wenyu Liu, and Ping Luo. Fairmot: On the fairness of detection and re-identification in multiple object tracking, 2020

    2020

  27. [35]

    CenterTrack: Tracking objects as points

    Xingyi Zhou, Vladlen Koltun, and Philipp Kr ¨ahenb¨uhl. CenterTrack: Tracking objects as points. In European Conference on Computer Vision (ECCV), pages 245–261, 2020

    2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.