Pith. sign in

REVIEW 4 major objections 4 minor 50 references

3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations

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

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

Pith's one-line read A compact tensorial representation stores neutral face appearance in triplanes and expression-dependent wrinkles as 1D feature lines decoded into opacity offsets, enabling 3D Gaussian head avatars that render at 300 FPS with about 10MB…

desk verdict The 10MB/300FPS head avatar is a real practical win, but the opacity-only dynamic model explains the LPIPS gap and the paper should own that limitation. read the letter →

arxiv 2504.14967 v1 pith:EMKUUW6B submitted 2025-04-21 cs.CV

Yating Wang , Xuan Wang , Ran Yi , Yanbo Fan , Jichen Hu , Jingcheng Zhu , Lizhuang Ma This is my paper · ORCID
classification cs.CV
keywords 3DGaussianSplattingheadavatardynamictexturetriplane1Dfeaturelinesopacityoffsetblendshapetensorialrepresentation
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

This paper aims to show that a 3D Gaussian head avatar can capture expression-dependent texture changes without paying a large cost in speed or storage. The proposed representation uses tri-planes to store the neutral face's appearance and lightweight 1D feature lines, one per expression blendshape, to store dynamic details; these lines are interpolated by the current expression coefficients and decoded into small opacity offsets applied on top of the neutral opacity. On the Nersemble dataset, the method improves PSNR over the GA, GHA, and GBS baselines on novel-view synthesis and self-reenactment while rendering at 300 FPS and storing roughly 10MB per subject. If the result holds, photorealistic animatable avatars become light enough for real-time video conferencing and mobile deployment.

What carries the argument

The central object is the compact tensorial representation: a triplane $T=\{T_{xy},T_{xz},T_{yz}\}$ in canonical space for neutral appearance, and per-blendshape 1D feature lines $L_i=(L_x,L_y,L_z)\in\mathbb{R}^{3\times n_{d2}\times n_s}$ for dynamic texture. Feature lines are interpolated by blendshape coefficients $\beta$ and jaw rotation bases via inverse distance weighting, projected onto axes and interpolated to give features $l(p)$, then decoded by a tiny MLP $\theta$ into an opacity offset $\Delta\alpha=\theta(l_b(p),l_r(p))$, with final opacity $\alpha=\alpha_c+\Delta\alpha$. This carries the argument by separating static appearance from expression-dependent dynamics in a way that compresses spatial redundancy (triplane) and expression redundancy (1D lines), enabling real-time rendering and low storage.

What would settle it

Train the avatar on a subject who exhibits strong expression-dependent color changes (e.g., skin reddening, glossy highlights) and test on a held-out expression; if the rendered images fail to show those color shifts while geometry matches, the opacity-only assumption is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that dynamic facial appearance can be factored into a static canonical part and a low-rank expression-dependent part: a triplane stores the view-dependent color of the neutral face, and per-blendshape 1D feature lines, interpolated by expression coefficients and jaw rotation bases, are decoded by a small MLP into an opacity offset added to the canonical opacity. This factorization lets the avatar reproduce wrinkles and other expression-specific details while keeping the model to about 10MB per subject and rendering at 300 FPS. Combined with an adaptive truncated opacity penalty that forces static triangles to have zero offset and a class-balanced sampling that oversamples rare large expressions, the method improves PSNR over GA, GHA, and GBS on novel views and self-reenactment on the Nersemble dataset.

Load-bearing premise

The method assumes every expression-dependent appearance change can be modeled as an opacity offset added to a fixed neutral triplane color, so if an expression changes surface color, specularity, or translucency, the model cannot capture it.

Editorial extensions

If this is right

  • If the representation is correct, photorealistic head avatars can be reduced to roughly 10MB per subject, making them practical for fast network transmission and real-time rendering in mobile video conferencing.
  • The opacity-offset factorization suggests that dynamic details for novel expressions can be generated by interpolating in feature-line space, so unseen blendshape combinations should render without per-subject retraining.
  • Using only the leading FLAME PCA blendshapes and a small number of jaw bases keeps storage and computation low, so adding more expressions to the model has modest cost.
  • The method's reliance on a tracked mesh means it inherits FLAME's coverage: complex hairstyles and mouth interiors remain outside the model's scope, and relighting is not supported.

