Pith. sign in

REVIEW 2 major objections 6 minor 60 references

Learning Efficient and Generalizable Human Representation with Human Gaussian Model

T0 review · 2 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a dual-layer graph linking per-frame 3D Gaussians to SMPL mesh vertices yields a generalizable, animatable human avatar from a video in a single feed-forward pass, the first such inference-time animatable Gaussian…

desk verdict Solid incremental architecture for feed-forward animatable avatars, but the evidence is thin and the single-view alignment is under-specified. read the letter →

arxiv 2507.18758 v1 pith:L5UP4D6P submitted 2025-07-24 cs.CV

Yifan Liu , Shengjun Zhang , Chensheng Dai , Yang Chen , Hao Liu , Chen Li , Yueqi Duan This is my paper · ORCID
classification cs.CV
keywords animatablehumanavatar3DGaussiansplattingfeed-forwardreconstructiontemporalaggregationSMPLgraphattentionnovelposeanimationgeneralizablemodel
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

Traditional animatable avatars require minutes to hours of per-person optimization, while recent feed-forward Gaussian predictors handle one frame at a time and cannot be reposed. The paper claims that a video's frames contain useful complementary evidence: attach every predicted 3D Gaussian to the nearest vertex of a shared human body mesh, aggregate information across frames at each vertex, and let neighboring vertices exchange messages, and the output is an SMPL-aligned avatar that can be driven into new poses in a single inference pass. On held-out multi-view scans the reported quality is higher than both per-frame feed-forward baselines and per-instance optimization methods, with new-view rendering above 120 frames per second. This would make poseable avatars from ordinary video practical for interactive applications.

What carries the argument

The central object is the Human Gaussian Graph (HGG), a dual-layer graph where the first-layer nodes are the 3D Gaussians predicted for every frame of the video and the second-layer nodes are the vertices of an SMPL mesh, a parametric human body model. Edges are defined in two ways: each Gaussian is connected to the SMPL vertex closest to it in the posed frame, and each mesh vertex is connected to its topological neighbors on the mesh. Information flows through two transformer operations: an intra-node operation in which a learnable query at each vertex attends to all Gaussians attached to that vertex across timesteps, and an inter-node operation in which queries at neighboring vertices exchange messages; the operations are stacked to widen the receptive field. The updated vertex queries are then used to refine the Gaussians of a chosen frame, yielding SMPL-aligned Gaussians that are reposed via linear blend skinning and rendered with Gaussian rasterization. This graph is what carries the argument: it converts temporal grouping into spatial grouping, so that cross-frame evidence is aggregated without expensive all-pairs attention between Gaussians.

What would settle it

Render a synthetic person with known ground-truth SMPL from motion capture, run the method with the true SMPL pose and with progressively perturbed SMPL estimates, and compare novel-pose PSNR; if perturbing the mesh by a few centimeters does not degrade output more than the per-frame baseline, the nearest-vertex edge construction is not doing the claimed work. Alternatively, test on a subject in loose flowing clothing: if temporal aggregation visibly merges cloth motion with body-surface Gaussians, the assignment assumption fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that temporal information in a video can be folded into a poseable avatar without any optimization by using a Human Gaussian Graph with two layers: Gaussian primitives from all frames and SMPL mesh vertices fixed across the timeline. Each Gaussian is attached to its nearest posed SMPL vertex; the graph then runs a cross-attention intra-node operation that lets a per-vertex query collect all Gaussians assigned to that vertex over the video, and an inter-node operation that passes messages between neighboring mesh vertices. The resulting vertex queries, applied to one frame's Gaussians, upgrade them into SMPL-aligned Gaussians that can be driven by new poses through linear blend skinning and rendered by Gaussian splatting. The paper reports that this yields the first animatable Gaussian avatars produced within inference time, surpassing per-frame feed-forward methods and optimization-based baselines on novel view synthesis and novel pose animation, with inference at roughly 0.9 seconds per frame for multi-view input and 9.7 seconds for monocular input.

Load-bearing premise

The method assumes each predicted 3D Gaussian sits close enough to the estimated surface of a parametric human body mesh that attaching it to the nearest mesh vertex groups genuinely related content across frames; if the estimated body shape or pose is wrong, or the Gaussian belongs to loose clothing or hair, the edges misassign and temporal aggregation mixes unrelated evidence.

