Pith. sign in

REVIEW 4 major objections 5 minor 43 references

An Uncertainty-aware DETR Enhancement Framework for Object Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Modeling predicted bounding boxes as 4D Gaussians with a Gromov-Wasserstein loss improves DETR detectors and yields per-box localization uncertainty.

desk verdict Solid DETR plug-in with honest ablations, but the theoretical core—Bayes Risk as trace of covariance—rests on an assumption the model doesn't satisfy; referee it, but expect the theory to be rewritten. read the letter →

arxiv 2507.14855 v1 pith:WNF6GEZE submitted 2025-07-20 cs.CV

Xingshu Chen , Sicheng Yu , Chong Cheng , Hao Wang , Ting Tian This is my paper
classification cs.CV
keywords objectdetectionDETRuncertaintyestimationGromov-WassersteindistanceGaussianboundingboxBayesriskregressionleukocyte
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper proposes a plug-in enhancement for DETR-style object detectors that treats each predicted bounding box not as four fixed numbers but as a four-dimensional Gaussian distribution with a learnable covariance matrix. The ground-truth box is modeled as a two-dimensional Gaussian derived from its inscribed ellipse, and the gap between the two distributions is minimized with the Gromov-Wasserstein distance rather than an IoU-style geometric term. From that probabilistic model the authors derive a Bayes Risk for $L_2$ regression that equals the trace of the predicted covariance, and they use it to reweight the decoder output embeddings, the classification loss, and the Hungarian matching cost. They also give an algorithm that converts the predicted covariance into a single localization-uncertainty score per box. On COCO the framework adds 0.1 to 1.4 AP across three DETR variants, and on the LISC and WBCDD leukocyte datasets it reports the highest AP among the compared detectors.

What carries the argument

The central object is the predicted 4D Gaussian $N_P(\boldsymbol\mu_P, \boldsymbol\Sigma_P)$ with learnable diagonal covariance, whose four coordinates are the box center, width, and height. The Gromov-Wasserstein distance between this 4D Gaussian and the 2D ground-truth Gaussian $N_g(\boldsymbol\mu_g, \boldsymbol\Sigma_g)$ provides a distributional regression loss; it is the mechanism that makes the covariance matrix trainable. The derived Bayes Risk, equal to the trace of $\boldsymbol\Sigma_P$, is the quantity that carries the uncertainty information into the three refinement modules. A confidence-interval algorithm over the predicted Gaussian converts the covariance into a single localization-uncertainty number per box.

What would settle it

On a held-out split, compute for every predicted box the empirical squared error per coordinate, average it, and compare it with the average predicted variance; if the average predicted variance does not bound or track the empirical error, or if the off-diagonal covariances are large, the central Bayes Risk identity and the uncertainty score are not measuring what the paper claims.

Watch

Extended reading notes

Core claim

The paper's central claim is that explicitly modeling bounding-box prediction uncertainty, rather than treating predictions as point values, is a generally useful enhancement for DETR-based detectors. Concretely, the authors claim that replacing part of the deterministic regression objective with a Gromov-Wasserstein distance between a 2D ground-truth Gaussian and a 4D predicted Gaussian, and then using the derived Bayes Risk $\mathrm{Risk}^* = \sigma^2_{\hat c_x} + \sigma^2_{\hat c_y} + \sigma^2_{\hat w} + \sigma^2_{\hat h}$ to refine output embeddings, classification loss, and matching cost, improves detection accuracy while also supplying an uncertainty estimate for each predicted box. The reported evidence is a consistent AP gain on COCO over H-DETR, DINO-DETR, and Relation-DETR, plus best-reported AP on the LISC and WBCDD leukocyte datasets. The authors further show that the uncertainty score inversely tracks a combined classification-IoU quality metric.

Load-bearing premise

The load-bearing assumption is that each predicted box coordinate is an independent Gaussian error centered on the true coordinate, with each true coordinate uniform on the unit interval; only under that setup does the Bayes Risk reduce to the trace of the predicted covariance, so the refinement and uncertainty scores inherit that assumption.

