Pith. sign in

REVIEW 4 major objections 6 minor 121 references

Few-Shot Multi-Human Neural Rendering Using Geometry Constraints

T0 review · 4 major / 6 minor · reviewed 2026-08-08 · deepseek-v4-flash

Pith's one-line read A neural rendering method reconstructs several people's surfaces and appearance from as few as five images by anchoring an implicit surface to pre-computed body-model meshes.

desk verdict Solid incremental paper on SMPL-prior sparse multi-human neural rendering; worth refereeing, but the SOTA claim overreaches given narrow baselines and unresolved SMPL bias. read the letter →

arxiv 2502.07140 v1 pith:AUPUCZDI submitted 2025-02-11 cs.CV cs.AIcs.GR

classification cs.CVcs.AIcs.GR
keywords neuralrenderingmulti-humanreconstructionsparseviewsimplicitsigneddistancefunctionsSMPLpriornovelviewsynthesisvolumetricgeometricregularization
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

The paper claims that the shape and appearance of a scene containing several people can be recovered from as few as five calibrated images, a regime where generic neural surface reconstruction loses correspondence and fuses people together. Its route is to treat an off-the-shelf SMPL body-model mesh for each person as a noisy geometric prior, initialize a signed-distance network from those meshes, and then refine geometry and radiance jointly with the images. Three auxiliary losses are proposed: an uncertainty-weighted SDF constraint that keeps the learned surface near the SMPL estimate where the images are ambiguous, a ray-consistency loss that regularizes unseen rays by interpolated patch rays, and a saturation loss that stabilizes training under variable illumination. On the CMU Panoptic and a synthetic multi-human dataset the method reports higher PSNR and SSIM than NeuS and VolSDF across 5, 10, 15 and 20 training views, together with lower Chamfer error.

What carries the argument

The load-bearing object is the SMPL mesh treated as a noisy signed-distance prior, encoded in the negative log-likelihood loss of Equation 6 with a predicted per-point uncertainty. Around it, the union of per-person SMPL bounding boxes in Equation 8 defines a foreground volume that replaces a single foreground sphere, making hierarchical sampling concentrate on the bodies and providing a clean foreground-background split without segmentation masks. The ray-consistency loss compares a rendered ray to an interpolated neighboring ray with an additional KL-divergence term on ray density, and the saturation loss adds an L1 term on HSV saturation values. Together these components carry the argument from coarse body-model fits to detailed multi-human surfaces.

What would settle it

Run the trained model on a multi-human scene where one person wears a large rigid backpack or holds a wide object, and compare Chamfer distance against ground-truth scans: if geometry quality collapses for that person while other people stay accurate, the body-model-prior-as-noise assumption is the point of failure.

Watch

Extended reading notes

Core claim

The paper establishes that a multi-human scene's implicit geometry can be recovered from sparse views by combining an SDF network pretrained on SMPL distance supervision with a foreground model made of the union of per-person bounding boxes rather than a single sphere. The SMPL meshes are not treated as ground truth: the network also outputs a per-point uncertainty, and the pretraining loss minimizes a Gaussian negative log-likelihood so that confident regions act as anchors while uncertain regions, such as clothing and hair, are free to move under photometric evidence. During fine-tuning an SDF loss with thresholds keeps the reconstruction consistent with the prior only where the predicted distance or uncertainty is small. The rendering is hybrid: points inside any person's box are handled by the SDF and radiance network, points outside by a background NeRF, which lets occlusion be reasoned about jointly instead of per person. The result claimed is state-of-the-art surface and novel-view quality on multi-human scenes with 5 to 20 views.

Load-bearing premise

The method assumes the body-model meshes computed from the sparse views are close enough to the real bodies that their geometry and bounding boxes contain every visible person.

Editorial extensions

If this is right

  • With as few as five training views, a multi-human scene can be rendered from novel viewpoints without segmentation masks or pre-scanned templates, because the union of SMPL boxes supplies the foreground-background split.
  • The number of people in view degrades the baselines much faster than the proposed method, so the approach is aimed specifically at crowded scenes rather than single-person capture.
  • The same trained network supports scene editing: rigidly transforming, scaling, or removing a person requires only transforming the points sampled inside that person's bounding box before rendering.
  • The performance gain over NeuS and VolSDF shrinks as the input becomes dense, indicating that the contribution is concentrated in the sparse-view regime.
  • Because all people are learned jointly through the union of boxes, the method avoids the per-person preprocessing and erroneous occlusion supervision that arise when running a single-human NeRF separately for each subject.

Reading between the lines

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

  • The per-point uncertainty output could be used as a confidence map to guide active camera selection in a capture system, suggesting which views would most reduce geometric ambiguity; the paper does not explore this.
  • Because the foreground is decomposed into independent person boxes, one could plausibly swap or re-compose people across scenes at render time, extending the editing results in the supplementary material into a broader re-composition tool.
  • The Gaussian-noise model assumes SMPL errors are unbiased, so systematic mis-fits such as heavy clothing or held objects may need an explicit bias term or an SMPL-free region; testing this would clarify the method's failure envelope.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes a few-shot multi-human neural rendering method that combines an implicit SDF representation with a pre-computed SMPL body-model prior. The method initializes the SDF network from SMPL distance fields with a Gaussian uncertainty model, defines per-person foreground bounding boxes for hybrid foreground/background rendering, and adds a ray-consistency loss and a saturation loss to regularize sparse-view optimization. Experiments on CMU Panoptic and on a synthetic MultiHuman-derived dataset compare against NeuS, VolSDF, InfoNeRF, and ARAH, reporting improved novel-view quality and geometry reconstruction from 5 to 20 training views.

Significance. If the reported results hold, the paper would make a useful practical contribution: it shows that a coarse SMPL prior can help sparse-view neural surface reconstruction in multi-human scenes, and the proposed bounding-box-based hybrid rendering is a plausible way to reduce empty-space sampling. The ablation study provides some evidence that each proposed component contributes. The significance is moderate rather than high, because the evaluation is narrow in both baselines and metrics, and the central geometric assumption about SMPL accuracy is not directly validated.