Editorial extensions

If this is right

  • From any short video of a person, a poseable avatar can be produced in a single forward pass, without per-person optimization, at roughly 0.9 seconds per frame for multi-view and 9.7 seconds for monocular input.
  • Because the avatar is anchored to a body mesh, it can be reposed by any pose parameter and rendered at over 120 FPS, enabling real-time interactive use.
  • The graph is a separate module on top of the single-frame Gaussian predictor; replacing that predictor with a stronger one improves results, so the method compounds with progress in per-frame reconstruction.
  • Novel-pose quality is limited by the accuracy of the estimated body pose and shape; correcting those estimates should narrow the reported novel-view/novel-pose gap.

Reading between the lines

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

  • The two-layer graph construction is not specific to humans; any deformable category with a parametric template or rigged mesh could adopt the same cross-frame aggregation, so the idea may transfer to hands, faces, or animals.
  • A likely improvement path is to replace hard nearest-vertex assignment with a learned soft assignment or additional per-vertex features, which would let the graph handle hair and loose clothing instead of leaving those regions to rely on nearby body vertices.
  • The bottleneck for monocular input is the quality of the generated pseudo-views; if a publicly available human-specific multi-view diffusion model emerges, the monocular and multi-view numbers should converge.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 6 minor

Summary. The paper proposes Human Gaussian Graph (HGG), a graph-based module that links per-frame Gaussian primitives predicted by a frozen feed-forward model (LGM) to shared SMPL mesh vertices. Via intra-node attention across frames and inter-node attention across mesh neighbors, the module produces SMPL-aligned Gaussians that can be re-posed for novel pose animation. The method operates in a single forward pass, supporting both multi-view and monocular video input. Experiments on MvHumanNet report state-of-the-art results in novel view synthesis and novel pose animation, with ablations showing contributions of each component.

Significance. If the results hold, the method provides a practical route to generalizable animatable avatars without per-instance optimization, which is an important step for real-time applications. The dual-layer graph formulation is a clean way to aggregate temporal information while maintaining mesh alignment, and the ablation study carefully isolates each module's contribution. The efficiency analysis in the supplementary is a useful contribution. The major caveats are the unaddressed coordinate-frame alignment in the monocular pipeline and the lack of statistical rigor in the evaluation, both of which need to be resolved before the quantitative claims are fully convincing.

major comments (2)
  1. [Section 3.2, Eq. (6)] The graph construction as written assumes that the Gaussian centers µ_t^m and the posed SMPL vertices v_t^i lie in the same Euclidean coordinate frame. In the multi-view setting this is plausible because MvHumanNet provides calibrated cameras and SMPL in a common world frame. In the single-view setting, however, the pipeline first synthesizes multi-view images with a fine-tuned Wonder3D and then runs LGM, whose Gaussian predictions live in the canonical coordinate frame induced by those generated views, while the SMPL pose p_t is estimated from the original monocular image. The paper never describes a canonicalization or root-alignment step between these two coordinate systems before Eq. (6) is applied. If the two frames are misaligned, the nearest-vertex assignment groups unrelated content and the intra-node operation (Eqs. 8–9) mixes Gaussians from different body parts; the ablation attributes 2.1 dB of the novel-pose gain to the intra-node operation, so this is load-bearing. Please specify the alignment procedure or add an explicit coordinate normalization step.
  2. [Section 4.2, Table 1] All quantitative comparisons are reported on a single set of 10 held-out MvHumanNet scans, with no error bars, per-scan variance, or cross-seed statistics. Given the small evaluation set, the claimed improvements (e.g., 1.6 dB over ExAvatar for single-view novel view, 2.3 dB for novel pose) may not be statistically reliable. Please report per-scan results or error bars, and state the number of random seeds used for the evaluation split selection.
minor comments (6)
  1. [Section 3.2] There are several typos: 'fisrt' should be 'first' and 'closet' should be 'closest' in the description of Eq. (6).
  2. [Section 3.3, Eq. (14)] B_n(V) is defined as the 'neighbor Gaussian group' but actually contains mesh vertices; rename it to 'neighbor vertex group' to avoid confusion.
  3. [Section 3.2, Eq. (7)] The threshold d0 for mesh-level edges is not specified in the implementation details; please provide its value or a sensitivity study.
  4. [Algorithm 1] The input line lists 'SMPL poses {p_t}_{i=1}^{N}' but the index should run t=1,...,T; please correct the notation.
  5. [Supplementary Section 7.3] IDOL and AniGS are mentioned without citations or clear definitions; please add references or a brief description.
  6. [Figure 1(b)] The 'Performance comparison' panel is referenced but the axes and units are not described in the caption; please clarify what is plotted.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the graph-construction and aggregation steps are defined by the paper's own equations, trained on a held-out split, and all supporting components are independently specified.

