arxiv: 2605.02784 · v1 · submitted 2026-05-04 · 💻 cs.CV

Recognition: 3 theorem links

· Lean Theorem

HumanSplatHMR: Closing the Loop Between Human Mesh Recovery and Gaussian Splatting Avatar

Yeheng Zong , Pou-Chun Kung , Yike Pan , Seth Isaacson , Yizhou Chen , Ram Vasudevan , Katherine A. Skinner

Authors on Pith no claims yet

Pith reviewed 2026-05-08 18:16 UTC · model grok-4.3

classification 💻 cs.CV

keywords human pose estimationgaussian splattingavatar reconstructiondifferentiable renderingnovel view synthesisjoint optimizationmesh recovery

0 comments

The pith

Joint optimization refines 3D human poses by routing rendering losses back through a Gaussian splatting avatar.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Existing methods for building human avatars from video either produce unreliable 3D geometry from single-view estimators or keep pose fixed while learning appearance, which limits generalization to new views and poses. HumanSplatHMR instead treats the avatar as a differentiable renderer and sends photometric, segmentation, and depth errors directly back to the pose parameters. This closes an optimization loop that starts from off-the-shelf mesh estimates and improves both the recovered global 3D trajectory and the final novel-view renderings. A reader would care because the approach works on ordinary video without motion-capture hardware or separate offline pose cleanup.

Core claim

HumanSplatHMR is a joint optimization framework that simultaneously refines 3D human poses and learns a Gaussian splatting avatar by backpropagating photometric, segmentation, and depth losses through a differentiable renderer to the pose parameters and global position. The method begins with human mesh estimates from a standard pose estimator rather than mocap or offline refinement, then uses the rendering losses to correct pose drift over time, yielding improved alignment and higher-fidelity novel-view and novel-pose synthesis.

What carries the argument

The differentiable renderer that propagates image-level losses from the Gaussian splatting avatar back to the underlying SMPL-style pose parameters, allowing the avatar reconstruction to serve as a supervisory signal for geometric pose refinement.

Load-bearing premise

Backpropagating photometric, segmentation, and depth losses through the renderer will stably improve global 3D poses without introducing instability or local minima that degrade the final avatar.

What would settle it

A set of in-the-wild videos in which the jointly optimized poses produce higher error against ground-truth motion capture or yield visibly worse novel-view renderings than the same avatar trained with fixed initial poses.

Figures

Figures reproduced from arXiv: 2605.02784 by Katherine A. Skinner, Pou-Chun Kung, Ram Vasudevan, Seth Isaacson, Yeheng Zong, Yike Pan, Yizhou Chen.

**Figure 1.** Figure 1: HumanSplatHMR advances both SMPL estimation and view at source ↗

**Figure 2.** Figure 2: Method Overview. HumanSplatHMR takes SMPL estimation as input and combines SMPL with Gaussians with the proposed view at source ↗

**Figure 3.** Figure 3: Raw SMPL estimation and photometric optimized view at source ↗

**Figure 4.** Figure 4: Cloth-Aware Mesh-Embedded Loss (CAMEL) illustration. CAMEL loosely couples the SMPL mesh with the Gaussian rep view at source ↗

**Figure 5.** Figure 5: SMPL estimation comparison on NeuMan dataset. HumanSplatHMR shows better SMPL refinement results than GART. view at source ↗

