Pith. sign in

REVIEW 6 major objections 6 minor 45 references

Depth-Aware Scoring and Hierarchical Alignment for Multiple Object Tracking

T0 review · 6 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Depth-aware, training-free tracker beats trained rivals on DanceTrack

desk verdict Genuinely new training-free depth+MOT scoring with credible partial results, undercut by missing normalization details and deferred benchmark numbers. read the letter →

arxiv 2506.00774 v1 pith:ANABIFY5 submitted 2025-06-01 cs.CV

Milad Khanchi , Maria Amer , Charalambos Poullis This is my paper
classification cs.CV
keywords multipleobjecttrackingmonoculardepthestimationzero-shotHierarchicalAlignmentScoredataassociationtracking-by-detectionpixel-levelMOT
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

DepthMOT aims to show that a multiple-object tracker can improve association accuracy without any training by adding two cheap cues to the standard bounding-box overlap score: monocular depth and pixel-level shape alignment. The paper argues that motion- and appearance-based scores alone confuse objects that look alike or overlap, whereas a depth histogram of each detection and a mask-level IoU between tracklet and detection disambiguate them. It reports state-of-the-art results on DanceTrack, improving HOTA by 1.97 points, AssA by 3.84 points, and IDF1 by 3.54 points over the trained DiffMOT baseline, and matching DiffMOT on SportsMOT. If correct, the framework would give a plug-in, training-free route to better tracking in crowded and occluded scenes.

What carries the argument

The load-bearing objects are the per-detection depth histogram and the Hierarchical Alignment Score (HAS). Zero-shot depth estimation (Depth Pro) produces a dense relative depth map per frame; for each bounding box the depth values are collected into a histogram, and cosine similarity between histograms of a tracklet and a new detection becomes the depth score $S_{\mathrm{depth}}$. HAS combines bounding-box IoU with segmentation-mask IoU from a promptable video segmentation model (SAM2), using the multiplicative-exponential form $S_{\mathrm{HAS}} = S_{\mathrm{IoU}}^{\mathrm{bbox}} \exp(S_{\mathrm{IoU}}^{\mathrm{seg}})$ so that fine-grained shape alignment matters most once boxes already overlap. The final matching cost sums HAS, directional motion similarity, depth similarity, and appearance similarity from a pretrained re-identification model, and is passed to a linear assignment solver.

What would settle it

Construct a sequence of two visually identical objects that cross paths while remaining at nearly the same distance from the camera; if the depth-histogram cosine similarity between the two different objects is comparable to the same-object similarity and the mask IoU also matches both boxes, the tracker should produce an identity switch at the crossing. Alternatively, recompute the reported DanceTrack results while varying the depth-histogram bin count; if HOTA changes by more than a few points, the claimed gains depend on an unspecified implementation detail.

Watch

Extended reading notes

Core claim

The paper's central claim is that a training-free association process can match or beat trained trackers by treating monocular depth as an independent decision matrix and by refining IoU with a Hierarchical Alignment Score. The depth score is obtained by running a zero-shot depth estimator on each frame, building a histogram of depth values inside every detection box, and computing cosine similarity between histograms across frames. The Hierarchical Alignment Score is defined as $S_{\mathrm{HAS}}(\hat{X},D) = S_{\mathrm{IoU}}^{\mathrm{bbox}}(\hat{X},D) \times \exp\bigl(S_{\mathrm{IoU}}^{\mathrm{seg}}(\hat{X},D)\bigr)$, so coarse bounding-box overlap dominates at first and pixel-level mask overlap, produced by a promptable video segmentation model, sharpens the score as alignment improves. These are added to directional motion and appearance scores, and the total matrix is solved as a linear assignment. On DanceTrack the paper reports HOTA 64.27, IDF1 66.54, and AssA 51.04, surpassing the trained DiffMOT tracker; on SportsMOT it reports HOTA 76.2, comparable to DiffMOT. The authors also claim this is the first MOT framework to use monocular depth as an independent association feature.