full rationale

The paper's central claim is that a Human Gaussian Graph, with per-frame predicted Gaussians as first-layer nodes and SMPL vertices as second-layer nodes, aggregates temporal information to produce SMPL-aligned, animatable Gaussians. The derivation chain is self-contained: per-frame Gaussians are obtained from a fixed, frozen LGM backbone (Eq. 4), the graph edges are defined geometrically by nearest-vertex assignment in Euclidean space (Eq. 6) and mesh adjacency (Eq. 7), and the intra-node / inter-node operations are standard attention modules (Eqs. 8-16). The final Gaussian update is a residual attention operation (Eq. 17). There is no step in which an output quantity is defined in terms of the target quantity it is supposed to predict. The model is trained on a large MvHumanNet training split (2,000 subjects, 120,000 poses) with a loss comparing rendered images to ground-truth multi-view images, and evaluated on 10 held-out scans. Ablations in Table 2 attribute specific PSNR gains to the intra-node and inter-node operations, which is empirical evidence rather than a construction-level equivalence. The only self-citation, reference [53] by overlapping authors, appears in a related-work sentence listing generalizable Gaussian methods; it is not used to justify any architectural choice, uniqueness claim, or result, and the HGG construction is fully described locally. The skeptic's coordinate-frame concern for the single-view setting is a potential correctness/alignment issue, not circularity: if LGM Gaussians and SMPL vertices were not in a common frame, Eq. 6 would be geometrically ill-founded and performance would degrade, but that would not make the prediction equivalent to its inputs. Similarly, the limitations noted in Section 5 and Supplementary Section 7.2 (Wonder3D failures, inaccurate SMPL parameters) are acknowledged error sources, not circular dependencies. Overall, no load-bearing reduction to fitted data or self-citation chain is present, so the appropriate circularity score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central result is empirical and rests on a stack of pretrained and borrowed components: SMPL and LBS for articulation, 3D Gaussian Splatting for rendering, a frozen LGM for per-frame Gaussian initialization, and a fine-tuned Wonder3D for monocular multi-view synthesis. None of these are derived in this paper. The graph architecture adds learned queries and attention operations trained on MvHumanNet. The main unstated choices are the mesh-neighbor threshold d0, the loss weights, and exact model dimensions, none of which are reported.

free parameters (4)
  • mesh neighbor threshold d0 = not reported
    Defines mesh-level edges in Eq. 7 and controls the inter-node receptive field; no value or ablation is given.
  • number of stacked graph blocks L = 6
    Ablation Table 2b shows PSNR grows with L; the final architecture uses 6 blocks, so the claim depends on this choice.
  • loss weights alpha1 and alpha2 = not reported
    The loss in Section 3.4 mixes MSE, LPIPS, and alpha MSE; weights are not specified, preventing exact reproduction.
  • per-vertex learnable queries q_n = learned on MvHumanNet
    These queries are the information collectors and human prior; their fitted values come from the training set, and the paper shows they add 0.57 dB on their own.
assumptions (5)
  • domain assumption SMPL is a valid shared body model: the same canonical vertices correspond across frames and LBS in Eq. 1 reposes them.
    The whole graph reorganizes Gaussians around a fixed SMPL vertex set; if per-frame SMPL fits are inconsistent or inaccurate, edges and reposing are wrong.
  • domain assumption 3D Gaussian Splatting (Eqs. 2 and 3) is an accurate differentiable renderer for these Gaussians.
    All supervision and quantitative metrics flow through 3DGS rasterization; the paper does not analyze splatting artifacts.
  • domain assumption The frozen LGM produces useful per-frame Gaussian bases (Eq. 4).
    Graph operations only refine LGM outputs; the supplementary fail cases show LGM and Wonder3D failures directly cause corrupted avatars.
  • domain assumption For monocular input, fine-tuned Wonder3D generates consistent multi-views.
    The monocular pipeline synthesizes four views before Gaussian prediction; the authors acknowledge a 3.4 dB monocular gap and document camera-consistency failures.
  • domain assumption Nearest-vertex Euclidean assignment (Eq. 6) links each Gaussian to the correct body part.
    Loose clothing, hair, and inaccuracies in SMPL fits can place Gaussians far from the true vertex, misdirecting temporal aggregation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Efficient and Generalizable Human Representation with Human Gaussian Model." pith.science (2026). https://pith.science/paper/L5UP4D6P

