Pith. sign in

REVIEW 4 major objections 6 minor 48 references

A New Teacher-Reviewer-Student Framework for Semi-supervised 2D Human Pose Estimation

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Reviewer networks that keep an exponential moving average of teacher and student weights, together with multi-level feature supervision and Keypoint-Mix augmentation, push semi-supervised 2D human pose estimation to a new state of the art.

desk verdict Solid incremental results for semi-supervised 2D HPE, but the algorithm as written omits the stop-gradient its own consistency losses need, and the core framework is a lightly-disclosed port of the authors' prior ECCV work. read the letter →

arxiv 2501.09565 v1 pith:UA3CCSBP submitted 2025-01-16 cs.CV

Wulian Yun , Mengshi Qi , Fei Peng , Huadong Ma This is my paper · ORCID
classification cs.CV
keywords semi-supervisedlearning2Dhumanposeestimationteacher-studentframeworkexponentialmovingaverageconsistencyregularizationdataaugmentationkeypointdetectionheatmap
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 argues that semi-supervised 2D human pose estimation can be advanced by giving the classic teacher-student setup a memory. It adds two reviewer networks that hold an exponential moving average of the teacher and student weights, so unlabeled-data supervision comes from both the current teacher and a smoothed version of its history. It also proposes supervising from the last two backbone stages instead of just the last, and a Keypoint-Mix augmentation that blends image patches around different keypoints. Experiments on COCO, MPII, and AI Challenger report state-of-the-art accuracy, including 50.9 AP on COCO with only 1K labeled images, up from 46.9 for the prior best method. The reader should care because the approach is a general recipe: any heatmap-based pose estimator can be plugged into the framework to reduce labeling cost.

What carries the argument

The central object is the Teacher-Reviewer-Student framework: two alternating teacher/student networks G and F plus two reviewer networks R1 and R2 updated by EMA ($\theta_{R1} = \alpha\theta_{R1} + (1-\alpha)\theta_G$, $\theta_{R2} = \beta\theta_{R2} + (1-\beta)\theta_F$). The reviewers supply historical parameter states as additional consistency targets in the unsupervised loss. Two supporting mechanisms carry the gains: Multi-level Feature Learning, which upsamples the last two backbone stages to estimate heatmaps and adds their losses, and Keypoint-Mix, an augmentation that blends patches around different predicted keypoints to create hard samples. The consistency loss uses the mapping $M_{e\to h}$ to align easy-augmented teacher and reviewer heatmaps with the hard-augmented student heatmap.

What would settle it

A direct test would be to run the method with Keypoint-Mix as the student's hard augmentation but compute the consistency loss only on keypoints that were not mixed; if AP stays the same or improves, the current loss is relying on misaligned regions. Alternatively, replace the reviewer networks with two frozen copies of the teacher from the previous epoch; if performance does not drop, the EMA-updated reviewers are not the cause of the gain.

Watch

Extended reading notes

Core claim

The central discovery is that retaining historical parameter information through reviewer networks, and using it as an extra consistency target for the student, materially improves semi-supervised 2D human pose estimation. In the proposed scheme, networks G and F alternate as teacher and student, and two reviewer networks R1 and R2 are updated from them by exponential moving average after every step. For unlabeled images, the student's prediction on a hard-augmented view is pulled toward both the teacher's and the reviewer's predictions on an easy-augmented view, after mapping them to a common coordinate space. The same supervision is enriched by estimating heatmaps from the last two backbone stages rather than only the deepest stage, and by Keypoint-Mix, which averages patches around randomly chosen keypoints and pastes the blend back, forcing the network to distinguish keypoints rather than relying on local appearance. The method reports gains over prior semi-supervised pose estimators on all tested benchmarks.

Load-bearing premise

The unsupervised loss assumes that the easy-augmented teacher and reviewer heatmaps can be meaningfully aligned to the hard-augmented student heatmap by the mapping $M_{e\to h}$; for Keypoint-Mix, which replaces keypoint regions with blended patches, no geometric transform is defined, so a misalignment would inject incorrect supervision into the student.

