Pith. sign in

REVIEW 3 major objections 5 minor 55 references

Gen4D: Synthesizing Humans and Scenes in the Wild

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims a fully automated pipeline can generate diverse, photorealistic 4D human animation datasets from text prompts and internet videos, demonstrated by SportPAL, a 583,403-frame synthetic sports dataset.

desk verdict A sensible synthetic-data pipeline and a large new sports dataset, but the experiments only test self-consistency, not real-world transfer. read the letter →

arxiv 2506.05397 v1 pith:MJYBFDKB submitted 2025-06-03 cs.GR cs.AI

Jerrin Bright , Zhibo Wang , Yuhao Chen , Sirisha Rambhatla , John Zelek , David Clausi This is my paper
classification cs.GRcs.AI
keywords 4DhumananimationGaussiansplattingsyntheticdataposeestimationtext-to-avatardiffusionmodelssportsdatasetSMPLX
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

Gen4D aims to remove the manual 3D modeling bottleneck from synthetic human data. It claims that text prompts, a pretrained pose estimator, and internet videos are enough to automatically produce diverse, photorealistic 4D human animations with aligned pose annotations. The paper builds SportPAL, a 583,403-frame synthetic sports dataset across baseball, ice hockey, and soccer, and shows that pose estimators trained on it transfer across sports. If the claim holds, large-scale training data for uncommon human-centric vision tasks can be manufactured on demand, without motion-capture studios or asset libraries.

What carries the argument

The load-bearing construction is the fully automated motion-to-scene pipeline. Motion extraction uses a pretrained human mesh recovery method to get SMPLX body parameters from internet videos; canonical avatars are optimized as 3D Gaussians supervised by a text-conditioned diffusion model through a dual-branch Score Distillation Sampling loss; scene composition deforms the Gaussians along the motion, rasterizes them from randomized viewpoints, and inpaints backgrounds with an illumination-consistent diffusion backbone that enforces physically grounded light transport for shadows and lighting. The dataset is the material expression: SportPAL, with 50 subjects, 2,012 clips, and 583,403 frames.

What would settle it

Record a set of expert sports movements with both the video-based pose estimator used by Gen4D and a synchronized marker-based motion-capture rig; if the recovered poses differ systematically on fast articulations or bulky gear, and a pose estimator trained on SportPAL inherits those errors when tested on real footage, the paper's claim of accurate image-pose pairs would be contradicted.

Watch

Extended reading notes

Core claim

The central claim is that existing synthetic human datasets are limited by rigid asset libraries and hand-crafted rendering pipelines, and that Gen4D overcomes this by chaining three automated stages: extracting SMPLX pose and shape sequences from internet videos of expert demonstrations; synthesizing varied canonical human avatars by optimizing Gaussian splatting parameters under a dual-branch Score Distillation Sampling loss conditioned on text prompts; and deforming, rasterizing, and compositing those avatars into diffusion-generated, illumination-consistent backgrounds. The paper asserts this produces accurate image-pose pairs even when the initial pose estimates do not perfectly match the source video, because the estimated poses become the ground truth that drives rendering. SportPAL then provides 2D and 3D keypoints, SMPLX parameters, segmentation masks, bounding boxes, and action labels for over half a million frames, and experiments show a token-based pose estimator trained on the dataset generalizes across baseball, ice hockey, and soccer.

Load-bearing premise

The pipeline assumes that per-frame body pose and shape estimates pulled from internet videos are reliable enough, after normalization, to serve as the ground truth that animates synthetic avatars, even for fast, occluded, or gear-heavy sports movements.

Editorial extensions

If this is right

  • If the pipeline works as claimed, generating a new synthetic human dataset reduces to writing prompts and collecting videos, so sports or niche activities with scarce real data can get large training sets quickly.
  • Because every frame is rendered rather than recorded, annotations such as 2D and 3D poses, SMPLX parameters, segmentation masks, bounding boxes, and action labels come for free and are perfectly aligned with the image.
  • Pose estimators pretrained on one sport can be fine-tuned on another with measured gains, so the data's diversity aids cross-domain transfer.
  • The same framework can be pointed at other in-the-wild human activities beyond the three demonstrated sports, since the motion and scene stages are not sport-specific.