Load-bearing premise

The load-bearing premise is that the zero-shot depth histogram inside a detection box is a stable and distinctive signature of that object across frames, despite changes in scale, partial occlusion, and depth-map noise.

Editorial extensions

If this is right

  • Depth as an independent cue should reduce identity switches when two objects have similar appearance but stand at different distances from the camera.
  • The HAS score should make association more robust to partial occlusion, because mask IoU can still match when bounding boxes overlap heavily.
  • Because no component is trained or fine-tuned, the framework can be dropped into existing tracking-by-detection pipelines without per-dataset training.
  • On the reported benchmarks, the gap over DiffMOT on DanceTrack suggests the largest gains appear under non-linear motion and frequent occlusion rather than linear pedestrian scenes.
  • The ablation indicates that depth and HAS contribute complementary gains: HAS alone raises HOTA from 54.78 to 60.62, and adding depth raises it further to 61.81 on the DanceTrack validation set.

Reading between the lines

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

  • A testable extension is to swap Depth Pro for any metric-depth estimator and check whether the histogram descriptor remains discriminative; the paper's argument implies the cue, not the specific model, is what matters.
  • One implicit consequence is that the depth histogram could serve as a weak ReID signature in appearance-free regimes, e.g., low-resolution or grayscale footage where appearance models degrade.
  • The method may inherit the limitations of the segmentation model: if the promptable segmentation fails on heavily truncated objects, the exponential mask term could mislead rather than disambiguate.
  • It would be worth measuring performance with the depth histogram's bin count and normalization varied, since the preprint does not fix or analyze those choices.
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

6 major / 6 minor

Summary. DepthMOT is a tracking-by-detection MOT framework that augments association with three cues on top of a non-linear Kalman filter with OC-SORT-style ORU: zero-shot monocular depth (Depth Pro), a Hierarchical Alignment Score (HAS) combining bounding-box IoU with SAM2 mask IoU via SHAS = SIoU_bbox * exp(SIoU_seg), and FastReID appearance embeddings with EMA updating. The total matching score (Eq. 7) is an unweighted sum of the HAS, angle, depth, and appearance components, solved with a linear assignment. Depth and appearance are applied only to high-confidence detections, while low-confidence detections are associated by IoU alone in a second stage. The paper reports gains over DiffMOT on DanceTrack (HOTA 64.27 vs 62.3, AssA 51.04 vs 47.2, IDF1 66.54 vs 63.0) and comparable results on SportsMOT, with MOT17/MOT20 results deferred to an external supplementary. An ablation on the DanceTrack validation set (Table 3) attributes the gains to HAS and depth.

Significance. Conditional significance is real but moderate. If the associations hold, the paper provides a clean, training-free depth augmentation that improves association metrics (HOTA/IDF1/AssA) on the non-linear-motion DanceTrack benchmark, a setting where IoU-only association is known to fail; the release of code and the use of the same YOLOX detector as DiffMOT make the comparison concrete and falsifiable. The paper also makes an explicit, checkable novelty claim (first use of monocular depth as an independent decision matrix) and is honest in the text that no single tracker wins on MOT17/MOT20. However, the gains are selective: MOTA and DetA drop on both benchmarks, the MOT17/MOT20 evidence is outside the manuscript, and the reproducibility gaps in the depth-histogram specification and in the calibration of Eq. (7) currently prevent verification of the central attribution of the gains to depth.