Editorial extensions

If this is right

  • Applying the reviewer mechanism to other teacher-student semi-supervised tasks, such as semantic segmentation or object detection, could yield similar gains from historical weight averaging.
  • The 50.9 AP at 1K labeled COCO images suggests that practical annotation budgets for pose estimation could be cut to a few thousand images without sacrificing accuracy.
  • Multi-level feature supervision can be added to any heatmap-based backbone with negligible overhead, since it only requires upsampling existing intermediate features.
  • Keypoint-Mix is a task-specific augmentation that may complement generic augmentations like Cutout and RandAugment in keypoint-focused models.
  • The alternating-role design allows lightweight and heavy models to co-train; the paper shows that a heavier teacher guiding a lighter student improves over a same-size pair.

Reading between the lines

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

  • An untested extension would be to replace the single EMA reviewer with a small ensemble of past checkpoints, which would test whether the gain comes specifically from exponentially decayed averaging or from access to multiple historical states.
  • Because Keypoint-Mix applies no geometric transform when blending patches, one could modify the consistency loss to ignore the heatmap locations of the mixed keypoints; if accuracy does not drop, the current loss is likely being hurt by misaligned supervision at those locations.
  • The reported gains use a moderate-capacity heatmap estimator; whether the same label-efficiency improvement holds with a stronger high-resolution backbone remains an open empirical question.
  • If the reviewer's benefit comes from smoothing over training history, then the framework might also improve stability in other consistency-based semi-supervised settings, but the paper only demonstrates this for 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

4 major / 6 minor

Summary. The paper proposes a Teacher-Reviewer-Student (TRS) framework for semi-supervised 2D human pose estimation. The method alternates two networks G and F between teacher and student roles, adds two reviewer networks R1 and R2 whose parameters are EMA averages of G and F, uses a Multi-level Feature Learning strategy that estimates heatmaps from the last two backbone stages, and introduces a Keypoint-Mix augmentation that blends image patches around detected keypoints. The unlabeled consistency losses in Eqs. (10)-(11) encourage the student prediction to match teacher and reviewer predictions under easy/hard augmentation. The paper reports state-of-the-art AP on COCO with 1K/5K/10K labeled images (e.g., 50.9 AP at 1K versus 46.9 for SSPCM), as well as PCKh@0.5 improvements on MPII and AI Challenger, with ablation studies over the proposed components.

Significance. If the reported gains hold, the method has practical value for low-annotation 2D pose estimation: the improvements over strong baselines such as Dual and SSPCM are consistent across datasets, backbones, and label budgets, and the ablations cover each proposed component. The paper also compares the Keypoint-Mix augmentation against several standard augmentations. However, the conceptual novelty is modest: the Teacher-Reviewer-Student design closely follows the authors' prior Teacher-Reference-Student architecture in [33], and the paper does not discuss what is new relative to that work beyond the pose-specific components. The reproducibility of the headline results is currently impaired by an inconsistent optimization statement in Algorithm 1, undefined terms in the central consistency losses, and missing hyperparameters. These issues are local rather than fatal, but they must be resolved before the empirical claims can be verified.