Reading between the lines

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

  • A natural extension, not argued in the paper, is that the three pipeline stages are modular enough to be upgraded independently: better motion sources or physics-constrained refinement could improve dataset realism without regenerating any avatars.
  • The dataset's self-consistency is guaranteed by construction, so its largest risk is not annotation noise but distribution shift from real sports imagery; a direct test would be to train on SportPAL and evaluate on real sports videos with held-out camera angles and athlete builds.
  • One testable extension is to run the same pipeline on motion-capture-verified sequences and compare downstream pose accuracy, isolating how much of the gain comes from pose fidelity versus image diversity.
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

3 major / 5 minor

Summary. The paper proposes Gen4D, a fully automated pipeline that extracts SMPLX pose and shape parameters from internet videos via a pretrained mesh-recovery method, optimizes canonical human Gaussians using dual-branch Score Distillation Sampling guided by text prompts, deforms and rasterizes these avatars under varied camera viewpoints, and composites human-aware backgrounds using an IC-Light diffusion module. Based on this pipeline, the authors introduce SportPAL, a synthetic dataset of over 583,000 frames spanning baseball, ice hockey, and soccer, with 2D/3D pose, SMPLX parameters, bounding boxes, segmentation masks, and action labels. The experimental section evaluates two 2D human pose estimators trained and tested on SportPAL and reports a cross-sport fine-tuning study.

Significance. If the central claims are supported, the pipeline would be a useful contribution: it removes the need for manual 3D modeling and scene design, sources motion from public internet videos, and produces a large-scale annotated synthetic dataset for an under-served domain. The paper gives a reasonably detailed description of the pipeline and provides a dataset statistics table. Its main strength is the proposed automation and scale. However, the evidence currently demonstrates only internal self-consistency of the renderer and annotation pipeline. The claimed photorealism, diversity, annotation accuracy, and utility for in-the-wild tasks are not measured against real data, and no code or data are released, so independent verification is impossible.

major comments (3)
  1. [§5.2, Table 2] The evaluation is entirely synthetic-to-synthetic and therefore does not support the central claim that SportPAL is a scalable foundation for in-the-wild human-centric vision tasks. Both the training and test sets are generated by the same Gen4D pipeline, so near-saturated AP15 scores (99.91 for ice hockey, 98.51 for soccer) reflect self-consistency of the renderer and annotation generator rather than transferable data quality. Please add a real-data transfer experiment on at least one public sports pose benchmark (e.g., AthletePose3D, SportsPose, ASPset-510, or WorldPose), and ideally a comparison with an existing synthetic dataset such as BEDLAM or AGORA under matched training conditions.
  2. [§3.1, §4] The annotations are not independent measurements. SMPLX parameters estimated by SMPLer-X from internet videos are normalized and used both as the pseudo-motion that drives rendering and as the ground truth in SportPAL. The paper's statement in §3.1 that perfect image–pose pairs will be generated is true only in the sense of internal consistency, not in the sense of correctness relative to real human motion. If SMPLer-X systematically errs on sports-specific articulation, occlusions, or bulky gear, that error is baked into every rendered frame and remains invisible in the synthetic test set. Please provide a validation study against independent motion capture or manual annotation on a subset of the three sports, and report per-joint error or failure modes.
  3. [§6, Future Work] The Future Work paragraph concedes that pseudo-ground-truth jitter and physical plausibility are not optimized, yet the abstract and introduction describe the generated sequences as 'lifelike' and 'photorealistic.' No experiment quantifies motion jitter, foot skating, or physical plausibility, and no perceptual study or image-quality metric supports the photorealism claim. Please add quantitative motion-quality metrics or a user study, or soften the claims accordingly.
minor comments (5)
  1. [Table 2] Table 2 cites 'TokenPose [17]', but reference [17] is Dwivedi et al. (TokenHMR), while the in-text citation for TokenPose is [29]; please correct the citation mismatch.
  2. [Table 3] In the Soccer row, '57110' should read '57,110' for readability, and the Total row should be checked for consistency with the per-sport frame counts.
  3. [Equation (2)] The symbols L1, L2, M, and Id are not defined precisely in the surrounding text; please define each variable next to the equation or in a notation list.
  4. [§4] The claim that SportPAL is 'the largest dataset featuring the most diverse range of actions from various sports' is not supported by a quantitative comparison with the datasets listed in Table 1; please add diversity or coverage metrics or qualify the claim.
  5. [§5.1] The SDS loss weights λ1 and λ2 and the background-diffusion weights λv and λic are fixed without an ablation or sensitivity analysis; please provide a brief study or at least a rationale for the chosen values.