**Figure 6.** Figure 6: Novel view rendering evaluation. HumanSplatHMR shows better rendering quality compared to GART [ view at source ↗

read the original abstract

Accurately recovering human pose and appearance from video is an essential component of scene reconstruction, with applications to motion capture, motion prediction, virtual reality, and digital twinning. Despite significant interest in building realistic human avatars from video, this paper demonstrates that existing methods do not accurately recover the 3D geometry of humans. ViT-based approaches are not consistently reliable and can overfit to 2D views, while NeRF- and Gaussian Splatting-based avatars treat pose and appearance separately, limiting rendering generalization to new poses. To resolve these shortcomings, this paper proposes HumanSplatHMR, a joint optimization framework that refines 3D human poses while simultaneously learning a high-fidelity avatar for novel-view and novel-pose synthesis. Our key insight is to close the loop between geometric pose estimation and differentiable rendering. Unlike prior human avatar methods that rely on accurate human pose obtained through motion capture systems or offline refinement, which are impractical in in-the-wild scenarios, our approach uses only human mesh estimates from a state-of-the-art human pose estimator to better reflect real-world conditions. Therefore, instead of using the human pose only as a deformation prior, HumanSplatHMR backpropagates photometric, segmentation, and depth losses through a differentiable renderer to the pose parameters and global position. This coupling refines the global 3D pose over time, improving accuracy and alignment while producing better renderings from novel views. Experiments show consistent improvements over pose recovery baselines that omit image-level refinement and avatar baselines that decouple pose estimation from avatar reconstruction.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper couples pose refinement to Gaussian splatting via backpropagated rendering losses, which is a direct fix for the usual decoupling but leaves the gradient stability question open.

read the letter

The core contribution is backpropagating photometric, segmentation, and depth losses from the rendered avatar straight to the SMPL pose parameters and root translation. This lets the initial ViT-based mesh estimates get refined during avatar training instead of staying fixed or being handled in a separate stage. It targets the practical case where you only have ordinary video and no mocap or offline pose cleanup, which matches real-world constraints better than most prior avatar pipelines. The approach is straightforward on paper and builds cleanly on existing differentiable renderers and Gaussian deformation fields. What stands out is the explicit claim that this joint pass improves both pose accuracy and novel-pose rendering quality over decoupled baselines. The experiments are described as showing consistent gains, which is the kind of result that matters for downstream uses like VR or digital twins. The soft spot is the optimization path itself. Small pose shifts can reorder splats or flip visibility, which tends to produce noisy or vanishing gradients through the deformation field. The abstract gives no indication of gradient clipping, pose regularization, or staged schedules that would keep the joint training stable, so it is not obvious whether the reported improvements generalize or depend on lucky initializations. Without seeing the full ablations, error bars, and any failure cases, it is hard to judge how load-bearing that assumption is. This is worth a look for anyone building video-driven human avatars or testing joint geometry-appearance optimization. A reader who already works with Gaussian splatting or HMR baselines will get the most out of the specific loss coupling and the in-the-wild starting point. It should go to peer review so the community can check whether the gradients actually behave and whether the quantitative edges hold under closer inspection.

Referee Report

2 major / 2 minor

Summary. The paper proposes HumanSplatHMR, a joint optimization framework for human mesh recovery and Gaussian Splatting avatar reconstruction from video. It starts from initial 3D poses provided by a ViT-based estimator and refines them by backpropagating photometric, segmentation, and depth losses through a differentiable renderer to the pose parameters and global translation, while simultaneously learning the avatar for improved novel-view and novel-pose synthesis. The central claim is that this closed-loop coupling yields more accurate poses and higher-quality renderings than decoupled baselines.

Significance. If the joint optimization reliably converges and produces measurable gains, the approach would address a practical limitation in in-the-wild avatar creation by removing the need for motion-capture or offline pose refinement. It could improve generalization in applications such as VR and digital twinning, provided the reported improvements hold under rigorous evaluation.

major comments (2)

[Method (optimization loop)] The central claim that backpropagating photometric/segmentation/depth losses through the differentiable renderer will refine global 3D poses without instability rests on the assumption of sufficiently smooth and informative gradients from the pose-conditioned Gaussian deformation field. Small pose perturbations can induce abrupt splat reordering or visibility changes, producing noisy or vanishing gradients; the manuscript provides no description of gradient clipping, pose regularization, or multi-stage schedules that would mitigate this risk.
[Experiments] The abstract asserts 'consistent improvements' over pose-recovery and decoupled-avatar baselines, yet the provided text contains no quantitative tables, error bars, ablation studies on loss weights, or dataset-specific metrics. Without these, it is impossible to determine whether the gains are statistically significant or robust to the free parameters (photometric, segmentation, and depth loss weights).