major comments (4)
  1. [Sec. 3.3, Eq. (8); Sec. 3.2, Eq. (6)] The hard foreground/background split in Eq. (8) assumes that every true surface point of a person lies inside the SMPL-derived bounding box plus margin δ=0.1. If an SMPL fit is systematically offset due to clothing, hair, object contact, or pose error, points beyond the margin are assigned to the background NeRF and are excluded from SDF and normal estimation, so the "complete geometry" claim in Sec. 4.2 cannot hold in general. The Gaussian uncertainty model in Eq. (6) treats the SMPL error as zero-mean noise; a systematic offset can be absorbed by inflating s_noise, which reduces the loss without moving the surface to the true location. The manuscript reports no evaluation of SMPL fit accuracy and no sensitivity analysis on δ; please add such an analysis or some form of soft/adaptive foreground assignment.
  2. [Sec. 4.3, Table 3; Supp. Table 5] The main geometry result is reported as one-way Chamfer distance from ground truth to reconstruction. One-way distance can be made artificially small by reconstructing only a subset of the surface, and it is not a sufficient metric to support the paper's claim of complete geometry. The supplementary table reports bidirectional Chamfer distances, which are considerably larger and sometimes close to the baselines (e.g., scene 1, 15 views: 0.154 vs 0.167). Please report bidirectional Chamfer or precision/recall/completeness in the main paper, and avoid the claim of an "order of magnitude less error" based on one-way values. Also, no quantitative geometry metric is reported for the real CMU Panoptic experiments, despite the geometry claim in Sec. 4.2.
  3. [Sec. 4.2-4.3, Tables 1 and 3] All quantitative results appear to be single runs on a single random selection of training views. Given the explicit statement that training views were randomly extracted, the reported improvements (which are often small, e.g., SSIM differences of 0.01-0.03 in Table 1) could be within view-selection noise. Please provide results over multiple random training-view selections with mean and standard deviation, or otherwise justify that the observed differences are statistically significant. This is directly load-bearing for the state-of-the-art claim.
  4. [Abstract; Sec. 4] The "state-of-the-art" claim is supported by only four sparse-view baselines, two of which (NeuS and VolSDF) are general-purpose implicit surface methods without human-specific priors. To make the claim proportionate, either restrict it to "compared with NeuS, VolSDF, InfoNeRF, and ARAH" or compare with recent multi-human sparse-view methods that also exploit body models (e.g., the methods cited as [30, 110, 115, 116]) and report computational cost. The current evidence supports an improvement over the four tested baselines, not a general state-of-the-art statement.
minor comments (6)
  1. [Sec. 3.4, Eq. (12) and Eq. (14)] Eq. (12) uses L_r for both the photometric reconstruction loss and the ray-consistency loss; the latter should be renamed (e.g., L_ray). In Eq. (14), C(r^*) is used but the text defines C(r_p^*) on an interpolated ray, and the density distribution P(·) is not fully defined.
  2. [Sec. 4.2] The text refers to "Tab. 6" when discussing the InfoNeRF comparison, but the actual table is Table 2; please fix the cross-reference.
  3. [Table 3 caption] The caption spells the metric as "LIPIS"; this should be LPIPS. Inconsistencies such as "Neus" versus "NeuS" also appear in Figure 2 and Table 4.
  4. [Fig. 1 caption] The caption says "SMLP bounding boxes"; this should be SMPL.
  5. [Sec. 4.2 and Supp. Sec. 8] The main text says training views were "randomly extracted" while the supplementary material says they were "uniformly sampled"; please clarify the protocol and report the random seed used.
  6. [Sec. 4.3 and Supp. Sec. 8] The synthetic dataset is described as having 29 cameras arranged in a "great circle" in the main text but as arranged in a "sphere" in the supplementary material; please reconcile these descriptions.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: SMPL prior is an external geometric input and all headline results are evaluated against external ground-truth geometry and images.

full rationale

The paper's derivation chain is self-contained in the relevant sense. The SMPL fits are obtained from off-the-shelf body estimation and are treated as a noisy external geometric prior: Section 3.2 'independently training fθ0 using off-the-shelf SMPL fittings,' and Section 3.3 defines bounding boxes directly from the same SMPL fits. These are inputs to the method, not quantities the paper claims to predict. Equation 6 models the SMPL distance field as Gaussian noise and learns an uncertainty, which is a regularizer rather than a renamed prediction. Equation 13 anchors the final SDF to the initial SMPL-initialized network only under thresholds, again a prior-supervision term. The final geometry is optimized with photometric losses against multi-view images and is evaluated against external ground truth: CMU Panoptic data for image metrics and synthetic MultiHuman meshes for Chamfer distance. The claimed state-of-the-art results are therefore externally falsifiable and do not reduce to the fitted SMPL values by construction. The self-citations in the related-work section (e.g., [48, 49, 65-70, 102]) are background context for generalizable sparse-view methods and are not load-bearing for the paper's central claim. The main risk identified by the skeptical analysis is that systematic SMPL fit errors could push true surface points outside the 0.1 m bounding-box margin, but that is a correctness and robustness concern, not circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the adequacy of SMPL fits as a geometric prior, the Gaussian noise model in Equation 6, the bounding-box decomposition in Equation 8, and the patch smoothness in Equation 14. Loss weights and thresholds are hand-set without sensitivity analysis.

free parameters (6)
  • Eikonal loss weight lambda_eik = not specified
    Hand-set hyperparameter in Equation 12; no sensitivity analysis reported.
  • SDF regularization weight lambda_sdf = not specified
    Hand-set loss weight in Equation 12 that controls how strongly the SMPL-based SDF prior is enforced.
  • Ray consistency loss weight lambda_r = not specified
    Hand-set loss weight in Equation 12 for the interpolation consistency regularizer.
  • Saturation loss weight lambda_s = not specified
    Hand-set loss weight in Equation 12 for HSV saturation matching.
  • SDF uncertainty thresholds xi0 and xi1 = xi0=0.2, xi1=0.5
    Hand-set thresholds in Equation 13 that decide when the SMPL prior is applied and when it is ignored.
  • Bounding box margin delta = 0.1
    Hand-set spatial margin around each SMPL bounding box in Section 3.3.
assumptions (5)
  • domain assumption NeuS SDF-based volume rendering formulation (Equations 1-5) renders color correctly for implicit surfaces.
    The method adopts NeuS without re-validating its assumptions in the multi-human setting.
  • domain assumption SMPL body fits from the input views are sufficiently accurate to initialize the SDF and define per-person boxes.
    Central geometric prior; if the fits are systematically wrong, the Gaussian noise model cannot correct the bias. Appears in Sections 3.2 and 3.3.
  • ad hoc to paper Per-point error between SMPL and the true surface is zero-mean Gaussian with variance s_noise^2 (Equation 6).
    Modeling choice without empirical validation; SMPL residuals are unlikely to be exactly Gaussian, especially around clothing and hair.
  • ad hoc to paper Each human is wholly contained in its SMPL-derived bounding box plus margin delta.
    Hard foreground/background split via b(p) in Equation 8; loose clothing or held objects could poke outside the box.
  • ad hoc to paper Interpolated rays on a patch should match observed neighboring rays in color and density (Equation 14).
    Smoothness assumption for sparse views, borrowed in spirit from InfoNeRF and RegNeRF, but not proven for occluded multi-human patches.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Few-Shot Multi-Human Neural Rendering Using Geometry Constraints." pith.science (2026). https://pith.science/paper/AUPUCZDI