Circularity Check

1 steps flagged · score 6.0 of 10

SportPAL's pose labels are the same parameters used to render the images, so Table 2's high AP is a self-consistency check; the in-the-wild utility claim is only partially supported.

  1. self definitional [Section 3.1 (Motion Extraction); evaluated in Section 5.2 / Table 2]
    "An important characteristic to note is that the accuracy of the extracted motion from the mesh recovery method [10] doesn’t need to perfectly match the observed pose of the person in the original video. Since our pipeline uses the estimated pose as the pseudo-motion to drive synthetic avatar animation (i.e., independent of the original viewpoint), we will still be able to generate perfect image–pose pairs in entirely new camera configurations for the corresponding rendered synthetic image."

    The SMPLer-X pose estimate is the input to the avatar deformation, and the same pose parameters are then projected back to 2D to create the annotation. Thus the 'perfect image–pose pairs' are perfect by construction: the image is a rendering of the label. When Table 2 trains and tests TokenPose/DETR on SportPAL, the test labels are exactly the parameters that generated the test images, so high AP (e.g., icehockey 99.91) measures the model's ability to invert the renderer, not the quality of the pseudo-groundtruth or its transfer to real sports data. Any SMPLer-X bias (bulky gear, skating articulation, occlusion) is baked into both the training labels and the rendered images and is therefore invisible in this internal evaluation.

full rationale

The pipeline's core generation stages (SMPLer-X pose extraction, SDS-based Gaussian optimization, background diffusion) are not mathematically circular in themselves: each stage consumes external pretrained models or losses, and no equation is algebraically rewritten as its own input. However, the quantitative support for the central dataset-utility claim is partially circular. Section 3.1 explicitly makes the pose labels and rendered images consistent by construction: the SMPLer-X estimate drives the avatar deformation, so the 'ground truth' 2D/3D pose equals the generator input. Table 2 then trains and tests 2D HPE models on these same self-consistent pairs; near-saturated AP therefore quantifies how well a model inverts the renderer, not how well the annotations match real human motion or transfer to real sports imagery. No external benchmark (e.g., AthletePose3D, SportsPose, ASPset-510, WorldPose) or real-data-trained baseline is used, and the Future Work paragraph acknowledges that pseudo-groundtruth jitter and physical plausibility are not optimized. The self-citations in the paper ([3], [8], [9]) are not load-bearing: the SMPL joint regressor is standard, and the cited prior work does not supply the validity of the pseudo-groundtruth or force the evaluation design. Overall, the paper is partially circular in its evaluation: the headline AP numbers are learnability/self-consistency checks on pairs that are perfect by construction, so the claim of a scalable foundation for in-the-wild tasks is supported only by internal consistency, not by external transfer evidence.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central contribution relies on several pretrained models treated as black boxes and on hand-set loss weights. No independent calibration data are shipped, so the calibration of pseudo ground truth and realism rests with the upstream models. No new physical or model entities are postulated; Gen4D is a software pipeline over existing representations.

free parameters (2)
  • SDS loss weights (lambda_1, lambda_2) = 0.5, 0.5
    Hand-set coefficients balancing RGB and depth SDS terms in Equation 1; no ablation or sensitivity analysis is reported.
  • Background diffusion weights (lambda_v, lambda_ic) = 1.0, 0.1
    Hand-set weights in Equation 2 for IC-Light illumination consistency; no sensitivity analysis is reported.
assumptions (5)
  • domain assumption SMPLX body model is a valid parametric human representation for pose and shape
    Section 3.1 and Section 4 use SMPLX parameters as both avatar initialization and dataset annotations; accuracy inherits SMPLX limitations.
  • domain assumption SMPLer-X mesh recovery from internet videos produces usable pseudo ground truth
    Section 3.1 treats SMPLer-X estimates as the motion script; if the estimator fails on sports poses, all annotations are biased.
  • domain assumption HumanGaussian's text-conditioned diffusion backbone yields diverse, realistic avatar appearance and geometry
    Section 3.2 uses RGB and depth SDS from [31]; avatar diversity depends on that model's text-to-3D capability.
  • domain assumption IC-Light background diffusion produces physically consistent lighting and shadows when compositing the rasterized avatar
    Section 3.3 relies on IC-Light [55] for relighting and shadow consistency; realism claims inherit its assumptions about light transport.
  • standard math 3D Gaussian splatting rasterization correctly renders the deformed avatar
    Section 3.3 uses standard 3DGS [27] projective blending; this is an established rendering tool.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Gen4D: Synthesizing Humans and Scenes in the Wild." pith.science (2026). https://pith.science/paper/MJYBFDKB