major comments (6)
  1. [Sec. 3.5, Eq. (7)] The total matching score sums four components with heterogeneous ranges: SHAS lies in [0, e] (since SIoU_bbox <= 1 and exp(SIoU_seg) <= e), while Sang, Sdepth, and Semb are cosine similarities in [-1, 1] or [0, 1]. The equal-weight sum therefore contains an implicit calibration: at high bounding-box overlap the HAS term can outweigh depth and appearance by a factor of two or more, and any change in histogram binning, embedding normalization, or a simple z-score would alter the relative contribution of the depth channel. Since the paper's headline claim is 'without any training nor fine-tuning', the authors must either report normalized components and their weights, or provide a sensitivity analysis showing that the Table 1 and Table 3 rankings are stable under renormalization of the four score types.
  2. [Sec. 3.3] The depth score is never specified operationally: the paper does not state the number of histogram bins, the binning strategy (fixed range vs per-frame percentile), the normalization (count, density, or L2), how background pixels inside the bounding box are masked out, or how Depth Pro's per-scene scale ambiguity is handled across frames. Because the depth histogram is the single new cue claimed to drive the association gains, this omission makes the central experiments unreproducible from the text; the statement that histograms 'yielded the most effective results' also implies a selection process that should be reported as part of the 'no fine-tuning' claim.
  3. [Table 3] The ablation table is unreadable as typeset: the checkmarks do not align with the column headers, the header row is merged ('DepthHOTA'), and the rows are concatenated, so a reader cannot determine which configuration produced 61.81 HOTA. More substantively, the table has no row that excludes both HAS and depth, and the apparent sequences of checkmarks suggest configurations whose comparison (e.g., 60.62 vs 60.45 vs 61.81) would need pairwise toggles of depth and HAS to attribute the gains; as printed, the core claim that 'the combination of both HAS and depth yields the highest accuracy' cannot be audited.
  4. [Tables 1 and 2; Sec. 4.1] The text states that DepthMOT 'outperforms' DiffMOT on DanceTrack, but DepthMOT is lower on MOTA (90.08 vs 92.8) and DetA (81.07 vs 82.5), and on SportsMOT it ties DiffMOT on HOTA (76.2) while losing on MOTA (95.9 vs 97.1) and DetA (88.5 vs 89.3). Since the abstract claims 'state-of-the-art results on challenging benchmarks', the systematic detection-side regression (possibly from the low-confidence second-stage association, which is pure IoU, or from depth failing on small or heavily occluded boxes) must be analyzed, and the claim should be qualified to association metrics rather than left implicit.
  5. [Sec. 3.4, Eq. (6) and Fig. 2 caption] The Fig. 2 caption states that the exponential term amplifies fine-grained segmentation 'particularly when bounding box overlap is low', but SHAS = SIoU_bbox * exp(SIoU_seg) vanishes as SIoU_bbox approaches 0, so pixel-level shape similarity cannot rescue a match when the bounding boxes do not overlap. The hierarchical narrative (shape influence grows as coarse alignment improves) is consistent with the product form only in the large-overlap regime; the caption and the text must be reconciled with the formula, or the formula changed, since this score is a claimed contribution.
  6. [Sec. 4.1, MOT17/MOT20] The breadth of the benchmark claim rests on a non-archival external link (sigport.org), and the sentence 'Ours achieves the lowest false positive rate' has no locale in the main text: it is unclear whether it refers to MOT17, MOT20, or both, and the supporting numbers are absent. The manuscript should include the MOT17/MOT20 tables (or cite a permanently archived supplementary) and state the scope of the claim; note that the paper itself concedes 'under MOT17 and MOT20 datasets (linear motion), no tracker performs best overall', which is a limitation that the abstract's unconditional phrasing does not reflect.
minor comments (6)
  1. [Sec. 3.1, Eq. (5)] State the range of the dynamic coefficient C explicitly: with T=0.95 and a confidence threshold of 0.6, C lies in [0.95, 1], so a new embedding receives weight at most 0.05; this near-frozen update deserves a sentence of justification and a check against the Deep OC-SORT reference.
  2. [Sec. 3.5 and Eq. (7)] The angle score Sang is referenced in Eq. (7) but never defined in the paper; give its formula (e.g., the direction cosine between tracklet velocity and observation displacement used in Deep OC-SORT).
  3. [Sec. 2 and Sec. 3.3] The novelty claim ('first MOT framework to incorporate 3D features (monocular depth) as an independent decision matrix') should be sharpened against Quach et al. [16] and Wang et al. [17], who already inject depth into the association step via a Kalman-filter state and a stereo-aware OCSORT variant; the distinction needs to be argued in the main text, not only asserted through the phrase 'independent decision matrix'.
  4. [Sec. 4] No runtime or hardware details are reported; since the paper argues an efficiency advantage over MOTRv2, report frames-per-second and the per-frame cost of Depth Pro and SAM2.
  5. [Sec. 4.2] The ablation is performed on the validation set and the final configuration is selected there; disclose this selection explicitly so the 'no fine-tuning' claim is unambiguous about validation-based design choices.
  6. [Throughout] Minor presentation items: the blue/bold highlighting convention described in Sec. 4 is not visible in the typeset tables, and the phrase 'without any training nor fine-tuning' should be 'without any training or fine-tuning'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DepthMOT's scores come from off-the-shelf pretrained models and its benchmark gains are externally measured, not derived from its own definitions.