major comments (4)
  1. [Algorithm 1 / §IV-B] Algorithm 1, line 13 updates θG, θF, θR1, and θR2 with the total loss L of Eq. (12). Because L includes L1un and L2un, the unlabeled consistency losses backpropagate into the teacher/reviewer networks through the terms Me→h(˜S^{uW}_j) unless those targets are explicitly detached. For example, L1un = Σ_W (¯S^{uF}_{V,j} − Me→h(˜S^{uW}_j))^2 has gradients with respect to θG and θR1 as well as θF. Minimizing this bidirectional objective drives all four networks to produce identical outputs on unlabeled data, which is the collapse mechanism studied in [21]. The prose in §IV-B says that only the student network's parameters are updated by each unlabeled loss, which contradicts line 13. Please state whether stop-gradient/detach is applied to the teacher and reviewer targets, correct Algorithm 1 accordingly (e.g., update θF by L1un and θG by L2un, with all four networks updated only by the supervised term and EMA), and confirm that the reported numbers were obtained with that corrected procedure.
  2. [§IV-C, Eqs. (10)-(11)] The stage index is inconsistent in the unlabeled losses. In Eq. (10), the student terms are written with V ∈ {z,p} but the teacher and reviewer terms are written as Me→h(˜S^{uG}_j) and Me→h(˜S^{uR1}_j), without V. The same issue appears in Eq. (11). Since Multi-level Feature Learning is a core contribution and the consistency loss must compare heatmaps from the same backbone stage, the definitions are incomplete: the paper must define ˜S^{uW}_{V,j} for W ∈ {G,R1,F,R2} and include V in both arguments of Eqs. (10)-(11) (or explain that the teacher/reviewer prediction is the fusion of stages, which would still need a formula).
  3. [§IV-D, Eq. (5)] The mapping Me→h is never defined. The text says only that it maps predictions for easy and hard augmented data to the same coordinate space. For random rotation and scaling a geometric mapping can be derived, but Keypoint-Mix replaces patches around K sampled keypoints with blended patches from other keypoint locations; no coordinate transformation is specified that relates the teacher's heatmap on the easy image to the student's heatmap on the Keypoint-Mixed image. As written, Eq. (5) and Eq. (10) are not well-defined for the hard augmentation that the method introduces. Please specify Me→h per augmentation type, or state clearly that Keypoint-Mix is applied only to the student input and that the teacher/reviewer predictions are left in the original coordinate frame, and justify that choice.
  4. [§V-B] Several training hyperparameters needed to reproduce the state-of-the-art results are missing. The EMA momenta η in Eq. (1) and α, β in Eq. (7) are never given numeric values; the batch composition (sizes b and c in Algorithm 1) is not specified; and the number K of keypoints used for Keypoint-Mix is not stated in the implementation details (Table VIII gives an ablation but no default). The learning-rate schedule for MPII and AI Challenger is also incomplete. Please provide a complete hyperparameter table and training schedule.
minor comments (6)
  1. [Table II] The arrow notation such as '50.9↑4.0' is not explained; the caption should define it as the improvement over the previous best method.
  2. [References] Reference [25] is cited as PseudoPose in Tables II and Table III, but the bibliography entry for [25] points to 'Spatial temporal graph convolutional networks for skeleton-based action recognition'; the citation needs to be corrected.
  3. [Table VI] The row with MFL and KM enabled but the reviewer removed reports 44.8 AP at 1K, which is lower than the reviewer-only baseline of 45.7 AP. This appears inconsistent with the claim that each component independently contributes to performance, and it should be explained in the text.
  4. [Algorithm 1 output and §V-B] Algorithm 1 returns θG and θF, but the experimental section states that the final accuracy is the average of the two networks; please clarify how the reported AP/PCKh values are computed from the two models.
  5. [Table IX caption] The caption contains the typo 'Tabel IX'; it should be 'Table IX'.
  6. [§IV-D] Figure 3 says the teacher network generates keypoint coordinates {x1, y1}, ..., {xn, yn} and patches are extracted around them, but the text does not specify whether the teacher's predicted coordinates are used directly or whether a confidence threshold is applied before sampling; please specify.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the reported gains are empirical, benchmark-anchored, and ablated on held-out validation; the same-author citation [33] is not load-bearing.

full rationale

The paper's central claim is the state-of-the-art AP on COCO, MPII, and AI Challenger, measured against held-out validation annotations. The total loss in Eq. (12) combines a supervised MSE against external ground truth (Eq. (9)) with unlabeled consistency losses (Eqs. (10), (11)); these are standard teacher-student regularizers, not definitions of the reported accuracy. The reviewer networks are EMA copies of the teacher/student models (Eq. (7)), a standard technique from Mean Teacher [36], and their contribution is empirically isolated in Table VI (removing reviewers drops AP from 50.9 to 44.8 at 1K). Multi-level Feature Learning and Keypoint-Mix are likewise ablated against external AP numbers, and the choices of two feature stages and K=5 are hyperparameter selections on validation, which is not circular. The only same-author reference, [33], appears in a general list of semi-supervised learning works in Related Work and is not invoked as a proof of the present framework's validity. The skeptic's concern about missing stop-gradient in Algorithm 1 is a correctness/reproducibility issue, not a pattern of circularity: it concerns training dynamics and potential collapse, not a prediction that is equivalent to its input by definition. Thus, no circular step is exhibited, and the proper circularity score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method relies on several hyperparameters that are either not reported (α, β, η, batch size, total epochs) or tuned on the validation set (λ=0.5, K=5, two MFL stages). The consistency loss assumes that aligning teacher and student predictions under different augmentations is a valid training signal, and that EMA-averaged reviewer parameters provide complementary supervision. No new physical entities are introduced.

