Pith. sign in

REVIEW 2 major objections 6 minor 64 references

A Coarse-to-Fine Human Pose Estimation Method based on Two-stage Distillation and Progressive Graph Neural Network

T0 review · 2 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A two-stage distillation pipeline lets a lightweight HRNet-W32 pose estimator match or beat much heavier models on COCO and CrowdPose by adding skeleton-aware losses and a progressive graph refinement network.

desk verdict A plausible but incremental combination of known distillation and GCN components, with a clean COCO gain, but the CrowdPose claim is compromised by ablations on the test set. read the letter →

arxiv 2508.11212 v1 pith:EOI6NAPB submitted 2025-08-15 cs.CV

Zhangjian Ji , Wenjin Zhang , Shaotong Qiao , Kai Feng , Yuhua Qian This is my paper
classification cs.CV
keywords humanposeestimationknowledgedistillationgraphconvolutionalnetworkskeleton-awarelossSimCCCrowdoccludedjointslightweightmodel
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 tries to show that a small pose-estimation model can keep most of the accuracy of a large teacher if the knowledge transfer happens in two stages: first mimic the teacher's features and joint positions while also enforcing limb-length structure, then let a graph network refine the coarse pose using image context. The authors build this around SimCC, a coordinate-classification pose model, with an HRNet-W48 teacher and an HRNet-W32 student. They report 76.2% AP on COCO val2017 and 68.7% AP on CrowdPose, improvements of 0.9 and 2.0 points over the same-backbone SimCC baseline. The larger gain on CrowdPose is presented as evidence that structural and contextual information helps most for occluded joints in crowded scenes. The practical point is that accurate pose estimation does not necessarily require a heavy model if the distillation itself encodes body topology.

What carries the argument

The load-bearing mechanism is the two-stage distillation objective combined with the IGP-GCN. The first stage uses a pose structure loss $L_{pos} = \frac{1}{N}\sum_n |P^{tea}_n - P^{stu}_n| + L_{cst}$, where $L_{cst}$ penalizes weighted differences in limb lengths between teacher and student skeletons, plus a feature distillation MSE loss, with a time-decay factor $\gamma(t) = 1 - (t-1)/T$ that gradually reduces the distillation penalty. The second stage is the Image-Guided Progressive Graph Convolutional Network: joints are nodes of an undirected body graph, and three residual GCN blocks consume the initial pose along with joint features extracted from low-to-high resolution maps, each bloc

What would settle it

Use CrowdPose's per-joint visibility labels and compare, for occluded joints only, the refined predictions against ground truth. If the student plus IGP-GCN is not more accurate on occluded joints than the teacher's own predictions, or if training the IGP-GCN with ground-truth targets instead of teacher targets produces a large gain, then the claimed mechanism of using image context and pose structure to recover hidden joints is not what drives the reported improvement.

Watch

Extended reading notes

Core claim

The central claim is that a coarse-to-fine, two-stage distillation scheme transfers pose knowledge more completely than standard feature or logit distillation alone. In the first stage, the student is trained with a feature-mimicking MSE loss, an L1 pose loss against the teacher, and a skeleton-aware loss on limb lengths, all scaled by a time-decaying weight that lets the student gradually shift from imitating the teacher to learning on its own. In the second stage, the frozen student's initial pose is fed into an Image-Guided Progressive Graph Convolutional Network (IGP-GCN), which refines joint positions using features from three resolution levels and is supervised progressively by the tea

Load-bearing premise

The argument assumes that the teacher model's predicted joint positions, including invisible ones, are trustworthy enough to supervise both the student and the graph refiner; if the teacher is systematically wrong on occluded joints, the whole pipeline learns those errors instead of true joint locations.

Editorial extensions

If this is right

  • A lightweight HRNet-W32 student with 31M parameters and 8.1 GFLOPs can reach 76.2% AP on COCO val2017, outperforming the same-backbone SimCC baseline by 0.9 AP.
  • On CrowdPose, the method gains 2.0 AP over SimCC, with larger margins on harder metrics such as AP75 and AR, suggesting occlusion-heavy scenes benefit most from structure-aware distillation.
  • The two-stage design means the first-stage student is already a complete pose estimator, so the IGP-GCN refiner can be added later to improve its outputs without retraining the backbone.
  • The ablation study shows skeleton-aware distillation contributes more than plain L1 pose distillation, supporting the claim that explicit limb-length constraints carry useful structural knowledge.
  • The reported CrowdPose result (68.7 AP) exceeds several heavier models, including HrHRNet (65.9 AP) with a larger input size, indicating the approach can shift accuracy gains from model size to training signal.