full rationale

The paper contains no derivation chain in which a predicted quantity is defined in terms of the result it is supposed to explain. Equation (7) combines four independently computed score matrices from YOLOX, FastReID, SAM2, and Depth Pro, and the paper explicitly states that it avoids training or fine-tuning any components. No parameters are fitted to the DanceTrack, SportsMOT, MOT17, or MOT20 benchmarks, so the reported HOTA, AssA, and IDF1 numbers are empirical evaluations against ground truth rather than consequences of the score formula. The Hierarchical Alignment Score in Eq. (6) is a new weighting of IoU and mask IoU, but it does not by construction determine any tracking metric. The reference list contains no self-citations, so no self-citation chain is load-bearing. Concerns about unnormalized score ranges in Eq. (7) are matters of reproducibility and score calibration, not circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method introduces no new physical entities. The key assumptions are the reliability of off-the-shelf zero-shot depth (Depth Pro) and segmentation (SAM2) models on the target scenes, the stability of depth histograms as an object signature, and the validity of summing four unnormalized affinity scores. The free parameters are mostly adopted from prior work or left unspecified, which creates reproducibility risk.

free parameters (4)
  • Detection confidence threshold = 0.6
    Used to split detections into high and low confidence groups (Section 3.5); adopted from ByteTrack [19], not tuned on the evaluated benchmarks, but it is a hand-set threshold that affects which objects enter the main association.
  • Equal weighting in matching score (Eq. 7) = 1 (implicit)
    The four terms (HAS, Sang, Sdepth, Semb) are summed with unit weights. No normalization or learned weighting is specified, so scale differences between IoU (0-1), cosine similarities (-1 to 1), and HAS (possibly >1) could bias the score.
  • Depth histogram bin count = not specified
    The number of bins used to vectorize depth values inside bounding boxes is never stated. This affects the cosine similarity Sdepth and is a free design choice.
  • EMA coefficient T = 0.95
    Used in Eq. 5 for tracklet embedding updates, following Deep OC-SORT [7]. It is a hand-set constant from prior work.
assumptions (5)
  • domain assumption Zero-shot monocular depth from Depth Pro is accurate enough for association.
    Section 3.3 relies on Depth Pro without calibration on target datasets; if depth errors are large or scale-inconsistent, the histogram similarity will not separate objects.
  • domain assumption SAM2 masks represent object shape well and propagate backward correctly.
    Section 3.2 uses SAM2 to generate tracklet masks and to propagate detections backward; this is a pretrained model, and its accuracy on sports scenes is assumed.
  • domain assumption The linear Kalman filter with ORU handles non-linear motion in DanceTrack.
    Section 3.1 adopts OC-SORT's ORU to recover from missed detections. The underlying motion model is still linear (constant velocity), which is a known limitation for non-linear motion.
  • domain assumption Depth histograms are stable for a given object across consecutive frames.
    The depth similarity Sdepth compares histograms between predicted tracklet states and new detections; if depth estimates flicker or vary with pose, the score loses power.
  • domain assumption The individual score components are independent and addable.
    Eq. 7 sums motion, depth, appearance, and HAS scores; there is no analysis of correlation between these cues or a principled fusion rule.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Depth-Aware Scoring and Hierarchical Alignment for Multiple Object Tracking." pith.science (2026). https://pith.science/paper/ANABIFY5