Editorial extensions

If this is right

  • DETR variants that add the Gaussian modeling, Gromov-Wasserstein loss, and Bayes Risk refinement should gain accuracy without architectural changes; the paper reports +0.1 to +1.4 AP on COCO val2017 over three variants.
  • The same plug-in transfers to domain-specific detection: the enhanced H-DETR reports +1.4 AP on LISC and +1.9 AP on WBCDD over the prior best leukocyte detector.
  • The framework supplies a per-box localization uncertainty that is low for high-quality predictions and high for poor ones, matching the reported inverse relationship to the combined metric $s \cdot u^{0.5}$.
  • The components are additive: the ablations attribute part of the gain to the Bayes Risk refinement modules, with the Gromov-Wasserstein loss as a prerequisite, and the remaining gains split between output embedding, matching cost, and classification loss.
  • The method also extends beyond DETR to one-stage detectors, with a VOC experiment showing YOLOv5 improved by 0.9 AP when the output embedding and classification loss are refined.

Reading between the lines

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

  • If the learned covariance truly tracks regression error, the same Bayes Risk scalar could serve as a query-ordering signal for DETR decoders beyond the three modules modified here, for example in iterative refinement stopping or active learning.
  • The trace identity depends on treating the four coordinates as independent and centered on the ground truth; testing whether off-diagonal covariance or systematic bias improves the uncertainty score would clarify whether the diagonal assumption is a real limitation.
  • The confidence-interval algorithm's division count $k$ trades inference time against uncertainty stability; a closed-form expected IoU under the predicted Gaussian might achieve similar calibration without sampling.
  • A direct comparison of predicted covariance with empirical error on held-out data would determine whether the reported uncertainty is aleatoric, epistemic, or a mix; the paper does not separate the two.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes an uncertainty-aware plug-in framework for DETR-based detectors. Bounding-box predictions are modeled as a 4D Gaussian with learnable diagonal covariance, and the Gromov-Wasserstein distance is added as a regression loss. A 'Bayes Risk' equal to the trace of the covariance (Eq. 5) is used to reweight decoder output embeddings, classification loss, and matching cost, and an algorithm based on 95% confidence intervals is proposed to output a localization-uncertainty score. Experiments on COCO val2017 report AP improvements of 0.1 to 1.4 points over three DETR variants, and the framework is reported to achieve state-of-the-art AP on LISC and WBCDD leukocyte detection. Ablations in Tables 3 to 5 isolate the contribution of the GW loss and each Bayes Risk refinement module.

Significance. If the central derivation were valid, the framework would be a broadly applicable, low-cost enhancement with a useful uncertainty output, and the cross-domain results on medical images would increase its practical value. The paper deserves credit for reporting ablations that separate the GW loss from the refinement modules, for including a computational-complexity analysis, and for posting a code page. However, the theoretical grounding of the main performance gain is compromised by the Bayes Risk derivation, and the uncertainty quantification is not calibrated in a quantitative sense. The empirical gains on some baselines are small enough that missing seeds and implementation details are consequential.

