Pith. sign in

REVIEW 4 major objections 5 minor 45 references

ReMP: Reusable Motion Prior for Multi-domain 3D Human Pose Estimation and Motion Inbetweening

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read A single motion prior, pretrained once on complete 3D body sequences, can be frozen and reused across depth cameras, LiDAR, and IMU sensors to reconstruct accurate human motion and fill missing frames.

desk verdict Promising method, but the synthetic CMU depth benchmark is contaminated because the frozen motion prior was pretrained on AMASS, which includes CMU. read the letter →

arxiv 2411.09435 v1 pith:YVCZGKF6 submitted 2024-11-13 cs.CV

classification cs.CV
keywords reusablemotionprior3DhumanposeestimationinbetweeningtransformerVAEsensormodalitytransferpointcloudIMUSMPL
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

ReMP is an attempt to show that a single motion prior, learned once from a large collection of complete 3D body sequences, can be reused across very different sensors instead of training a separate model per sensor. The paper trains a transformer-based variational autoencoder on complete SMPL sequences to capture temporal dynamics, then freezes that prior and attaches only a small input encoder and a latent mapper for each new modality. It reports consistent accuracy gains over per-task baselines on depth point clouds, LiDAR scans, and IMU sensor data, including real-world datasets, and shows the prior makes pose estimation much more data-efficient. If the claim holds, it means a shared spatio-temporal body model can serve as a practical foundation for multi-domain 3D human pose estimation and motion completion.

What carries the argument

The load-bearing object is the latent space of the motion prior, which encodes a sequence of motion parameters $M_{1:T}$ (6D rotations of 24 joints plus dimension-expanded incremental root translation) into Gaussian latents per time step. Training minimizes reconstruction losses over pose, angular velocity, translation, translation velocity, joints, and vertices, together with a KL term, while random temporal masking forces the transformer to learn temporal context. On reuse, the latent mapper produces a distribution $p'_\psi(z_t|I'_t)$ from encoded sensor features, and the KL divergence between this distribution and the prior's distribution is minimized together with the same reconstruction losses. This lets a new sensor be adapted by training only the small encoder and mapper, and the same temporal mask used in training can be adjusted to perform motion inbetweening.

What would settle it

Train ReMP and a from-scratch baseline on a real multi-sensor capture whose noise and occlusion differ sharply from the synthetic recipe—for instance, long-range LiDAR with heavy dropout or IMUs worn under loose clothing—and compare held-out pose error; if the frozen-prior model no longer beats the from-scratch baseline, the reported reusability is an artifact of the simulation rather than a property of the prior.

Watch

Extended reading notes

Core claim

The paper's central claim is that a frozen spatio-temporal motion prior, trained on complete parametric mesh sequences from a large motion archive, transfers to multiple sensing modalities and outperforms methods that train separate pose-estimation models. The prior is a transformer VAE with a continuous latent sequence; it is trained by reconstructing SMPL pose parameters, translations, velocities, joints, and vertices, with random frame masking that also enables motion inbetweening. At reuse time, the first-stage networks are frozen, and each task adds only a lightweight input encoder (PointNet for point clouds, an MLP for IMU) plus a transformer-based latent mapper that aligns the sensor features to the prior's latent distribution. The paper demonstrates superior pose, joint, and mesh metrics on synthetic and real depth data, on the SLOPER4D LiDAR dataset, and on TotalCapture IMU data, and shows that the prior retains strong performance when training data is reduced to a quarter, indicating a large data-efficiency benefit.

Load-bearing premise

The load-bearing assumption is that the synthetic sensor data generated from AMASS meshes—1,024 depth points, 256 LiDAR points, six simulated IMUs—faithfully mimics real depth, LiDAR, and IMU noise and occlusion, so the frozen prior transfers to real measurements without fine-tuning.

Editorial extensions

If this is right

  • One frozen prior can be attached to new sensor modalities by training only a lightweight input encoder and latent mapper, so adding a sensor does not require retraining the motion model.
  • Training data requirements drop substantially: ReMP keeps its edge even when the training set is cut to one quarter, making pose estimation feasible with limited labeled data.
  • The same temporal-masking mechanism that builds the prior also performs motion inbetweening, so missing frames can be filled without a separate interpolation model.
  • Models trained only on synthetic sensor data transfer to real depth, LiDAR, and IMU captures without fine-tuning, which points toward simulation-driven training pipelines.