minor comments (2)

[Method] Clarify the exact parameterization of the Gaussian deformation field (how SMPL pose parameters map to per-Gaussian rotations, positions, and opacities) and whether any additional regularization is applied to the root translation.
[Introduction] The abstract mentions 'human mesh estimates from a state-of-the-art human pose estimator' but does not name the specific model or its training data; this detail should be added for reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive feedback. We address each major comment point by point below and will revise the manuscript to incorporate the suggested clarifications and additional analyses.

read point-by-point responses

Referee: [Method (optimization loop)] The central claim that backpropagating photometric/segmentation/depth losses through the differentiable renderer will refine global 3D poses without instability rests on the assumption of sufficiently smooth and informative gradients from the pose-conditioned Gaussian deformation field. Small pose perturbations can induce abrupt splat reordering or visibility changes, producing noisy or vanishing gradients; the manuscript provides no description of gradient clipping, pose regularization, or multi-stage schedules that would mitigate this risk.

Authors: We agree that gradient stability is essential for reliable joint optimization and that the manuscript should explicitly address potential issues arising from splat reordering and visibility changes. The current version describes the overall backpropagation through the differentiable renderer but does not detail the stabilization mechanisms. In the revision we will add a new subsection under the optimization framework that specifies: (1) a multi-stage schedule that first optimizes avatar parameters with fixed poses before jointly refining poses, (2) an L2 regularization term on pose deltas to discourage large perturbations, and (3) gradient clipping applied to the pose and translation gradients. These additions will make the convergence behavior transparent and directly respond to the concern. revision: yes
Referee: [Experiments] The abstract asserts 'consistent improvements' over pose-recovery and decoupled-avatar baselines, yet the provided text contains no quantitative tables, error bars, ablation studies on loss weights, or dataset-specific metrics. Without these, it is impossible to determine whether the gains are statistically significant or robust to the free parameters (photometric, segmentation, and depth loss weights).

Authors: The referee is correct that the version provided for review lacked the full quantitative results. The manuscript text supplied to the referee contained only the high-level claim in the abstract and a brief statement in the experiments paragraph. We will expand the Experiments section with: (i) full tables reporting MPJPE, PA-MPJPE, PSNR, SSIM, and LPIPS on multiple datasets with standard deviations across three random seeds, (ii) ablation tables varying the photometric, segmentation, and depth loss weights, and (iii) statistical significance tests (paired t-tests) against the baselines. These additions will allow readers to assess both the magnitude and robustness of the reported gains. revision: yes

Circularity Check

0 steps flagged

No circularity: joint optimization uses external losses and initial poses from independent estimator

full rationale

The paper's core claim is a joint optimization that starts from initial human mesh estimates produced by an external state-of-the-art pose estimator and then back-propagates standard photometric, segmentation, and depth losses (derived directly from input video frames) through a differentiable renderer to refine pose parameters and learn the Gaussian avatar. This is a conventional end-to-end training loop with no reduction of the output to a self-defined quantity, no fitted parameter renamed as a prediction, and no load-bearing self-citation or ansatz imported from prior author work. The derivation chain remains self-contained against external video data and does not collapse to its inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that differentiable rendering can provide useful gradients for pose refinement and that initial mesh estimates are sufficiently close to allow convergence.

free parameters (1)

loss weights for photometric, segmentation, and depth terms
These balancing hyperparameters are typically fitted or chosen by hand in such optimization frameworks.

axioms (1)

domain assumption Differentiable renderer produces accurate gradients for pose parameters
Invoked when backpropagating losses to refine global 3D pose.

