Pith. sign in

REVIEW 4 major objections 5 minor 52 references

3D Human Pose Estimation via Spatial Graph Order Attention and Temporal Body Aware Transformer

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read The paper claims a graph-order attention GCN plus a central-frame-aware transformer reaches 27.0 mm MPJPE on Human3.6m and 24.7 mm on MPI-INF-3DHP, showing that dynamic per-joint neighborhood weighting and central-position-biased temporal…

desk verdict Solid benchmark numbers and a clean ablation, but the paper's own table-text contradictions on drop rates and the CPN average make the headline SOTA claims hard to verify as written. read the letter →

arxiv 2505.01003 v1 pith:LFKBP2Z6 submitted 2025-05-02 cs.CV

classification cs.CV
keywords 3Dhumanposeestimationgraphconvolutionalnetworkorderattentiontransformertemporalmodelingself-attentionMPJPEskeletonsequence
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 2D-to-3D human pose lifting improves when the spatial model can choose, per joint, which graph neighborhood radius is most informative, and when the temporal model tracks each joint's motion separately while focusing on the frame being predicted. It proposes a Multi-Order GCN with a Graph Order Attention (GOA) module that learns per-joint weights over graph orders, followed by a Body Aware Transformer with Joints Weighted Attention (JWA) and a Body-Centred Multi-Head Attention (BCMA) that applies a learnable exponential centrality mask. If the claim holds, static graph topologies and uniform temporal attention are bottlenecks, and these two inductive biases are worth adopting. The reported results on three benchmarks support the claim: 27.0 mm MPJPE on Human3.6m with ground-truth 2D input, 21.2 mm P-MPJPE under Protocol #2, 24.7 mm on MPI-INF-3DHP, and 10.4 mm on HumanEva-I.

What carries the argument

The load-bearing machinery is a pair of attention modules. Graph Order Attention (GOA) learns per-joint attention weights over multiple graph orders, with weights O_tt = softmax(tanh(Q+K)W_o), then computes a weighted sum of order features; this lets the model use a different effective neighborhood radius for different joints. Body-Centred Multi-Head Attention (BCMA) adds a learnable exponential centrality mask to the self-attention logits, P_scl = exp(-F_pos * W_F * (Pos - 0.5)^2), so attention concentrates on the central sequence frame and decays toward the edges. A third module, Joints Weighted Attention (JWA), applies a per-joint softmax over frames to capture each joint's local temporal evolution before the global attention.

What would settle it

Train the same architecture on Human3.6m with 324 frames and ground-truth 2D input, but replace BCMA with vanilla multi-head attention while keeping JWA and GOA fixed; the paper's 9-frame ablation shows a 38.5 vs 35.8 mm difference, so a full-protocol replication where the gap nearly vanishes would contradict the claim that central-frame awareness drives the improvement.

Watch

Extended reading notes

Core claim

The central claim is that representing each skeleton by several graph orders and letting an attention module select the most representative order for each joint, then modeling the temporal sequence with per-joint attention and central-frame-aware self-attention, yields state-of-the-art or competitive 3D pose accuracy. The GOA module computes, for each joint, a softmax over the R+1 graph orders of tanh(Q+K)W_o and sums the order features by those weights, replacing both the static highest-order graph and the simple concatenation of all orders. The BCMA scales vanilla self-attention logits by a learnable exponential decay centered at the central frame, P_scl = exp(-F_pos * W_F * (Pos - 0.5)^2), so attention peaks on the frame whose 3D pose is being predicted. The paper argues, and supports by ablations, that each component contributes: removing GOA raises MPJPE from 35.8 to 37.0 on a 9-frame setting, removing JWA raises it to 38.6, and replacing BCMA with vanilla multi-head attention raises it to 38.5.

Load-bearing premise

The method assumes that the fixed central-frame prior (P_c = 0.5 with a learnable exponential decay) and the per-dataset tuned input lengths and frame-drop rates are the right inductive biases for temporal pose lifting; if these are overspecified to the benchmarks, the reported gains, especially the 24.7 mm MPI-INF-3DHP result, may not transfer to new sequences or settings.

Editorial extensions

If this is right

  • If the claim is correct, static graph topologies and uniform temporal attention are suboptimal inductive biases for pose lifting; learned per-joint neighborhood radii and central-frame-aware attention are better defaults.
  • The reported 24.7 mm MPJPE on MPI-INF-3DHP suggests that a temporal many-to-one model with long input sequences and controlled frame dropping can beat 3D-reconstruction methods on mid-size datasets.
  • The architecture keeps its parameter count stable with respect to input length (14.45M parameters at both 27 and 324 frames), which is a practical advantage for deploying on variable-length video.
  • The central-frame exponential mask could apply to any sequence-to-one prediction task where the output aligns with a specific temporal position, not only pose estimation.