Reading between the lines

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

  • If the claim generalizes, it suggests that human motion, unlike language or images, is constrained enough that a pretrained body model may be built from a relatively modest motion archive rather than internet-scale data.
  • One natural testable extension is to swap the input encoder for other sparse sensors (radar, WiFi, pressure mats) and see whether the same frozen prior keeps working, which would strengthen the 'reusable' claim.
  • The paper leaves shape outside the prior and estimates it separately, so a stress test would be subjects with unusual body proportions or clothing, where a single prior may need to be coupled with a richer shape model.
  • Because the prior is trained without action labels, it may transfer across unseen activities more readily than action-conditioned priors, but this is an untested consequence of the paper's design choice.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. ReMP is a two-stage framework for 3D human pose estimation and motion inbetweening. In the first stage, a transformer-based variational autoencoder is trained on AMASS SMPL sequences to obtain a frozen spatio-temporal motion prior. In the second stage, a modality-specific input encoder and a latent mapper are trained to map depth point clouds, LiDAR scans, or IMU data into the prior's latent space, with an additional shape estimator. Experiments report pose/joint/mesh errors on synthetic CMU and real B-MHAD depth data, SLOPER4D LiDAR data, and TotalCapture IMU data, together with a data-efficiency study (Fig. 7) and qualitative motion inbetweening results.

Significance. If its empirical claims hold, ReMP is a meaningful step toward a reusable 3D motion prior: a single frozen prior, transferred across three sensor modalities, improves or matches strong baselines on real datasets (B-MHAD, SLOPER4D, TotalCapture) and shows data efficiency in Fig. 7. The real-data evaluations are a particular strength, as is the practical design that avoids per-modality heavy optimization. The confidence in the central claim is limited by the synthetic CMU/AMASS test overlap and by the absence of statistical guarantees for several close table entries.

major comments (4)
  1. [Section 4, Synthetic Dataset Generation / Depth Point Cloud] The statement that the synthetic CMU dataset 'was not seen during training' is not correct for the full ReMP pipeline. AMASS contains CMU MoCap motions, and the motion prior in Phase 1 is trained on AMASS; therefore the transformer decoder and prior have already been fit to CMU sequences. The Table 1 gains on synthetic CMU (Pose 4.90 vs 5.43, Joint 9.89 vs 11.15, Mesh 11.16 vs 12.62) may partly reflect memorization of the test distribution rather than a reusable prior. The B-MHAD real-data result is not affected by this overlap, but the depth-modality claim should be re-evaluated by excluding CMU/AMASS test sequences from prior pretraining, by evaluating on a held-out non-AMASS motion corpus, or by relying primarily on B-MHAD.
  2. [Tables 2 and 3] The claim that ReMP 'consistently outperforms' baselines is not supported by all reported numbers: LiDAR pose error is marginally worse than Jang et al. (8.58 vs 8.57) and IMU angular error is worse than PIP (12.07 vs 12.04). Since no error bars or multiple-seed results are reported, single-run differences of this size are not evidence of consistent superiority. Please add repeated-seed statistics or qualify the claim as 'outperforms on most metrics'.
  3. [Section 4.2, Motion Inbetweening] Motion inbetweening is presented as a contribution but has no quantitative evaluation. Figure 6 compares only with SLERP, and no error metric on the interpolated frames (e.g., pose or joint error against ground truth) or comparison with a learned baseline is reported. Please add quantitative inbetweening results, ideally on held-out AMASS sequences with known endpoints.
  4. [Section 4.3, Figure 7] The data-efficiency experiment does not state whether the frozen motion prior is used as-is for every dataset-size setting, nor how the reduced training splits relate to the synthetic AMASS splits used elsewhere. Since the abstract claims significantly improved training efficiency, please specify the protocol (e.g., fixed prior with only the mapper retrained, and the same test split across all sizes).
minor comments (5)
  1. [Figure 6 caption] The caption says 'which shows superior results compared to LERP' but the figure compares ReMP with SLERP; please clarify the comparison.
  2. [Section 3.1, Eq. (2)] The reconstruction loss weights w_s are not specified in the paper; provide the values in the main text or supplementary for reproducibility.
  3. [Section 4.1, Depth] The sentence 'ReMP consistently outperforms all baselines across every metric' is true for Table 1, but the wording should be qualified when discussing Tables 2 and 3.
  4. [Tables 1-3] Please add a note on how many random seeds are averaged and report standard deviations, or state explicitly that results are from a single seed.
  5. [Section 4, Synthetic Dataset Generation] The details of synthetic sensor generation are deferred to the supplementary material; since the supplementary is not available in this arXiv version, please include the key parameters (noise levels, point sampling strategy, IMU placement) in the main text or make the supplementary accessible.