Reading between the lines

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

  • Because the IGP-GCN is supervised entirely by the teacher's final output, the reported occlusion gains may partly reflect the student learning the teacher's systematic errors on invisible joints; an oracle-supervision variant that uses ground-truth joints for occluded cases would separate learned structure from inherited bias.
  • The refiner is trained after the backbone is frozen, so the same IGP-GCN could likely be attached to other lightweight pose estimators as a model-agnostic plug-in, without retraining them from scratch.
  • The first-stage weight-decay schedule effectively turns distillation into a curriculum (imitate the teacher early, then rely on the task loss), a principle that could transfer to other coordinate-classification tasks beyond human pose estimation.
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

2 major / 6 minor

Summary. The paper proposes a coarse-to-fine, two-stage knowledge distillation framework for lightweight top-down 2D human pose estimation. A SimCC-based student with an HRNet-W32 backbone is trained in the first stage under an HRNet-W48 teacher using feature-distillation MSE, a pose ℓ1 loss, an anatomical edge-constraint loss, and a time-decayed weighting schedule. In the second stage, the student is frozen and an Image-Guided Progressive Graph Convolutional Network (IGP-GCN) refines the initial pose; the GCN's three progressive outputs are trained with ℓ1 losses against the teacher's final pose, masked by ground-truth joint visibility. Experiments are reported on COCO val2017, COCO test-dev2017, and CrowdPose test. The method reaches 76.2% AP on COCO val and 68.7% AP on CrowdPose, exceeding the same-backbone SimCC baseline by 0.9 and 2.0 points respectively.

Significance. If the reported results are reliable, the paper demonstrates a useful lightweight pose estimator (HRNet-W32, 31M params) whose accuracy approaches much heavier models, and it provides a clear loss-level decomposition of the proposed distillation gains. The loss formulations in Eqs. (1)-(7) are comprehensible, and the same-backbone comparisons against SimCC in Tables I and III are appropriate. The visualizations give qualitative support to the claim that the method helps in crowded and occluded scenes. However, the CrowdPose evidence is weakened by a serious evaluation-protocol flaw: ablations and hyperparameter selection are performed on the CrowdPose test set. In addition, the second-stage supervision target is exclusively the teacher's output, and the paper does not examine whether that teacher is reliable precisely on the occluded joints that the method claims to improve. These issues make the main empirical claim not fully supported in its current form.

major comments (2)
  1. [Section IV-A and Table IV] The manuscript explicitly states that results are reported on the CrowdPose-test set for ablation studies, and Table IV uses that set to select among six schemes and to tune α, β, δ, λ, ξ and the γ(t) schedule. This is test-set model selection. Because the reported 2.0% AP gain over the SimCC baseline on CrowdPose is the paper's strongest claim, the improvement may be inflated by selection bias and does not by itself demonstrate generalization. Please rerun the ablations and hyperparameter selection on the CrowdPose validation split, freeze the choices there, and report CrowdPose-test results only for the final configuration.
  2. [Section III-B.2, Eq. (6)] The IGP-GCN is trained solely toward the teacher's final output P^tea, with a ground-truth visibility mask M. The paper's stated motivation is to locate invisible/occluded joints in crowded scenes, but it provides no analysis of the teacher's accuracy on those joints. If the HRNet-W48 teacher makes systematic errors on occluded joints, those errors are directly propagated through Eq. (6) to the student and the GCN, so the CrowdPose gain could partly reflect learning the teacher's mistakes. Please report teacher accuracy by occlusion level (or per-joint AP for occluded joints), and consider comparing the current teacher-only target with ground-truth supervision or a teacher/ground-truth blend to verify the supervision choice.
minor comments (6)
  1. [Throughout] There are numerous typos and grammatical errors: "fucus" (Introduction), "seconde" (Section III-B), "Inspirited" (Section III-B), "krypoints" (Eq. 3 context), "Ap score" (Table IV paragraph), and "PHE model" in Section IV-C. A careful proofreading pass is needed.
  2. [Section IV-A] Reference [56] is cited for the MS COCO keypoint dataset but [56] is the ResNet paper. Please cite the original COCO dataset paper (Lin et al., ECCV 2014) and re-check all dataset references.
  3. [Section II-A] The text says "Papandreou et al. [32] presented" but reference [32] is the EfficientHRNet paper. The correct reference for Papandreou et al. is likely [30]. Please fix the citation mapping.
  4. [Section I] The organization paragraph states that Section III introduces "the human anatomical keypoints constraint model and how to plug it into the existing bottom-up and top-down methods," but Section III actually presents the two-stage distillation framework and does not describe plugging into bottom-up methods. Update the description.
  5. [Section IV-A] The sentence defining AP50 as "the recall scores at OKS = 0.50" is inaccurate; AP50 is average precision at OKS=0.5, while AR is the recall metric. Please correct the terminology.
  6. [Eq. (3)] The edge weights ω_{i,j} are defined only by reference to the authors' prior work [18]. Since [18] is an arXiv preprint and the weight definition affects L_cst, please provide the explicit definition in the paper or an appendix to make the method self-contained.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: distillation is trained toward a pre-trained teacher and measured against ground-truth AP; the sole self-citation ([18] skeleton loss) is component-level and empirically ablated, not a load-bearing self-referential derivation.