@misc{pith2026250605397,
  author       = {Pith},
  title        = {Pith review of: Gen4D: Synthesizing Humans and Scenes in the Wild},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MJYBFDKB}},
  note         = {Machine review of arXiv:2506.05397}
}
read the original abstract

Lack of input data for in-the-wild activities often results in low performance across various computer vision tasks. This challenge is particularly pronounced in uncommon human-centric domains like sports, where real-world data collection is complex and impractical. While synthetic datasets offer a promising alternative, existing approaches typically suffer from limited diversity in human appearance, motion, and scene composition due to their reliance on rigid asset libraries and hand-crafted rendering pipelines. To address this, we introduce Gen4D, a fully automated pipeline for generating diverse and photorealistic 4D human animations. Gen4D integrates expert-driven motion encoding, prompt-guided avatar generation using diffusion-based Gaussian splatting, and human-aware background synthesis to produce highly varied and lifelike human sequences. Based on Gen4D, we present SportPAL, a large-scale synthetic dataset spanning three sports: baseball, icehockey, and soccer. Together, Gen4D and SportPAL provide a scalable foundation for constructing synthetic datasets tailored to in-the-wild human-centric vision tasks, with no need for manual 3D modeling or scene design.

Figures

Figures reproduced from arXiv: 2506.05397 by the authors.

Figure 1
Figure 1. Overview of the SportPAL dataset generated using our proposed framework, Gen4D. (a) Examples of di￾verse canonical avatars synthesized via diffusion-guided prompt modeling, illustrating variation in clothing, body shape, and appearance. (b) Final synthetic frames rasterized with motion-driven avatar animation and human pose-aware backgrounds, demonstrating realistic lighting, shadows, and foot-ground interaction. Ab… view at source ↗
Figure 2
Figure 2. The pipeline of Gen4D. Our method consists of three stages: Section 3.1 Motion Extraction extracts 3D hu￾man motion from expert demonstrations; Section 3.2 Canonical Human Gaussians generates diverse human avatars via diffusion-guided Gaussian optimization; and Section 3.3 Scene Composition deforms avatars with motion, rasterizes them from varied viewpoints, and synthesizes realistic backgrounds using a scene-condit… view at source ↗
Figure 3
Figure 3. Prompt modeling pipeline of Gen4D. Diverse prompt templates are used to synthesize a variety of 3D canonical avatars, illustrated here from five different camera views. ∇θLSDS = λ1 · Eϵxc ,t  wt (ϵϕ (xc; pc, Phuman) − ϵxc ) ∂xc ∂θ  + λ2 · Eϵdc ,t  wt (ϵϕ (dc; pc, Phuman) − ϵdc ) ∂dc ∂θ  (1) where λ1 and λ2 are coefficients that balance the effects between RGB and depth. We compute LSDS according to our parsed ra… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative examples of synthetic images. Using Gen4D, we applied diverse prompt modeling to generate images across the three sports categories included in the SportPAL dataset. each frame, ensuring that the deformation of the Gaussian cloud closely follows the articul…
Figure 5
Figure 5. Figure 5: Qualitative visualizations of pose estimation results generated by TokenPose [29]. Groundtruth and predicted keypoints are shown in red and blue, respectively. Pose Estimation Performance on SportPAL [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 42 canonical work pages

  1. [1]

    Adobe fuse cc.https://www.adobe

    Adobe Systems. Adobe fuse cc.https://www.adobe. com/es/products/fuse.html. Accessed: 2025-05- 16

    work page 2025

  2. [2]