@misc{pith2026250600774,
  author       = {Pith},
  title        = {Pith review of: Depth-Aware Scoring and Hierarchical Alignment for Multiple Object Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ANABIFY5}},
  note         = {Machine review of arXiv:2506.00774}
}
read the original abstract

Current motion-based multiple object tracking (MOT) approaches rely heavily on Intersection-over-Union (IoU) for object association. Without using 3D features, they are ineffective in scenarios with occlusions or visually similar objects. To address this, our paper presents a novel depth-aware framework for MOT. We estimate depth using a zero-shot approach and incorporate it as an independent feature in the association process. Additionally, we introduce a Hierarchical Alignment Score that refines IoU by integrating both coarse bounding box overlap and fine-grained (pixel-level) alignment to improve association accuracy without requiring additional learnable parameters. To our knowledge, this is the first MOT framework to incorporate 3D features (monocular depth) as an independent decision matrix in the association step. Our framework achieves state-of-the-art results on challenging benchmarks without any training nor fine-tuning. The code is available at https://github.com/Milad-Khanchi/DepthMOT

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

45 extracted references · 39 canonical work pages

  1. [16]

    Unifying Short and Long- Term Tracking with Graph Hierarchies,

    O. Cetintas, G. Bras ´o, and L. Leal-Taix ´e, “Unifying Short and Long- Term Tracking with Graph Hierarchies,” in IEEE Conf. Comput. Vis. Pattern Recog., 2023, pp. 22 877–22 887

    work page 2023

  2. [1]

    Chal- lenges include noisy observations, occlusion, rapid motion, or similar objects

    INTRODUCTION Multiple object tracking (MOT) [1, 2, 3] involves detect- ing objects in video frames and continuously tracking them across time, requiring the simultaneous resolution of object detection, data association, and trajectory prediction. Chal- lenges include noisy observations, occlusion, rapid motion, or similar objects. Recent research has focu...

    work page

  3. [2]

    Depth-Aware Scoring and Hierarchical Alignment for Multiple Object Tracking

    RELATED WORKS MOT approaches can be broadly categorized into: joint detection-ReIdentification (JDR) [4, 5] and tracking-by- detection (TBD) [1, 2, 6, 7, 8]. Tracking can be performed frame-wise [2, 6, 7, 8], where the model processes each frame sequentially, utilizing information from previous frames. Al- ternatively, it can be conducted at the sequence ...

    work page Pith review arXiv 2025

  4. [3]

    For object detection, we use the known model YOLOX [13]

    METHODOLOGY In a TBD paradigm, objects are initially detected in indi- vidual frames and subsequently associated across time in a frame-wise manner. For object detection, we use the known model YOLOX [13]. We perform object association in two steps. In the first step, we categorize detected objects based on their detection confidence ( DF ) as in [19]: hi...

    work page

  5. [4]

    This allows us to retrieve the segmentation of each newly detected object in the previous frame, even if that object was not explicitly tracked before

    For objects newly detected in the current frame t, SAM2 uses the bounding box of each detected object in frame t to propagate the segmentation backward to find the correspond- ing object mask in frame t − 1. This allows us to retrieve the segmentation of each newly detected object in the previous frame, even if that object was not explicitly tracked befor...

    work page

  6. [5]

    HOTA assesses both detection and association accuracy

    EXPERIMENTAL RESULTS Main performance metrics for MOT are Higher Order Track- ing Accuracy (HOTA), ID-based F1 Score (IDF1), and Asso- ciation Accuracy (AssA) [29]. HOTA assesses both detection and association accuracy. IDF1 and AssA primarily evaluate association performance. Multi-Object Tracking Accuracy (MOTA) metric focuses on detection accuracy. In ...