full rationale

The paper's chain is empirical, not derivational: a pre-trained SimCC HRNet-W48 teacher supervises a W32 student via feature and pose-structure losses (Eqs. 1-5), then a frozen student feeds an IGP-GCN refined under teacher supervision (Eqs. 6-7), and the reported numbers are OKS/AP against ground-truth labels. No equation defines the reported gain in terms of itself: the teacher is an external model ([19]) trained on the same benchmark but not on the reported test labels, and the student's final outputs are never substituted for ground truth in evaluation. The only notable self-citation is [18] for the skeleton constraint loss in Eq. (3); that loss is specified in the paper (up to the edge weights taken from [18]) and its contribution is tested by the scheme-3 vs scheme-4 ablation rather than assumed from the citation, so it is not load-bearing circularity. Section IV-A's statement 'We report the results on the CrowdPose-test set for ablation studies' is a genuine validation-protocol weakness -- selecting hyperparameters and ablations on the test set risks overfitting and inflating the 2.0% CrowdPose gain -- but this is a statistical-validity concern, not a circularity: the evaluation still compares against held-out labels. The use of OPEC-Net's IGP-GCN [20] is also externally attributed. Overall, the central claim has independent empirical content; score reflects only the minor self-citation, not circular reasoning.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard distillation and graph-convolution assumptions. The free parameters are the loss weights and schedule, all chosen by hand or borrowed from prior work. No new entities such as particles, forces, or latent factors are introduced. The most fragile axiom is the reliability of teacher targets on occluded joints, which is never validated.

free parameters (7)
  • alpha (feature distillation weight) = 5e-5
    Set by hand in Section IV-B to balance feature MSE loss against the KL loss.
  • beta (pose structure loss weight) = 0.1
    Set by hand in Section IV-B for the Lcst term.
  • weight-decay schedule gamma(t) = 1 - (t-1)/T with T=210
    Borrowed from TADF [55]; the decay rate is a design choice that affects how strongly the student is tied to the teacher.
  • delta (LGCN1 weight) = 0.3
    Set by hand in Section IV-B for the first progressive GCN output.
  • lambda (LGCN2 weight) = 0.5
    Set by hand in Section IV-B for the second progressive GCN output.
  • xi (LGCN3 weight) = 1.0
    Set by hand in Section IV-B for the third progressive GCN output.
  • edge weights omega_ij in Lcst = defined as in [18]
    The structure loss uses per-edge weights taken from the authors' earlier paper; they are not reported here and may be fixed or learned.
assumptions (5)
  • domain assumption The fixed human skeleton graph G=(V,E) with standard limb adjacency captures the joint structure needed for pose refinement.
    The paper builds the IGP-GCN on this graph (Section III-B.1) and the structure loss is defined over its edges; if the topology is wrong, the transferred structural information is misleading.
  • domain assumption The teacher model (SimCC with HRNet-W48) provides reliable and useful soft targets for occluded joints.
    Both distillation stages supervise the student using teacher outputs (Eq. 5 and Eq. 6), but the paper never evaluates teacher accuracy on occluded joints.
  • domain assumption The IGP-GCN module from OPEC-Net [20] works as described when plugged into this distillation pipeline.
    The paper refers to [20] for the GCN design and does not specify or ablate the module's internal graph convolution layers, attention blocks, or image feature fusion.
  • standard math Binary mask M indicates which joints have ground-truth labels and only those joints are used in the second-stage loss.
    Eq. 6 uses a mask to ignore unlabeled joints, which is standard practice for pose estimation on COCO and CrowdPose.
  • domain assumption The COCO keypoint and CrowdPose benchmarks are appropriate measures of pose estimation quality, and the standard OKS-based AP reflects real-world usefulness.
    The paper reports only these benchmarks and does not test on other datasets, so the claimed generality depends on this assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Coarse-to-Fine Human Pose Estimation Method based on Two-stage Distillation and Progressive Graph Neural Network." pith.science (2026). https://pith.science/paper/EOI6NAPB