Reading between the lines

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

  • The GOA principle is not obviously limited to skeletons: any graph regression problem where the informative neighborhood radius varies per node could benefit from the same learned order weighting, e.g., hand pose, mesh, or scene graph estimation.
  • The BCMA mask is a soft prior rather than a hard constraint; replacing the fixed P_c=0.5 with a learnable center or a per-dataset tuned center could make the approach adapt to outputs not exactly at the sequence middle.
  • The very large MPI-INF-3DHP margin over prior temporal methods may be partly due to the specific input length (81) and frame-drop schedule rather than the attention modules alone; a testable extension is to apply the same schedule to PoseFormerV2 to isolate the architectural contribution.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes a 2D-to-3D human pose estimation method combining a Multi-Order GCN with a learnable Graph Order Attention (GOA) module for spatial modeling, and a temporal Body Aware Transformer (BAT) that integrates Joints Weighted Attention (JWA) and Body-Centred Multi-Head Attention (BCMA), where BCMA adds a learnable central-frame bias. The method is evaluated on Human3.6m, MPI-INF-3DHP, and HumanEva-I, reporting competitive or state-of-the-art MPJPE/P-MPJPE numbers, with an ablation study (Table 5) showing that each proposed module contributes to the final accuracy. The main claims are that GOA dynamically selects the most informative graph order per joint, JWA captures per-joint temporal evolution, and BCMA focuses attention on the central frame aligned with the 3D output.

Significance. If the reported results are correct and reproducible, the method is a solid contribution: it combines GCN and Transformer strengths with two simple, well-motivated attention mechanisms, and it outperforms several recent temporal methods on standard benchmarks. The paper provides an explicit ablation study (Table 5) and a parameter/frame-length analysis (Table 7), and the code is promised publicly, which would support verification. The central-frame inductive bias is a reasonable design choice for many-to-one temporal pose estimation, not a circular argument. However, the current manuscript contains internal inconsistencies in experimental settings and an underspecified equation for JWA, which currently prevent the reader from reproducing the claimed MPI-INF-3DHP state-of-the-art result.

major comments (4)
  1. [§4.2.2 vs §4.4.3 / Table 6] The frame-dropping rate for MPI-INF-3DHP is stated inconsistently: §4.2.2 says the model uses 81 input frames with a drop rate of 2, while §4.4.3 and Table 6 state a drop rate of 3. Similarly, for Human3.6m, §4.2.2 says a drop rate of 3, but Table 6 says 4. Since the dropping rate determines which central frame is predicted, this directly affects the reported MPJPE values, including the headline 24.7 mm on MPI-INF-3DHP in Table 3. The discrepancy must be resolved with an unambiguous statement of the exact preprocessing for each dataset, and the reported numbers should be re-verified under that stated configuration.
  2. [§4.3 / Table 1] The text claims on the CPN protocol that the method achieves 'an average of 43.0, outperforming MHFormer' , but Table 1 lists the Ours row as 43.5, whereas MHFormer is 43.0 and GLA-GCN is 44.4. The number 43.0 in the text appears to be MHFormer's average, not the proposed method's. This is a direct numerical inconsistency in a headline comparison and must be corrected; the claim of outperformance on the CPN row is not supported by the table as printed.
  3. [§3.2.1, Eq. (8)] The JWA operation is not dimensionally well-defined as written. Eq. (8) defines Jtt = softmax( ̃S W_J ), with ̃S ∈ ℝ^{T×J×D} and W_J ∈ ℝ^{T×J}. The product ̃S W_J is not a valid matrix multiplication for these shapes, and the subsequent element-wise multiplication in Eq. (10) is also ambiguous because Jtt (T×J) and ̃S (T×J×D) have different ranks. The authors should specify the intended reshaping or the actual parameter dimensions (e.g., W_J ∈ ℝ^D or a per-joint linear map), or provide a PyTorch-style shape trace of the JWA module.
  4. [§3.2.2 / §4.2.2] The paper states that the output 3D pose aligns with the central 2D pose in the input sequence, but the input lengths used include even values (T=324 for Human3.6m). For an even-length sequence there is no exact central frame; the evaluation must select one of the two middle frames, and this selection is not specified. This matters because a one-frame shift can change MPJPE by several millimeters and interacts with the frame-dropping rate. The authors should define exactly how the central frame index is computed for both even and odd T, and how this is applied during training and evaluation.
minor comments (5)
  1. [§4.4.4 / Table 7] The sentence 'the number of frames does not impact computational complexity' is imprecise: while the number of parameters is independent of T, FLOPs and memory usage scale with the sequence length in the Transformer encoder. Please clarify that the stability refers to parameter count, not to overall computational cost.
  2. [§4.4.3] The paragraph on Human3.6m ends with a stray fragment 'rate.' that should be removed.
  3. [§4.1.1 / Table 2] Table 2 is labeled 'Protocol #2 (MPJPE)' in the caption, but Protocol #2 is P-MPJPE (procrustes-aligned MPJPE) as defined in §4.1.1; the table heading should be corrected for consistency.
  4. [§3.2.2, Eq. (14)] In Eq. (14), the multiplication of the T×T attention matrix QK^T by the vector P_scl is undefined in terms of broadcasting; the authors should specify whether P_scl is applied as a row/column mask or as a diagonal scaling, with the exact broadcasting rule.
  5. [§4.3] The sentence 'it shows second second-best performance on Discussion and Waiting' contains a duplicated 'second'; please edit for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the method is an empirical architecture composed of external GCN/Transformer building blocks, evaluated against external benchmarks; reported inconsistencies are reproducibility issues, not circular reasoning.