Reading between the lines

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

  • A natural extension the paper leaves implicit is to allow the decoder to output RGB offsets rather than only opacity offsets; this would capture expression-dependent color changes such as blushing or specular shifts that the current model cannot represent.
  • The same triplane-plus-feature-lines pattern could transfer to other blendshape-driven articulated objects, such as hands or bodies, wherever a low-dimensional basis spans the deformation space.
  • The reported 300 FPS is measured on a high-end RTX 4090; a testable extension is to profile the same model on edge or mobile GPUs to see whether the real-time claim survives on consumer hardware.
  • The class-balanced sampling based on vertex-displacement clustering could be reused for any dataset with skewed expression distributions, independent of the tensorial representation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes a compact 3D Gaussian head avatar representation that stores neutral appearance in a triplane and per-blendshape dynamic texture as 1D feature lines decoded into opacity offsets. The method binds Gaussians to a FLAME mesh, renders with 3DGS, and introduces an adaptive truncated opacity penalty and class-balanced sampling. On nine Nersemble subjects, it reports ~10MB storage, 300 FPS, and higher PSNR than GA, GHA, GBS, and INSTA on novel-view and self-reenactment tasks, while using less storage than all baselines.

Significance. If confirmed, the method is practically significant: it demonstrates that a very compact tensorial representation can achieve real-time, low-storage head avatars with better PSNR than established baselines. The strengths are the clear empirical validation over nine subjects with multiple baselines, a plausible compression rationale, and ablations of the proposed components. The main weakness is that dynamic appearance is limited to opacity modulation, which the paper does not disclose as a limitation, and the reported LPIPS is consistently worse than Gaussian Avatars, so the perceptual-quality claim is not fully established.

major comments (4)
  1. [Sec. 3.3 (final-opacity formula)] Since the RGB color of each splat is obtained from the canonical triplane and is independent of expression, the only expression-dependent appearance channel is the opacity offset. This restricts dynamic appearance to reweighting of fixed colors; any expression-induced color change (e.g., skin-tone changes, specular variations) is unrepresentable. The paper's title and abstract claim 'expressive dynamic appearances,' and the limitations paragraph in Sec. 4.5 does not mention this restriction. Please either add a color-offset variant to the ablations (e.g., a 1D feature line decoded into an RGB offset) to test whether the opacity-only design is sufficient, or explicitly acknowledge this as a limitation and temper the claims accordingly.
  2. [Table 1 and Sec. 4.3] The reported LPIPS is worse than GA on both tasks (0.0594 vs 0.0514 novel view; 0.0772 vs 0.0667 self-reenactment) and worse than GHA on novel view. The text only highlights PSNR improvements and does not discuss this perceptual-metric regression. Because LPIPS is a standard perceptual measure, the claim of 'high-fidelity dynamic appearances' needs qualification. Please provide statistical significance tests, qualitative analysis of failure modes, or an explanation of the trade-off.
  3. [Table 2] The ablation table is formatted ambiguously; the checkmark columns for trip/fl/penalty/resample are not properly aligned in the text, and the verbal description ('the first two rows', 'the first and third rows') does not clearly map to the table rows. This makes it impossible to verify which component combinations are compared and the contributions claimed in Sec. 4.4. Please reformat the table so each row explicitly lists the components present.
  4. [Table 1 FPS column] The rendering resolution used for the FPS measurement is not specified for GA, GBS, or the proposed method; the supplementary says GHA is tested at 1024x1024, but if other methods are tested at 802x550, the comparison is not at matched resolution. Please report FPS at identical resolutions (or state the resolution for each method) to support the real-time claim.
minor comments (4)
  1. [Eq. (2)] The threshold tau in the adaptive truncated opacity penalty is never given a value in the text or supplementary; please report it and ideally a sensitivity analysis.
  2. [Sec. 3.3] The statement 'Using only the leading expression coefficients achieves similar results' is unsupported; either cite an ablation or remove it.
  3. [Sec. 2.1] '10M storage' should be '10MB storage' for consistency with Table 1.
  4. [Supplementary Table 4] The 'Others' variant (position/rotation/scale offsets) achieves higher PSNR on novel-view synthesis (36.39 vs 35.16), suggesting the opacity-only choice may not dominate in all conditions; please discuss this trade-off.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: end-to-end supervision against ground truth and external baselines makes the evaluation self-contained.