@misc{pith2026250207140,
  author       = {Pith},
  title        = {Pith review of: Few-Shot Multi-Human Neural Rendering Using Geometry Constraints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AUPUCZDI}},
  note         = {Machine review of arXiv:2502.07140}
}
read the original abstract

We present a method for recovering the shape and radiance of a scene consisting of multiple people given solely a few images. Multi-human scenes are complex due to additional occlusion and clutter. For single-human settings, existing approaches using implicit neural representations have achieved impressive results that deliver accurate geometry and appearance. However, it remains challenging to extend these methods for estimating multiple humans from sparse views. We propose a neural implicit reconstruction method that addresses the inherent challenges of this task through the following contributions: First, we propose to use geometry constraints by exploiting pre-computed meshes using a human body model (SMPL). Specifically, we regularize the signed distances using the SMPL mesh and leverage bounding boxes for improved rendering. Second, we propose a ray regularization scheme to minimize rendering inconsistencies, and a saturation regularization for robust optimization in variable illumination. Extensive experiments on both real and synthetic datasets demonstrate the benefits of our approach and show state-of-the-art performance against existing neural reconstruction methods.

Figures

Figures reproduced from arXiv: 2502.07140 by the authors.

Figure 1
Figure 1. Overview. We address the multi-human implicit shape and appearance learning problem by initializing the geometry using SMPL (Sec. 3.2), along with uncertainty-based SDF supervision and novel photometric regularizations designed to compensate for the lack of views (Sec. 3.4). We also model the foreground (Union of SMLP bounding boxes) and remainder of the scene seperatelty (Sec. 3.3). complement of the inner volume. … view at source ↗
Figure 2
Figure 2. Qualitative comparison against NeuS [90] and VolSDF [100] of synthesised novel views and reconstructed normal images of multiple humans on CMU Panoptic dataset [40, 81], using 5/10/15/20 training views. Scene Method PSNR↑ SSIM↑ LPIPS↓ 5 10 15 20 5 10 15 20 5 10 15 20 NeuS 17.83 18.84 19.39 21.97 0.62 0.67 0.69 0.55 0.74 0.51 0.49 0.45 1 VolSDF 17.50 18.08 19.51 22.31 0.64 0.61 0.67 0.71 0.61 0.54 0.51 0.48 Ours 18.4… view at source ↗
Figure 3
Figure 3. Quantitative comparison of average PSNR ( [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Comparison against [91] from 5 training views. PSNRs for the 3 examples are respectively: 26.97/29.56, 27.48/33.66, 24.36/30.56 (ARAH/Ours). Scene Method PSNR↑ SSIM↑ LPIPS↓ InfoNeRF 14.64 0.50 0.64 1 NeuS w/ info 17.98 0.65 0.58 Ours 18.41 0.67 0.55 InfoNeRF 14.21 0.49…
Figure 5
Figure 5. Figure 5: Qualitative comparison of synthesised novel views and reconstructed normal images on the synthetic dataset (MultiHuman [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Ablation study on CMU Panoptic dataset [ [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Qualitative results for the editing application. We show synthesised novel views and reconstructed normal images of multiple [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparisons against NeuS [90] and VolSDF [100] for different number of people in the scene. Showing synthesised novel views and reconstructed normal images on 5 scenes from CMU Panoptic dataset [40, 81], using 20 training views [PITH_FULL_IMAGE:figures/ful…
Figure 12
Figure 12. Figure 12: Illustration of our losses. Rays without ground-truth are [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]
Figure 10
Figure 10. Figure 10: Comparison against single human method ARAH [ [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: Network architecture (Section 3 of the main submis￾sion). p is a sampled point along a ray. γ is the positional encoding [59, 85]. n(p) is the gradient of predicted sdf w.r.t the input point p. v is the direction of the ray, and f1 the rasterized depth feature describ…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

121 extracted references · 73 canonical work pages

  1. [1]

    Neural point-based graph- ics

    Kara-Ali Aliev, Artem Sevastopolsky, Maria Kolos, Dmitry Ulyanov, and Victor Lempitsky. Neural point-based graph- ics. In European Conference on Computer Vision , pages 696–712. Springer, 2020. 2

  2. [2]

    Detailed human avatars from monocular video

    Thiemo Alldieck, Marcus Magnor, Weipeng Xu, Christian Theobalt, and Gerard Pons-Moll. Detailed human avatars from monocular video. In 2018 International Conference on 3D Vision (3DV), pages 98–109. IEEE, 2018. 1

  3. [3]

    Video based reconstruc- tion of 3d people models

    Thiemo Alldieck, Marcus Magnor, Weipeng Xu, Christian Theobalt, and Gerard Pons-Moll. Video based reconstruc- tion of 3d people models. In Proceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition, pages 8387–8397, 2018. 2

  4. [4]

    Photorealistic monocular 3d reconstruction of hu- mans wearing clothing

    Thiemo Alldieck, Mihai Zanfir, and Cristian Sminchis- escu. Photorealistic monocular 3d reconstruction of hu- mans wearing clothing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 1506–1515, 2022. 2

  5. [5]

    Neural rgb-d surface reconstruction

    Dejan Azinovi ´c, Ricardo Martin-Brualla, Dan B Goldman, Matthias Nießner, and Justus Thies. Neural rgb-d surface reconstruction. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , pages 6290–6301, 2022. 4

  6. [6]

    Detailed human shape and pose from images

    Alexandru O Balan, Leonid Sigal, Michael J Black, James E Davis, and Horst W Haussecker. Detailed human shape and pose from images. In 2007 IEEE Conference on Computer Vision and Pattern Recognition , pages 1–8. IEEE, 2007. 2

  7. [7]

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

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

  8. [8]

    Poco: Point con- volution for surface reconstruction

    Alexandre Boulch and Renaud Marlet. Poco: Point con- volution for surface reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6302–6314, 2022. 2

Show all 121 references
  1. [9]

    Dy- namic surface function networks for clothed human bodies

    Andrei Burov, Matthias Nießner, and Justus Thies. Dy- namic surface function networks for clothed human bodies. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 10754–10764, 2021. 2

  2. [10]

    Free-viewpoint video of human ac- tors

    Joel Carranza, Christian Theobalt, Marcus A Magnor, and Hans-Peter Seidel. Free-viewpoint video of human ac- tors. ACM transactions on graphics (TOG), 22(3):569–577,