full rationale

The paper's central derivation is an architecture and an empirical evaluation, not a closed-form derivation of a prediction from its own inputs. The proposed modules (MO-GCN + GOA, JWA, BCMA) are defined on the input skeleton sequence and learned end-to-end; none of the reported quantities (MPJPE on Human3.6m, MPI-INF-3DHP, HumanEva-I) is defined in terms of another reported quantity, nor is any parameter fitted to a subset of data and then relabeled as a prediction of a closely related quantity. The BCMA central bias (P_c=0.5 with a learnable decay) is an explicit inductive bias justified by the task convention that the output 3D pose aligns with the central frame; that is a design prior, not a circular reduction. The method builds on external prior work (Zou et al. [52] for multi-order GCN, Zheng et al. [50] for the transformer backbone, Pavllo et al. [29] for training and loss), and the authors do not invoke any self-citation as a load-bearing premise. The benchmark comparisons are against externally published methods, making the SOTA claims externally falsifiable rather than self-referential. There are internal inconsistencies (the text in §4.3 says an average of 43.0 while Table 1 reads 43.5; §4.2.2 and Table 6 give different frame-drop rates for Human3.6m and MPI-INF-3DHP) and reproducibility concerns around per-dataset hyperparameter tuning and exact preprocessing, but these are correctness and verification issues, not circularity. Therefore the circularity score is 0.

Assumptions & free parameters 7 free parameters · 4 assumptions · 3 invented entities

The central claim depends on standard deep learning assumptions (dataset labels, CPN detections), the task setup (central-frame output), and a set of hand-tuned hyperparameters (graph order R, input length, drop rate, learning rate, batch size, number of transformer iterations). No new physical entities are introduced; the new modules are internal and have no independent falsifiable handles.

free parameters (7)
  • Graph order count R = 3
    Chosen by ablation on HumanEva-I (Fig. 5); peak performance at R=3.
  • Feature dimension D = 32
    Chosen by hand to balance parameters and feature diversity (Section 4.2.1).
  • Input sequence length T = 324 (Human3.6m), 81 (MPI-INF-3DHP), 5 (HumanEva-I)
    Tuned per dataset; Table 6 reports the best accuracy for these values.
  • Frame dropping rate = 3 or 4 (Human3.6m), 2 or 3 (MPI-INF-3DHP), 0 (HumanEva-I)
    Used to mitigate overfitting; text and Table 6 disagree on the exact values.
  • Learning rate = 0.0003
    Determined by trial-and-error (Section 4.2.1).
  • Batch size = 16
    Smaller batch size found to mitigate overfitting (Section 4.2.2).
  • Number of BAT iterations = 6 (Human3.6m), 8 (MPI-INF-3DHP and HumanEva-I)
    Chosen by experiments; affects parameter count and accuracy (Table 7).
assumptions (4)
  • domain assumption The skeleton can be represented as a graph with joints as nodes and bones as edges, and high-order adjacency matrices encode k-hop connections.
    Used throughout Section 3.1; follows the graph definition in [52].
  • domain assumption The output 3D pose is aligned with the central frame of the input 2D pose sequence.
    Task setup following [17,49]; motivates the BCMA central bias.
  • domain assumption CPN 2D detections and ground truth 2D poses are reliable inputs for evaluation.
    Used in Section 4.1.1; standard protocol in the field.
  • standard math Self-attention (Vaswani et al.) and GCN (Kipf and Welling) are effective building blocks for sequence and graph modeling.
    Base architectures adopted from the literature without proof.
invented entities (3)
  • Graph Order Attention (GOA) module
    purpose: Dynamically weights features from different graph orders for each joint
    New module proposed; only evaluated on benchmarks, no independent falsifiable prediction outside the paper.
  • Joints Weighted Attention (JWA)
    purpose: Applies frame-wise attention per joint across time
    New module proposed; internal component without separate verification.
  • Body-Centred Multi-Head Attention (BCMA)
    purpose: Adds central-frame positional bias to self-attention
    New module proposed; only tested as part of the full architecture.