free parameters (6)
  • λ (total loss weight) = 0.5
    Set in Section V-B; trades supervised and unsupervised losses.
  • α (EMA momentum for R1) = not reported
    Used in Eq (7); not specified in experiments.
  • β (EMA momentum for R2) = not reported
    Used in Eq (7); not specified in experiments.
  • η (EMA momentum for teacher update) = not reported
    Used in Eq (1); not specified in experiments.
  • K (Keypoint-Mix number of keypoints) = 5
    In Table VIII, best on validation at K=5.
  • MFL stage count = 2 (last two stages)
    In Table VII, best on validation.
assumptions (4)
  • domain assumption Heatmap-based estimation with a SimpleBaseline backbone is an appropriate base model for semi-supervised HPE.
    The paper builds on SimpleBaseline [24] without questioning this choice.
  • domain assumption Consistency regularization between teacher and student predictions on differently augmented inputs improves learning from unlabeled data.
    Core assumption of the unsupervised loss in Eq (5), (6), (10), (11).
  • domain assumption EMA-averaged reviewer parameters provide complementary and useful supervision signals.
    The central premise of the Teacher-Reviewer-Student design in Section IV-B.
  • standard math MSE loss and EMA updates are used as defined.
    Standard operations; no proof given, consistent with common usage.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A New Teacher-Reviewer-Student Framework for Semi-supervised 2D Human Pose Estimation." pith.science (2026). https://pith.science/paper/UA3CCSBP

@misc{pith2026250109565,
  author       = {Pith},
  title        = {Pith review of: A New Teacher-Reviewer-Student Framework for Semi-supervised 2D Human Pose Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UA3CCSBP}},
  note         = {Machine review of arXiv:2501.09565}
}
read the original abstract

Conventional 2D human pose estimation methods typically require extensive labeled annotations, which are both labor-intensive and expensive. In contrast, semi-supervised 2D human pose estimation can alleviate the above problems by leveraging a large amount of unlabeled data along with a small portion of labeled data. Existing semi-supervised 2D human pose estimation methods update the network through backpropagation, ignoring crucial historical information from the previous training process. Therefore, we propose a novel semi-supervised 2D human pose estimation method by utilizing a newly designed Teacher-Reviewer-Student framework. Specifically, we first mimic the phenomenon that human beings constantly review previous knowledge for consolidation to design our framework, in which the teacher predicts results to guide the student's learning and the reviewer stores important historical parameters to provide additional supervision signals. Secondly, we introduce a Multi-level Feature Learning strategy, which utilizes the outputs from different stages of the backbone to estimate the heatmap to guide network training, enriching the supervisory information while effectively capturing keypoint relationships. Finally, we design a data augmentation strategy, i.e., Keypoint-Mix, to perturb pose information by mixing different keypoints, thus enhancing the network's ability to discern keypoints. Extensive experiments on publicly available datasets, demonstrate our method achieves significant improvements compared to the existing methods.

Figures

Figures reproduced from arXiv: 2501.09565 by the authors.

Figure 1
Figure 1. Illustrations of our proposed Teacher-Reviewer-Student framework for semi-supervised 2D HPE task. Unlike fully supervised methods that rely solely on labeled data for pose estimation, our semi-supervised method utilizes both labeled and unlabeled data to estimate human pose. Furthermore, we propose the reviewer network based on the teacher-student framework to provide additional supervisory signals. model’s predicti… view at source ↗
Figure 2
Figure 2. Overview of our framework. Our method comprises network [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Illustration of data augmentation strategy Keypoint-Mix. Unlabeled [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of our method and other semi-supervised 2D HPE methods Dual [21] and SSPCM [23] on COCO [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Heatmap visualization of two samples from COCO dataset. The [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Heatmap visualization of two samples on COCO dataset. Arranged [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 40 canonical work pages

  1. [21]

    An empirical study of the collapsing problem in semi-supervised 2d human pose estimation,

    R. Xie, C. Wang, W. Zeng, and Y . Wang, “An empirical study of the collapsing problem in semi-supervised 2d human pose estimation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision(ICCV), 2021, pp. 11 220–11 229

    work page 2021

  2. [33]

    Semi-supervised teacher-reference- student architecture for action quality assessment,

    W. Yun, M. Qi, F. Peng, and H. Ma, “Semi-supervised teacher-reference- student architecture for action quality assessment,” in European Confer- ence on Computer Vision . Springer, 2025, pp. 161–178

    work page 2025

  3. [1]