  3. [11]

    Unsuper- vised inference of signed distance functions from single sparse point clouds without learning priors

    Chao Chen, Zhizhong Han, and Yu-Shen Liu. Unsuper- vised inference of signed distance functions from single sparse point clouds without learning priors. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 2

  4. [12]

    Animatable neural ra- diance fields from monocular rgb videos

    Jianchuan Chen, Ying Zhang, Di Kang, Xuefei Zhe, Lin- chao Bao, Xu Jia, and Huchuan Lu. Animatable neural ra- diance fields from monocular rgb videos. arXiv preprint arXiv:2106.13629, 2021. 1

  5. [13]

    Implicit functions in feature space for 3d shape reconstruc- tion and completion

    Julian Chibane, Thiemo Alldieck, and Gerard Pons-Moll. Implicit functions in feature space for 3d shape reconstruc- tion and completion. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 6970–6981, 2020. 2

  6. [14]

    Learning to estimate robust 3d human mesh from in-the-wild crowded scenes

    Hongsuk Choi, Gyeongsik Moon, JoonKyu Park, and Ky- oung Mu Lee. Learning to estimate robust 3d human mesh from in-the-wild crowded scenes. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1475–1484, 2022. 1, 2

  7. [15]

    Monocular expres- sive body regression through body-driven attention

    Vasileios Choutas, Georgios Pavlakos, Timo Bolkart, Dim- itrios Tzionas, and Michael J Black. Monocular expres- sive body regression through body-driven attention. In European Conference on Computer Vision , pages 20–40. Springer, 2020. 1, 2

  8. [16]

    High-quality streamable free- viewpoint video

    Alvaro Collet, Ming Chuang, Pat Sweeney, Don Gillett, Dennis Evseev, David Calabrese, Hugues Hoppe, Adam Kirk, and Steve Sullivan. High-quality streamable free- viewpoint video. ACM Transactions on Graphics (ToG) , 34(4):1–13, 2015. 1, 2

  9. [17]

    Performance capture from sparse multi-view video

    Edilson De Aguiar, Carsten Stoll, Christian Theobalt, Naveed Ahmed, Hans-Peter Seidel, and Sebastian Thrun. Performance capture from sparse multi-view video. InACM SIGGRAPH 2008 papers, pages 1–10. 2008. 2

  10. [18]

    Depth-supervised nerf: Fewer views and faster training for free

    Kangle Deng, Andrew Liu, Jun-Yan Zhu, and Deva Ra- manan. Depth-supervised nerf: Fewer views and faster training for free. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , pages 12882–12891, 2022. 2, 3

  11. [19]

    Pina: Learning a personalized implicit neural avatar from a single rgb-d video sequence

    Zijian Dong, Chen Guo, Jie Song, Xu Chen, Andreas Geiger, and Otmar Hilliges. Pina: Learning a personalized implicit neural avatar from a single rgb-d video sequence. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20470–20480, 2022. 1

  12. [20]

    Shape-aware multi-person pose estimation from multi-view images

    Zijian Dong, Jie Song, Xu Chen, Chen Guo, and Otmar Hilliges. Shape-aware multi-person pose estimation from multi-view images. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision, pages 11158– 11168, 2021. 2

  13. [21]

    Fusion4d: Real-time performance capture of challenging scenes

    Mingsong Dou, Sameh Khamis, Yury Degtyarev, Philip Davidson, Sean Ryan Fanello, Adarsh Kowdle, Sergio Orts Escolano, Christoph Rhemann, David Kim, Jonathan Tay- lor, et al. Fusion4d: Real-time performance capture of challenging scenes. ACM Transactions on Graphics (ToG), 35(4)...

  14. [22]

    Three- dimensional reconstruction of human interactions

    Mihai Fieraru, Mihai Zanfir, Elisabeta Oneata, Alin-Ionut Popa, Vlad Olaru, and Cristian Sminchisescu. Three- dimensional reconstruction of human interactions. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7214–7223, 2020. 1, 2

  15. [23]

    Motion capture using joint skeleton tracking and surface estimation

    Juergen Gall, Carsten Stoll, Edilson De Aguiar, Christian Theobalt, Bodo Rosenhahn, and Hans-Peter Seidel. Motion capture using joint skeleton tracking and surface estimation. In 2009 IEEE Conference on Computer Vision and Pattern Recognition, pages 1746–1753. Ieee, 2009. 2

  16. [24]

    Implicit geometric regularization for learn- ing shapes

    Amos Gropp, Lior Yariv, Niv Haim, Matan Atzmon, and Yaron Lipman. Implicit geometric regularization for learn- ing shapes. arXiv preprint arXiv:2002.10099, 2020. 4

  17. [25]

    Holopose: Holis- tic 3d human reconstruction in-the-wild

    Riza Alp Guler and Iasonas Kokkinos. Holopose: Holis- tic 3d human reconstruction in-the-wild. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10884–10894, 2019. 2

  18. [26]

    The re- lightables: V olumetric performance capture of humans with realistic relighting

    Kaiwen Guo, Peter Lincoln, Philip Davidson, Jay Busch, Xueming Yu, Matt Whalen, Geoff Harvey, Sergio Orts- Escolano, Rohit Pandey, Jason Dourgarian, et al. The re- lightables: V olumetric performance capture of humans with realistic relighting. ACM Transactions on Graphics (To...

  19. [27]

    Geo-pifu: Geometry and pixel aligned implicit functions for single-view human reconstruction

    Tong He, John Collomosse, Hailin Jin, and Stefano Soatto. Geo-pifu: Geometry and pixel aligned implicit functions for single-view human reconstruction. Advances in Neural Information Processing Systems, 33:9276–9287, 2020. 2

  20. [28]

    Arch++: Animation-ready clothed human re- construction revisited

    Tong He, Yuanlu Xu, Shunsuke Saito, Stefano Soatto, and Tony Tung. Arch++: Animation-ready clothed human re- construction revisited. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision, pages 11046– 11056, 2021. 2

  21. [29]

    Image quality metrics: Psnr vs

    Alain Hore and Djemel Ziou. Image quality metrics: Psnr vs. ssim. In 2010 20th international conference on pattern recognition, pages 2366–2369. IEEE, 2010. 5

  22. [30]

    Dynamic multi-person mesh recovery from uncali- brated multi-view cameras

    Buzhen Huang, Yuan Shu, Tianshu Zhang, and Yangang Wang. Dynamic multi-person mesh recovery from uncali- brated multi-view cameras. In 2021 International Confer- ence on 3D Vision (3DV), pages 710–720. IEEE, 2021. 1, 2

  23. [31]

    Neural kernel surface re- construction

    Jiahui Huang, Zan Gojcic, Matan Atzmon, Or Litany, Sanja Fidler, and Francis Williams. Neural kernel surface re- construction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4369– 4379, 2023. 2

  24. [32]

    Towards accurate marker-less hu- man shape and pose estimation over time

    Yinghao Huang, Federica Bogo, Christoph Lassner, Angjoo Kanazawa, Peter V Gehler, Javier Romero, Ijaz Akhter, and Michael J Black. Towards accurate marker-less hu- man shape and pose estimation over time. In 2017 inter- national conference on 3D vision (3DV) , pages 421–430. I...