how reviews work

0 comments
Cite this review

Pith. "Pith review of 3D Human Pose Estimation via Spatial Graph Order Attention and Temporal Body Aware Transformer." pith.science (2026). https://pith.science/paper/LFKBP2Z6

@misc{pith2026250501003,
  author       = {Pith},
  title        = {Pith review of: 3D Human Pose Estimation via Spatial Graph Order Attention and Temporal Body Aware Transformer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LFKBP2Z6}},
  note         = {Machine review of arXiv:2505.01003}
}
read the original abstract

Nowadays, Transformers and Graph Convolutional Networks (GCNs) are the prevailing techniques for 3D human pose estimation. However, Transformer-based methods either ignore the spatial neighborhood relationships between the joints when used for skeleton representations or disregard the local temporal patterns of the local joint movements in skeleton sequence modeling, while GCN-based methods often neglect the need for pose-specific representations. To address these problems, we propose a new method that exploits the graph modeling capability of GCN to represent each skeleton with multiple graphs of different orders, incorporated with a newly introduced Graph Order Attention module that dynamically emphasizes the most representative orders for each joint. The resulting spatial features of the sequence are further processed using a proposed temporal Body Aware Transformer that models the global body feature dependencies in the sequence with awareness of the local inter-skeleton feature dependencies of joints. Given that our 3D pose output aligns with the central 2D pose in the sequence, we improve the self-attention mechanism to be aware of the central pose while diminishing its focus gradually towards the first and the last poses. Extensive experiments on Human3.6m, MPIINF-3DHP, and HumanEva-I datasets demonstrate the effectiveness of the proposed method. Code and models are made available on Github.

Figures

Figures reproduced from arXiv: 2505.01003 by the authors.

Figure 1
Figure 1. Framework of the proposed method. Multiple-order GCN (MO-GCN) generates features of various graph orders for each skeleton. The graph Order Attention (GOA) module assigns weights to various orders for each joint within each pose. The temporal Body Aware Transformer (BAT) captures local attention among body joints using Joints Weighted Attention (JWA) and incorporates global-centered attention using Body-Centred Mult… view at source ↗
Figure 2
Figure 2. Visual illustration of the proposed Graph Order Attention module. For each skeleton in the input sequence, each joint is represented with features of the highest attention weight of the same frame along different orders. Big joints represent high attention weights. 3.1. Spatial GCN 3.1.1. Multiple-Order GCN (MO-GCN) Let the 2D skeleton sequence 𝑆 = {𝑆1 , , ..., 𝑆𝑇 } gen￾erated by off-the-shelf 2D detection model, wh… view at source ↗
Figure 3
Figure 3. Illustration of the distance calculation between the central frame and the left/right poses in the sequence. 𝑃 𝑜𝑠 = [𝑃1 , … , 𝑃𝑐 ,… , 𝑃𝑇 ], 𝐷𝑖𝑠 = [𝐷𝑖𝑠1 , …, 𝐷𝑖𝑠𝑐 , …, 𝐷𝑖𝑠𝑇 ], and 𝐹𝑝𝑜𝑠 ∗ W𝐹 correspond to those defined in equations 12 and 13. respectively, where: 𝑄 = 𝑂𝑟𝑑𝑆𝑖 W𝑞 (3) 𝐾 = 𝑂𝑟𝑑𝑆𝑖 W𝑘 We define a learnable order weighted attention 𝑂𝑡𝑡 as: 𝑂𝑡𝑡 = 𝑠𝑜𝑓 𝑡𝑚𝑎𝑥(tanh(𝑄 + 𝐾) W𝑜 ) (4) Where 𝑂𝑡𝑡 ∈ ℝ𝐽×(𝑅+1) represents the … view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Detailed structure of GOA, JWA, and BCMA. T: Number of frames, J: Number of joints, R: Number of the highest order, D: Features dimension. Pscl: Scaling position vector in Equation 13. The colors match the components in [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Ablation study on the effect of the number of orders R on the performance using 9 frames of 2D ground truth (GT) joints from HumanEva-I dataset on 𝑃 𝑟𝑜𝑡𝑜𝑐𝑜𝑙#1 for 8xBAT. 𝑅 = 0 to 𝑅 = 3 the number of connections between joints increases from 15 to 68, and the performanc…
Figure 6
Figure 6. Figure 6: Qualitative comparison Poseformer [50] and the ground truth for subjects S9 and S11 on 6 actions of Human3.6m dataset using 2D ground truth (GT) joints as input. order, which enhances the efficacy of spatial skeleton repre￾sentation. This allows the model to capture co…
Figure 7
Figure 7. Figure 7: Qualitative comparison with MHFormer [17] in the wild. Results obtained from the model trained on the Human3.6m indoor dataset [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 9
Figure 9. Figure 9: Failure pose estimation frames in case of fast movement with blurring. Acknowledgement Aouaidjia Kamel and Chongsheng Zhang are supported in part by the National Natural Science Foundation of China (No.62250410371) and the Henan Provincial Key R&D Project (No.232102211…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 49 canonical work pages