    Multi- context attention for human pose estimation,

    X. Chu, W. Yang, W. Ouyang, C. Ma, A. L. Yuille, and X. Wang, “Multi- context attention for human pose estimation,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2017, pp. 5669– 5678

    work page 2017

  4. [2]

    Deep high-resolution representation learning for visual recognition,

    J. Wang, K. Sun, T. Cheng, B. Jiang, C. Deng, Y . Zhao, D. Liu, Y . Mu, M. Tan, X. Wang, W. Liu, and B. Xiao, “Deep high-resolution representation learning for visual recognition,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 43, no. 10, pp. 3349– 3364, 2021

    2021

  5. [3]

    Simple baselines for human pose estimation and tracking,

    B. Xiao, H. Wu, and Y . Wei, “Simple baselines for human pose estimation and tracking,” in Computer Vision – ECCV 2018 , V . Ferrari, M. Hebert, C. Sminchisescu, and Y . Weiss, Eds. Cham: Springer International Publishing, 2018, pp. 472–487

    work page 2018

  6. [4]

    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 (CVPR), June 2023, pp. 2163–2172

    work page 2023

  7. [5]

    Distribution-aware coordinate representation for human pose estimation,

    F. Zhang, X. Zhu, H. Dai, M. Ye, and C. Zhu, “Distribution-aware coordinate representation for human pose estimation,” in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2020, pp. 7091–7100

    work page 2020

  8. [6]

    Skeleton-based action recognition through contrasting two-stream spatial-temporal networks,

    C. Pang, X. Lu, and L. Lyu, “Skeleton-based action recognition through contrasting two-stream spatial-temporal networks,” IEEE Transactions on Multimedia, vol. 25, pp. 8699–8711, 2023

    work page 2023

Show all 48 references
  1. [7]

    stagnet: An attentive semantic rnn for group activity recognition,

    M. Qi, J. Qin, A. Li, Y . Wang, J. Luo, and L. Van Gool, “stagnet: An attentive semantic rnn for group activity recognition,” in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 101–117

    2018

  2. [8]

    Pose- guided attention learning for cloth-changing person re-identification,

    X. Liu, K. Liu, J. Guo, P. Zhao, Y . Quan, and Q. Miao, “Pose- guided attention learning for cloth-changing person re-identification,” IEEE Transactions on Multimedia , vol. 26, pp. 5490–5498, 2024

    2024

  3. [9]

    Mutual distillation learning for person re-identification,

    H. Fu, K. Cui, C. Wang, M. Qi, and H. Ma, “Mutual distillation learning for person re-identification,” IEEE Transactions on Multimedia , 2024

    2024

  4. [10]

    Exploiting temporal correlations for 3d human pose estimation,

    R. Wang, X. Ying, and B. Xing, “Exploiting temporal correlations for 3d human pose estimation,” IEEE Transactions on Multimedia , vol. 26, pp. 4527–4539, 2024

    2024

  5. [11]

    Frame-padded multiscale transformer for monocular 3d human pose estimation,

    Y . Zhong, G. Yang, D. Zhong, X. Yang, and S. Wang, “Frame-padded multiscale transformer for monocular 3d human pose estimation,” IEEE Transactions on Multimedia , vol. 26, pp. 6191–6201, 2024

    2024

  6. [12]

    Deeppose: Human pose estimation via deep neural networks,

    A. Toshev and C. Szegedy, “Deeppose: Human pose estimation via deep neural networks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2014, pp. 1653–1660

    2014

  7. [13]

    Directpose: Direct end-to-end multi- person pose estimation,

    Z. Tian, H. Chen, and C. Shen, “Directpose: Direct end-to-end multi- person pose estimation,” arXiv preprint arXiv:1911.07451 , 2019

    1911 arXiv

  8. [14]

    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 Computer Vision – ECCV 2022 , S. Avidan, G. Brostow, M. Ciss ´e, G. M. Farinella, and T. Hassner, Eds. Cham: Spri...