@misc{pith2026250718758,
  author       = {Pith},
  title        = {Pith review of: Learning Efficient and Generalizable Human Representation with Human Gaussian Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L5UP4D6P}},
  note         = {Machine review of arXiv:2507.18758}
}
read the original abstract

Modeling animatable human avatars from videos is a long-standing and challenging problem. While conventional methods require per-instance optimization, recent feed-forward methods have been proposed to generate 3D Gaussians with a learnable network. However, these methods predict Gaussians for each frame independently, without fully capturing the relations of Gaussians from different timestamps. To address this, we propose Human Gaussian Graph to model the connection between predicted Gaussians and human SMPL mesh, so that we can leverage information from all frames to recover an animatable human representation. Specifically, the Human Gaussian Graph contains dual layers where Gaussians are the first layer nodes and mesh vertices serve as the second layer nodes. Based on this structure, we further propose the intra-node operation to aggregate various Gaussians connected to one mesh vertex, and inter-node operation to support message passing among mesh node neighbors. Experimental results on novel view synthesis and novel pose animation demonstrate the efficiency and generalization of our method.

Figures

Figures reproduced from arXiv: 2507.18758 by the authors.

Figure 1
Figure 1. Our method achieves state-of-the-art rendering quality while maintaining remarkable fast run-time performance. (a) Qualitative [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview. Given an input human video, our goal is to build high-fidelity animatable Gaussian representations within inference [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison of ours against GART [ [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Qualitative results for novel pose animation. We evaluated each module in our proposed Human Gaussian Graph and analyzed [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Fail cases. (a) Wonder3D fails to generate reasonable [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: More qualitative results on novel pose animation. The human avatar in various poses indicates the high-quality of our recon [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Visualization on Deepfashion and MVHumanNet. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: More qualitative results on novel view synthesis. The novel views in multiple directions indicate the high-quality and potential [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 40 canonical work pages

  1. [1]

    Single-image 3d human digitization with shape-guided diffusion

    Badour AlBahar, Shunsuke Saito, Hung-Yu Tseng, Changil Kim, Johannes Kopf, and Jia-Bin Huang. Single-image 3d human digitization with shape-guided diffusion. InSIG- GRAPH Asia 2023 Conference Papers, 2023. 2

    work page 2023

  2. [2]

    Video based reconstruction of 3d people models

    Thiemo Alldieck, Marcus Magnor, Weipeng Xu, Christian Theobalt, and Gerard Pons-Moll. Video based reconstruction of 3d people models. InCVPR, pages 8387–8397, 2018. 2

    work page 2018

  3. [3]

    Detailed human avatars from monocular video

    Thiemo Alldieck, Marcus Magnor, Weipeng Xu, Christian Theobalt, and Gerard Pons-Moll. Detailed human avatars from monocular video. In2018 International Conference on 3D Vision (3DV), pages 98–109. IEEE, 2018. 2

    work page 2018

  4. [4]

    Photorealistic monocular 3d reconstruction of humans wear- ing clothing

    Thiemo Alldieck, Mihai Zanfir, and Cristian Sminchisescu. Photorealistic monocular 3d reconstruction of humans wear- ing clothing. InCVPR, 2022. 1, 2

    work page 2022

  5. [5]

    pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction.arXiv preprint arXiv:2312.12337, 2023

    David Charatan, Sizhe Li, Andrea Tagliasacchi, and Vincent Sitzmann. pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction.arXiv preprint arXiv:2312.12337, 2023. 2

    arXiv 2023

  6. [6]

    Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images.arXiv preprint arXiv:2403.14627, 2024

    Yuedong Chen, Haofei Xu, Chuanxia Zheng, Bohan Zhuang, Marc Pollefeys, Andreas Geiger, Tat-Jen Cham, and Jianfei Cai. Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images.arXiv preprint arXiv:2403.14627, 2024. 2

    arXiv 2024

  7. [7]