  25. [33]

    Deep volumetric video from very sparse multi- view performance capture

    Zeng Huang, Tianye Li, Weikai Chen, Yajie Zhao, Jun Xing, Chloe LeGendre, Linjie Luo, Chongyang Ma, and Hao Li. Deep volumetric video from very sparse multi- view performance capture. In Proceedings of the European Conference on Computer Vision (ECCV) , pages 336–354,

  26. [34]

    Arch: Animatable reconstruction of clothed humans

    Zeng Huang, Yuanlu Xu, Christoph Lassner, Hao Li, and Tony Tung. Arch: Animatable reconstruction of clothed humans. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 3093– 3102, 2020. 2

  27. [35]

    Neural mesh-based graphics

    Shubhendu Jena, Franck Multon, and Adnane Boukhayma. Neural mesh-based graphics. In European Conference on Computer Vision, pages 739–757. Springer, 2022. 2

  28. [36]

    Geotransfer: Generalizable few-shot multi-view re- construction via transfer learning

    Shubhendu Jena, Franck Multon, and Adnane Boukhayma. Geotransfer: Generalizable few-shot multi-view re- construction via transfer learning. arXiv preprint arXiv:2408.14724, 2024. 2

  29. [37]

    Selfrecon: Self reconstruction your digital avatar from monocular video

    Boyi Jiang, Yang Hong, Hujun Bao, and Juyong Zhang. Selfrecon: Self reconstruction your digital avatar from monocular video. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , pages 5605–5615, 2022. 1

  30. [38]

    Coherent reconstruction of multiple humans from a single image

    Wen Jiang, Nikos Kolotouros, Georgios Pavlakos, Xiaowei Zhou, and Kostas Daniilidis. Coherent reconstruction of multiple humans from a single image. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5579–5588, 2020. 1, 2

  31. [39]

    Panoptic studio: A massively multiview system for social motion capture

    Hanbyul Joo, Hao Liu, Lei Tan, Lin Gui, Bart Nabbe, Iain Matthews, Takeo Kanade, Shohei Nobuhara, and Yaser Sheikh. Panoptic studio: A massively multiview system for social motion capture. In Proceedings of the IEEE Interna- tional Conference on Computer Vision , pages 3334–3342,

  32. [40]

    Panoptic studio: A massively multiview system for social interaction capture

    Hanbyul Joo, Tomas Simon, Xulong Li, Hao Liu, Lei Tan, Lin Gui, Sean Banerjee, Timothy Scott Godisart, Bart Nabbe, Iain Matthews, Takeo Kanade, Shohei Nobuhara, and Yaser Sheikh. Panoptic studio: A massively multiview system for social interaction capture. IEEE Transactions on...

  33. [41]

    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. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 7122–7131, 2018. 1, 2

  34. [42]

    Neural lumigraph ren- dering

    Petr Kellnhofer, Lars C Jebe, Andrew Jones, Ryan Spicer, Kari Pulli, and Gordon Wetzstein. Neural lumigraph ren- dering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 4287– 4297, 2021. 2

  35. [43]

    Infonerf: Ray entropy minimization for few-shot neural volume ren- dering

    Mijeong Kim, Seonguk Seo, and Bohyung Han. Infonerf: Ray entropy minimization for few-shot neural volume ren- dering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 12912– 12921, 2022. 2, 5, 7

  36. [44]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 ,

  37. [45]

    Muhammed Kocabas, Nikos Athanasiou, and Michael J. Black. VIBE: Video inference for human body pose and shape estimation. In 2020 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR 2020), pages 5252–5262, Piscataway, NJ, June 2020. IEEE. 1, 2

  38. [46]

    Neural human performer: Learning generalizable radiance fields for human performance rendering.Advances in Neural Information Processing Systems, 34, 2021

    Youngjoong Kwon, Dahun Kim, Duygu Ceylan, and Henry Fuchs. Neural human performer: Learning generalizable radiance fields for human performance rendering.Advances in Neural Information Processing Systems, 34, 2021. 1, 2

  39. [47]

    Shape reconstruction using volume sweeping and learned photoconsistency

    Vincent Leroy, Jean-S ´ebastien Franco, and Edmond Boyer. Shape reconstruction using volume sweeping and learned photoconsistency. In Proceedings of the European Confer- ence on Computer Vision (ECCV) , pages 781–796, 2018. 2

  40. [48]

    Learn- ing generalizable light field networks from few images

    Qian Li, Franck Multon, and Adnane Boukhayma. Learn- ing generalizable light field networks from few images. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages 1–5. IEEE, 2023. 2

  41. [49]

    Regu- larizing neural radiance fields from sparse rgb-d inputs

    Qian Li, Franck Multon, and Adnane Boukhayma. Regu- larizing neural radiance fields from sparse rgb-d inputs. In 2023 IEEE International Conference on Image Processing (ICIP), pages 2320–2324. IEEE, 2023. 2

  42. [50]

    Shape-aware human pose and shape reconstruction using multi-view images

    Junbang Liang and Ming C Lin. Shape-aware human pose and shape reconstruction using multi-view images. In Pro- ceedings of the IEEE/CVF International Conference on Computer Vision, pages 4352–4362, 2019. 2

  43. [51]

    Dynamic plane convolutional occupancy networks

    Stefan Lionar, Daniil Emtsev, Dusan Svilarkovic, and Songyou Peng. Dynamic plane convolutional occupancy networks. In Proceedings of the IEEE/CVF Winter Con- ference on Applications of Computer Vision , pages 1829– 1838, 2021. 2

  44. [52]

    Neural actor: Neural free-view synthesis of human actors with pose con- trol

    Lingjie Liu, Marc Habermann, Viktor Rudnev, Kripasindhu Sarkar, Jiatao Gu, and Christian Theobalt. Neural actor: Neural free-view synthesis of human actors with pose con- trol. ACM Transactions on Graphics (TOG) , 40(6):1–16,

  45. [53]

    Recent advances of monocular 2d and 3d human pose estimation: A deep learning perspective