@misc{pith2026250811212,
  author       = {Pith},
  title        = {Pith review of: A Coarse-to-Fine Human Pose Estimation Method based on Two-stage Distillation and Progressive Graph Neural Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EOI6NAPB}},
  note         = {Machine review of arXiv:2508.11212}
}
read the original abstract

Human pose estimation has been widely applied in the human-centric understanding and generation, but most existing state-of-the-art human pose estimation methods require heavy computational resources for accurate predictions. In order to obtain an accurate, robust yet lightweight human pose estimator, one feasible way is to transfer pose knowledge from a powerful teacher model to a less-parameterized student model by knowledge distillation. However, the traditional knowledge distillation framework does not fully explore the contextual information among human joints. Thus, in this paper, we propose a novel coarse-to-fine two-stage knowledge distillation framework for human pose estimation. In the first-stage distillation, we introduce the human joints structure loss to mine the structural information among human joints so as to transfer high-level semantic knowledge from the teacher model to the student model. In the second-stage distillation, we utilize an Image-Guided Progressive Graph Convolutional Network (IGP-GCN) to refine the initial human pose obtained from the first-stage distillation and supervise the training of the IGP-GCN in the progressive way by the final output pose of teacher model. The extensive experiments on the benchmark dataset: COCO keypoint and CrowdPose datasets, show that our proposed method performs favorably against lots of the existing state-of-the-art human pose estimation methods, especially for the more complex CrowdPose dataset, the performance improvement of our model is more significant.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

64 extracted references · 59 canonical work pages

  1. [1]

    2d/3d pose estimation and action recognition using multitask deep learning,

    D. C. Luvizon, D. Picard, and H. Tabia, “2d/3d pose estimation and action recognition using multitask deep learning,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 5137–5146. 1

    work page 2018

  2. [2]

    Grab: A dataset of whole-body human grasping of objects,

    O. Taheri, N. Ghorbani, M. J. Black, and D. Tzionas, “Grab: A dataset of whole-body human grasping of objects,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16 . Springer, 2020, pp. 581–600. 1, 2

    work page 2020

  3. [3]

    Generating holistic 3d human motion from speech,

    H. Yi, H. Liang, Y . Liu, Q. Cao, Y . Wen, T. Bolkart, D. Tao, and M. J. Black, “Generating holistic 3d human motion from speech,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 469–480. 1

    work page 2023

  4. [4]

    Deep high-resolution repre- sentation learning for human pose estimation,

    K. Sun, B. Xiao, D. Liu, and J. Wang, “Deep high-resolution repre- sentation learning for human pose estimation,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2019, pp. 5693–

    work page 2019

  5. [5]

    Higherhrnet: Scale-aware representation learning for bottom-up human pose estimation,

    B. Cheng, B. Xiao, J. Wang, H. Shi, T. S. Huang, and L. Zhang, “Higherhrnet: Scale-aware representation learning for bottom-up human pose estimation,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 5385–5394. 1, 2, 5, 8

    work page 2020

  6. [6]

    Vitpose: Simple vision transformer baselines for human pose estimation,

    Y . Xu, J. Zhang, Q. Zhang, and D. Tao, “Vitpose: Simple vision transformer baselines for human pose estimation,” Advances in Neural Information Processing Systems , vol. 35, pp. 38 571–38 584, 2022. 1, 3, 7

    work page 2022

  7. [7]