pith-pipeline@v0.9.0 · 5607 in / 1234 out tokens · 51048 ms · 2026-05-08T18:16:17.541804+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith.Cost / Foundation.AlphaCoordinateFixation J_uniquely_calibrated_via_higher_derivative unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

HumanSplatHMR backpropagates photometric, segmentation, and depth losses through a differentiable renderer to the pose parameters and global position.
IndisputableMonolith.Foundation.BranchSelection branch_selection unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

L = L_color + λ_depth L_depth + λ_CAMEL L_CAMEL (with sub-weights λ1..λ4 inside CAMEL)
IndisputableMonolith.Cost.FunctionalEquation washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Loss terms include log(s_t/s_n) − log τ flatness regularization and Huber penalty on Gaussian-to-depth distances.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

37 extracted references · 3 canonical work pages · 1 internal anchor

[1]

Keep it smpl: Automatic estimation of 3d human pose and shape from a single image

Federica Bogo, Angjoo Kanazawa, Christoph Lassner, Peter Gehler, Javier Romero, and Michael J Black. Keep it smpl: Automatic estimation of 3d human pose and shape from a single image. InEuropean conference on computer vision, pages 561–578. Springer, 2016. 2, 4, 6

2016
[2]

Meva: A large-scale multiview, multimodal video dataset for activity detection

Kellie Corona, Katie Osterdahl, Roderic Collins, and An- thony Hoogs. Meva: A large-scale multiview, multimodal video dataset for activity detection. InProceedings of the IEEE/CVF winter conference on applications of computer vision, pages 1060–1068, 2021. 2

2021
[3]

Tokenhmr: Advancing human mesh re- covery with a tokenized pose representation

Sai Kumar Dwivedi, Yu Sun, Priyanka Patel, Yao Feng, and Michael J Black. Tokenhmr: Advancing human mesh re- covery with a tokenized pose representation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1323–1333, 2024. 2

2024
[4]

Humans in 4d: Re- constructing and tracking humans with transformers

Shubham Goel, Georgios Pavlakos, Jathushan Rajasegaran, Angjoo Kanazawa, and Jitendra Malik. Humans in 4d: Re- constructing and tracking humans with transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 14783–14794, 2023. 1, 2, 4, 6, 8

2023
[5]

Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering

Antoine Gu ´edon and Vincent Lepetit. Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5354–5363, 2024. 4

2024
[6]

Gauhuman: Articu- lated gaussian splatting from monocular human videos

Shoukang Hu, Tao Hu, and Ziwei Liu. Gauhuman: Articu- lated gaussian splatting from monocular human videos. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 20418–20431, 2024. 2, 3, 4, 7

2024
[7]

2d gaussian splatting for geometrically ac- curate radiance fields

Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically ac- curate radiance fields. InACM SIGGRAPH 2024 conference papers, pages 1–11, 2024. 4

2024
[8]

Robust estimation of a location parameter

Peter J Huber. Robust estimation of a location parameter. In Breakthroughs in statistics: Methodology and distribution, pages 492–518. Springer, 1992. 5

1992
[9]

Catalin Ionescu, Dragos Papava, Vlad Olaru, and Cristian Sminchisescu. Human3. 6m: Large scale datasets and pre- dictive methods for 3d human sensing in natural environ- ments.IEEE transactions on pattern analysis and machine intelligence, 36(7):1325–1339, 2013. 6

2013
[10]

Fast automatic skinning transformations

Alec Jacobson, Ilya Baran, Ladislav Kavan, Jovan Popovi ´c, and Olga Sorkine. Fast automatic skinning transformations. ACM Transactions on Graphics (ToG), 31(4):1–10, 2012. 2

2012
[11]

In- stantavatar: Learning avatars from monocular video in 60 seconds

Tianjian Jiang, Xu Chen, Jie Song, and Otmar Hilliges. In- stantavatar: Learning avatars from monocular video in 60 seconds. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16922– 16932, 2023. 2, 6