  1. [1]

    2272–2281

    Cai,Y.,Ge,L.,Liu,J.,Cai,J.,Cham,T.J.,Yuan,J.,Thalmann,N.M., 2019.Exploitingspatial-temporalrelationshipsfor3dposeestimation via graph convolutional networks, in: Proceedings of the IEEE/CVF international conference on computer vision, pp. 2272–2281. Kamel et al.: Preprint submitted to Elsevier Page 14 of 16

  2. [2]

    Anatomy-aware 3d human pose estimation with bone-based pose decomposition

    Chen, T., Fang, C., Shen, X., Zhu, Y., Chen, Z., Luo, J., 2021. Anatomy-aware 3d human pose estimation with bone-based pose decomposition. IEEETransactionsonCircuitsandSystemsforVideo Technology 32, 198–209

  3. [3]

    Cascaded pyramid network for multi-person pose estimation, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    Chen, Y., Wang, Z., Peng, Y., Zhang, Z., Yu, G., Sun, J., 2018. Cascaded pyramid network for multi-person pose estimation, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 7103–7112

  4. [4]

    3d human pose estimation using spatio-temporal networks with explicit occlusion training, in: Proceedings of the AAAI Conference on Artificial In- telligence, pp

    Cheng, Y., Yang, B., Wang, B., Tan, R.T., 2020. 3d human pose estimation using spatio-temporal networks with explicit occlusion training, in: Proceedings of the AAAI Conference on Artificial In- telligence, pp. 10631–10638

  5. [5]

    Occlusion- aware networks for 3d human pose estimation in video, in: Proceed- ings of the IEEE/CVF international conference on computer vision, pp

    Cheng,Y.,Yang,B.,Wang,B.,Yan,W.,Tan,R.T.,2019. Occlusion- aware networks for 3d human pose estimation in video, in: Proceed- ings of the IEEE/CVF international conference on computer vision, pp. 723–732

  6. [6]

    Choi, H., Moon, G., Lee, K.M., 2020. Pose2mesh: Graph con- volutional network for 3d human pose and mesh recovery from a 2d human pose, in: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part VII 16, Springer. pp. 769–787

  7. [7]

    Optimizingnetworkstruc- ture for 3d human pose estimation, in: Proceedings of the IEEE/CVF international conference on computer vision, pp

    Ci,H.,Wang,C.,Ma,X.,Wang,Y.,2019. Optimizingnetworkstruc- ture for 3d human pose estimation, in: Proceedings of the IEEE/CVF international conference on computer vision, pp. 2262–2271

  8. [8]

    Learningpose grammartoencodehumanbodyconfigurationfor3dposeestimation, in: Proceedings of the AAAI conference on artificial intelligence

    Fang,H.S.,Xu,Y.,Wang,W.,Liu,X.,Zhu,S.C.,2018. Learningpose grammartoencodehumanbodyconfigurationfor3dposeestimation, in: Proceedings of the AAAI conference on artificial intelligence

Show all 52 references
  1. [9]

    Diff- pose:Towardmorereliable3dposeestimation,in:Proceedingsofthe IEEE/CVFConferenceonComputerVisionandPatternRecognition, pp

    Gong,J.,Foo,L.G.,Fan,Z.,Ke,Q.,Rahmani,H.,Liu,J.,2023. Diff- pose:Towardmorereliable3dposeestimation,in:Proceedingsofthe IEEE/CVFConferenceonComputerVisionandPatternRecognition, pp. 13041–13051

  2. [10]

    Hu,W.,Zhang,C.,Zhan,F.,Zhang,L.,Wong,T.T.,2021.Conditional directedgraphconvolutionfor3dhumanposeestimation,in:Proceed- ings of the 29th ACM International Conference on Multimedia, pp. 602–611

  3. [11]

    Hu- man3.6m: Large scale datasets and predictive methods for 3d human sensing in natural environments

    Ionescu, C., Papava, D., Olaru, V., Sminchisescu, C., 2013. Hu- man3.6m: Large scale datasets and predictive methods for 3d human sensing in natural environments. IEEE transactions on pattern analy- sis and machine intelligence 36, 1325–1339

  4. [12]

    Kang, S., Jeon, S., Woo, W., 2023. Ips: Integrating pose with speech for enhancement of body pose estimation in vr remote collaboration, in: 2023 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW), IEEE. pp. 637–638