    Lite- hrnet: A lightweight high-resolution network,

    C. Yu, B. Xiao, C. Gao, L. Yuan, L. Zhang, N. Sang, and J. Wang, “Lite- hrnet: A lightweight high-resolution network,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 10 440– 10 450. 1

    work page 2021

  8. [8]

    Dite-hrnet: Dynamic lightweight high-resolution network for human pose estimation,

    Q. Li, Z. Zhang, F. Xiao, F. Zhang, and B. Bhanu, “Dite-hrnet: Dynamic lightweight high-resolution network for human pose estimation,” in Pro- ceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI, 2022, pp. 1095–1101. 1

    work page 2022

Show all 64 references
  1. [9]

    Multiposenet: Fast multi-person pose estimation using pose residual network,

    M. Kocabas, S. Karagoz, and E. Akbas, “Multiposenet: Fast multi-person pose estimation using pose residual network,” in Proceedings of the European conference on computer vision (ECCV) , 2018, pp. 417–433. 1

    2018

  2. [10]

    Better teacher better student: Dynamic prior knowledge for knowledge distillation,

    M. Zong, Z. Qiu, X. Ma, K. Yang, C. Liu, J. Hou, S. Yi, and W. Ouyang, “Better teacher better student: Dynamic prior knowledge for knowledge distillation,” in The Eleventh International Conference on Learning Representations, 2022. 1

    2022

  3. [11]

    Damo- yolo: A report on real-time object detection design,

    X. Xu, Y . Jiang, W. Chen, Y . Huang, Y . Zhang, and X. Sun, “Damo- yolo: A report on real-time object detection design,” arXiv preprint arXiv:2211.15444, 2022. 1

    2022 arXiv

  4. [12]

    Transkd: Transformer knowledge distillation for efficient seman- tic segmentation,

    R. Liu, K. Yang, A. Roitberg, J. Zhang, K. Peng, H. Liu, and R. Stiefel- hagen, “Transkd: Transformer knowledge distillation for efficient seman- tic segmentation,” arXiv preprint arXiv:2202.13393 , 2022. 1

    2022 arXiv

  5. [13]

    Distilpose: Tokenized pose regression with heatmap distillation,

    S. Ye, Y . Zhang, J. Hu, L. Cao, S. Zhang, L. Shen, J. Wang, S. Ding, and R. Ji, “Distilpose: Tokenized pose regression with heatmap distillation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 2163–2172. 1, 3, 6, 7

    2023

  6. [14]

    Online knowledge dis- tillation for efficient pose estimation,

    Z. Li, J. Ye, M. Song, Y . Huang, and Z. Pan, “Online knowledge dis- tillation for efficient pose estimation,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 11 740–11 750. 1, 3, 6, 7

    2021

  7. [15]

    Effective whole-body pose estimation with two-stages distillation,

    Z. Yang, A. Zeng, C. Yuan, and Y . Li, “Effective whole-body pose estimation with two-stages distillation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 4210–4220. 1, 3

    2023

  8. [16]

    Dope: Distillation of part experts for whole-body 3d pose estimation in the wild,

    P. Weinzaepfel, R. Br ´egier, H. Combaluzier, V . Leroy, and G. Rogez, “Dope: Distillation of part experts for whole-body 3d pose estimation in the wild,” in Computer Vision–ECCV 2020: 16th European Confer- ence, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXVI 16 . Spr...

    2020

  9. [17]

    Structure-aware human pose estimation with graph convolutional net- works,

    Y . Bin, Z.-M. Chen, X.-S. Wei, X. Chen, C. Gao, and N. Sang, “Structure-aware human pose estimation with graph convolutional net- works,” Pattern Recognition, vol. 106, p. 107410, 2020. 1, 3

    2020

  10. [18]

    2d human pose estimation with explicit anatomical keypoints structure constraints,

    Z. Ji, Z. Wang, M. Zhang, Y . Chen, and Y . Qian, “2d human pose estimation with explicit anatomical keypoints structure constraints,” arXiv preprint arXiv:2212.02163 , 2022. 1, 4

    2022 arXiv

  11. [19]

    Simcc: A simple coordinate classification perspective for human pose estimation,

    Y . Li, S. Yang, P. Liu, S. Zhang, Y . Wang, Z. Wang, W. Yang, and S.-T. Xia, “Simcc: A simple coordinate classification perspective for human pose estimation,” in European Conference on Computer Vision . Springer, 2022, pp. 89–106. 1, 3, 6, 7, 8