    work page

  7. [6]

    CONCLUSION We proposed the DepthMOT framework for MOT, which incorporates depth information alongside visual and motion features. Additionally, we introduced a hierarchical align- ment score (HAS) that prioritizes bounding boxes during the initial association phase and progressively balances pixel- level segments as matching improves. DepthMOT achieves st...

    work page

  8. [7]

    Lost and Found: Overcoming Detector Failures in Online Multi-object Tracking,

    L. Vaquero, Y . Xu, X. Alameda-Pineda, V . M. Brea, and M. Mucientes, “Lost and Found: Overcoming Detector Failures in Online Multi-object Tracking,” in Eur . Conf. Comput. Vis. Springer, 2024, pp. 448–466

    work page 2024

Show all 45 references
  1. [8]

    Diff- Mot: 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, “Diff- Mot: A Real-time Diffusion-based Multiple Object Tracker with Non- linear Prediction,” in IEEE Conf. Comput. Vis. Pattern Recog. , 2024, pp. 19 321–19 330

    2024

  2. [9]

    A Confidence-Aware Match- ing Strategy for Generalized Multi-Object Tracking,

    K. Shim, J. Hwang, K. Ko, and C. Kim, “A Confidence-Aware Match- ing Strategy for Generalized Multi-Object Tracking,” inIEEE Int. Conf. Image Process. IEEE, 2024, pp. 4042–4048

    2024

  3. [10]

    FairMot: On the Fairness of Detection and Re-identification in Multiple Object Track- ing,

    Y . Zhang, C. Wang, X. Wang, W. Zeng, and W. Liu, “FairMot: On the Fairness of Detection and Re-identification in Multiple Object Track- ing,” Int. J. Comput. Vis., vol. 129, pp. 3069–3087, 2021

    2021

  4. [11]

    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 Trans. Pattern Anal. Mach. Intell. , vol. 45, no. 6, pp. 7820–7835, 2022

    2022

  5. [12]

    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 IEEE Conf. Comput. Vis. Pattern Recog., 2023, pp. 9686–9696

    2023

  6. [13]

    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,” inIEEE Int. Conf. Image Process. IEEE, 2023, pp. 3025–3029

    2023

  7. [14]

    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 IEEE Conf. Comput. Vis. Pattern Recog., 2022, pp. 8844–8854

    2022

  8. [15]

    Learning a Neural Solver for Multiple Object Tracking,

    G. Bras ´o and L. Leal-Taix ´e, “Learning a Neural Solver for Multiple Object Tracking,” in IEEE Conf. Comput. Vis. Pattern Recog. , 2020, pp. 6247–6257

    2020

  9. [17]

    Multi-Object Tracking and Segmentation Via Neural Message Passing,

    G. Bras ´o, O. Cetintas, and L. Leal-Taix ´e, “Multi-Object Tracking and Segmentation Via Neural Message Passing,” Int. J. Comput. Vis. , vol. 130, no. 12, pp. 3035–3053, 2022

    2022

  10. [18]

    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 IEEE Conf. Comput. Vis. Pattern Recog., 2023, pp. 22 056–22 065

    2023

  11. [19]

    Yolox: Ex- ceeding yolo series in 2021,

    G. Zheng, L. Songtao, W. Feng, L. Zeming, S. Jian et al., “Yolox: Ex- ceeding yolo series in 2021,” arXiv preprint arXiv:2107.08430, 2021

    2021 arXiv

  12. [20]

    MeMOTR: Long-Term Memory-Augmented Transformer for Multi-Object Tracking,

    R. Gao and L. Wang, “MeMOTR: Long-Term Memory-Augmented Transformer for Multi-Object Tracking,” in Int. Conf. Comput. Vis. , 2023, pp. 9901–9910