    2022

  9. [15]

    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

    2021

  10. [16]

    Deep high-resolution representa- tion learning for human pose estimation,

    K. Sun, B. Xiao, D. Liu, and J. Wang, “Deep high-resolution representa- tion learning for human pose estimation,” in 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2019, pp. 5686– 5696

    2019

  11. [17]

    Pytorch: An imperative style, high-performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga et al. , “Pytorch: An imperative style, high-performance deep learning library,” Advances in neural information processing systems , vol. 32, pp. 8026–8037, 2019

    2019

  12. [18]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” Computer Science, vol. abs/1412.6980, p. 6, 2014

    2014 arXiv

  13. [19]

    2d human pose estimation: New benchmark and state of the art analysis,

    M. Andriluka, L. Pishchulin, P. Gehler, and B. Schiele, “2d human pose estimation: New benchmark and state of the art analysis,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, 2014, pp. 3686–3693

    2014

  14. [20]

    Microsoft coco: Common objects in con- text,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll´ar, and C. L. Zitnick, “Microsoft coco: Common objects in con- text,” in Computer Vision – ECCV 2014 , D. Fleet, T. Pajdla, B. Schiele, and T. Tuytelaars, Eds. Cham: Springer International Publishing, 2...

    2014

  15. [22]

    Large-scale datasets for going deeper in image understanding,

    J. Wu, H. Zheng, B. Zhao, Y . Li, B. Yan, R. Liang, W. Wang, S. Zhou, G. Lin, Y . Fu, Y . Wang, and Y . Wang, “Large-scale datasets for going deeper in image understanding,” in Proceedings of the IEEE International Conference on Multimedia and Expo (ICME) , 2019, pp. 1480–1485

    2019

  16. [23]

    Semi- supervised 2d human pose estimation driven by position inconsistency pseudo label correction module,

    L. Huang, Y . Li, H. Tian, Y . Yang, X. Li, W. Deng, and J. Ye, “Semi- supervised 2d human pose estimation driven by position inconsistency pseudo label correction module,” in Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR), June 2023...

    2023

  17. [24]

    Simple baselines for human pose estimation and tracking,

    B. Xiao, H. Wu, and Y . Wei, “Simple baselines for human pose estimation and tracking,” in Proceedings of the European Conference on Computer Vision (ECCV) , 2018, pp. 466–481

    2018

  18. [25]

    Spatial temporal graph convolutional networks for skeleton-based action recognition,

    S. Yan, Y . Xiong, and D. Lin, “Spatial temporal graph convolutional networks for skeleton-based action recognition,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 32, no. 1, 2018

    2018

  19. [26]

    Data distillation: Towards omni-supervised learning,

    I. Radosavovic, P. Doll ´ar, R. Girshick, G. Gkioxari, and K. He, “Data distillation: Towards omni-supervised learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 4119–4128

    2018

  20. [27]

    Stacked hourglass networks for human pose estimation,

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

    2016

  21. [28]

    Multi-scale structure-aware network for human pose estimation,

    L. Ke, M.-C. Chang, H. Qi, and S. Lyu, “Multi-scale structure-aware network for human pose estimation,” in Proceedings of the European Conference on Computer Vision (ECCV) , 2018, pp. 713–728

    2018

  22. [29]

    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 Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 5693–5703

    2019

  23. [30]

    The devil is in the details: Delving into unbiased data processing for human pose estimation,

    J. Huang, Z. Zhu, F. Guo, and G. Huang, “The devil is in the details: Delving into unbiased data processing for human pose estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 5700–5709

    2020

  24. [31]

    Human pose estimation with spatial contextual information,

    H. Zhang, H. Ouyang, S. Liu, X. Qi, X. Shen, R. Yang, and J. Jia, “Human pose estimation with spatial contextual information,” arXiv preprint arXiv:1901.01760, 2019

    1901 arXiv

  25. [32]

    Decenternet: Bottom-up human pose estimation via decentral- ized pose representation,

    T. Wang, L. Jin, Z. Wang, X. Fan, Y . Cheng, Y . Teng, J. Xing, and J. Zhao, “Decenternet: Bottom-up human pose estimation via decentral- ized pose representation,” in Proceedings of the 31st ACM International Conference on Multimedia , 2023, pp. 1798–1808

    2023

  26. [34]

    Ke-gan: Knowledge embedded generative adversarial networks for semi-supervised scene parsing,

    M. Qi, Y . Wang, J. Qin, and A. Li, “Ke-gan: Knowledge embedded generative adversarial networks for semi-supervised scene parsing,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 5237–5246

    2019

  27. [35]

    Weakly-supervised temporal action localization by inferring salient snippet-feature,