    Mixamo: 3d character animation.https: //www.mixamo.com/

    Adobe Systems. Mixamo: 3d character animation.https: //www.mixamo.com/. Accessed: 2025-05-16

    work page 2025

  3. [3]

    Seeing beyond the crop: Using language priors for out-of-bounding box key- point prediction.Advances in Neural Information Processing Systems, 37:102897–102918, 2024

    Bavesh Balaji, Jerrin Bright, Yuhao Chen, Sirisha Ramb- hatla, John Zelek, and David Clausi. Seeing beyond the crop: Using language priors for out-of-bounding box key- point prediction.Advances in Neural Information Processing Systems, 37:102897–102918, 2024

    work page 2024

  4. [4]

    Seamless human motion composition with blended posi- tional encodings

    German Barquero, Sergio Escalera, and Cristina Palmero. Seamless human motion composition with blended posi- tional encodings. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 457–469, 2024

    work page 2024

  5. [5]

    Hspace: Synthetic parametric humans animated in complex environments.arXiv preprint arXiv:2112.12867, 2021

    Eduard Gabriel Bazavan, Andrei Zanfir, Mihai Zanfir, William T Freeman, Rahul Sukthankar, and Cristian Smin- chisescu. Hspace: Synthetic parametric humans animated in complex environments.arXiv preprint arXiv:2112.12867, 2021

    arXiv 2021

  6. [6]

    Bedlam: A synthetic dataset of bodies exhibit- ing detailed lifelike animated motion

    Michael J Black, Priyanka Patel, Joachim Tesch, and Jin- long Yang. Bedlam: A synthetic dataset of bodies exhibit- ing detailed lifelike animated motion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8726–8737, 2023

    work page 2023

  7. [7]

    Blender - a 3d modelling and rendering package.https://www.blender.org/

    Blender Foundation. Blender - a 3d modelling and rendering package.https://www.blender.org/. Accessed: 2025-05-16

    work page 2025

  8. [8]

    Distribution and depth- aware transformers for 3d human mesh recovery.arXiv preprint arXiv:2403.09063, 2024

    Jerrin Bright, Bavesh Balaji, Harish Prakash, Yuhao Chen, David A Clausi, and John Zelek. Distribution and depth- aware transformers for 3d human mesh recovery.arXiv preprint arXiv:2403.09063, 2024

    arXiv 2024

Show all 55 references
  1. [9]

    Mitigating mo- tion blur for robust 3d baseball player pose modeling for pitch analysis, 2023

    Jerrin Bright, Yuhao Chen, and John Zelek. Mitigating mo- tion blur for robust 3d baseball player pose modeling for pitch analysis, 2023

    2023

  2. [10]

    SMPLer-X: Scaling up expressive human pose and shape estimation

    Zhongang Cai, Wanqi Yin, Ailing Zeng, Chen Wei, Qing- ping Sun, Wang Yanjun, Hui En Pang, Haiyi Mei, Mingyuan Zhang, Lei Zhang, Chen Change Loy, Lei Yang, and Zi- wei Liu. SMPLer-X: Scaling up expressive human pose and shape estimation. InAdvances in Neural Information Pro- ce...

    2023

  3. [11]

    Playing for 3d human re- covery.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    Zhongang Cai, Mingyuan Zhang, Jiawei Ren, Chen Wei, Daxuan Ren, Zhengyu Lin, Haiyu Zhao, Lei Yang, Chen Change Loy, and Ziwei Liu. Playing for 3d human re- covery.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  4. [12]

    End-to- end object detection with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to- end object detection with transformers. InEuropean confer- ence on computer vision, pages 213–229. Springer, 2020

    2020

  5. [13]

    Easi3r: Estimating disentangled motion from dust3r without training.arXiv preprint arXiv:2503.24391, 2025

    Xingyu Chen, Yue Chen, Yuliang Xiu, Andreas Geiger, and Anpei Chen. Easi3r: Estimating disentangled motion from dust3r without training.arXiv preprint arXiv:2503.24391, 2025

    2025

  6. [14]

    Subramanian Chidambaram, Rahul Jain, Sai Swarup Reddy, Asim Unmesh, and Karthik Ramani. Annotatexr: An ex- tended reality workflow for automating data annotation to support computer vision applications.Journal of Computing and Information Science in Engineering, 24(12), 2024

    2024

  7. [15]

    Pose2mesh: Graph convolutional network for 3d human pose and mesh recovery from a 2d human pose

    Hongsuk Choi, Gyeongsik Moon, and Kyoung Mu Lee. Pose2mesh: Graph convolutional network for 3d human pose and mesh recovery from a 2d human pose. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part VII 16, pages 769–787. Sp...