    2023

  13. [21]

    DeconfuseTrack: Dealing with Confusion for Multi-Object Tracking,

    C. Huang, S. Han, M. He, W. Zheng, and Y . Wei, “DeconfuseTrack: Dealing with Confusion for Multi-Object Tracking,” in IEEE Conf. Comput. Vis. Pattern Recog., 2024, pp. 19 290–19 299

    2024

  14. [22]

    Depth Perspective-Aware Multiple Object Tracking,

    K. G. Quach, P. Nguyen, C. N. Duong, T. D. Bui, and K. Luu, “Depth Perspective-Aware Multiple Object Tracking,” inEng. Appl. AI Swarm Intell. Springer, 2024, pp. 181–205

    2024

  15. [23]

    Robust multiple obsta- cle tracking method based on depth aware OCSORT for agricultural robots,

    J. Wang, H. Zheng, Y . Yu, Y . He, and Y . Liu, “Robust multiple obsta- cle tracking method based on depth aware OCSORT for agricultural robots,” Comput. Electron. Agric., vol. 217, p. 108580, 2024

    2024

  16. [24]

    DET: Depth-Enhanced Tracker to Mitigate Severe Occlusion and Homogeneous Appearance Problems for In- door Multiple-Object Tracking,

    C.-J. Liu and T.-N. Lin, “DET: Depth-Enhanced Tracker to Mitigate Severe Occlusion and Homogeneous Appearance Problems for In- door Multiple-Object Tracking,” IEEE Access, vol. 10, pp. 8287–8304, 2022

    2022

  17. [25]

    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 Eur . Conf. Comput. Vis. Springer, 2022, pp. 1–21

    2022

  18. [26]

    Fastreid: A Pytorch Toolbox for General Instance Re-identification,

    L. He, X. Liao, W. Liu, X. Liu, P. Cheng, and T. Mei, “Fastreid: A Pytorch Toolbox for General Instance Re-identification,” in ACM Int. Conf. Multimedia, 2023, pp. 9664–9667

    2023

  19. [27]

    BoT-SORT: Ro- bust Associations Multi-Pedestrian Tracking,

    N. Aharon, R. Orfaig, and B.-Z. Bobrovsky, “BoT-SORT: Ro- bust Associations Multi-Pedestrian Tracking,” arXiv preprint arXiv:2206.14651, 2022

    2022 arXiv

  20. [28]

    StrongSORT: Make DeepSORT Great Again,

    Y . Du, Z. Zhao, Y . Song, Y . Zhao, F. Su, T. Gong, and H. Meng, “StrongSORT: Make DeepSORT Great Again,” IEEE Trans. Multime- dia, vol. 25, pp. 8725–8737, 2023

    2023

  21. [29]

    Learning the What and How of Annotation in Video Object Segmentation,

    T. Delatolas, V . Kalogeiton, and D. P. Papadopoulos, “Learning the What and How of Annotation in Video Object Segmentation,” in IEEE Winter Conf. Appl. Comput. Vis., 2024, pp. 6951–6961

    2024

  22. [30]

    SAM 2: Segment Anything in Images and Videos,

    N. Ravi, V . Gabeur, Y .-T. Hu, R. Hu, C. Ryali, T. Ma, H. Khedr, R. R¨adle, C. Rolland, L. Gustafson et al., “SAM 2: Segment Anything in Images and Videos,” arXiv preprint arXiv:2408.00714, 2024

    2024 arXiv

  23. [31]

    To- wards Zero-Shot Scale-Aware Monocular Depth Estimation,

    V . Guizilini, I. Vasiljevic, D. Chen, R. Ambrus ,, and A. Gaidon, “To- wards Zero-Shot Scale-Aware Monocular Depth Estimation,” in IEEE Conf. Comput. Vis. Pattern Recog., 2023, pp. 9233–9243

    2023

  24. [32]