full rationale

The paper's derivation is self-contained. The core components (static triplane appearance, 1D feature-line opacity offsets, adaptive truncated penalty, class-balanced sampling) are trained end-to-end by comparing rendered images against ground truth with L1 and D-SSIM losses (Sec. 3.4), and the reported PSNR/SSIM/LPIPS numbers are computed on held-out viewpoints and expressions against the same ground truth. No reported quantity is defined in terms of the fitted parameters: the equation alpha = alpha_c + theta(l_b(p), l_r(p)) is a model ansatz, not a re-statement of the metric. It restricts the representable dynamic appearance to opacity changes over fixed canonical colors, which is a correctness limitation that the paper only partially discloses in Sec. 4.5, but it does not make the evaluation circular. The references to the authors' own prior work ([1] and [40]) appear only in related-work context and are not load-bearing: mesh binding, FLAME fitting, jaw-basis interpolation, and the loss terms are attributed to external or earlier works ([29], [22], [16], [18]), and the ablations compare design choices on the same benchmark. Storage and FPS figures are direct measurements of the implemented system. No self-definitional prediction, fitted-input-as-prediction, or self-citation chain forces the central result; the comparisons against GA, GHA, GBS, INSTA and the ablation tables provide external, independent evidence.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical or conceptual entities. It relies on hand-set architectural hyperparameters (PCA bases, jaw bases, cluster count, weights, threshold, training length) and on domain assumptions about FLAME tracking accuracy, opacity-only dynamics, and local texture consistency. These choices determine the reported storage/quality trade-off and are not derived from first principles.

free parameters (6)
  • Number of expression PCA bases used for feature lines = 80
    The paper states 'Using only the leading expression coefficients achieves similar results' and assigns feature lines to the first 80 of the 100 FLAME PCA bases (Sec. 3.3 and Implementation Details). This hand-chosen truncation affects storage and dynamic coverage.
  • Number of jaw rotation linear bases = 16 per subject
    Extracted from training videos via farthest point sampling (Sec. 3.3 and Supp. Sec. 5). The number 16 is fixed by the authors; different counts would change the dynamic jaw texture model.
  • Cluster count for class-balanced sampling = 16
    Sec. 3.4 states frames are clustered into n=16 classes. The choice is empirical and affects which expressions are oversampled during training.
  • Eye vertex weight w = 2
    Sec. 3.4: 'We set w = 2 for eye vertices and w = 1 for other vertices'. This weighting modifies the mesh-displacement similarity metric used for clustering and is chosen empirically.
  • Opacity penalty threshold tau = not disclosed
    Sec. 3.4, Eq. 2 uses a threshold tau to decide static vs dynamic triangles, described as 'empirically set'. The value is never reported, making the penalty's contribution irreproducible.
  • Training iterations = 600000
    Sec. 4.2: trained for 600k iterations with the Adam optimizer. This is a chosen schedule; shorter training would change results and storage-performance trade-offs.
assumptions (4)
  • domain assumption FLAME mesh tracking is sufficiently accurate to bind Gaussian splats and define canonical and deformed space transformations.
    The whole pipeline (Sec. 3.1, 3.3) relies on tracked FLAME meshes; if tracking is noisy, splat positions and opacity offsets are computed at wrong locations.
  • domain assumption Expression-dependent appearance is fully captured by opacity offsets relative to a fixed neutral triplane; dynamic color changes are unnecessary.
    Sec. 3.3 defines final opacity as alpha_c + Delta_alpha and keeps the triplane RGB fixed. The supplementary ablation (Tab. 4) rejects position/rotation/scale offsets but does not test color offsets, so the opacity-only claim is an unverified assumption.
  • domain assumption Neighboring Gaussian splats share appearance and motion, so low-rank tensorial encodings (triplanes and 1D feature lines) represent the texture field without meaningful loss.
    Motivates the compact representation in Sec. 3.2. If local consistency is violated, the projection and interpolation would blur details, partially reflected in the worse LPIPS scores.
  • domain assumption The Nersemble dataset with nine subjects is representative enough to evaluate generalization to novel expressions and identities.
    All experiments use Nersemble subjects with pre-tracked FLAME parameters; no cross-dataset or cross-identity robustness test is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of 3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations." pith.science (2026). https://pith.science/paper/EMKUUW6B

@misc{pith2026250414967,
  author       = {Pith},
  title        = {Pith review of: 3D Gaussian Head Avatars with Expressive Dynamic Appearances by Compact Tensorial Representations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EMKUUW6B}},
  note         = {Machine review of arXiv:2504.14967}
}
read the original abstract