    Objaverse: A universe of annotated 3d objects

    Matt Deitke, Dustin Schwenk, Jordi Salvador, Luca Weihs, Oscar Michel, Eli VanderBilt, Ludwig Schmidt, Kiana Ehsani, Aniruddha Kembhavi, and Ali Farhadi. Objaverse: A universe of annotated 3d objects. InCVPR, 2023. 1

    work page 2023

  8. [8]

    Humans in 4d: Re- constructing and tracking humans with transformers.arXiv preprint arXiv:2305.20091, 2023

    Shubham Goel, Georgios Pavlakos, Jathushan Rajasegaran, Angjoo Kanazawa, and Jitendra Malik. Humans in 4d: Re- constructing and tracking humans with transformers.arXiv preprint arXiv:2305.20091, 2023. 2

    arXiv 2023

Show all 60 references
  1. [9]

    Livecap: Real-time human performance capture from monocular video.ACM Transactions On Graphics (TOG), 38(2):1–17, 2019

    Marc Habermann, Weipeng Xu, Michael Zollhoefer, Ger- ard Pons-Moll, and Christian Theobalt. Livecap: Real-time human performance capture from monocular video.ACM Transactions On Graphics (TOG), 38(2):1–17, 2019. 2

    2019

  2. [10]

    High-fidelity 3d hu- man digitization from single 2k resolution images

    Sang-Hun Han, Min-Gyu Park, Ju Hong Yoon, Ju-Mi Kang, Young-Jae Park, and Hae-Gon Jeon. High-fidelity 3d hu- man digitization from single 2k resolution images. InCVPR,

  3. [11]

    Collomosse, Hailin Jin, and Stefano Soatto

    Tong He, John P. Collomosse, Hailin Jin, and Stefano Soatto. Geo-PIFu: Geometry and pixel aligned implicit functions for single-view human reconstruction. InNeurIPS, 2020. 2

    2020

  4. [12]

    Gaussianavatar: Towards realistic human avatar modeling from a single video via animatable 3d gaussians

    Liangxiao Hu, Hongwen Zhang, Yuxiang Zhang, Boyao Zhou, Boning Liu, Shengping Zhang, and Liqiang Nie. Gaussianavatar: Towards realistic human avatar modeling from a single video via animatable 3d gaussians. InCVPR,

  5. [13]

    Sherf: Generalizable human nerf from a single image

    Shoukang Hu, Fangzhou Hong, Liang Pan, Haiyi Mei, Lei Yang, and Ziwei Liu. Sherf: Generalizable human nerf from a single image. InICCV, 2023. 1

    2023

  6. [14]

    One-shot implicit animatable avatars with model- based priors

    Yangyi Huang, Hongwei Yi, Weiyang Liu, Haofan Wang, Boxi Wu, Wenxiao Wang, Binbin Lin, Debing Zhang, and Deng Cai. One-shot implicit animatable avatars with model- based priors. InICCV, 2023. 1

    2023

  7. [15]

    Splatarmor: Articulated gaussian splatting for animat- able humans from monocular rgb videos.arXiv preprint arXiv:2311.10812, 2023

    Rohit Jena, Ganesh Subramanian Iyer, Siddharth Choud- hary, Brandon Smith, Pratik Chaudhari, and James Gee. Splatarmor: Articulated gaussian splatting for animat- able humans from monocular rgb videos.arXiv preprint arXiv:2311.10812, 2023. 2

    2023 arXiv

  8. [16]

    Instant-nvr: Instant neural volumetric ren- dering for human-object interactions from monocular rgbd stream

    Yuheng Jiang, Kaixin Yao, Zhuo Su, Zhehao Shen, Haimin Luo, and Lan Xu. Instant-nvr: Instant neural volumetric ren- dering for human-object interactions from monocular rgbd stream. InCVPR, 2023. 2

    2023

  9. [17]

    3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics (TOG), 2023

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkuehler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics (TOG), 2023. 3

    2023

  10. [18]

    3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics, 42 (4), 2023

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics, 42 (4), 2023. 1, 2

    2023

  11. [19]

    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. InCVPR, pages 5253–5263, 2020. 2

    2020

  12. [20]

    Gart: Gaussian articulated template mod- els

    Jiahui Lei, Yufu Wang, Georgios Pavlakos, Lingjie Liu, and Kostas Daniilidis. Gart: Gaussian articulated template mod- els. InCVPR, 2024. 2, 5, 6, 7