    2020

  8. [16]

    Acquiring the reflectance field of a human face

    Paul Debevec, Tim Hawkins, Chris Tchou, Haarm-Pieter Duiker, Westley Sarokin, and Mark Sagar. Acquiring the reflectance field of a human face. InProceedings of the 27th annual conference on Computer graphics and interac- tive techniques, pages 145–156, 2000

    2000

  9. [17]

    Tokenhmr: Advancing human mesh recov- ery with a tokenized pose representation

    Sai Kumar Dwivedi, Yu Sun, Priyanka Patel, Yao Feng, and Michael J Black. Tokenhmr: Advancing human mesh recov- ery with a tokenized pose representation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1323–1333, 2024

    2024

  10. [18]

    Unreal engine.https://www

    Epic Games. Unreal engine.https://www. unrealengine.com, 2020. Accessed: 2025-05-16

    2020

  11. [19]

    Panoptop: A framework for gen- erating viewpoint-invariant human pose estimation datasets

    Nicola Garau, Giulia Martinelli, Piotr Br ´odka, Niccolo Bis- agno, and Nicola Conci. Panoptop: A framework for gen- erating viewpoint-invariant human pose estimation datasets. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 234–242, 2021

    2021

  12. [20]

    Gaussianavatar: To- wards realistic human avatar modeling from a single video via animatable 3d gaussians, 11 2023

    Liangxiao Hu and Hongwen Zhang. Gaussianavatar: To- wards realistic human avatar modeling from a single video via animatable 3d gaussians, 11 2023

    2023

  13. [21]

    Gauhuman: Articulated gaus- sian splatting from monocular human videos

    Shoukang Hu and Ziwei Liu. Gauhuman: Articulated gaus- sian splatting from monocular human videos. 2023

    2023

  14. [22]

    Sportspose: A dynamic 3d sports pose dataset

    Christian Keilstrup Ingwersen, Christian Mikkelstrup, Janus Nørtoft Jensen, Morten Rieger Hannemose, and An- ders Bjorholm Dahl. Sportspose: A dynamic 3d sports pose dataset. InProceedings of the IEEE/CVF International Workshop on Computer Vision in Sports, 2023

    2023

  15. [23]

    Human3.6m: Large scale datasets and predic- tive methods for 3d human sensing in natural environments

    Catalin Ionescu, Dragos Papava, Vlad Olaru, and Cristian Sminchisescu. Human3.6m: Large scale datasets and predic- tive methods for 3d human sensing in natural environments. IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 36(7):1325–1339, jul 2014

    2014

  16. [24]

    World- pose: A world cup dataset for global 3d human pose esti- mation

    Tianjian Jiang, Johsan Billingham, Sebastian M ¨uksch, Juan Zarate, Nicolas Evans, Martin R Oswald, Marc Polleyfeys, Otmar Hilliges, Manuel Kaufmann, and Jie Song. World- pose: A world cup dataset for global 3d human pose esti- mation. InEuropean Conference on Computer Vision,...

    2024

  17. [25]

    Clustered pose and nonlinear appearance models for human pose estimation

    Sam Johnson and Mark Everingham. Clustered pose and nonlinear appearance models for human pose estimation. In bmvc, volume 2, page 5. Aberystwyth, UK, 2010

    2010

  18. [26]

    End-to-end recovery of human shape and pose

    Angjoo Kanazawa, Michael J Black, David W Jacobs, and Jitendra Malik. End-to-end recovery of human shape and pose. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 7122–7131, 2018

    2018

  19. [27]

    3d gaussian splatting for real-time ra- diance field rendering.ACM Trans

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkuehler, and George Drettakis. 3d gaussian splatting for real-time ra- diance field rendering.ACM Trans. Graph., 42(4), jul 2023

    2023

  20. [28]

    Ross, and Angjoo Kanazawa

    Ruilong Li, Sha Yang, David A. Ross, and Angjoo Kanazawa. Ai choreographer: Music conditioned 3d dance generation with aist++.2021 IEEE/CVF International Con- ference on Computer Vision (ICCV), pages 13381–13392, 2021