Recent studies have combined 3D Gaussian and 3D Morphable Models (3DMM) to construct high-quality 3D head avatars. In this line of research, existing methods either fail to capture the dynamic textures or incur significant overhead in terms of runtime speed or storage space. To this end, we propose a novel method that addresses all the aforementioned demands. In specific, we introduce an expressive and compact representation that encodes texture-related attributes of the 3D Gaussians in the tensorial format. We store appearance of neutral expression in static tri-planes, and represents dynamic texture details for different expressions using lightweight 1D feature lines, which are then decoded into opacity offset relative to the neutral face. We further propose adaptive truncated opacity penalty and class-balanced sampling to improve generalization across different expressions. Experiments show this design enables accurate face dynamic details capturing while maintains real-time rendering and significantly reduces storage costs, thus broadening the applicability to more scenarios.

Figures

Figures reproduced from arXiv: 2504.14967 by the authors.

Figure 1
Figure 1. Our method improves rendering quality while ensuring [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Our goal is to reconstruct 3DGS head avatar with dynamic details, ensuring real-time rendering and minimized storage. We [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison with baseline methods on novel view synthesis task. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison with baseline methods on self-reenactment task. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Cross-identity reenactment of head avatars. We use the expression and pose of the source subject on the far right to drive the [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Lop helps prevent artifacts in self-reenactment. to unknown expressions distribution, achieving better self￾reenactment performance. Adaptive truncated opacity offset penalty. Ablation stud￾ies are conducted to evaluate the adaptive truncated opacity offset penalty, as…
Figure 8
Figure 8. Figure 8: Cluster center for expression balanced sampling of sub [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 7
Figure 7. Figure 7: Basis jaw rotation extracted from all frames from videos [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 9
Figure 9. Figure 9: Rendering results of extreme viewpoints and expressions. Extrapolated viewpoints are in the red box. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 28 canonical work pages

  1. [1]

    High-fidelity fa- cial avatar reconstruction from monocular video with gen- erative priors

    Yunpeng Bai, Yanbo Fan, Xuan Wang, Yong Zhang, Jingx- iang Sun, Chun Yuan, and Ying Shan. High-fidelity fa- cial avatar reconstruction from monocular video with gen- erative priors. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 4541– 4551, 2023. 2

    work page 2023

  2. [2]

    A morphable model for the synthesis of 3d faces

    V olker Blanz and Thomas Vetter. A morphable model for the synthesis of 3d faces. 1999. 1, 2

    work page 1999

  3. [3]

    Chan, Connor Z

    Eric R. Chan, Connor Z. Lin, Matthew A. Chan, Koki Nagano, Boxiao Pan, Shalini De Mello, Orazio Gallo, Leonidas Guibas, Jonathan Tremblay, Sameh Khamis, Tero Karras, and Gordon Wetzstein. Efficient geometry-aware 3D generative adversarial networks. In CVPR, 2022. 2

    2022

  4. [4]

    Tensorf: Tensorial radiance fields

    Anpei Chen, Zexiang Xu, Andreas Geiger, Jingyi Yu, and Hao Su. Tensorf: Tensorial radiance fields. InEuropean con- ference on computer vision, pages 333–350. Springer, 2022. 2

    2022

  5. [5]

    Implicit neural head synthesis via controllable local deformation fields

    Chuhan Chen, Matthew O’Toole, Gaurav Bharaj, and Pablo Garrido. Implicit neural head synthesis via controllable local deformation fields. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 416–426, 2023. 2

    work page 2023

  6. [6]

    Monogaus- sianavatar: Monocular gaussian point-based head avatar

    Yufan Chen, Lizhen Wang, Qijing Li, Hongjiang Xiao, Shengping Zhang, Hongxun Yao, and Yebin Liu. Monogaus- sianavatar: Monocular gaussian point-based head avatar. In ACM SIGGRAPH 2024 Conference Papers, pages 1–9, 2024. 2

    work page 2024

  7. [7]

    Generalizable and animatable gaussian head avatar

    Xuangeng Chu and Tatsuya Harada. Generalizable and animatable gaussian head avatar. arXiv preprint arXiv:2410.07971, 2024. 2

    arXiv 2024

  8. [8]

