Pith. sign in

REVIEW 4 major objections 5 minor 154 references

UVFaceFusion: Fast Multi-view Topologically Consistent Face Reconstruction in the Wild via UV-space Neural Fusion

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Multi-view face reconstruction becomes a learnable UV-space fusion problem, and the learned fusion beats the heuristic it replaces.

desk verdict A clean UV-space neural fusion system for fixed-topology face reconstruction, but the headline SOTA claim is riding on Chamfer gaps smaller than one fifth of the per-subject standard deviation. read the letter →

arxiv 2607.18798 v1 pith:5MCEPQA2 submitted 2026-07-21 cs.CV cs.GR

Xin Ming , Yuxuan Han , Junhai Yong , Feng Xu This is my paper
classification cs.CVcs.GR
keywords multi-viewfacereconstructionfixedtopologyUV-spacefusionneuralpointmapcompletionin-the-wildgeneralizationdigitalavatarsmask-awareaggregation
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 tries to establish that fixed-topology face reconstruction from casual multi-view photos no longer needs hand-designed topological optimization. Its proposal is to convert each view's predicted 3D geometry into partial maps in a shared facial UV coordinate space, then train a mask-aware network to complete and fuse those maps into one coherent surface. The claim is that this geometry-to-geometry fusion is accurate enough to beat both the prior heuristic fusion approach and end-to-end multi-view methods on public benchmarks, while inheriting the wild-image robustness of the frozen geometry backbone. If true, it matters because it separates the hard problem of 3D estimation from the bookkeeping of topology, making high-quality avatars from everyday phone captures practical.

What carries the argument

Canonical UV-space geometric fusion: each view's pixel-aligned 3D point map is forward-splatted into a shared UV face atlas using dense per-pixel UV correspondences, then per-view similarity alignment on sparse template vertices places the partial maps in a common frame. A shared-weight encoder interprets each partial map, a valid-mask-weighted average pools cross-view evidence, and a decoder completes the missing texels; the final fixed-topology mesh is sampled from the completed UV point map at template UV coordinates.

What would settle it

Construct a multi-view test set with a face mask or another strong occluder and compare against a high-quality scan; the paper's stated limitation predicts the occluder is absorbed into the mesh. A sharper test: swap the dense-correspondence backbone for a deliberately worse one while keeping the point maps fixed—if accuracy does not degrade, the claimed role of UV lifting is not load-bearing.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the topological-fusion stage of multi-view face reconstruction can be learned instead of engineered. Each view is lifted into a canonical UV parameterization by dense correspondences, lightly re-aligned to a template, and encoded with a shared network; the encoded features are combined with mask-weighted averaging, and a decoder fills in unobserved regions. The authors claim this yields a complete, fixed-topology mesh that outperforms heuristic bundle adjustment and several end-to-end competitors, using only a modest studio dataset for training, and does so in under three seconds for sixteen views.

Load-bearing premise

The pipeline inherits the per-view 3D point maps and the UV correspondences from two frozen pretrained models, and assumes those are accurate enough that a UV-space fusion network can correct the remaining noise; if point-map noise is correlated with expression or identity, or if correspondences fail under occlusion, the fused result cannot recover what was never observed.

Editorial extensions

If this is right

  • A fixed-topology face mesh with 16 views is produced in under 3 seconds on one GPU, roughly 3× faster than the heuristic-fusion baseline.
  • Chamfer errors drop on all four public benchmarks (e.g., mean error from 1.18 to 1.12 mm on one dataset and 0.98 to 0.91 mm on another), and the gains hold with 4 input views.
  • Training only on studio data transfers to in-the-wild captures because the fusion operates on pure geometry in a canonical space, not on appearance.
  • The ablation shows learned feature-space fusion beats raw coordinate-space averaging, so the fusion network is doing real denoising and completion rather than mere smoothing.
  • The same formulation degrades gracefully to a single input view, reconstructing a complete mesh from one partial UV map and a learned prior.