major comments (4)
  1. [3.3 and A.5 (Eq. 5)] The Bayes Risk identity Risk*=trace(Sigma_P) is derived under the assumption that each predicted coordinate is centered at the ground-truth coordinate: Section 3.1 sets mu_i = {c_x, c_y, w, h}, and Appendix A.5 uses the conditional model p(x|c_x)=N(c_x, sigma^2). In the actual detector, the predictive distribution is centered at the model output \hat{R}, not at the ground truth R. For the paper's own Gaussian model, E[||\hat{R} - R||_2^2] = trace(Sigma_P) + ||mu_P - R||_2^2, so trace(Sigma_P) is not the Bayes Risk of the L2 regression loss whenever mu_P differs from R. Since the normalized risk vector T in Eq. (7) is the sole input to the reweighting operations in Eqs. (8), (10), and (11), and Table 3 attributes +0.8 AP to BRRM, the central performance claim currently rests on an unsupported identity. The derivation must be replaced with one that integrates over the actual predictive distribution (with mean \hat{R}) and a posterior over R, or the reweighting must be presented as an empirical heuristic backed by calibration evidence.
  2. [Algorithm 1 and Section 5.3] Algorithm 1 is underspecified: after dividing each of the four 95% confidence intervals into k bins, it says 'Form k bounding boxes R_i = (c_x_i, c_y_i, w_i, h_i)' without specifying how the bins are paired; if each interval is independently divided there are k^4 possible boxes, and choosing only the diagonal pairing is an unstated modeling decision that changes the output uncertainty. Moreover, the claimed reliability of the uncertainty estimate is supported only by the qualitative heatmap in Figure 6; no calibration curve, coverage rate, or correlation statistic is provided. Because the confidence intervals are built from the same sigma values whose meaning is called into question by the Bayes Risk issue, the uncertainty contribution needs either a quantitative calibration study or a clearly heuristic framing.
  3. [3.1 and 4.2] The manuscript never states how the per-query covariance Sigma_P is produced. Section 3.1 only says sigma_i is a learnable parameter with 0<sigma_i^2<=1, while Eq. (7) defines a per-query vector T and Eqs. (8) to (11) apply it per query. If the variances are not input-dependent outputs of a network head, the refinement modules cannot actually be query-adaptive; if they are, the architecture, initialization, and constraints need to be specified. This is a reproducibility gap for the central method.
  4. [Table 1] All experiments appear to be single runs with no seeds reported. For Relation-DETR the reported improvements are +0.2 AP and +0.1 AP, which are within the usual run-to-run variation of 12-epoch COCO training, so the claim that the method 'consistently improves' all evaluated DETR variants is not statistically supported. At minimum, two or three seeds with means and standard deviations should be reported for the main comparisons, or released checkpoints should allow independent verification.
minor comments (5)
  1. [4.2 (Eq. 11)] Equation (11) uses Risk* in the exponent of u but Risk*/4 in the exponent of s and in Eq. (7); this normalization inconsistency is likely a typo and should be corrected.
  2. [Appendix A.5] The derivation assumes each ground-truth coordinate has a U(0,1) prior without stating this assumption in Section 3.3; the prior is also dimensionally unusual for width and height coordinates and should be stated explicitly when the Bayes Risk is introduced.
  3. [References [11] and [12]] References [11] and [12] are the same paper (He et al., 2019) and should be merged.
  4. [Table 1 caption] The caption notes that Relation-DETR uses its default classification loss without the Bayes Risk modification, but Section 5.2 does not discuss the implication; the text should be explicit that for Relation-DETR the complete proposed method was not applied.
  5. [Figure 6] Figure 6 is described as a heatmap but no colorbar or axis labels are shown, making it hard to read the claimed inverse relationship.

Circularity Check

1 steps flagged · score 3.0 of 10

The 'Bayes Risk' identity (Eq. 5) is self-definitional: it assumes the predicted box mean equals the ground truth, so trace(Sigma_P) is an identity under the model's own assumption rather than a derived lower bound; the empirical AP gains remain externally benchmarked.

  1. self definitional [Section 3.1, Eq. 5 (Section 3.3), Appendix A.5]
    "We assume that each component of R̂ follows a 1D Gaussian distribution N_i(μ_i, σ_i^2), where i ={ĉ_x, ĉ_y, ŵ, ĥ}, μ_i ={c_x,c_y,w,h}, and 0 < σ_i^2 ≤ 1. ... Risk* = σ^2_ĉx + σ^2_ĉy + σ^2_ŵ + σ^2_ĥ. (5) ... which is just the trace of ΣP. ... Notice that for the predicted bounding box R̂ = (ĉ_x, ĉ_y, ŵ, ĥ), it follows a 4D Gaussian distribution N_P(μP, ΣP) with μP = [c_x, c_y, w, h]^T."