Circularity Check

2 steps flagged · score 5.0 of 10

Synthetic CMU depth evaluation is contaminated: the frozen motion prior was pretrained on AMASS, which includes the CMU motion sequences used to make the synthetic test set.

  1. fitted input called prediction [Section 4 (Synthetic Dataset Generation) and Section 4.1 (Depth Point Cloud evaluation)]
    "We utilize the AMASS dataset [26] for both motion prior training and the reusing phases of our method. ... Firstly, we test ReMP on the synthetic CMU [17] dataset, which is part of the larger synthetic AMASS dataset we generated. This dataset was not seen during training."

    The claim that the synthetic CMU test set was not seen during training is true only for the reusing phase, not for the motion-prior phase. The frozen transformer encoder/decoder and the VAE prior were trained on the full AMASS corpus, and the synthetic CMU test is generated from AMASS sequences. ReMP's prior therefore already encodes the exact CMU motions it is asked to reconstruct; the latent mapper and input encoder only need to route depth point clouds into latent regions the prior has memorized. This partially forces the Table 1 gains on synthetic CMU (e.g., pose 4.90 vs. 5.43 for Jang et al.), while baselines have no access to this prior. The Figure 7 data-efficiency experiment uses the same contaminated test set.

  2. other [Section 4.3, Figure 7]
    "To further demonstrate the effectiveness of ReMP, we conducted an experiment to evaluate its data efficiency compared to the baselines on synthetic CMU [17] dataset."

    This data-efficiency comparison inherits the same contamination as the synthetic CMU depth test: the evaluation set is drawn from the same motion corpus used to pretrain the motion prior. The efficiency advantage shown in Figure 7 may therefore reflect prior memorization of CMU motion rather than sample-efficient reuse of a general prior. It does not directly affect the LiDAR or IMU conclusions, but it is a second load-bearing result built on the same leaked test set.

full rationale

The paper is not circular in a definitional sense: ReMP is a learned representation, not a mathematical derivation, and the central evaluations include external real-world datasets (B-MHAD, SLOPER4D, TotalCapture) that are not part of AMASS. The self-citations to Jang et al. and Neural Marionette are contextual or baseline comparisons, not load-bearing evidence for the main claim. However, the synthetic CMU depth evaluation is a genuine partial-circularity problem: the paper states this test set was 'not seen during training,' but the motion prior was pretrained on AMASS, which contains the CMU motion sequences from which the test set was generated. On synthetic CMU, the prior has already been fit to the target motion content, so the reported improvements over baselines are partially forced by pretraining exposure. The data-efficiency experiment in Section 4.3 uses the same contaminated set. Because the external real-data results remain independent support, the overall circularity is moderate rather than total.

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

The central claim rests on the sufficiency of AMASS, the fidelity of synthetic sensor data, and the validity of KL-based latent transfer. Several loss weights and architecture hyperparameters are not reported, which limits exact reproduction.