  5. [13]

    Semi-supervised classification with graph convolutional networks

    Kipf, T.N., Welling, M., 2016. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907

  6. [14]

    Propagating lstm: 3d pose estimation based on joint interdependency, in: Proceedings of the European conference on computer vision (ECCV), pp

    Lee, K., Lee, I., Lee, S., 2018. Propagating lstm: 3d pose estimation based on joint interdependency, in: Proceedings of the European conference on computer vision (ECCV), pp. 119–135

  7. [15]

    Maximum-margin structured learning with deep networks for 3d human pose estimation, in: Pro- ceedingsoftheIEEEinternationalconferenceoncomputervision,pp

    Li, S., Zhang, W., Chan, A.B., 2015. Maximum-margin structured learning with deep networks for 3d human pose estimation, in: Pro- ceedingsoftheIEEEinternationalconferenceoncomputervision,pp. 2848–2856

  8. [16]

    Exploiting temporal contexts with strided transformer for 3d human pose estimation

    Li, W., Liu, H., Ding, R., Liu, M., Wang, P., Yang, W., 2022a. Exploiting temporal contexts with strided transformer for 3d human pose estimation. IEEE Transactions on Multimedia 25, 1282–1293

  9. [17]

    Mh- former: Multi-hypothesis transformer for 3d human pose estimation, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Li, W., Liu, H., Tang, H., Wang, P., Van Gool, L., 2022b. Mh- former: Multi-hypothesis transformer for 3d human pose estimation, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 13147–13156

  10. [18]

    arXiv preprint arXiv:1908.08289

    Lin,J.,Lee,G.H.,2019.Trajectoryspacefactorizationfordeepvideo- based 3d human pose estimation. arXiv preprint arXiv:1908.08289

  11. [19]

    End-to-end human pose and mesh reconstruction with transformers, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp

    Lin, K., Wang, L., Liu, Z., 2021. End-to-end human pose and mesh reconstruction with transformers, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 1954– 1963

  12. [20]

    A graph attention spatio-temporal convolutional network for 3d human pose estimation in video, in: 2021 IEEE international conference on robotics and automation (ICRA), IEEE

    Liu,J.,Rojas,J.,Li,Y.,Liang,Z.,Guan,Y.,Xi,N.,Zhu,H.,2021. A graph attention spatio-temporal convolutional network for 3d human pose estimation in video, in: 2021 IEEE international conference on robotics and automation (ICRA), IEEE. pp. 3374–3380

  13. [21]

    Attentionmechanismexploitstemporalcontexts:Real-time3dhuman posereconstruction,in:ProceedingsoftheIEEE/CVFConferenceon Computer Vision and Pattern Recognition, pp

    Liu, R., Shen, J., Wang, H., Chen, C., Cheung, S.c., Asari, V., 2020. Attentionmechanismexploitstemporalcontexts:Real-time3dhuman posereconstruction,in:ProceedingsoftheIEEE/CVFConferenceon Computer Vision and Pattern Recognition, pp. 5064–5073

  14. [22]

    Jointformer: Single-frame lifting transformer with error predictionandrefinementfor3dhumanposeestimation,in:202226th International Conference on Pattern Recognition (ICPR), IEEE

    Lutz,S.,Blythman,R.,Ghosal,K.,Moynihan,M.,Simms,C.,Smolic, A., 2022. Jointformer: Single-frame lifting transformer with error predictionandrefinementfor3dhumanposeestimation,in:202226th International Conference on Pattern Recognition (ICPR), IEEE. pp. 1156–1163

  15. [23]

    A simple yet effectivebaselinefor3dhumanposeestimation,in:Proceedingsofthe IEEE international conference on computer vision, pp

    Martinez, J., Hossain, R., Romero, J., Little, J.J., 2017. A simple yet effectivebaselinefor3dhumanposeestimation,in:Proceedingsofthe IEEE international conference on computer vision, pp. 2640–2649

  16. [24]

    Monocular3dhumanposeestimationinthewild usingimprovedcnnsupervision,in:2017internationalconferenceon 3D vision (3DV), IEEE

    Mehta, D., Rhodin, H., Casas, D., Fua, P., Sotnychenko, O., Xu, W., Theobalt,C.,2017. Monocular3dhumanposeestimationinthewild usingimprovedcnnsupervision,in:2017internationalconferenceon 3D vision (3DV), IEEE. pp. 506–516

  17. [25]

    Motion imitation of a humanoid robot via pose estimation, in: 2023 35th Chinese Control and Decision Conference (CCDC), IEEE

    Meng, S., Qiu, S., Liang, T., Ren, Q., 2023. Motion imitation of a humanoid robot via pose estimation, in: 2023 35th Chinese Control and Decision Conference (CCDC), IEEE. pp. 1526–1532