    Zero-Shot Metric Depth with a Field-of-View Conditioned Diffusion Model,

    S. Saxena, J. Hur, C. Herrmann, D. Sun, and D. J. Fleet, “Zero-Shot Metric Depth with a Field-of-View Conditioned Diffusion Model,” arXiv preprint arXiv:2312.13252, 2023

    2023 arXiv

  25. [33]

    Depth Pro: Sharp Monocular Metric Depth in Less Than a Second,

    A. Bochkovskii, A. Delaunoy, H. Germain, M. Santos, Y . Zhou, S. R. Richter, and V . Koltun, “Depth Pro: Sharp Monocular Metric Depth in Less Than a Second,” arXiv preprint arXiv:2410.02073, 2024

    2024 arXiv

  26. [34]

    Simple online and realtime tracking,

    A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, “Simple online and realtime tracking,” in IEEE Int. Conf. Image Process. IEEE, 2016, pp. 3464–3468

    2016

  27. [35]

    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,” Int. J. Comput. Vis., vol. 129, pp. 548–578, 2021

    2021

  28. [36]

    Tracking Objects as Points,

    X. Zhou, V . Koltun, and P. Kr ¨ahenb¨uhl, “Tracking Objects as Points,” in Eur . Conf. Comput. Vis. Springer, 2020, pp. 474–490

    2020

  29. [37]

    Track To Detect and Segment: An Online Multi-Object Tracker,

    J. Wu, J. Cao, L. Song, Y . Wang, M. Yang, and J. Yuan, “Track To Detect and Segment: An Online Multi-Object Tracker,” in IEEE Conf. Comput. Vis. Pattern Recog., 2021, pp. 12 352–12 361

    2021

  30. [38]

    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

    2012 arXiv

  31. [39]

    Diffusion- Track: Diffusion Model for Multi-Object Tracking,

    R. Luo, Z. Song, L. Ma, J. Wei, W. Yang, and M. Yang, “Diffusion- Track: Diffusion Model for Multi-Object Tracking,” in AAAI, vol. 38, no. 5, 2024, pp. 3991–3999

    2024

  32. [40]

    Simple Cues Lead to a Strong Multi-Object Tracker,

    J. Seidenschwarz, G. Bras ´o, V . C. Serrano, I. Elezi, and L. Leal-Taix´e, “Simple Cues Lead to a Strong Multi-Object Tracker,” in IEEE Conf. Comput. Vis. Pattern Recog., 2023, pp. 13 813–13 823

    2023

  33. [41]

    MotionTrack: Learning motion predictor for multiple object track- ing,

    C. Xiao, Q. Cao, Y . Zhong, L. Lan, X. Zhang, Z. Luo, and D. Tao, “MotionTrack: Learning motion predictor for multiple object track- ing,” Neural Networks, vol. 179, p. 106539, 2024

    2024

  34. [42]

    Towards Gen- eralizable Multi-Object Tracking,

    Z. Qin, L. Wang, S. Zhou, P. Fu, G. Hua, and W. Tang, “Towards Gen- eralizable Multi-Object Tracking,” in IEEE Conf. Comput. Vis. Pattern Recog., 2024, pp. 18 995–19 004

    2024

  35. [43]

    Hard to Track Objects With Irregular Motions and Similar Appearances? Make It Easier by Buffering the Matching Space,

    F. Yang, S. Odashima, S. Masui, and S. Jiang, “Hard to Track Objects With Irregular Motions and Similar Appearances? Make It Easier by Buffering the Matching Space,” in IEEE Winter Conf. Appl. Comput. Vis., 2023, pp. 4799–4808

    2023

  36. [44]

    Global Tracking Trans- formers,

    X. Zhou, T. Yin, V . Koltun, and P. Kr¨ahenb¨uhl, “Global Tracking Trans- formers,” in IEEE Conf. Comput. Vis. Pattern Recog. , 2022, pp. 8771– 8780

    2022

  37. [45]

    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 Int. Conf. Comput. Vis., 2023, pp. 9921–9931

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.