    2021

  21. [29]

    Tokenpose: Learning keypoint tokens for human pose estimation

    Yanjie Li, Shoukui Zhang, Zhicheng Wang, Sen Yang, Wankou Yang, Shu-Tao Xia, and Erjin Zhou. Tokenpose: Learning keypoint tokens for human pose estimation. In Proceedings of the IEEE/CVF International conference on computer vision, pages 11313–11322, 2021

    2021

  22. [30]

    Microsoft coco: Common objects in context

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

    2014

  23. [31]

    Humangaus- sian: Text-driven 3d human generation with gaussian splat- ting

    Xian Liu, Xiaohang Zhan, Jiaxiang Tang, Ying Shan, Gang Zeng, Dahua Lin, Xihui Liu, and Ziwei Liu. Humangaus- sian: Text-driven 3d human generation with gaussian splat- ting. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 6646–6657, 2024

    2024

  24. [32]

    Black.SMPL: A Skinned Multi- Person Linear Model

    Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, and Michael J. Black.SMPL: A Skinned Multi- Person Linear Model. Association for Computing Machin- ery, New York, NY , USA, 1 edition, 2023

    2023

  25. [33]

    Amass: Archive of motion capture as surface shapes

    Naureen Mahmood, Nima Ghorbani, Nikolaus F Troje, Ger- ard Pons-Moll, and Michael J Black. Amass: Archive of motion capture as surface shapes. InProceedings of the IEEE/CVF international conference on computer vision, pages 5442–5451, 2019

    2019

  26. [34]

    Makehuman: Open source 3d human creator.http://www.makehumancommunity

    MakeHuman Community. Makehuman: Open source 3d human creator.http://www.makehumancommunity. org/. Accessed: 2025-05-16

    2025

  27. [35]

    Monocular 3d human pose estimation in the wild using improved cnn supervision

    Dushyant Mehta, Helge Rhodin, Dan Casas, Pascal Fua, Oleksandr Sotnychenko, Weipeng Xu, and Christian Theobalt. Monocular 3d human pose estimation in the wild using improved cnn supervision. In3D Vision (3DV), 2017 Fifth International Conference on. IEEE, 2017

    2017

  28. [36]

    Efficient anno- tation of image data sets for computer vision applications

    Julia Moehrmann and Gunther Heidemann. Efficient anno- tation of image data sets for computer vision applications. InProceedings of the 1st International Workshop on Visual Interfaces for Ground Truth Collection in Computer Vision Applications, pages 1–6, 2012

    2012

  29. [37]

    ASPset: An outdoor sports pose video dataset with 3D keypoint annotations.Image and Vision Computing, page 104196, 2021

    Aiden Nibali, Joshua Millward, Zhen He, and Stuart Mor- gan. ASPset: An outdoor sports pose video dataset with 3D keypoint annotations.Image and Vision Computing, page 104196, 2021

    2021

  30. [38]

    Agora: Avatars in geography optimized for regression analysis

    Priyanka Patel, Chun-Hao P Huang, Joachim Tesch, David T Hoffmann, Shashank Tripathi, and Michael J Black. Agora: Avatars in geography optimized for regression analysis. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 13468–13478, 2021

    2021

  31. [39]

    3dpeople: Mod- eling the geometry of dressed humans

    Albert Pumarola, Jordi Sanchez-Riera, Gary Choi, Alberto Sanfeliu, and Francesc Moreno-Noguer. 3dpeople: Mod- eling the geometry of dressed humans. InProceedings of the IEEE/CVF international conference on computer vision, pages 2242–2251, 2019

    2019

  32. [40]

    An- novate: Revolutionizing data annotation with automated la- beling technique.VF AST Transactions on Software Engi- neering, 12(2):24–30, 2024

    Farheen Qazi, Muhammad Naseem, Sonish Aslam, Zainab Attaria, Muhammad Ali Jan, and Syed Salman Junaid. An- novate: Revolutionizing data annotation with automated la- beling technique.VF AST Transactions on Software Engi- neering, 12(2):24–30, 2024

    2024

  33. [41]