    Equation (5) is obtained by positing that the predicted box distribution's mean equals the ground-truth vector (μP = R). Under that assumption E||R̂−R||² = tr(ΣP) is an algebraic identity, so calling it the Bayes Risk / 'theoretical lower bound of the regression loss' is true only by construction. For an actual detector the expected squared error is tr(ΣP) + ||μP − R||², so the trace is not a lower bound on the model's real localization loss whenever the predicted mean is biased. The paper then uses this trace to reweight output embeddings, classification loss, and matching cost (Eqs. 8, 10, 11) and to build the confidence-interval uncertainty of Algorithm 1, with no calibration argument connecting tr(ΣP) to true localization error.

full rationale

The central empirical claim is benchmarked against external baselines (COCO, LISC, WBCDD) and does not reduce to a fitted constant, so the paper is not wholly circular. The GW distance is taken from the external reference [6], and no load-bearing self-citation chain or uniqueness theorem is invoked. The main circular/definitional issue is confined to the Bayes Risk derivation: the paper defines the predicted Gaussian's mean to be the ground-truth box, which makes Eq. (5) an identity rather than a derived lower bound; the subsequent risk-based module reweighting and uncertainty quantification inherit this unvalidated identification. This is a real self-definitional step in the theoretical justification, but it does not by itself force the measured AP improvements, which are obtained by direct comparison with prior detectors. A separate correctness concern, not itself circularity, is that the trace-based risk is not calibrated to actual localization error unless the predicted mean is unbiased; the paper supplies only a qualitative heatmap (Fig. 6) as validation.

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

The paper adds no new physical entities. Its theoretical contribution rests on modeling choices: a 2D Gaussian for ground truth, a diagonal 4D Gaussian for predictions, a uniform prior in the Bayes Risk proof, and a quoted GW formula. The learnable variances and hand-selected loss weights are free parameters that the central uncertainty claims depend on.

free parameters (5)
  • learnable variances sigma^2 for predicted bbox coordinates = not reported
    Learned during training; all Bayes Risk refinement and uncertainty estimates depend on them.
  • loss weights lambda_iou, lambda_L1, lambda_gw = not reported
    Hand-selected weights in Eq. (6); no sensitivity analysis is reported.
  • k, number of divisions in Algorithm 1 = 300
    Selected in A.7 as a trade-off between inference time and uncertainty estimate robustness.
  • top-5 count in Algorithm 1 = 5
    Averaging the top-5 IoUs is an arbitrary design choice with no derivation.
  • exponent 4 and 1+Risk* terms in matching cost = not reported
    Ad hoc design choices in Eq. (11); no derivation or sensitivity study.
assumptions (5)
  • domain assumption Ground-truth bbox can be represented as a 2D Gaussian via its inscribed ellipse.
    Section 3.1, Eq. (1) and Eq. (2): the Gaussian surface is back-projected from the inscribed ellipse.
  • domain assumption Predicted bbox components are independent 1D Gaussians whose means are the ground-truth coordinates, with learnable variances forming a diagonal 4D covariance.
    Section 3.1: the paper explicitly assumes independence and sets mu_i = {cx, cy, w, h}.
  • ad hoc to paper Each ground-truth bbox coordinate has an independent uniform U(0,1) prior for the Bayes Risk derivation.
    Appendix A.5 states this assumption; it is not justified from the detection setting and is not used elsewhere.
  • domain assumption The analytical Gromov-Wasserstein formula from Delon et al. applies to the 2D-vs-4D Gaussian pair with Euclidean ground cost.
    Appendix A.3 quotes the formula from [6] without deriving the hetero-dimensional extension or discussing mean dependence.
  • domain assumption L2 loss is the relevant bounding-box regression loss for computing Bayes Risk.
    Appendix A.5 chooses L2 loss and derives Risk* = trace(Sigma_P); the main loss in Eq. (6) is a different combination of IoU, L1, and GW terms.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Uncertainty-aware DETR Enhancement Framework for Object Detection." pith.science (2026). https://pith.science/paper/WNF6GEZE