    Wu Liu and Tao Mei. Recent advances of monocular 2d and 3d human pose estimation: A deep learning perspective. ACM Computing Surveys (CSUR), 2022. 1, 2

  46. [54]

    Markerless motion capture of multiple characters using multiview image seg- mentation

    Yebin Liu, Juergen Gall, Carsten Stoll, Qionghai Dai, Hans- Peter Seidel, and Christian Theobalt. Markerless motion capture of multiple characters using multiview image seg- mentation. IEEE transactions on pattern analysis and ma- chine intelligence, 35(11):2720–2735, 2013. 2

  47. [55]

    Markerless motion capture of in- teracting characters using multi-view image segmentation

    Yebin Liu, Carsten Stoll, Juergen Gall, Hans-Peter Seidel, and Christian Theobalt. Markerless motion capture of in- teracting characters using multi-view image segmentation. In CVPR 2011, pages 1249–1256. Ieee, 2011. 1, 2

  48. [56]

    SparseNeuS: Fast generalizable neural sur- face reconstruction from sparse views

    Xiaoxiao Long, Cheng Lin, Peng Wang, Taku Komura, and Wenping Wang. SparseNeuS: Fast generalizable neural sur- face reconstruction from sparse views. ECCV, 2022. 2, 3

  49. [57]

    Smpl: A skinned multi-person linear model

    Matthew Loper, Naureen Mahmood, Javier Romero, Ger- ard Pons-Moll, and Michael J Black. Smpl: A skinned multi-person linear model. ACM transactions on graphics (TOG), 34(6):1–16, 2015. 1, 2

  50. [58]

    Nerf in the wild: Neural radiance fields for uncon- strained photo collections

    Ricardo Martin-Brualla, Noha Radwan, Mehdi SM Sajjadi, Jonathan T Barron, Alexey Dosovitskiy, and Daniel Duck- worth. Nerf in the wild: Neural radiance fields for uncon- strained photo collections. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni...

  51. [59]

    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. In European conference on computer vision, pages 405–421. Springer, 2020. 1, 2, 3, 4, 5, 14, 16

  52. [60]

    On self-contact and human pose

    Lea Muller, Ahmed AA Osman, Siyu Tang, Chun-Hao P Huang, and Michael J Black. On self-contact and human pose. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 9990–9999,

  53. [61]

    Multi-person implicit reconstruction from a single image

    Armin Mustafa, Akin Caliskan, Lourdes Agapito, and Adrian Hilton. Multi-person implicit reconstruction from a single image. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , pages 14474–14483, 2021. 1, 2

  54. [62]

    Regnerf: Regularizing neural radiance fields for view syn- thesis from sparse inputs

    Michael Niemeyer, Jonathan T Barron, Ben Mildenhall, Mehdi SM Sajjadi, Andreas Geiger, and Noha Radwan. Regnerf: Regularizing neural radiance fields for view syn- thesis from sparse inputs. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , ...

  55. [63]

    Differentiable volumetric rendering: Learning implicit 3d representations without 3d supervi- sion

    Michael Niemeyer, Lars Mescheder, Michael Oechsle, and Andreas Geiger. Differentiable volumetric rendering: Learning implicit 3d representations without 3d supervi- sion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 3504–3515,

  56. [64]

    Unisurf: Unifying neural implicit surfaces and radiance fields for multi-view reconstruction

    Michael Oechsle, Songyou Peng, and Andreas Geiger. Unisurf: Unifying neural implicit surfaces and radiance fields for multi-view reconstruction. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 5589–5599, 2021. 2

  57. [65]

    Few’zero level set’-shot learning of shape signed distance functions in fea- ture space

    Amine Ouasfi and Adnane Boukhayma. Few’zero level set’-shot learning of shape signed distance functions in fea- ture space. In ECCV, 2022. 2

  58. [66]

    Few-shot unsu- pervised implicit neural shape representation learning with spatial adversaries

    Amine Ouasfi and Adnane Boukhayma. Few-shot unsu- pervised implicit neural shape representation learning with spatial adversaries. arXiv preprint arXiv:2408.15114, 2024. 2

  59. [67]

    Mixing-denoising generalizable occupancy networks

    Amine Ouasfi and Adnane Boukhayma. Mixing-denoising generalizable occupancy networks. 3DV, 2024. 2

  60. [68]

    Robustifying gen- eralizable implicit shape networks with a tunable non- parametric model

    Amine Ouasfi and Adnane Boukhayma. Robustifying gen- eralizable implicit shape networks with a tunable non- parametric model. Advances in Neural Information Pro- cessing Systems, 36, 2024. 2

  61. [69]

    Unsupervised oc- cupancy learning from sparse point cloud

    Amine Ouasfi and Adnane Boukhayma. Unsupervised oc- cupancy learning from sparse point cloud. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 21729–21739, 2024. 2

  62. [70]

    Toward robust neural reconstruction from sparse point sets, 2024

    Amine Ouasfi, Shubhendu Jena, Eric Marchand, and Ad- nane Boukhayma. Toward robust neural reconstruction from sparse point sets, 2024. 2

  63. [71]

    Deepsdf: Learning continuous signed distance functions for shape representa- tion

    Jeong Joon Park, Peter Florence, Julian Straub, Richard Newcombe, and Steven Lovegrove. Deepsdf: Learning continuous signed distance functions for shape representa- tion. In Proceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, pages 165–174, 2019. 1, 3

  64. [72]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems , 32,

  65. [73]

    An- imatable neural radiance fields for modeling dynamic hu- man bodies

    Sida Peng, Junting Dong, Qianqian Wang, Shangzhan Zhang, Qing Shuai, Xiaowei Zhou, and Hujun Bao. An- imatable neural radiance fields for modeling dynamic hu- man bodies. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision , pages 14314– 14323, 2021. 1, 2

  66. [74]

    Shape as points: A differentiable poisson solver

    Songyou Peng, Chiyu Jiang, Yiyi Liao, Michael Niemeyer, Marc Pollefeys, and Andreas Geiger. Shape as points: A differentiable poisson solver. Advances in Neural Informa- tion Processing Systems, 34:13032–13044, 2021. 2

  67. [75]

    Convolutional occupancy networks

    Songyou Peng, Michael Niemeyer, Lars Mescheder, Marc Pollefeys, and Andreas Geiger. Convolutional occupancy networks. In European Conference on Computer Vision , pages 523–540. Springer, 2020. 2

  68. [76]

    Neural body: 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: Implicit neural representations with structured latent codes for novel view synthesis of dynamic humans. In Proceed- ings of the IEEE/CVF Conference on Computer Vision a...