free parameters (7)
  • reconstruction loss weights w_s for S={theta, Delta theta, x, Delta x, J, V} = not reported
    Eq. (2) defines a weighted sum, but the weights are not listed; the central training objective depends on them.
  • KL weights wprior_KL and wreuse_KL = not reported
    Eqs. (4) and (7) use these weights; their values affect the balance between reconstruction and prior matching.
  • shape loss weight w_beta = not reported
    Eq. (7) includes this weight for shape estimation; the value is not given.
  • sequence length and framerate = 40 frames at 10 fps
    Section 4 states these values; temporal context and comparability across tasks depend on them.
  • sensor simulation densities and IMU count = 1,024 depth points; 256 LiDAR points; six IMUs
    Section 4 defines these; they determine task difficulty and realism of synthetic-to-real transfer.
  • random temporal masking ratio = not reported
    Section 3.1 uses random masking to promote temporal context and inbetweening, but the masking ratio is not specified.
  • latent dimension D_z and transformer sizes = not reported
    Architecture hyperparameters are not listed in the main text, limiting exact reproduction.
assumptions (5)
  • domain assumption AMASS contains about 11,000 SMPL motion sequences that are sufficient to learn a reusable motion prior.
    Invoked in Sections 1 and 4 as the sole source for pretraining; if AMASS is not representative, the prior inherits the bias.
  • domain assumption Synthetic depth, LiDAR, and IMU data generated from SMPL meshes accurately model real sensor measurements.
    Section 4 'Synthetic Dataset Generation' creates the training data; real-dataset transfer depends on this fidelity.
  • domain assumption The latent distribution p_psi learned by the VAE is a valid target for the latent mapper, so KL matching transfers temporal dynamics.
    Eq. (5) in Section 3.2 forces the mapped distribution to match the prior; if the prior latent space is not informative, the reuse mechanism fails.
  • ad hoc to paper Random temporal masking during pretraining makes the same model capable of motion inbetweening.
    Section 3.1 says masking enables inbetweening, and Section 4.2 uses it qualitatively, but no explicit inbetweening objective or quantitative result is given.
  • standard math Standard transformer attention, VAE reparameterization, and SMPL differentiable skinning behave as assumed.
    Foundational tools cited from references [18, 23, 38]; no new mathematical claims are made.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ReMP: Reusable Motion Prior for Multi-domain 3D Human Pose Estimation and Motion Inbetweening." pith.science (2026). https://pith.science/paper/YVCZGKF6

@misc{pith2026241109435,
  author       = {Pith},
  title        = {Pith review of: ReMP: Reusable Motion Prior for Multi-domain 3D Human Pose Estimation and Motion Inbetweening},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YVCZGKF6}},
  note         = {Machine review of arXiv:2411.09435}
}
read the original abstract

We present Reusable Motion prior (ReMP), an effective motion prior that can accurately track the temporal evolution of motion in various downstream tasks. Inspired by the success of foundation models, we argue that a robust spatio-temporal motion prior can encapsulate underlying 3D dynamics applicable to various sensor modalities. We learn the rich motion prior from a sequence of complete parametric models of posed human body shape. Our prior can easily estimate poses in missing frames or noisy measurements despite significant occlusion by employing a temporal attention mechanism. More interestingly, our prior can guide the system with incomplete and challenging input measurements to quickly extract critical information to estimate the sequence of poses, significantly improving the training efficiency for mesh sequence recovery. ReMP consistently outperforms the baseline method on diverse and practical 3D motion data, including depth point clouds, LiDAR scans, and IMU sensor data. Project page is available in https://hojunjang17.github.io/ReMP.

Figures

Figures reproduced from arXiv: 2411.09435 by the authors.

Figure 1
Figure 1. We extract rich motion priors from the large-scale mo [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall pipeline of our method consists of two parts: (a) training motion prior and (b) reusing pretrained prior. In the motion [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Results of ReMP and the baselines on synthetic CMU [ [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Pose estimation results of ReMP and the baselines on SLOPER4D dataset [ [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Motion reconstruction results of ReMP and the baselines from IMU sensor data on TotalCapture dataset [ [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Motion inbetweening results of ReMP and SLERP when [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 37 canonical work pages

  1. [1]

    Neural marionette: Unsupervised learning of motion skeleton and latent dynamics from vol- umetric video

    Jinseok Bae, Hojun Jang, Cheol-Hui Min, Hyungun Choi, and Young Min Kim. Neural marionette: Unsupervised learning of motion skeleton and latent dynamics from vol- umetric video. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 86–94, 2022. 8

  2. [2]

    Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D. Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Lan- guage models are few-shot learners. Advances in Neural In- formation Processing Systems, 33:1877–1901, 2020. 1

  3. [3]

    Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Gi- ancarlo Baldan, and Oscar Beijbom

    Holger Caesar, Varun Bankiti, Alex H. Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Gi- ancarlo Baldan, and Oscar Beijbom. nuscenes: A multi- modal dataset for autonomous driving. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2020. 3

  4. [4]

    Emerg- ing properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing properties in self-supervised vision transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021. 1

  5. [5]

    Sloper4d: A scene-aware dataset for global 4d human pose estimation in urban environments

    Yudi Dai, Yitai Lin, Xiping Lin, Chenglu Wen, Lan Xu, Hongwei Yi, Siqi Shen, Yuexin Ma, and Cheng Wang. Sloper4d: A scene-aware dataset for global 4d human pose estimation in urban environments. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 682–692, June 2023. 3, 6, 7

  6. [6]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018. 1

  7. [7]

    Human-m3: A multi-view multi-modal dataset for 3d human pose estimation in outdoor scenes

    Bohao Fan, Siqi Wang, Wenzhao Zheng, Jianjiang Feng, and Jie Zhou. Human-m3: A multi-view multi-modal dataset for 3d human pose estimation in outdoor scenes. arXiv preprint arXiv:2308.00628, 2023. 3

  8. [8]

    A point set generation network for 3d object reconstruction from a sin- gle image

    Haoqiang Fan, Hao Su, and Leonidas J Guibas. A point set generation network for 3d object reconstruction from a sin- gle image. In Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR) , pages 605– 613, 2017. 5

Show all 45 references
  1. [9]

    Bottom-up human pose estimation via disentan- gled keypoint regression

    Zigang Geng, Ke Sun, Bin Xiao, Zhaoxiang Zhang, and Jing- dong Wang. Bottom-up human pose estimation via disentan- gled keypoint regression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 14676–14686, 2021. 1

  2. [10]

    Kim, Bryan C

    Thibault Groueix, Matthew Fisher, Vladimir G. Kim, Bryan C. Russell, and Mathieu Aubry. 3d-coded: 3d cor- respondences by deep deformation. In Proceedings of the European Conference on Computer Vision (ECCV), Septem- ber 2018. 3

  3. [11]

    Momask: Generative masked mod- eling of 3d human motions

    Chuan Guo, Yuxuan Mu, Muhammad Gohar Javed, Sen Wang, and Li Cheng. Momask: Generative masked mod- eling of 3d human motions. 2023. 2

  4. [12]

    Abdelrahman, and Ayoub Al- Hamadi

    Thorsten Hempel, Ahmed A. Abdelrahman, and Ayoub Al- Hamadi. 6d rotation representation for unconstrained head pose estimation. In 2022 IEEE International Conference on Image Processing (ICIP), pages 2496–2500, 2022. 3

  5. [13]

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

    Junjie Huang, Zheng Zhu, Feng Guo, and Guan 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), pages 5700–5709, 2020. 1

  6. [14]

    Black, Otmar Hilliges, and Gerard Pons-Moll

    Yinghao Huang, Manuel Kaufmann, Emre Aksan, Michael J. Black, Otmar Hilliges, and Gerard Pons-Moll. Deep inertial poser: Learning to reconstruct human pose from sparse iner- tial measurements in real time.ACM Transactions on Graph- ics, (Proc. SIGGRAPH Asia) , 37:185:1–185:15,...

  7. [15]

    Dynamic mesh recovery from partial point cloud se- quence

    Hojun Jang, Minkwan Kim, Jinseok Bae, and Young Min Kim. Dynamic mesh recovery from partial point cloud se- quence. In Proceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV) , pages 15074–15084, October 2023. 2, 3, 5, 6

  8. [16]

    Winkler, and C

    Yifeng Jiang, Yuting Ye, Deepak Gopinath, Jungdam Won, Alexander W. Winkler, and C. Karen Liu. Transformer in- ertial poser: Real-time human motion reconstruction from sparse imus with simultaneous terrain generation. In SIG- GRAPH Asia 2022 Conference Papers , SA ’22 Conferen...