@misc{pith2026250714855,
  author       = {Pith},
  title        = {Pith review of: An Uncertainty-aware DETR Enhancement Framework for Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WNF6GEZE}},
  note         = {Machine review of arXiv:2507.14855}
}
read the original abstract

This paper investigates the problem of object detection with a focus on improving both the localization accuracy of bounding boxes and explicitly modeling prediction uncertainty. Conventional detectors rely on deterministic bounding box regression, ignoring uncertainty in predictions and limiting model robustness. In this paper, we propose an uncertainty-aware enhancement framework for DETR-based object detectors. We model bounding boxes as multivariate Gaussian distributions and incorporate the Gromov-Wasserstein distance into the loss function to better align the predicted and ground-truth distributions. Building on this, we derive a Bayes Risk formulation to filter high-risk information and improve detection reliability. We also propose a simple algorithm to quantify localization uncertainty via confidence intervals. Experiments on the COCO benchmark show that our method can be effectively integrated into existing DETR variants, enhancing their performance. We further extend our framework to leukocyte detection tasks, achieving state-of-the-art results on the LISC and WBCDD datasets. These results confirm the scalability of our framework across both general and domain-specific detection tasks. Code page: https://github.com/ParadiseforAndaChen/An-Uncertainty-aware-DETR-Enhancement-Framework-for-Object-Detection.

Figures

Figures reproduced from arXiv: 2507.14855 by the authors.

Figure 1
Figure 1. The left side shows two examples (Ex. a, Ex. b) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Portion of the entire whole Gaussian surface that [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The diagram illustrates our approach to modeling bounding boxes and integrating it into DETR-based frameworks. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparison of the results after integrating our [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Heatmap between the Combined Metric and local [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Density distribution of predicted classification [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Density distribution of predicted classification [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: The density distribution of IoU before and after [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

43 extracted references · 24 canonical work pages

  1. [1]

    Zhi Cai, Songtao Liu, Guodong Wang, Zheng Ge, Xiangyu Zhang, and Di Huang

  2. [2]

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexan- der Kirillov, and Sergey Zagoruyko. 2020. End-to-end object detection with transformers. In European conference on computer vision . Springer, 213–229

    2020

  3. [3]

    Yifei Chen, Chenyan Zhang, Ben Chen, Yiyu Huang, Yifei Sun, Changmiao Wang, Xianjun Fu, Yuxing Dai, Feiwei Qin, Yong Peng, et al. 2024. Accurate leukocyte detection based on deformable-DETR and multi-level feature fusion for aiding diagnosis of blood diseases.Computers in biology and medicine 170 (2024), 107917

    work page 2024

  4. [4]

    Jiwoong Choi, Dayoung Chun, Hyun Kim, and Hyuk-Jae Lee. 2019. Gaussian yolov3: An accurate and fast object detector using localization uncertainty for autonomous driving. In Proceedings of the IEEE/CVF International conference on computer vision. 502–511

    work page 2019

  5. [5]

    Xiyang Dai, Yinpeng Chen, Jianwei Yang, Pengchuan Zhang, Lu Yuan, and Lei Zhang. 2021. Dynamic detr: End-to-end object detection with dynamic attention. In Proceedings of the IEEE/CVF international conference on computer vision . 2988– 2997

    work page 2021

  6. [6]

    Julie Delon, Agnes Desolneux, and Antoine Salmona. 2022. Gromov–Wasserstein distances between Gaussian distributions. Journal of Applied Probability 59, 4 (2022), 1178–1198

    work page 2022

  7. [7]

    Zhora Gevorgyan. 2022. SIoU loss: More powerful learning for bounding box regression. arXiv preprint arXiv:2205.12740 (2022)

    arXiv 2022

  8. [8]