    W. Yun, M. Qi, C. Wang, and H. Ma, “Weakly-supervised temporal action localization by inferring salient snippet-feature,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 7, 2024, pp. 6908–6916

    2024

  28. [36]

    Mean teachers are better role mod- els: Weight-averaged consistency targets improve semi-supervised deep learning results,

    A. Tarvainen and H. Valpola, “Mean teachers are better role mod- els: Weight-averaged consistency targets improve semi-supervised deep learning results,” in Proceedings of the 31st International Conference on Neural Information Processing Systems , 2017, p. 1195–1204

    2017

  29. [37]

    Fixmatch: Simplifying semi- supervised learning with consistency and confidence,

    K. Sohn, D. Berthelot, N. Carlini, Z. Zhang, H. Zhang, C. A. Raffel, E. D. Cubuk, A. Kurakin, and C.-L. Li, “Fixmatch: Simplifying semi- supervised learning with consistency and confidence,” in Advances in Neural Information Processing Systems , vol. 33, 2020, pp. 596–608

    2020

  30. [38]

    Self-training with noisy student improves imagenet classification,

    Q. Xie, M.-T. Luong, E. Hovy, and Q. V . Le, “Self-training with noisy student improves imagenet classification,” in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2020, pp. 10 684– 10 695

    2020

  31. [39]

    Unsupervised data augmentation for consistency training,

    Q. Xie, Z. Dai, E. Hovy, M.-T. Luong, and Q. V . Le, “Unsupervised data augmentation for consistency training,” in Proceedings of the 34th International Conference on Neural Information Processing Systems , 2020

    2020

  32. [40]

    Unbiased teacher v2: Semi-supervised object detection for anchor-free and anchor-based detectors,

    Y .-C. Liu, C.-Y . Ma, and Z. Kira, “Unbiased teacher v2: Semi-supervised object detection for anchor-free and anchor-based detectors,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 9819–9828

    2022

  33. [41]

    Bootstrap your own latent-a new approach to self-supervised learning,

    J.-B. Grill, F. Strub, F. Altch ´e, C. Tallec, P. Richemond, E. Buchatskaya, C. Doersch, B. Avila Pires, Z. Guo, M. Gheshlaghi Azar et al., “Bootstrap your own latent-a new approach to self-supervised learning,” Advances in neural information processing systems, vol. 33, pp. 2...

    2020

  34. [42]

    Momentum contrast for unsupervised visual representation learning,

    K. He, H. Fan, Y . Wu, S. Xie, and R. Girshick, “Momentum contrast for unsupervised visual representation learning,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 9729–9738

    2020

  35. [43]

    2d human pose estimation: New benchmark and state of the art analysis,

    M. Andriluka, L. Pishchulin, P. Gehler, and B. Schiele, “2d human pose estimation: New benchmark and state of the art analysis,” in Proceedings of the IEEE Conference on computer Vision and Pattern Recognition , 2014, pp. 3686–3693

    2014

  36. [44]

    Improved regularization of convolutional neural networks with cutout,

    T. DeVries and G. W. Taylor, “Improved regularization of convolutional neural networks with cutout,” arXiv preprint arXiv:1708.04552 , 2017

    2017 arXiv

  37. [45]

    Semi-supervised semantic segmentation with mutual knowledge distillation,

    J. Yuan, J. Ge, Z. Wang, and Y . Liu, “Semi-supervised semantic segmentation with mutual knowledge distillation,” in Proceedings of the 31st ACM International Conference on Multimedia, 2023, pp. 5436–5444

    2023

  38. [46]

    mixup: Beyond empirical risk minimization,

    H. Zhang, M. Cisse, Y . N. Dauphin, and D. Lopez-Paz, “mixup: Beyond empirical risk minimization,” arXiv preprint arXiv:1710.09412 , 2017

    2017 arXiv

  39. [47]

    Cutmix: Reg- ularization strategy to train strong classifiers with localizable features,

    S. Yun, D. Han, S. J. Oh, S. Chun, J. Choe, and Y . Yoo, “Cutmix: Reg- ularization strategy to train strong classifiers with localizable features,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 6023–6032

    2019

  40. [48]

    Randaugment: Practical automated data augmentation with a reduced search space,

    E. D. Cubuk, B. Zoph, J. Shlens, and Q. V . Le, “Randaugment: Practical automated data augmentation with a reduced search space,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, 2020, pp. 702–703

    2020

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.