  69. [77]

    Dense depth priors for neural radiance fields from sparse input views

    Barbara Roessle, Jonathan T Barron, Ben Mildenhall, Pratul P Srinivasan, and Matthias Nießner. Dense depth priors for neural radiance fields from sparse input views. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition , pages 12892–12901, 2022. 2, 3

  70. [78]

    Pifu: Pixel-aligned implicit function for high-resolution clothed human digiti- zation

    Shunsuke Saito, Zeng Huang, Ryota Natsume, Shigeo Mor- ishima, Angjoo Kanazawa, and Hao Li. Pifu: Pixel-aligned implicit function for high-resolution clothed human digiti- zation. In Proceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 2304–2314, 2019. 2

  71. [79]

    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. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 84–93, 2020. 2

  72. [80]

    Novel view synthe- sis of human interactions from sparse multi-view videos

    Qing Shuai, Chen Geng, Qi Fang, Sida Peng, Wenhao Shen, Xiaowei Zhou, and Hujun Bao. Novel view synthe- sis of human interactions from sparse multi-view videos. In ACM SIGGRAPH 2022 Conference Proceedings, pages 1–10, 2022. 2

  73. [81]

    Hand key- point detection in single images using multiview bootstrap- ping

    Tomas Simon, Hanbyul Joo, and Yaser Sheikh. Hand key- point detection in single images using multiview bootstrap- ping. CVPR, 2017. 2, 5, 6, 7, 8, 9, 14, 15

  74. [82]

    Surface capture for performance-based animation

    Jonathan Starck and Adrian Hilton. Surface capture for performance-based animation. IEEE computer graphics and applications, 27(3):21–31, 2007. 2

  75. [83]

    Monocular, one-stage, regression of mul- tiple 3d people

    Yu Sun, Qian Bao, Wu Liu, Yili Fu, Michael J Black, and Tao Mei. Monocular, one-stage, regression of mul- tiple 3d people. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision , pages 11179– 11188, 2021. 1, 2

  76. [84]

    Putting people in their place: Monocular regression of 3d people in depth

    Yu Sun, Wu Liu, Qian Bao, Yili Fu, Tao Mei, and Michael J Black. Putting people in their place: Monocular regression of 3d people in depth. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 13243–13252, 2022. 1, 2

  77. [85]

    Fourier fea- tures let networks learn high frequency functions in low di- mensional domains

    Matthew Tancik, Pratul Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan Barron, and Ren Ng. Fourier fea- tures let networks learn high frequency functions in low di- mensional domains. Advances in Neural Information ...

  78. [86]

    De- ferred neural rendering: Image synthesis using neural tex- tures

    Justus Thies, Michael Zollh ¨ofer, and Matthias Nießner. De- ferred neural rendering: Image synthesis using neural tex- tures. Acm Transactions on Graphics (TOG) , 38(4):1–12,

  79. [87]

    Body size and depth disambiguation in multi-person reconstruction from single images

    Nicolas Ugrinovic, Adria Ruiz, Antonio Agudo, Alberto Sanfeliu, and Francesc Moreno-Noguer. Body size and depth disambiguation in multi-person reconstruction from single images. In 2021 International Conference on 3D Vi- sion (3DV), pages 53–63. IEEE, 2021. 1, 2

  80. [88]

    Articulated mesh animation from multi-view sil- houettes

    Daniel Vlasic, Ilya Baran, Wojciech Matusik, and Jovan Popovi´c. Articulated mesh animation from multi-view sil- houettes. In ACM SIGGRAPH 2008 papers , pages 1–9

  81. [89]

    Dynamic shape capture using multi-view photometric stereo

    Daniel Vlasic, Pieter Peers, Ilya Baran, Paul Debevec, Jo- van Popovi ´c, Szymon Rusinkiewicz, and Wojciech Ma- tusik. Dynamic shape capture using multi-view photometric stereo. In ACM SIGGRAPH Asia 2009 papers, pages 1–11

  82. [90]

    Neus: Learning neural implicit surfaces by volume rendering for multi-view recon- struction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view recon- struction. arXiv preprint arXiv:2106.10689, 2021. 1, 2, 3, 5, 6, 8, 14, 15

  83. [91]

    Arah: Animatable volume rendering of articulated human sdfs

    Shaofei Wang, Katja Schwarz, Andreas Geiger, and Siyu Tang. Arah: Animatable volume rendering of articulated human sdfs. In European Conference on Computer Vision,

  84. [92]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image pro- cessing, 13(4):600–612, 2004. 5

  85. [93]

    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. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , pages 162...

  86. [94]

    Neural fields as learnable kernels for 3d reconstruction

    Francis Williams, Zan Gojcic, Sameh Khamis, Denis Zorin, Joan Bruna, Sanja Fidler, and Or Litany. Neural fields as learnable kernels for 3d reconstruction. In CVPR, 2022. 2

  87. [95]

    Neural splines: Fitting 3d surfaces with infinitely- wide neural networks

    Francis Williams, Matthew Trager, Joan Bruna, and Denis Zorin. Neural splines: Fitting 3d surfaces with infinitely- wide neural networks. In CVPR, 2021. 2

  88. [96]

    On-set performance capture of multiple actors with a stereo camera

    Chenglei Wu, Carsten Stoll, Levi Valgaerts, and Christian Theobalt. On-set performance capture of multiple actors with a stereo camera. ACM Transactions on Graphics (TOG), 32(6):1–11, 2013. 1, 2

  89. [97]

    Full body performance capture under uncontrolled and vary- ing illumination: A shading-based approach

    Chenglei Wu, Kiran Varanasi, and Christian Theobalt. Full body performance capture under uncontrolled and vary- ing illumination: A shading-based approach. In European Conference on Computer Vision, pages 757–770. Springer,

  90. [98]

    Multi- view neural human rendering

    Minye Wu, Yuehao Wang, Qiang Hu, and Jingyi Yu. Multi- view neural human rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1682–1691, 2020. 2

  91. [99]

    H-nerf: Neural radiance fields for rendering and temporal reconstruction of humans in motion.Advances in Neural In- formation Processing Systems, 34:14955–14966, 2021

    Hongyi Xu, Thiemo Alldieck, and Cristian Sminchisescu. H-nerf: Neural radiance fields for rendering and temporal reconstruction of humans in motion.Advances in Neural In- formation Processing Systems, 34:14955–14966, 2021. 1, 2

  92. [100]

    V olume rendering of neural implicit surfaces

    Lior Yariv, Jiatao Gu, Yoni Kasten, and Yaron Lipman. V olume rendering of neural implicit surfaces. Advances in Neural Information Processing Systems, 34:4805–4815,