    2024

  13. [21]

    Tianye Li, Timo Bolkart, Michael. J. Black, Hao Li, and Javier Romero. Learning a model of facial shape and ex- pression from 4D scans.ACM Transactions on Graphics, (Proc. SIGGRAPH Asia), 36(6):194:1–194:17, 2017. 2

    2017

  14. [22]

    Gva: Reconstructing vivid 3d gaussian avatars from monocular videos.arXiv preprint arXiv:2402.16607, 2024

    Xinqi Liu, Chenming Wu, Jialun Liu, Xing Liu, Chen Zhao, Haocheng Feng, Errui Ding, and Jingdong Wang. Gva: Reconstructing vivid 3d gaussian avatars from monocular videos.arXiv preprint arXiv:2402.16607, 2024. 2

    2024 arXiv

  15. [23]

    Wonder3d: Single image to 3d using cross-domain diffusion,

    Xiaoxiao Long, Yuan-Chen Guo, Cheng Lin, Yuan Liu, Zhiyang Dou, Lingjie Liu, Yuexin Ma, Song-Hai Zhang, Marc Habermann, Christian Theobalt, and Wenping Wang. Wonder3d: Single image to 3d using cross-domain diffusion,

  16. [24]

    Smpl: A skinned multi- person linear model.ACM transactions on graphics (TOG),

    Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, and Michael J Black. Smpl: A skinned multi- person linear model.ACM transactions on graphics (TOG),

  17. [25]

    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. InSeminal Graphics Papers: Pushing the Boundaries, Volume 2, pages 851–866. 2023. 2, 3

    2023

  18. [26]

    Occupancy networks: Learning 3d reconstruction in function space

    Lars Mescheder, Michael Oechsle, Michael Niemeyer, Se- bastian Nowozin, and Andreas Geiger. Occupancy networks: Learning 3d reconstruction in function space. InCVPR,

  19. [27]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis

    Ben Mildenhall, Pratul Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. InECCV, 2020. 1

    2020

  20. [28]

    Expressive whole-body 3d gaussian avatar

    Gyeongsik Moon, Takaaki Shiratori, and Shunsuke Saito. Expressive whole-body 3d gaussian avatar. InECCV, 2024. 2, 5, 6, 7, 8

    2024

  21. [29]

    Humansplat: Generalizable single-image human gaussian splatting with structure priors.arXiv preprint arXiv:2406.12459, 2024

    Panwang Pan, Zhuo Su, Chenguo Lin, Zhen Fan, Yongjie Zhang, Zeming Li, Tingting Shen, Yadong Mu, and Yebin Liu. Humansplat: Generalizable single-image human gaussian splatting with structure priors.arXiv preprint arXiv:2406.12459, 2024. 2, 8

    2024 arXiv

  22. [30]

    Deepsdf: Learning con- tinuous signed distance functions for shape representation

    Jeong Joon Park, Peter Florence, Julian Straub, Richard Newcombe, and Steven Lovegrove. Deepsdf: Learning con- tinuous signed distance functions for shape representation. InCVPR, 2019. 1

    2019

  23. [31]

    Expressive body capture: 3d hands, face, and body from a single image

    Georgios Pavlakos, Vasileios Choutas, Nima Ghorbani, Timo Bolkart, Ahmed AA Osman, Dimitrios Tzionas, and Michael J Black. Expressive body capture: 3d hands, face, and body from a single image. InCVPR, pages 10975– 10985, 2019. 2

    2019

  24. [32]

    Georgios Pavlakos, Vasileios Choutas, Nima Ghorbani, Timo Bolkart, Ahmed A. A. Osman, Dimitrios Tzionas, and Michael J. Black. Expressive body capture: 3D hands, face, and body from a single image. InCVPR, 2019. 2

    2019

  25. [33]

    Gaus- sianavatars: Photorealistic head avatars with rigged 3d gaus- sians

    Shenhan Qian, Tobias Kirschstein, Liam Schoneveld, Davide Davoli, Simon Giebenhain, and Matthias Nießner. Gaus- sianavatars: Photorealistic head avatars with rigged 3d gaus- sians. InCVPR, pages 20299–20309, 2024. 2

    2024

  26. [34]

    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. 2024. 2

    2024

  27. [35]