    Headgas: Real-time animatable head avatars via 3d gaus- sian splatting

    Helisa Dhamo, Yinyu Nie, Arthur Moreau, Jifei Song, Richard Shaw, Yiren Zhou, and Eduardo P ´erez-Pellitero. Headgas: Real-time animatable head avatars via 3d gaus- sian splatting. In European Conference on Computer Vision, pages 459–476. Springer, 2025. 2

    work page 2025

Show all 50 references
  1. [9]

    K-planes: Explicit radiance fields in space, time, and appearance

    Sara Fridovich-Keil, Giacomo Meanti, Frederik Rahbæk Warburg, Benjamin Recht, and Angjoo Kanazawa. K-planes: Explicit radiance fields in space, time, and appearance. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 12479–12488, 2023. 2

    2023

  2. [10]

    Dynamic neural radiance fields for monocular 4d facial avatar reconstruction

    Guy Gafni, Justus Thies, Michael Zollhofer, and Matthias Nießner. Dynamic neural radiance fields for monocular 4d facial avatar reconstruction. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 8649–8658, 2021. 2

    2021

  3. [11]

    Reconstructing personalized se- mantic facial nerf models from monocular video

    Xuan Gao, Chenglai Zhong, Jun Xiang, Yang Hong, Yudong Guo, and Juyong Zhang. Reconstructing personalized se- mantic facial nerf models from monocular video. ACM Transactions on Graphics (TOG), 41(6):1–12, 2022. 2

    2022

  4. [12]

    Learning neural parametric head models

    Simon Giebenhain, Tobias Kirschstein, Markos Georgopou- los, Martin R ¨unz, Lourdes Agapito, and Matthias Nießner. Learning neural parametric head models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21003–21012, 2023. 2

    2023

  5. [13]

    Npga: Neural parametric gaussian avatars

    Simon Giebenhain, Tobias Kirschstein, Martin R ¨unz, Lour- des Agapito, and Matthias Nießner. Npga: Neural parametric gaussian avatars. arXiv preprint arXiv:2405.19331, 2024. 2

    2024 arXiv

  6. [14]

    Headnerf: A real-time nerf-based parametric head model

    Yang Hong, Bo Peng, Haiyao Xiao, Ligang Liu, and Juy- ong Zhang. Headnerf: A real-time nerf-based parametric head model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20374– 20384, 2022. 2

    2022

  7. [15]

    Sc-gs: Sparse-controlled gaussian splatting for editable dynamic scenes

    Yi-Hua Huang, Yang-Tian Sun, Ziyi Yang, Xiaoyang Lyu, Yan-Pei Cao, and Xiaojuan Qi. Sc-gs: Sparse-controlled gaussian splatting for editable dynamic scenes. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4220–4230, 2024. 2

    2024

  8. [16]

    Metrics for 3d rotations: Comparison and anal- ysis

    Du Q Huynh. Metrics for 3d rotations: Comparison and anal- ysis. Journal of Mathematical Imaging and Vision, 2009. 4

    2009

  9. [17]

    Hifi4g: High-fidelity human performance rendering via compact gaussian splatting

    Yuheng Jiang, Zhehao Shen, Penghao Wang, Zhuo Su, Yu Hong, Yingliang Zhang, Jingyi Yu, and Lan Xu. Hifi4g: High-fidelity human performance rendering via compact gaussian splatting. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 19...

    2024

  10. [18]

    3d gaussian splatting for real-time radiance field rendering

    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,

  11. [19]

    Nersemble: Multi-view radi- ance field reconstruction of human heads.ACM Transactions on Graphics (TOG), 42(4):1–14, 2023

    Tobias Kirschstein, Shenhan Qian, Simon Giebenhain, Tim Walter, and Matthias Nießner. Nersemble: Multi-view radi- ance field reconstruction of human heads.ACM Transactions on Graphics (TOG), 42(4):1–14, 2023. 2

    2023

  12. [20]

    Compact 3d gaussian representation for radiance field

    Joo Chan Lee, Daniel Rho, Xiangyu Sun, Jong Hwan Ko, and Eunbyung Park. Compact 3d gaussian representation for radiance field. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21719– 21728, 2024. 2

    2024

  13. [21]

    Learning a model of facial shape and expression from 4d scans

    Tianye Li, Timo Bolkart, Michael J Black, Hao Li, and Javier Romero. Learning a model of facial shape and expression from 4d scans. ACM Trans. Graph., 36(6):194–1, 2017. 3

    2017

  14. [22]