  93. [101]

    Multiview neu- ral surface reconstruction by disentangling geometry and appearance

    Lior Yariv, Yoni Kasten, Dror Moran, Meirav Galun, Matan Atzmon, Basri Ronen, and Yaron Lipman. Multiview neu- ral surface reconstruction by disentangling geometry and appearance. Advances in Neural Information Processing Systems, 33:2492–2502, 2020. 1, 2, 3, 4, 5

  94. [102]

    Sparsecraft: Few-shot neural reconstruction through stere- opsis guided geometric linearization

    Mae Younes, Amine Ouasfi, and Adnane Boukhayma. Sparsecraft: Few-shot neural reconstruction through stere- opsis guided geometric linearization. In European Confer- ence on Computer Vision, pages 37–56. Springer, 2025. 2

  95. [103]

    pixelnerf: Neural radiance fields from one or few images

    Alex Yu, Vickie Ye, Matthew Tancik, and Angjoo Kanazawa. pixelnerf: Neural radiance fields from one or few images. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4578– 4587, 2021. 2

  96. [104]

    Bodyfusion: Real-time capture of human motion and sur- face geometry using a single depth camera

    Tao Yu, Kaiwen Guo, Feng Xu, Yuan Dong, Zhaoqi Su, Jianhui Zhao, Jianguo Li, Qionghai Dai, and Yebin Liu. Bodyfusion: Real-time capture of human motion and sur- face geometry using a single depth camera. In Proceedings of the IEEE International Conference on Computer Vision, p...

  97. [105]

    Function4d: Real-time human vol- umetric capture from very sparse consumer rgbd sensors

    Tao Yu, Zerong Zheng, Kaiwen Guo, Pengpeng Liu, Qiong- hai Dai, and Yebin Liu. Function4d: Real-time human vol- umetric capture from very sparse consumer rgbd sensors. In IEEE Conference on Computer Vision and Pattern Recog- nition (CVPR2021), June 2021. 7, 16

  98. [106]

    Dou- blefusion: Real-time capture of human performances with inner body shapes from a single depth sensor

    Tao Yu, Zerong Zheng, Kaiwen Guo, Jianhui Zhao, Qiong- hai Dai, Hao Li, Gerard Pons-Moll, and Yebin Liu. Dou- blefusion: Real-time capture of human performances with inner body shapes from a single depth sensor. In Proceed- ings of the IEEE conference on computer vision and pa...

  99. [107]

    Glamr: Global occlusion-aware human mesh recovery with dynamic cameras

    Ye Yuan, Umar Iqbal, Pavlo Molchanov, Kris Kitani, and Jan Kautz. Glamr: Global occlusion-aware human mesh recovery with dynamic cameras. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. 2

  100. [108]

    Monocular 3d pose and shape estimation of mul- tiple people in natural scenes-the importance of multiple scene constraints

    Andrei Zanfir, Elisabeta Marinoiu, and Cristian Sminchis- escu. Monocular 3d pose and shape estimation of mul- tiple people in natural scenes-the importance of multiple scene constraints. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2...

  101. [109]

    Deep network for the integrated 3d sensing of multiple people in natural images

    Andrei Zanfir, Elisabeta Marinoiu, Mihai Zanfir, Alin-Ionut Popa, and Cristian Sminchisescu. Deep network for the integrated 3d sensing of multiple people in natural images. Advances in Neural Information Processing Systems , 31,

  102. [110]

    Editable free-viewpoint video using a lay- ered neural representation

    Jiakai Zhang, Xinhang Liu, Xinyi Ye, Fuqiang Zhao, Yan- shun Zhang, Minye Wu, Yingliang Zhang, Lan Xu, and Jingyi Yu. Editable free-viewpoint video using a lay- ered neural representation. ACM Transactions on Graphics (TOG), 40(4):1–18, 2021. 1

  103. [111]

    Body meshes as points

    Jianfeng Zhang, Dongdong Yu, Jun Hao Liew, Xuecheng Nie, and Jiashi Feng. Body meshes as points. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 546–556, 2021. 1, 2

  104. [112]

    Perceiving 3d human-object spatial arrangements from a single image in the wild

    Jason Y Zhang, Sam Pepose, Hanbyul Joo, Deva Ramanan, Jitendra Malik, and Angjoo Kanazawa. Perceiving 3d human-object spatial arrangements from a single image in the wild. InEuropean conference on computer vision, pages 34–51. Springer, 2020. 2

  105. [113]

    Nerf++: Analyzing and improving neural radiance fields

    Kai Zhang, Gernot Riegler, Noah Snavely, and Vladlen Koltun. Nerf++: Analyzing and improving neural radiance fields. arXiv preprint arXiv:2010.07492, 2020. 3

  106. [114]

    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. 5

  107. [115]

    Lightweight multi-person total motion cap- ture using sparse multi-view cameras

    Yuxiang Zhang, Zhe Li, Liang An, Mengcheng Li, Tao Yu, and Yebin Liu. Lightweight multi-person total motion cap- ture using sparse multi-view cameras. In Proceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 5560–5569, 2021. 1, 2

  108. [116]

    Deepmulticap: Per- formance capture of multiple characters using sparse mul- tiview cameras

    Yang Zheng, Ruizhi Shao, Yuxiang Zhang, Tao Yu, Zerong Zheng, Qionghai Dai, and Yebin Liu. Deepmulticap: Per- formance capture of multiple characters using sparse mul- tiview cameras. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision , pages 6239–6249,

  109. [118]

    1, 2, 7, 8, 9, 16 Few-Shot Multi-Human Neural Rendering Using Geometry Constraints Supplementary Material

  110. [119]

    We show here how our method can be used to perform post-learning scene editing without any addi- tional training

    Additional Results Scene editing. We show here how our method can be used to perform post-learning scene editing without any addi- tional training. Thanks to the human bounding-box-based modeling of the foreground scene, it is straightforward to rigidly transform or omit each ...

  111. [120]

    11 shows the architecture of our network in more detail (Section 3 in the main submission)

    Implementation Details Fig. 11 shows the architecture of our network in more detail (Section 3 in the main submission). The geometry MLP has 8 layers of width 256, with a skip connection from the input to the 4th layer. The radiance MLP consists of additional 4 layers of width...

  112. [121]

    CMU Panoptic [40, 81]

    Datasets We provide here additional details on the evaluation datasets used in Section 4 from the main paper. CMU Panoptic [40, 81]. Our experiments were per- formed on five different scenes from the CMU Panoptic dataset [40, 81], where each scene includes originally 30 views ...

  113. [2021]

    1, 2, 5, 6, 8, 14, 15

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.