    Javier Romero, Dimitrios Tzionas, and Michael J. Black. Embodied hands: Modeling and capturing hands and bod- ies together.ACM Transactions on Graphics, (Proc. SIG- GRAPH Asia), 36(6), 2017. 2

    2017

  28. [36]

    Pifu: Pixel-aligned implicit function for high-resolution clothed human digitiza- tion

    Shunsuke Saito, Zeng Huang, Ryota Natsume, Shigeo Mor- ishima, Angjoo Kanazawa, and Hao Li. Pifu: Pixel-aligned implicit function for high-resolution clothed human digitiza- tion. InICCV, 2019. 1

    2019

  29. [37]

    PIFu: Pixel-aligned implicit function for high-resolution clothed human digitiza- tion

    Shunsuke Saito, Zeng Huang, Ryota Natsume, Shigeo Mor- ishima, Hao Li, and Angjoo Kanazawa. PIFu: Pixel-aligned implicit function for high-resolution clothed human digitiza- tion. InICCV, 2019. 2

    2019

  30. [38]

    Pifuhd: Multi-level pixel-aligned implicit function for high-resolution 3d human digitization

    Shunsuke Saito, Tomas Simon, Jason Saragih, and Hanbyul Joo. Pifuhd: Multi-level pixel-aligned implicit function for high-resolution 3d human digitization. InCVPR, 2020. 1

    2020

  31. [39]

    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. InCVPR, 2024. 2

    2024

  32. [40]

    Monocular, one-stage, regression of multiple 3d people

    Yu Sun, Qian Bao, Wu Liu, Yili Fu, Michael J Black, and Tao Mei. Monocular, one-stage, regression of multiple 3d people. InICCV, pages 11179–11188, 2021. 2

    2021

  33. [41]

    Trace: 5d temporal regression of avatars with dynamic cam- eras in 3d environments

    Yu Sun, Qian Bao, Wu Liu, Tao Mei, and Michael J Black. Trace: 5d temporal regression of avatars with dynamic cam- eras in 3d environments. InCVPR, pages 8856–8866, 2023. 2

    2023

  34. [42]

    Haha: Highly articulated gaussian human avatars with textured mesh prior.arXiv preprint arXiv:2404.01053,

    David Svitov, Pietro Morerio, Lourdes Agapito, and Alessio Del Bue. Haha: Highly articulated gaussian human avatars with textured mesh prior.arXiv preprint arXiv:2404.01053,

    arXiv

  35. [43]

    Lgm: Large multi-view gaussian model for high-resolution 3d content creation.ECCV, 2024

    Jiaxiang Tang, Zhaoxi Chen, Xiaokang Chen, Tengfei Wang, Gang Zeng, and Ziwei Liu. Lgm: Large multi-view gaussian model for high-resolution 3d content creation.ECCV, 2024. 2, 5, 6, 7, 8

    2024

  36. [44]

    Refit: Recurrent fitting network for 3d human recovery

    Yufu Wang and Kostas Daniilidis. Refit: Recurrent fitting network for 3d human recovery. InICCV, pages 14644– 14654, 2023. 2

    2023

  37. [45]

    Image quality assessment: from error visibility to structural similarity.IEEE Transactions on Image Process- ing (TIP), 2004

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity.IEEE Transactions on Image Process- ing (TIP), 2004. 6

    2004

  38. [46]

    Template-free single-view 3d human digitalization with diffusion-guided lrm.arXiv preprint arXiv:2401.12175, 2024

    Zhenzhen Weng, Jingyuan Liu, Hao Tan, Zhan Xu, Yang Zhou, Serena Yeung-Levy, and Jimei Yang. Template-free single-view 3d human digitalization with diffusion-guided lrm.arXiv preprint arXiv:2401.12175, 2024. 2

    2024 arXiv

  39. [47]

    4d gaussian splatting for real-time dynamic scene rendering

    Guanjun Wu, Taoran Yi, Jiemin Fang, Lingxi Xie, Xiaopeng Zhang, Wei Wei, Wenyu Liu, Qi Tian, and Xinggang Wang. 4d gaussian splatting for real-time dynamic scene rendering. arXiv preprint arXiv:2310.08528, 2023. 6

    2023 arXiv

  40. [48]

    4d gaussian splatting for real-time dynamic scene rendering

    Guanjun Wu, Taoran Yi, Jiemin Fang, Lingxi Xie, Xiaopeng Zhang, Wei Wei, Wenyu Liu, Qi Tian, and Xinggang Wang. 4d gaussian splatting for real-time dynamic scene rendering. InCVPR, pages 20310–20320, 2024. 2, 5, 7