  9. [17]

    Panoptic studio: A massively multiview system for social motion capture

    Hanbyul Joo, Hao Liu, Lei Tan, Lin Gui, Bart Nabbe, Iain Matthews, Takeo Kanade, Shohei Nobuhara, and Yaser Sheikh. Panoptic studio: A massively multiview system for social motion capture. InThe IEEE International Conference on Computer Vision (ICCV), 2015. 5, 6, 8

  10. [18]

    Auto-encoding varia- tional bayes

    Diederik P Kingma and Max Welling. Auto-encoding varia- tional bayes. arXiv preprint arXiv:1312.6114, 2013. 3

  11. [19]

    Berg, Wan-Yen Lo, et al

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C. Berg, Wan-Yen Lo, et al. Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 4015–4026,

  12. [20]

    Li- darcap: Long-range marker-less 3d human motion capture with lidar point clouds

    Jialian Li, Jingyi Zhang, Zhiyong Wang, Siqi Shen, Chenglu Wen, Yuexin Ma, Lan Xu, Jingyi Yu, and Cheng Wang. Li- darcap: Long-range marker-less 3d human motion capture with lidar point clouds. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognitio...

  13. [21]

    Character controllers using motion vaes.ACM Trans

    Hung Yu Ling, Fabio Zinno, George Cheng, and Michiel van de Panne. Character controllers using motion vaes.ACM Trans. Graph., 39(4), 2020. 2

  14. [22]

    V otehmr: Occlusion- aware voting network for robust 3d human mesh recovery from partial point clouds

    Guanze Liu, Yu Rong, and Lu Sheng. V otehmr: Occlusion- aware voting network for robust 3d human mesh recovery from partial point clouds. In Proceedings of the 29th ACM International Conference on Multimedia , pages 955–964,

  15. [23]

    Matthew Loper, Naureen Mahmood, Javier Romero, Ger- ard Pons-Moll, and Michael J. Black. SMPL: A skinned multi-person linear model. ACM Trans. Graphics (Proc. SIGGRAPH Asia), 34(6):248:1–248:16, Oct. 2015. 1, 3

  16. [24]

    Posegpt: Quantization-based 3d human mo- 9 tion generation and forecasting

    Thomas Lucas*, Fabien Baradel*, Philippe Weinzaepfel, and Gr´egory Rogez. Posegpt: Quantization-based 3d human mo- 9 tion generation and forecasting. In European Conference on Computer Vision (ECCV), 2022. 2

  17. [25]

    Rethinking the heatmap regres- sion for bottom-up human pose estimation

    Zhengxiong Luo, Zhicheng Wang, Yan Huang, Liang Wang, Tieniu Tan, and Erjin Zhou. Rethinking the heatmap regres- sion for bottom-up human pose estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 13264–13273, 2021. 1

  18. [26]

    Troje, Ger- ard Pons-Moll, and Michael J

    Naureen Mahmood, Nima Ghorbani, Nikolaus F. Troje, Ger- ard Pons-Moll, and Michael J. Black. AMASS: Archive of motion capture as surface shapes. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 5442–5451, Oct. 2019. 1, 2, 5, 6

  19. [27]

    Berkeley mhad: A comprehensive multimodal human action database

    Ferda Ofli, Rizwan Ahmed Chaudhry, Gregorij Kurillo, Ren´e Vidal, and Ruzena Bajcsy. Berkeley mhad: A comprehensive multimodal human action database. 2013 IEEE Workshop on Applications of Computer Vision (WACV) , pages 53–60,

  20. [28]

    Training language models to follow instructions with human feed- back

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Car- roll Wainwright, Pamela Mishkin, Chong Zhang, Sand- hini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feed- back. Advances in Neural Information Processing Systems , ...

  21. [29]

    Black, and G ¨ul Varol

    Mathis Petrovich, Michael J. Black, and G ¨ul Varol. Action- conditioned 3d human motion synthesis with transformer vae. In Proceedings of the IEEE/CVF International Confer- ence on Computer Vision (ICCV), pages 10985–10995, Oc- tober 2021. 2

  22. [30]

    Qi, Hao Su, Kaichun Mo, and Leonidas J

    Charles R. Qi, Hao Su, Kaichun Mo, and Leonidas J. Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 652–660, 2017. 4

  23. [31]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In International Conference on Machine Learning...