    3dgs-avatar: Animatable avatars via deformable 3d gaussian splatting

    Zhiyin Qian, Shaofei Wang, Marko Mihajlovic, Andreas Geiger, and Siyu Tang. 3dgs-avatar: Animatable avatars via deformable 3d gaussian splatting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5020–5030, 2024

    2024

  34. [42]

    Civilian american and european surface anthropometry resource (caesar), final re- port

    Kathleen M Robinette, Sherri Blackwell, Hein Daanen, Mark Boehmer, and Scott Fleming. Civilian american and european surface anthropometry resource (caesar), final re- port. volume 1. summary.(No Title), 2002

    2002

  35. [43]

    Text-driven mo- tion generation: Overview, challenges and directions, 2025

    Ali Rida Sahili, Najett Neji, and Hedi Tabia. Text-driven mo- tion generation: Overview, challenges and directions, 2025

    2025

  36. [44]

    Human motion diffusion as a generative prior

    Yonatan Shafir, Guy Tevet, Roy Kapon, and Amit H Bermano. Human motion diffusion as a generative prior. arXiv preprint arXiv:2303.01418, 2023

    2023 arXiv

  37. [45]

    Leonid Sigal, Alexandru O Balan, and Michael J Black. Hu- maneva: Synchronized video and motion capture dataset and baseline algorithm for evaluation of articulated human mo- tion.International journal of computer vision, 87(1):4–27, 2010

    2010

  38. [46]

    Seitz, and Richard Szeliski.Photo Tourism: Exploring Photo Collections in 3D

    Noah Snavely, Steven M. Seitz, and Richard Szeliski.Photo Tourism: Exploring Photo Collections in 3D. Association for Computing Machinery, New York, NY , USA, 1 edition, 2023

    2023

  39. [47]

    Human motion diffu- sion model

    Guy Tevet, Sigal Raab, Brian Gordon, Yoni Shafir, Daniel Cohen-or, and Amit Haim Bermano. Human motion diffu- sion model. InThe Eleventh International Conference on Learning Representations, 2023

    2023

  40. [48]

    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. In2017 British Machine Vision Conference (BMVC), 2017

    2017

  41. [49]

    Cmu graphics lab motion cap- ture database

    Carnegie Mellon University. Cmu graphics lab motion cap- ture database

  42. [50]

    Learning from synthetic humans

    Gul Varol, Javier Romero, Xavier Martin, Naureen Mah- mood, Michael J Black, Ivan Laptev, and Cordelia Schmid. Learning from synthetic humans. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 109–117, 2017

    2017

  43. [51]

    Recovering ac- curate 3d human pose in the wild using imus and a moving camera

    Timo V on Marcard, Roberto Henschel, Michael J Black, Bodo Rosenhahn, and Gerard Pons-Moll. Recovering ac- curate 3d human pose in the wild using imus and a moving camera. InProceedings of the European conference on com- puter vision (ECCV), pages 601–617, 2018

    2018

  44. [52]

    Ghum & ghuml: Generative 3d human shape and articulated pose models

    Hongyi Xu, Eduard Gabriel Bazavan, Andrei Zanfir, William T Freeman, Rahul Sukthankar, and Cristian Smin- chisescu. Ghum & ghuml: Generative 3d human shape and articulated pose models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6...

    2020

  45. [53]

    Athletepose3d: A benchmark dataset for 3d human pose estimation and kinematic validation in ath- letic movements.arXiv preprint arXiv:2503.07499, 2025

    Calvin Yeung, Tomohiro Suzuki, Ryota Tanaka, Zhuoer Yin, and Keisuke Fujii. Athletepose3d: A benchmark dataset for 3d human pose estimation and kinematic validation in ath- letic movements.arXiv preprint arXiv:2503.07499, 2025

    2025 arXiv

  46. [54]

    Monst3r: A simple approach for estimat- ing geometry in the presence of motion.arXiv preprint arxiv:2410.03825, 2024

    Junyi Zhang, Charles Herrmann, Junhwa Hur, Varun Jam- pani, Trevor Darrell, Forrester Cole, Deqing Sun, and Ming- Hsuan Yang. Monst3r: A simple approach for estimat- ing geometry in the presence of motion.arXiv preprint arxiv:2410.03825, 2024

    arXiv

  47. [55]

    Ic-light github page, 2024

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Ic-light github page, 2024

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.