    2022

  12. [20]

    Peeking into occluded joints: A novel framework for crowd pose estimation,

    L. Qiu, X. Zhang, Y . Li, G. Li, X. Wu, Z. Xiong, X. Han, and S. Cui, “Peeking into occluded joints: A novel framework for crowd pose estimation,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XIX

    2020

  13. [21]

    Springer, 2020, pp. 488–504. 2, 3, 5, 7

    2020

  14. [22]

    Auto-deeplab: Hierarchical neural architecture search for semantic image segmentation,

    C. Liu, L. Chen, F. Schroff, H. Adam, W. Hua, A. L. Yuille, and L. Fei-Fei, “Auto-deeplab: Hierarchical neural architecture search for semantic image segmentation,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2019, pp. 82–92. 2

    2019

  15. [23]

    Realtime multi-person 2d pose estimation using part affinity fields,

    Z. Cao, T. Simon, S. Wei, and Y . Sheikh, “Realtime multi-person 2d pose estimation using part affinity fields,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2017, pp. 1302–1310. 2

    2017

  16. [24]

    Associative embedding: End-to- end learning for joint detection and grouping,

    A. Newell, Z. Huang, and J. Deng, “Associative embedding: End-to- end learning for joint detection and grouping,” in Advances in Neural Information Processing Systems (NIPS) , 2017, pp. 2277–2287. 2

    2017

  17. [25]

    Stacked hourglass networks for human pose estimation,

    A. Newell, K. Yang, and J. Deng, “Stacked hourglass networks for human pose estimation,” in 14th European Conference on Computer Vision (ECCV), 2016, pp. 483–499. 2

    2016

  18. [26]

    Bottom-up pose estimation of mul- tiple person with bounding box constraint,

    M. Li, Z. Zhou, J. Li, and X. Liu, “Bottom-up pose estimation of mul- tiple person with bounding box constraint,” in 2018 24th international conference on pattern recognition (ICPR) . IEEE, 2018, pp. 115–120. 2

    2018

  19. [27]

    Deepcut: Joint subset partition and labeling for multi person pose estimation,

    L. Pishchulin, E. Insafutdinov, S. Tang, B. Andres, M. Andriluka, P. V . Gehler, and B. Schiele, “Deepcut: Joint subset partition and labeling for multi person pose estimation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , June 2016. 2

    2016

  20. [28]

    Deep network for the integrated 3d sensing of multiple people in natural images,

    A. Zanfir, E. Marinoiu, M. Zanfir, A.-I. Popa, and C. Sminchisescu, “Deep network for the integrated 3d sensing of multiple people in natural images,” Advances in neural information processing systems , vol. 31,

  21. [29]

    Mask R-CNN,

    K. He, G. Gkioxari, P. Doll ´ar, and R. B. Girshick, “Mask R-CNN,” in IEEE International Conference on Computer Vision (ICCV) , 2017, pp. 2980–2988. 2, 8

    2017

  22. [30]

    Rmpe: Regional multi-person pose estimation,

    H.-S. Fang, S. Xie, Y .-W. Tai, and C. Lu, “Rmpe: Regional multi-person pose estimation,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 2334–2343. 2, 7, 8

    2017

  23. [31]

    Towards accurate multi-person pose estimation in the JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, NOVEMBER 2022 10 wild,

    G. Papandreou, T. Zhu, N. Kanazawa, A. Toshev, J. Tompson, C. Bregler, and K. Murphy, “Towards accurate multi-person pose estimation in the JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, NOVEMBER 2022 10 wild,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR...

    2022

  24. [32]

    Simple baselines for human pose estimation and tracking,

    B. Xiao, H. Wu, and Y . Wei, “Simple baselines for human pose estimation and tracking,” in 15th European Conference on Computer Vision(ECCV), V . Ferrari, M. Hebert, C. Sminchisescu, and Y . Weiss, Eds., 2018, pp. 472–487. 2, 6, 7, 8

    2018

  25. [33]

    Effi- cienthrnet,

    C. Neff, A. Sheth, S. Furgurson, J. Middleton, and H. Tabkhi, “Effi- cienthrnet,” J. Real Time Image Process., vol. 18, no. 4, pp. 1037–1049,

  26. [34]