2023
[12]

Neuman: Neural human radiance field from a single video

Wei Jiang, Kwang Moo Yi, Golnoosh Samei, Oncel Tuzel, and Anurag Ranjan. Neuman: Neural human radiance field from a single video. InEuropean Conference on Computer Vision, pages 402–418. Springer, 2022. 6

2022
[13]

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. 2, 6

2018
[14]

Learning 3d human dynamics from video

Angjoo Kanazawa, Jason Y Zhang, Panna Felsen, and Jiten- dra Malik. Learning 3d human dynamics from video. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 5614–5623, 2019. 2

2019
[15]

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

Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Trans. Graph., 42(4):139–1,
[16]

Autosplat: Constrained gaussian splatting for autonomous driving scene reconstruction

Mustafa Khan, Hamidreza Fazlali, Dhruv Sharma, Tongtong Cao, Dongfeng Bai, Yuan Ren, and Bingbing Liu. Autosplat: Constrained gaussian splatting for autonomous driving scene reconstruction. In2025 IEEE International Conference on Robotics and Automation (ICRA), pages 8315–8321. IEEE,
[17]

Vibe: Video inference for human body pose and shape estimation

Muhammed Kocabas, Nikos Athanasiou, and Michael J Black. Vibe: Video inference for human body pose and shape estimation. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 5253–5263, 2020. 2

2020
[18]

Hugs: Human gaussian splats

Muhammed Kocabas, Jen-Hao Rick Chang, James Gabriel, Oncel Tuzel, and Anurag Ranjan. Hugs: Human gaussian splats. InProceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, pages 505–515, 2024. 2, 3, 4, 7

2024
[19]

Learning to reconstruct 3d human pose and shape via model-fitting in the loop

Nikos Kolotouros, Georgios Pavlakos, Michael J Black, and Kostas Daniilidis. Learning to reconstruct 3d human pose and shape via model-fitting in the loop. InProceedings of the IEEE/CVF international conference on computer vision, pages 2252–2261, 2019. 2

2019
[20]

Sad-gs: Shape-aligned depth- supervised gaussian splatting

Pou-Chun Kung, Seth Isaacson, Ram Vasudevan, and Katherine A Skinner. Sad-gs: Shape-aligned depth- supervised gaussian splatting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2842–2851, 2024. 4

2024
[21]

Gart: Gaussian articulated template mod- els

Jiahui Lei, Yufu Wang, Georgios Pavlakos, Lingjie Liu, and Kostas Daniilidis. Gart: Gaussian articulated template mod- els. InProceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, pages 19876–19887,
[22]

Splatface: Gaus- sian splat face reconstruction leveraging an optimizable sur- face

Jiahao Luo, Jing Liu, and James Davis. Splatface: Gaus- sian splat face reconstruction leveraging an optimizable sur- face. In2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 774–783. IEEE, 2025. 4

2025
[23]

Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021

Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021. 2

2021
[24]

ihuman: Instant animatable digital humans from monocular videos

Pramish Paudel, Anubhav Khanal, Danda Pani Paudel, Jyoti Tandukar, and Ajad Chhatkuli. ihuman: Instant animatable digital humans from monocular videos. InEuropean Con- ference on Computer Vision, pages 304–323. Springer, 2024. 2, 3, 7

2024
[25]

Neural body: 9 Implicit neural representations with structured latent codes for novel view synthesis of dynamic humans

Sida Peng, Yuanqing Zhang, Yinghao Xu, Qianqian Wang, Qing Shuai, Hujun Bao, and Xiaowei Zhou. Neural body: 9 Implicit neural representations with structured latent codes for novel view synthesis of dynamic humans. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9054–9063, 2021. 2

2021
[26]