    2024

  41. [49]

    Mvhumannet: A large- scale dataset of multi-view daily dressing human captures

    Zhangyang Xiong, Chenghong Li, Kenkun Liu, Hongjie Liao, Jianqiao Hu, Junyi Zhu, Shuliang Ning, Lingteng Qiu, Chongjie Wang, Shijie Wang, et al. Mvhumannet: A large- scale dataset of multi-view daily dressing human captures. InCVPR, pages 19801–19811, 2024. 5, 6

    2024

  42. [50]

    Grm: Large gaussian reconstruction model for ef- ficient 3d reconstruction and generation.arXiv preprint arXiv:2403.14621, 2024

    Yinghao Xu, Zifan Shi, Wang Yifan, Hansheng Chen, Ceyuan Yang, Sida Peng, Yujun Shen, and Gordon Wet- zstein. Grm: Large gaussian reconstruction model for ef- ficient 3d reconstruction and generation.arXiv preprint arXiv:2403.14621, 2024. 2

    2024 arXiv

  43. [51]

    Gs-lrm: Large recon- struction model for 3d gaussian splatting

    Kai Zhang, Sai Bi, Hao Tan, Yuanbo Xiangli, Nanxuan Zhao, Kalyan Sunkavalli, and Zexiang Xu. Gs-lrm: Large recon- struction model for 3d gaussian splatting. InECCV, 2025. 2

    2025

  44. [52]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InCVPR, 2018. 6

    2018

  45. [53]

    Gaussian graph network: Learning efficient and generatlizable gaussian representations from multi-view images.NeurIPS, 2024

    Shengjun Zhang, Xin Fei, Fangfu Liu, Haixu Song, and Yueqi Duan. Gaussian graph network: Learning efficient and generatlizable gaussian representations from multi-view images.NeurIPS, 2024. 2

    2024

  46. [54]

    Global-correlated 3d-decoupling transformer for clothed avatar reconstruction

    Zechuan Zhang, Li Sun, Zongxin Yang, Ling Chen, and Yi Yang. Global-correlated 3d-decoupling transformer for clothed avatar reconstruction. InNeurIPS, 2023. 2

    2023

  47. [55]

    Gps- gaussian: Generalizable pixel-wise 3d gaussian splatting for real-time human novel view synthesis

    Shunyuan Zheng, Boyao Zhou, Ruizhi Shao, Boning Liu, Shengping Zhang, Liqiang Nie, and Yebin Liu. Gps- gaussian: Generalizable pixel-wise 3d gaussian splatting for real-time human novel view synthesis. InCVPR, 2024. 2, 5, 6, 7, 8 Learning Efficient and Generalizable Human Repr...

    2024

  48. [56]

    Implementation Details Fine-tuning multi-view diffusion.We utilize the Won- der3D [23] as the multi-view diffusion model. As the Won- der3D model is designed for objects and trained on Obja- verse [7], it has limited knowledge about humans and di- rectly applying it into our a...

  49. [57]

    Evaluation split.Our evaluation split is separated from the training split

    We trained our model on 8×Nvidia A800 GPUs, and it converges in 18 hours. Evaluation split.Our evaluation split is separated from the training split. We randomly selected 10 scans in the MvHu- manNet as the evaluation split. Their IDs are listed as fol- lows: 200102, 200114, 2...

  50. [58]

    More Analysis 7.1. Efficiency Though processing large amount of information across the frames, HGG is highly-efficient thanks to the design of (a) RGB failure (b) Camera failure LGM Wonder3D Wonder3D LGM Ground TruthMulti-viewLGM Gaussians Figure 5. Fail cases. (a) Wonder3D fa...

  51. [59]

    More Visualization Results We present more visualization results for both novel view synthesis and novel pose animation in Figure 6 and 8

  52. [60]

    To address GT IDOL Ours (a) Deepfashion (b) MvHumanNet Figure 7

    Broader Impacts Our model’s capacity to generate high-quality 3D animat- able avatars raises substantial privacy risks. To address GT IDOL Ours (a) Deepfashion (b) MvHumanNet Figure 7. Visualization on Deepfashion and MVHumanNet. these, the creation of ethical guidelines and l...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.