    Faster R-CNN: towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. B. Girshick, and J. Sun, “Faster R-CNN: towards real-time object detection with region proposal networks,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 39, no. 6, pp. 1137–1149, 2017. 2

    2017

  27. [35]

    Cascaded pyramid network for multi-person pose estimation,

    Y . Chen, Z. Wang, Y . Peng, Z. Zhang, G. Yu, and J. Sun, “Cascaded pyramid network for multi-person pose estimation,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2018, pp. 7103–

    2018

  28. [36]

    Crowdpose: Efficient crowded scenes pose estimation and a new benchmark,

    J. Li, C. Wang, H. Zhu, Y . Mao, H.-S. Fang, and C. Lu, “Crowdpose: Efficient crowded scenes pose estimation and a new benchmark,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 10 863–10 872. 2, 6, 8

    2019

  29. [37]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv preprint arXiv:1503.02531 , 2015. 3

    2015 arXiv

  30. [38]

    Learning efficient object detection models with knowledge distillation,

    G. Chen, W. Choi, X. Yu, T. Han, and M. Chandraker, “Learning efficient object detection models with knowledge distillation,” Advances in neural information processing systems , vol. 30, 2017. 3

    2017

  31. [39]

    Masked generative distillation,

    Z. Yang, Z. Li, M. Shao, D. Shi, Z. Yuan, and C. Yuan, “Masked generative distillation,” in European Conference on Computer Vision . Springer, 2022, pp. 53–69. 3

    2022

  32. [40]

    Bevdistill: Cross-modal bev distillation for multi-view 3d object detection,

    Z. Chen, Z. Li, S. Zhang, L. Fang, Q. Jiang, and F. Zhao, “Bevdistill: Cross-modal bev distillation for multi-view 3d object detection,” in The Eleventh International Conference on Learning Representations , 2022. 3

    2022

  33. [41]

    Knowledge adaptation for efficient semantic segmentation,

    T. He, C. Shen, Z. Tian, D. Gong, C. Sun, and Y . Yan, “Knowledge adaptation for efficient semantic segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2019, pp. 578–587. 3

    2019

  34. [42]

    Structured knowledge distillation for semantic segmentation,

    Y . Liu, K. Chen, C. Liu, Z. Qin, Z. Luo, and J. Wang, “Structured knowledge distillation for semantic segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 2604–2613. 3

    2019

  35. [43]

    On distillation of guided diffusion models,

    C. Meng, R. Rombach, R. Gao, D. Kingma, S. Ermon, J. Ho, and T. Sal- imans, “On distillation of guided diffusion models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 14 297–14 306. 3

    2023

  36. [44]

    Dynamic kernel distillation for efficient pose estimation in videos,

    X. Nie, Y . Li, L. Luo, N. Zhang, and J. Feng, “Dynamic kernel distillation for efficient pose estimation in videos,” in 2019 IEEE/CVF International Conference on Computer Vision (ICCV) , 2019, pp. 6941–

    2019

  37. [45]

    Fast human pose estimation,

    F. Zhang, X. Zhu, and M. Ye, “Fast human pose estimation,” in Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 3517–3526. 3, 6, 7

    2019

  38. [46]

    Integral knowledge distillation for multi-person pose estimation,

    X. Xu, Q. Zou, X. Lin, Y . Huang, and Y . Tian, “Integral knowledge distillation for multi-person pose estimation,” IEEE Signal Processing Letters, vol. 27, pp. 436–440, 2020. 3

    2020

  39. [47]

    Sdpose: Tokenized pose estimation via circulation- guide self-distillation,

    S. Chen, Y . Zhang, S. Huang, R. Yi, K. Fan, R. Zhang, P. Chen, J. Wang, S. Ding, and L. Ma, “Sdpose: Tokenized pose estimation via circulation- guide self-distillation,” arXiv preprint arXiv:2404.03518 , 2024. 3, 6, 7

    2024 arXiv

  40. [48]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” in International Conference on Learning Rep- resentations, 2016. 3

    2016

  41. [49]

    Link prediction based on graph neural net- works,

    M. Zhang and Y . Chen, “Link prediction based on graph neural net- works,” Advances in neural information processing systems , vol. 31,

  42. [50]

    Dgcn: Dynamic graph convolutional network for efficient multi-person pose estimation,