Unidepthv2: Universal monocular metric depth estimation made simpler

Luigi Piccinelli, Christos Sakaridis, Yung-Hsu Yang, Mat- tia Segu, Siyuan Li, Wim Abbeloos, and Luc Van Gool. Unidepthv2: Universal monocular metric depth estimation made simpler.arXiv preprint arXiv:2502.20110, 2025. 5, 6

work page arXiv 2025
[27]

SAM 2: Segment Anything in Images and Videos

Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman R¨adle, Chloe Rolland, Laura Gustafson, et al. Sam 2: Segment anything in images and videos.arXiv preprint arXiv:2408.00714, 2024. 5, 6

work page internal anchor Pith review arXiv 2024
[28]

Splattingavatar: Realistic real-time human avatars with mesh-embedded gaussian splatting

Zhijing Shao, Zhaolong Wang, Zhuang Li, Duotun Wang, Xiangru Lin, Yu Zhang, Mingming Fan, and Zeyu Wang. Splattingavatar: Realistic real-time human avatars with mesh-embedded gaussian splatting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1606–1616, 2024. 2, 3, 4, 7

2024
[29]

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. 6

2018
[30]

Bovik, H.R

Zhou Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli. Image quality assessment: from error visibility to structural similarity.IEEE Transactions on Image Processing, 13(4): 600–612, 2004. 5

2004
[31]

Gomavatar: Efficient an- imatable human modeling from monocular video using gaussians-on-mesh

Jing Wen, Xiaoming Zhao, Zhongzheng Ren, Alexander G Schwing, and Shenlong Wang. Gomavatar: Efficient an- imatable human modeling from monocular video using gaussians-on-mesh. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 2059–2069, 2024. 2, 3, 4, 7

2059
[32]

Hu- mannerf: Free-viewpoint rendering of moving people from monocular video

Chung-Yi Weng, Brian Curless, Pratul P Srinivasan, Jonathan T Barron, and Ira Kemelmacher-Shlizerman. Hu- mannerf: Free-viewpoint rendering of moving people from monocular video. InProceedings of the IEEE/CVF con- ference on computer vision and pattern Recognition, pages 16210–16220, 2022. 2

2022
[33]

Reconstructing humans with a biome- chanically accurate skeleton

Yan Xia, Xiaowei Zhou, Etienne V ouga, Qixing Huang, and Georgios Pavlakos. Reconstructing humans with a biome- chanically accurate skeleton. InProceedings of the Com- puter Vision and Pattern Recognition Conference, pages 5355–5365, 2025. 2

2025
[34]

Animatable nerf dynamic detail enhancement based on residual deformation field with progressive training

Menglei Yang, Yuhang Han, Shenhao Zhang, and Xiaohui Zhang. Animatable nerf dynamic detail enhancement based on residual deformation field with progressive training. In 2025 5th International Conference on Computer Graphics, Image and Virtualization (ICCGIV), pages 161–165. IEEE,

2025
[35]

Learnable smplify: A neural solution for optimization-free human pose inverse kinematics.arXiv preprint arXiv:2508.13562, 2025

Yuchen Yang, Linfeng Dong, Wei Wang, Zhihang Zhong, and Xiao Sun. Learnable smplify: A neural solution for optimization-free human pose inverse kinematics.arXiv preprint arXiv:2508.13562, 2025. 2

work page arXiv 2025
[36]

Decoupling human and camera motion from videos in the wild

Vickie Ye, Georgios Pavlakos, Jitendra Malik, and Angjoo Kanazawa. Decoupling human and camera motion from videos in the wild. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 21222–21232, 2023. 2

2023
[37]

Film and television animation pro- duction technology based on expression transfer and virtual digital human.Scalable Computing: Practice and Experi- ence, 25(6):5560–5567, 2024

Ning Zhang and Belei Pu. Film and television animation pro- duction technology based on expression transfer and virtual digital human.Scalable Computing: Practice and Experi- ence, 25(6):5560–5567, 2024. 2 10

2024