    Yecai Guo and Mengyao Zhang. 2023. Blood cell detection method based on improved YOLOv5. IEEE Access 11 (2023), 67987–67995

    work page 2023

Show all 43 references
  1. [9]

    Ali Harakeh, Michael Smart, and Steven L Waslander. 2020. Bayesod: A bayesian approach for uncertainty estimation in deep object detectors. In 2020 IEEE Inter- national Conference on Robotics and Automation (ICRA) . IEEE, 87–93

    2020

  2. [10]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition . 770–778

    2016

  3. [12]

    Yihui He, Chenchen Zhu, Jianren Wang, Marios Savvides, and Xiangyu Zhang

  4. [13]

    Xiuquan Hou, Meiqin Liu, Senlin Zhang, Ping Wei, Badong Chen, and Xuguang Lan. 2025. Relation detr: Exploring explicit position relation prior for object detection. In European Conference on Computer Vision . Springer, 89–105

    2025

  5. [14]

    Ding Jia, Yuhui Yuan, Haodi He, Xiaopei Wu, Haojun Yu, Weihong Lin, Lei Sun, Chao Zhang, and Han Hu. 2023. Detrs with hybrid matching. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 19702–19712

    2023

  6. [15]

    In Proceedings of the ieee/cvf conference on computer vision and pattern recognition

    Bounding box regression with uncertainty for accurate object detection. In Proceedings of the ieee/cvf conference on computer vision and pattern recognition . 2888–2897

  7. [16]

    Youngwan Lee, Joong-won Hwang, Hyung-Il Kim, Kimin Yun, Yongjin Kwon, Yuseok Bae, and Sung Ju Hwang. 2022. Localization uncertainty estimation for anchor-free object detection. In European Conference on Computer Vision . Springer, 27–42

    2022

  8. [17]

    Xiang Li, Wenhai Wang, Xiaolin Hu, Jun Li, Jinhui Tang, and Jian Yang. 2021. Generalized focal loss v2: Learning reliable localization quality estimation for dense object detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 11632–11641

    2021

  9. [18]

    Balaji Lakshminarayanan, Alexander Pritzel, and Charles Blundell. 2017. Simple and scalable predictive uncertainty estimation using deep ensembles. Advances in neural information processing systems 30 (2017)

    2017

  10. [19]

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C Lawrence Zitnick. 2014. Microsoft coco: Common objects in context. In Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proce...

    2014

  11. [20]

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C Berg. 2016. Ssd: Single shot multibox detec- tor. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part...

    2016

  12. [21]

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. 2017. Focal loss for dense object detection. In Proceedings of the IEEE international conference on computer vision . 2980–2988

    2017

  13. [22]

    Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, and Jingdong Wang. 2021. Conditional detr for fast training convergence. In Proceedings of the IEEE/CVF international conference on computer vision . 3651– 3660

    2021

  14. [23]

    Yifan Pu, Weicong Liang, Yiduo Hao, Yuhui Yuan, Yukang Yang, Chao Zhang, Han Hu, and Gao Huang. 2024. Rank-DETR for high quality object detection. Advances in Neural Information Processing Systems 36 (2024)

    2024

  15. [24]

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. 2021. Swin transformer: Hierarchical vision transformer us- ing shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision. 10012–10022

    2021

  16. [25]

    J Redmon. 2016. You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition

    2016

  17. [26]

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. 2016. Faster R-CNN: To- wards real-time object detection with region proposal networks.IEEE transactions on pattern analysis and machine intelligence 39, 6 (2016), 1137–1149

    2016

  18. [27]

    Heqian Qiu, Hongliang Li, Qingbo Wu, and Hengcan Shi. 2020. Offset bin classification network for accurate object detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 13188–13197

    2020

  19. [28]

    Seyed Hamid Rezatofighi and Hamid Soltanian-Zadeh. 2011. Automatic recogni- tion of five types of white blood cells in peripheral blood. Computerized Medical Imaging and Graphics 35, 4 (2011), 333–343