    Z. Qiu, K. Qiu, J. Fu, and D. Fu, “Dgcn: Dynamic graph convolutional network for efficient multi-person pose estimation,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 34, no. 07, 2020, pp. 11 924–11 931. 3

    2020

  43. [51]

    Robust pose estimation in crowded scenes with direct pose-level inference,

    D. Wang, S. Zhang, and G. Hua, “Robust pose estimation in crowded scenes with direct pose-level inference,” Advances in Neural Information Processing Systems, vol. 34, pp. 6278–6289, 2021. 3

    2021

  44. [52]

    Hierarchical graph neural network for human pose estimation,

    G. Zheng, Z. Zhao, Z. Zhang, and Y . Yang, “Hierarchical graph neural network for human pose estimation,” in 2023 IEEE International Conference on Multimedia and Expo (ICME) . IEEE, 2023, pp. 2663–

    2023

  45. [53]

    Objects as points,

    X. Zhou, D. Wang, and P. Kr ¨ahenb¨uhl, “Objects as points,” CoRR, vol. abs/1904.07850, 2019. 3

    1904 arXiv

  46. [54]

    Bottom-up human pose estimation via disentangled keypoint regression,

    Z. Geng, K. Sun, B. Xiao, Z. Zhang, and J. Wang, “Bottom-up human pose estimation via disentangled keypoint regression,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 14 676–14 686. 3, 8

    2021

  47. [55]

    Point-set anchors for object detection, instance segmentation and pose estimation,

    F. Wei, X. Sun, H. Li, J. Wang, and S. Lin, “Point-set anchors for object detection, instance segmentation and pose estimation,” in 16th European Conference on Computer Vision (ECCV) , 2020, pp. 527–544. 3

    2020

  48. [56]

    Distilling object detectors with task adaptive regularization,

    R. Sun, F. Tang, X. Zhang, H. Xiong, and Q. Tian, “Distilling object detectors with task adaptive regularization,” arXiv preprint arXiv:2006.13108, 2020. 4

    2006 arXiv

  49. [57]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770–778. 5

    2016

  50. [58]

    Yolov3: An incremental improvement,

    J. Redmon and A. Farhadi, “Yolov3: An incremental improvement,” arXiv preprint arXiv:1804.02767 , 2018. 6

    2018 arXiv

  51. [59]

    Distribution-aware coor- dinate representation for human pose estimation,

    F. Zhang, X. Zhu, H. Dai, M. Ye, and C. Zhu, “Distribution-aware coor- dinate representation for human pose estimation,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2020, pp. 7091–

    2020

  52. [60]

    Transpose: Keypoint localiza- tion via transformer,

    S. Yang, Z. Quan, M. Nie, and W. Yang, “Transpose: Keypoint localiza- tion via transformer,” in 2021 IEEE/CVF International Conference on Computer Vision (ICCV) , 2021, pp. 11 782–11 792. 7

    2021

  53. [61]

    Tf- pose: Direct human pose estimation with transformers,

    W. Mao, Y . Ge, C. Shen, Z. Tian, X. Wang, and Z. Wang, “Tf- pose: Direct human pose estimation with transformers,” arXiv preprint arXiv:2103.15320, 2021. 7

    2021 arXiv

  54. [62]

    Poseur: Direct human pose regression with transformers,

    W. Mao, Y . Ge, C. Shen, Z. Tian, X. Wang, Z. Wang, and A. v. den Hengel, “Poseur: Direct human pose regression with transformers,” in Computer Vision – ECCV 2022 . Cham: Springer Nature Switzerland, 2022, pp. 72–88. 7

    2022

  55. [63]

    Human pose regression with residual log-likelihood estimation,

    J. Li, S. Bian, A. Zeng, C. Wang, B. Pang, W. Liu, and C. Lu, “Human pose regression with residual log-likelihood estimation,” in 2021 IEEE/CVF International Conference on Computer Vision (ICCV) , 2021, pp. 11 005–11 014. 7 Zhangjian Ji received the B.S. degree from Wuhan Univ...

    2021

  56. [2022]

    degree in engineering from Shanxi University, China, in 2025

    received the M.E. degree in engineering from Shanxi University, China, in 2025. His research interests include computer vision, human-computer interaction, etc. Shaotong Qiao received the B.S. degree in engi- neering from Shandong Jianzhu University, China, in 2020. He is curr...

    2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.