    Posevocab: Learning joint-structured pose embeddings for human avatar modeling

    Zhe Li, Zerong Zheng, Yuxiao Liu, Boyao Zhou, and Yebin Liu. Posevocab: Learning joint-structured pose embeddings for human avatar modeling. In ACM SIGGRAPH 2023 Con- ference Proceedings, pages 1–11, 2023. 4

    2023

  15. [23]

    Spacetime gaus- sian feature splatting for real-time dynamic view synthesis

    Zhan Li, Zhang Chen, Zhong Li, and Yi Xu. Spacetime gaus- sian feature splatting for real-time dynamic view synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8508–8520, 2024. 2

    2024

  16. [24]

    3d gaussian blendshapes for head avatar animation

    Shengjie Ma, Yanlin Weng, Tianjia Shao, and Kun Zhou. 3d gaussian blendshapes for head avatar animation. In ACM SIGGRAPH 2024 Conference Papers, pages 1–10, 2024. 2, 6

    2024

  17. [25]

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

    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. 1, 2

    2021

  18. [26]

    Instant neural graphics primitives with a mul- tiresolution hash encoding

    Thomas M ¨uller, Alex Evans, Christoph Schied, and Alexan- der Keller. Instant neural graphics primitives with a mul- tiresolution hash encoding. ACM transactions on graphics (TOG), 41(4):1–15, 2022. 2

    2022

  19. [27]

    Compact3d: Com- pressing gaussian splat radiance field models with vector quantization

    KL Navaneet, Kossar Pourahmadi Meibodi, Soroush Abbasi Koohpayegani, and Hamed Pirsiavash. Compact3d: Com- pressing gaussian splat radiance field models with vector quantization. arXiv preprint arXiv:2311.18159, 2023. 2

    2023 arXiv

  20. [28]

    Compressed 3d gaussian splatting for accelerated novel view synthesis

    Simon Niedermayr, Josef Stumpfegger, and R ¨udiger West- ermann. Compressed 3d gaussian splatting for accelerated novel view synthesis. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 10349–10358, 2024. 2

    2024

  21. [29]

    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. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 2...

  22. [30]

    Relightable gaussian codec avatars

    Shunsuke Saito, Gabriel Schwartz, Tomas Simon, Junxuan Li, and Giljoo Nam. Relightable gaussian codec avatars. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 130–141, 2024. 2

    2024

  23. [31]

    Tensor4d: Efficient neural 4d decomposition for high-fidelity dynamic reconstruction and rendering

    Ruizhi Shao, Zerong Zheng, Hanzhang Tu, Boning Liu, Hongwen Zhang, and Yebin Liu. Tensor4d: Efficient neural 4d decomposition for high-fidelity dynamic reconstruction and rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 166...

    2023

  24. [32]

    Face2face: Real-time face capture and reenactment of rgb videos

    Justus Thies, Michael Zollhofer, Marc Stamminger, Chris- tian Theobalt, and Matthias Nießner. Face2face: Real-time face capture and reenactment of rgb videos. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 2387–2395, 2016. 1, 2

    2016

  25. [33]

    Neural trajectory fields for dynamic novel view syn- thesis

    Chaoyang Wang, Ben Eckart, Simon Lucey, and Orazio Gallo. Neural trajectory fields for dynamic novel view syn- thesis. arXiv preprint arXiv:2105.05994, 2021. 2

    2021 arXiv

  26. [34]

    Flashavatar: High-fidelity head avatar with efficient gaussian embedding

    Jun Xiang, Xuan Gao, Yudong Guo, and Juyong Zhang. Flashavatar: High-fidelity head avatar with efficient gaussian embedding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 1802– 1812, 2024. 2

    2024

  27. [35]

    Avatarmav: Fast 3d head avatar reconstruction using motion-aware neural voxels

    Yuelang Xu, Lizhen Wang, Xiaochen Zhao, Hongwen Zhang, and Yebin Liu. Avatarmav: Fast 3d head avatar reconstruction using motion-aware neural voxels. In ACM SIGGRAPH 2023 Conference Proceedings , pages 1–10,

    2023

  28. [36]

    Artist-friendly re- lightable and animatable neural heads

    Yingyan Xu, Prashanth Chandran, Sebastian Weiss, Markus Gross, Gaspard Zoss, and Derek Bradley. Artist-friendly re- lightable and animatable neural heads. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2457–2467, 2024. 2

    2024

  29. [37]