  18. [26]

    Newell, A., Yang, K., Deng, J., 2016. Stacked hourglass networks for human pose estimation, in: Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part VIII 14, Springer. pp. 483–499

  19. [27]

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

    Pang, C., Lu, X., Lyu, L., 2023. Skeleton-based action recognition through contrasting two-stream spatial-temporal networks. IEEE Transactions on Multimedia

  20. [28]

    Ordinal depth super- vision for 3d human pose estimation, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    Pavlakos, G., Zhou, X., Daniilidis, K., 2018. Ordinal depth super- vision for 3d human pose estimation, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 7307– 7316

  21. [29]

    3d humanposeestimationinvideowithtemporalconvolutionsandsemi- supervised training, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp

    Pavllo, D., Feichtenhofer, C., Grangier, D., Auli, M., 2019. 3d humanposeestimationinvideowithtemporalconvolutionsandsemi- supervised training, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 7753–7762

  22. [30]

    Higher-order implicit fairing networks for 3d human pose estimation

    Quan, J., Hamza, A.B., 2021. Higher-order implicit fairing networks for 3d human pose estimation. arXiv preprint arXiv:2111.00950

  23. [31]

    Higher-order implicit fairing networks for 3d human pose estimation, in: British Machine Vision Conference

    Quan, J., Quan, J., 2021. Higher-order implicit fairing networks for 3d human pose estimation, in: British Machine Vision Conference. URL: https://api.semanticscholar.org/CorpusID:240354750

  24. [32]

    Shan,W.,Liu,Z.,Zhang,X.,Wang,S.,Ma,S.,Gao,W.,2022.P-stmo: Pre-trained spatial temporal many-to-one model for 3d human pose estimation, in: European Conference on Computer Vision, Springer. pp. 461–478

  25. [33]

    Improving robustness and accuracy via relative information encoding in 3d hu- man pose estimation, in: Proceedings of the 29th ACM International Conference on Multimedia, pp

    Shan, W., Lu, H., Wang, S., Zhang, X., Gao, W., 2021. Improving robustness and accuracy via relative information encoding in 3d hu- man pose estimation, in: Proceedings of the 29th ACM International Conference on Multimedia, pp. 3446–3454

  26. [34]

    Humaneva: Synchronized video and motion capture dataset and baseline algorithm for evalua- tion of articulated human motion

    Sigal, L., Balan, A.O., Black, M.J., 2010. Humaneva: Synchronized video and motion capture dataset and baseline algorithm for evalua- tion of articulated human motion. International journal of computer vision 87, 4–27

  27. [35]

    Compositional human poseregression,in:ProceedingsoftheIEEEinternationalconference on computer vision, pp

    Sun, X., Shang, J., Liang, S., Wei, Y., 2017. Compositional human poseregression,in:ProceedingsoftheIEEEinternationalconference on computer vision, pp. 2602–2611

  28. [36]

    Attention is all you need

    Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez, A.N., Kaiser, Ł., Polosukhin, I., 2017. Attention is all you need. Advances in neural information processing systems 30

  29. [37]

    Motion projection consistency based 3d human pose estimation with virtual bones from monocular videos

    Wang, G., Zeng, H., Wang, Z., Liu, Z., Wang, H., 2022. Motion projection consistency based 3d human pose estimation with virtual bones from monocular videos. IEEE Transactions on Cognitive and Developmental Systems

  30. [38]

    Motion guided 3d pose estimation from videos, in: European Conference on Computer Vision, Springer

    Wang, J., Yan, S., Xiong, Y., Lin, D., 2020. Motion guided 3d pose estimation from videos, in: European Conference on Computer Vision, Springer. pp. 764–780

  31. [39]

    Drpose3d:Depthrankingin3dhumanposeestimation.arXivpreprint arXiv:1805.08973

    Wang, M., Chen, X., Liu, W., Qian, C., Lin, L., Ma, L., 2018. Drpose3d:Depthrankingin3dhumanposeestimation.arXivpreprint arXiv:1805.08973 . Kamel et al.: Preprint submitted to Elsevier Page 15 of 16

  32. [40]

    High-order localconnectionnetworkfor3dhumanposeestimationbasedongcn

    Wu, W., Zhou, D., Zhang, Q., Dong, J., Wei, X., 2022. High-order localconnectionnetworkfor3dhumanposeestimationbasedongcn. Applied Intelligence 52, 15690–15702

  33. [41]

    Deep kinematics analysis for monocular 3d human pose estimation, in: Proceedings of the IEEE/CVF Conference on computer vision and Pattern recognition, pp