  24. [32]

    Davis Rempe, Tolga Birdal, Aaron Hertzmann, Jimei Yang, Srinath Sridhar, and Leonidas J. Guibas. Humor: 3d human motion model for robust pose estimation. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 11488–11499, October 2021. 2, 3

  25. [33]

    Deep high-resolution representation learning for human pose es- timation

    Ke Sun, Bin Xiao, Dong Liu, and Jingdong Wang. Deep high-resolution representation learning for human pose es- timation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 5693–5703, 2019. 1

  26. [34]

    Scalability in perception for autonomous driving: Waymo open dataset

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, Vijay Vasudevan, Wei Han, Jiquan Ngiam, Hang Zhao, Aleksei Timofeev, Scott Et- tinger, Maxim Krivokon, Amy Gao, Aditya Joshi, Yu Zhan...

  27. [35]

    Total capture: 3d human pose estimation fusing video and inertial sensors

    Matt Trumble, Andrew Gilbert, Charles Malleson, Adrian Hilton, and John Collomosse. Total capture: 3d human pose estimation fusing video and inertial sensors. In 2017 British Machine Vision Conference (BMVC), 2017. 6, 7

  28. [36]

    Neural discrete representation learning

    Aaron Van Den Oord, Oriol Vinyals, et al. Neural discrete representation learning. Advances in Neural Information Processing Systems, 30, 2017. 2

  29. [37]

    Black, Ivan Laptev, and Cordelia Schmid

    G ¨ul Varol, Javier Romero, Xavier Martin, Naureen Mah- mood, Michael J. Black, Ivan Laptev, and Cordelia Schmid. Learning from synthetic humans. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017. 5

  30. [38]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Ł ukasz Kaiser, and Il- lia Polosukhin. Attention is all you need. In I. Guyon, U. V on Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vish- wanathan, and R. Garnett, editors,Advances in N...

  31. [39]

    Black, and Gerard Pons-Moll

    Timo V on Marcard, Bodo Rosenhahn, Michael J. Black, and Gerard Pons-Moll. Sparse inertial poser: Automatic 3d hu- man pose estimation from sparse imus. In Computer Graph- ics Forum, volume 36, pages 349–360. Wiley Online Library,

  32. [40]

    Phys- ical inertial poser (pip): Physics-aware real-time human mo- tion tracking from sparse inertial sensors

    Xinyu Yi, Yuxiao Zhou, Marc Habermann, Soshi Shimada, Vladislav Golyanik, Christian Theobalt, and Feng Xu. Phys- ical inertial poser (pip): Physics-aware real-time human mo- tion tracking from sparse inertial sensors. In Proceedings of the IEEE/CVF Conference on Computer Visio...

  33. [41]

    Transpose: Real-time 3d human translation and pose estimation with six inertial sensors

    Xinyu Yi, Yuxiao Zhou, and Feng Xu. Transpose: Real-time 3d human translation and pose estimation with six inertial sensors. ACM Transactions on Graphics, 40(4), 08 2021. 3, 6

  34. [42]

    Distribution-aware coordinate representation for human pose estimation

    Feng Zhang, Xiatian Zhu, Hanbin Dai, Mao Ye, and Ce Zhu. Distribution-aware coordinate representation for human pose estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 7093–7102, 2020. 1

  35. [43]

    Microsoft kinect sensor and its effect

    Zhengyou Zhang. Microsoft kinect sensor and its effect. IEEE multimedia, 19(2):4–10, 2012. 5

  36. [44]

    Open3D: A modern library for 3D data processing

    Qian-Yi Zhou, Jaesik Park, and Vladlen Koltun. Open3D: A modern library for 3D data processing. arXiv:1801.09847,

  37. [45]

    Self-supervised 3d human mesh recovery from noisy point clouds

    Xinxin Zuo, Sen Wang, Qiang Sun, Minglun Gong, and Li Cheng. Self-supervised 3d human mesh recovery from noisy point clouds. arXiv preprint arXiv:2107.07539, 2021. 3, 5, 6 10

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.