    Gaussian head avatar: Ultra high-fidelity head avatar via dynamic gaussians

    Yuelang Xu, Benwang Chen, Zhe Li, Hongwen Zhang, Lizhen Wang, Zerong Zheng, and Yebin Liu. Gaussian head avatar: Ultra high-fidelity head avatar via dynamic gaussians. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1931–1941, 2024. 2, 6

    1931

  30. [38]

    3d gaussian parametric head model

    Yuelang Xu, Lizhen Wang, Zerong Zheng, Zhaoqi Su, and Yebin Liu. 3d gaussian parametric head model. In European Conference on Computer Vision , pages 129–147. Springer,

  31. [39]

    Learning object-compositional neural radiance field for ed- itable scene rendering

    Bangbang Yang, Yinda Zhang, Yinghao Xu, Yijin Li, Han Zhou, Hujun Bao, Guofeng Zhang, and Zhaopeng Cui. Learning object-compositional neural radiance field for ed- itable scene rendering. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision , pages 1377...

    2021

  32. [40]

    Nofa: Nerf-based one-shot facial avatar recon- struction

    Wangbo Yu, Yanbo Fan, Yong Zhang, Xuan Wang, Fei Yin, Yunpeng Bai, Yan-Pei Cao, Ying Shan, Yang Wu, Zhongqian Sun, et al. Nofa: Nerf-based one-shot facial avatar recon- struction. In ACM SIGGRAPH 2023 conference proceedings, pages 1–12, 2023. 2

    2023

  33. [41]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 586–595, 2018. 6

    2018

  34. [42]

    Headgap: Few-shot 3d head avatar via generalizable gaussian priors

    Xiaozheng Zheng, Chao Wen, Zhaohu Li, Weiyi Zhang, Zhuo Su, Xu Chang, Yang Zhao, Zheng Lv, Xiaoyuan Zhang, Yongjie Zhang, et al. Headgap: Few-shot 3d head avatar via generalizable gaussian priors. arXiv preprint arXiv:2408.06019, 2024. 2

    2024 arXiv

  35. [43]

    Mofanerf: Morphable facial neural radiance field

    Yiyu Zhuang, Hao Zhu, Xusen Sun, and Xun Cao. Mofanerf: Morphable facial neural radiance field. In European confer- ence on computer vision, pages 268–285. Springer, 2022. 2

    2022

  36. [44]

    Towards metrical reconstruction of human faces

    Wojciech Zielonka, Timo Bolkart, and Justus Thies. Towards metrical reconstruction of human faces. In European confer- ence on computer vision, pages 250–269. Springer, 2022. 1

    2022

  37. [45]

    Instant volumetric head avatars

    Wojciech Zielonka, Timo Bolkart, and Justus Thies. Instant volumetric head avatars. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 4574–4584, 2023. 2

    2023

  38. [46]

    Triplane meets gaussian splatting: Fast and generalizable single-view 3d reconstruction with transformers

    Zi-Xin Zou, Zhipeng Yu, Yuan-Chen Guo, Yangguang Li, Ding Liang, Yan-Pei Cao, and Song-Hai Zhang. Triplane meets gaussian splatting: Fast and generalizable single-view 3d reconstruction with transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern...

    2024

  39. [47]

    Implementation Details Jaw Pose Linear Bases. We use farthest point sampling to extract linear jaw pose bases from jaw poses of each frames in videos, in order to unify dynamic textures due to linear blendshape and non-linear jaw rotation to linearly- interpolated feature line...

  40. [48]

    Baselines We conduct comparative experiments with three baseline methods: GA, GHA, and GBS

    Comparison Details 6.1. Baselines We conduct comparative experiments with three baseline methods: GA, GHA, and GBS. To ensure fair comparisons, we align the inputs, including image resolution and pre- tracked mesh. GHA. Both GHA and our method utilize multi-view videos from th...

    2020

  41. [49]

    INSTA [45] is a NeRF based head avatar method which enables fast training and inference

    More Experiments NeRF head avatar . INSTA [45] is a NeRF based head avatar method which enables fast training and inference. INSTA relies on FLAME mesh to guide NeRF to move cor- rectly, which warps points according to the nearest mesh triangle directly. Novel View SynthesisSe...

  42. [50]

    These issues could have profound societal implications

    Ethical Considerations The generation of artificial portrait videos using our method poses risks, including the spread of false information, and erosion of trust in media credibility. These issues could have profound societal implications. Addressing this chal- lenge requires ...

Pith tools

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