    Xu, J., Yu, Z., Ni, B., Yang, J., Yang, X., Zhang, W., 2020. Deep kinematics analysis for monocular 3d human pose estimation, in: Proceedings of the IEEE/CVF Conference on computer vision and Pattern recognition, pp. 899–908

  34. [42]

    Graph stacked hourglass networks for 3d humanposeestimation,in:ProceedingsoftheIEEE/CVFconference on computer vision and pattern recognition, pp

    Xu, T., Takano, W., 2021. Graph stacked hourglass networks for 3d humanposeestimation,in:ProceedingsoftheIEEE/CVFconference on computer vision and pattern recognition, pp. 16105–16114

  35. [43]

    Spatial temporal graph convolu- tionalnetworksforskeleton-basedactionrecognition,in:Proceedings of the AAAI conference on artificial intelligence

    Yan, S., Xiong, Y., Lin, D., 2018. Spatial temporal graph convolu- tionalnetworksforskeleton-basedactionrecognition,in:Proceedings of the AAAI conference on artificial intelligence

  36. [44]

    Gla-gcn: Global-local adaptive graph convolutional network for 3d humanposeestimationfrommonocularvideo,in:Proceedingsofthe IEEE/CVF International Conference on Computer Vision, pp

    Yu, B.X., Zhang, Z., Liu, Y., Zhong, S.h., Liu, Y., Chen, C.W., 2023. Gla-gcn: Global-local adaptive graph convolutional network for 3d humanposeestimationfrommonocularvideo,in:Proceedingsofthe IEEE/CVF International Conference on Computer Vision, pp. 8818– 8829

  37. [45]

    Zeng, A., Sun, X., Huang, F., Liu, M., Xu, Q., Lin, S., 2020. Srnet: Improving generalization in 3d human pose estimation with a split- and-recombineapproach,in:ComputerVision–ECCV2020:16thEu- ropeanConference,Glasgow,UK,August23–28,2020,Proceedings, Part XIV 16, Springer. pp. 507–523

  38. [46]

    Mixste: Seq2seqmixedspatio-temporalencoderfor3dhumanposeestimation in video, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp

    Zhang, J., Tu, Z., Yang, J., Chen, Y., Yuan, J., 2022. Mixste: Seq2seqmixedspatio-temporalencoderfor3dhumanposeestimation in video, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 13232–13242

  39. [47]

    Zhang,Y.,Lu,Y.,Liu,B.,Zhao,Z.,Chu,Q.,Yu,N.,2023. Evopose: Arecursivetransformerfor3dhumanposeestimationwithkinematic structure priors, in: ICASSP 2023-2023 IEEE International Confer- ence on Acoustics, Speech and Signal Processing (ICASSP), IEEE. pp. 1–5

  40. [48]

    Semanticgraphconvolutionalnetworksfor3dhumanposeregression, in:ProceedingsoftheIEEE/CVFconferenceoncomputervisionand pattern recognition, pp

    Zhao, L., Peng, X., Tian, Y., Kapadia, M., Metaxas, D.N., 2019. Semanticgraphconvolutionalnetworksfor3dhumanposeregression, in:ProceedingsoftheIEEE/CVFconferenceoncomputervisionand pattern recognition, pp. 3425–3435

  41. [49]

    Pose- formerv2: Exploring frequency domain for efficient and robust 3d humanposeestimation,in:ProceedingsoftheIEEE/CVFConference on Computer Vision and Pattern Recognition, pp

    Zhao, Q., Zheng, C., Liu, M., Wang, P., Chen, C., 2023. Pose- formerv2: Exploring frequency domain for efficient and robust 3d humanposeestimation,in:ProceedingsoftheIEEE/CVFConference on Computer Vision and Pattern Recognition, pp. 8877–8886

  42. [50]

    3dhumanposeestimationwithspatialandtemporaltransformers,in: ProceedingsoftheIEEE/CVFInternationalConferenceonComputer Vision, pp

    Zheng,C.,Zhu,S.,Mendieta,M.,Yang,T.,Chen,C.,Ding,Z.,2021. 3dhumanposeestimationwithspatialandtemporaltransformers,in: ProceedingsoftheIEEE/CVFInternationalConferenceonComputer Vision, pp. 11656–11665

  43. [51]

    Hemlets posh: learning part-centric heatmap triplets for 3d human pose and shape estimation

    Zhou, K., Han, X., Jiang, N., Jia, K., Lu, J., 2021. Hemlets posh: learning part-centric heatmap triplets for 3d human pose and shape estimation. IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 3000–3014

  44. [52]

    High-order graph convolutional networks for 3d human pose estimation., in: BMVC

    Zou, Z., Liu, K., 0003, L.W., Tang, W., 2020. High-order graph convolutional networks for 3d human pose estimation., in: BMVC. Kamel et al.: Preprint submitted to Elsevier Page 16 of 16

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.