    2011

  20. [29]

    Zhi Tian, Chunhua Shen, Hao Chen, and Tong He. 2020. FCOS: A simple and strong anchor-free object detector. IEEE transactions on pattern analysis and machine intelligence 44, 4 (2020), 1922–1933

    2020

  21. [30]

    Hamid Rezatofighi, Nathan Tsoi, JunYoung Gwak, Amir Sadeghian, Ian Reid, and Silvio Savarese. 2019. Generalized intersection over union: A metric and a loss for bounding box regression. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 658–666

    2019

  22. [31]

    Ao Wang, Hui Chen, Lihao Liu, Kai Chen, Zijia Lin, Jungong Han, and Guiguang Ding. 2024. Yolov10: Real-time end-to-end object detection. arXiv preprint arXiv:2405.14458 (2024)

    2024 arXiv

  23. [32]

    Chengcheng Wang, Wei He, Ying Nie, Jianyuan Guo, Chuanjian Liu, Yunhe Wang, and Kai Han. 2024. Gold-YOLO: Efficient object detector via gather-and-distribute mechanism. Advances in Neural Information Processing Systems 36 (2024)

    2024

  24. [33]

    A Vaswani. 2017. Attention is all you need. Advances in Neural Information Processing Systems (2017)

    2017

  25. [34]

    Chien-Yao Wang, I-Hau Yeh, and Hong-Yuan Mark Liao. 2025. Yolov9: Learning what you want to learn using programmable gradient information. In European conference on computer vision . Springer, 1–21

    2025

  26. [35]

    Jinwang Wang, Chang Xu, Wen Yang, and Lei Yu. 2021. A normalized Gaussian Wasserstein distance for tiny object detection. arXiv preprint arXiv:2110.13389 (2021)

    2021 arXiv

  27. [36]

    Chien-Yao Wang, Alexey Bochkovskiy, and Hong-Yuan Mark Liao. 2023. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. In Proceedings of the IEEE/CVF conference on computer vision and pattern recogni- tion. 7464–7475

    2023

  28. [37]

    Xue Yang, Junchi Yan, Qi Ming, Wentao Wang, Xiaopeng Zhang, and Qi Tian

  29. [38]

    Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel M Ni, and Heung-Yeung Shum. 2022. Dino: Detr with improved denoising anchor boxes for end-to-end object detection. arXiv preprint arXiv:2203.03605 (2022)

    2022 arXiv

  30. [39]

    Fanxin Xu, Xiangkui Li, Hang Yang, Yali Wang, and Wei Xiang. 2022. TE-YOLOF: Tiny and efficient YOLOF for blood cell detection. Biomedical Signal Processing and Control 73 (2022), 103416

    2022

  31. [40]

    Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, and Jifeng Dai. 2020. Deformable detr: Deformable transformers for end-to-end object detection.arXiv preprint arXiv:2010.04159 (2020). 9 Table 6: Effect of GIoU, Wasserstein distance and Gromov-Wasserstein distance. Here ...

    2020 arXiv

  32. [43]

    Zhaohui Zheng, Ping Wang, Wei Liu, Jinze Li, Rongguang Ye, and Dongwei Ren

  33. [2020]

    In Proceedings of the AAAI conference on artificial intelligence , Vol

    Distance-IoU loss: Faster and better learning for bounding box regression. In Proceedings of the AAAI conference on artificial intelligence , Vol. 34. 12993–13000

  34. [2021]

    In International conference on machine learning

    Rethinking rotated object detection with gaussian wasserstein distance loss. In International conference on machine learning . PMLR, 11830–11841

  35. [2023]

    arXiv preprint arXiv:2304.07527 (2023)

    Align-detr: Improving detr with simple iou-aware bce loss. arXiv preprint arXiv:2304.07527 (2023)

    2023 arXiv

Pith tools

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