Reading between the lines

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

  • If the decoupling holds, improvements to the frozen geometry and correspondence backbones should transfer directly to this system with no retraining, making the fusion network a swappable head that can sit on top of stronger estimators.
  • The same canonical-UV fusion recipe likely applies to any object class with a stable template parameterization, such as ears, hands, or full heads; the expensive ingredient is a registered training set.
  • A testable extension would be to feed per-texel confidence from the geometry backbone into the mask-weighted averaging, potentially reducing reliance on a fixed threshold and improving occlusion handling.
  • The monocular results suggest the network learns a genuine shape prior in UV space; this could be probed explicitly by measuring how much completed geometry is recalled from training identities versus inferred from the observed image.
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 / 5 minor

Summary. UVFaceFusion proposes a feed-forward, fixed-topology face reconstruction pipeline for in-the-wild multi-view images. It uses frozen VGGT point maps and Pixel3DMM UV correspondences, lifts each view's geometry into a canonical UV space via bilinear splatting and per-view Umeyama alignment, then trains a mask-aware convolutional encoder/decoder to fuse the partial UV-space point maps and predict a complete UV-space point map from which a fixed-topology mesh is sampled. The method is trained only on Ava-256 and evaluated on H3DS, NeRSemble, EmoTalk3D, and RenderMe-360, where it reports lower mean Chamfer distances than VGGTFace and other baselines, as well as runtime below 3 seconds for 16 views. The paper also includes ablations of the neural fusion module and Laplacian loss, a user study, and supplementary 4-view and monocular results.

Significance. If the reported benchmark numbers are reliable, the contribution is significant: it replaces a hand-crafted bundle-adjustment fusion stage with a learnable UV-space aggregation, decouples general geometry estimation from topology construction, and demonstrates data efficiency by training on a single studio dataset while generalizing to casual captures. The speed and fixed-topology output are practically useful for avatar pipelines. The paper also provides an honest limitations section and makes code available. However, the central quantitative claim currently rests on very small Chamfer-distance differences without statistical support, and part of the comparison is quoted from the authors' own prior work under potentially different protocols. The architecture is plausible and the ablations point in the expected direction, but the evidence as presented is not yet sufficient to establish state-of-the-art accuracy.

major comments (4)
  1. [Table 1, §4.1] The main SOTA claim is supported by mean Chamfer differences of only 0.06–0.15 mm between UVFaceFusion and VGGTFace (e.g., NeRSemble 0.91 vs 0.98; RenderMe-360 0.93 vs 1.08), while the reported per-dataset standard deviations are 0.65–0.76 mm. These gaps are below 0.2σ. No confidence intervals, per-subject error lists, or significance tests are provided. The H3DS row lacks even a standard deviation. Since the paper's central thesis is that learned neural fusion outperforms heuristic BA, the reader cannot tell whether the differences are real. Please report per-subject paired errors, bootstrap or paired-test confidence intervals, and the number of subjects used for each dataset. This applies also to the 'w/o Neural Fusion' ablation, whose 0.10–0.12 mm gaps may not be significant.
  2. [§4.1, Table 1] The evaluation protocol is not unified. For H3DS and NeRSemble, the paper says it 'follows the protocol used by prior work and quotes the numbers reported in SIRA++ and VGGTFace.' The primary competitor VGGTFace is the authors' own earlier system, and quoting its numbers rather than re-running it under an identical protocol introduces possible systematic differences in alignment, mesh sampling, subject splits, and metric code. For EmoTalk3D and RenderMe-360, ground truth is reconstructed with Metashape, which is a different source of supervision. To substantiate the SOTA claim, the authors should re-run VGGTFace (and ideally the other multi-view baselines) under exactly the same evaluation pipeline, with the same subject selection and alignment procedure, and report the results side-by-side.
  3. [§4.6, §3.4] The limitations section acknowledges that bangs and occluders such as masks can be absorbed into the reconstructed face mesh because neither VGGT point maps nor registered training meshes separate non-face geometry. This directly affects the interpretation of the quantitative metrics. If the training ground-truth meshes contain hair/forehead geometry or if the benchmark ground truth is restricted to a face mask, the Chamfer distance may not penalize non-face geometry consistently. Please clarify how the Ava-256 registration and each benchmark's ground truth define the face region, and discuss whether the reported Chamfer numbers include or exclude hair/occluder regions. A metric that ignores these artifacts could inflate the apparent accuracy of a method that silently blends occluders into the surface.
  4. [§3.2, Eqs. (11)–(12)] A load-bearing assumption is that per-view Umeyama alignment to the template, using a sparse set of sampled template vertices, is sufficient to correct residual cross-view misalignment in VGGT point maps. If the sampled vertices are themselves corrupted by VGGT noise or Pixel3DMM correspondence errors, the canonicalization step can inject a systematic distortion into every view before fusion. The paper does not provide any sensitivity analysis for this step. Please quantify the effect: e.g., compare with and without canonicalization, measure residual alignment error on the sparse vertices, and show failure cases where UV correspondence is poor. This would make the correctness-risk of the frozen-backbone assumption explicit rather than implicit.
minor comments (5)
  1. [§4.1] The loss weights λ_pm, λ_v, λ_lap in Eq. (20) are never given numerical values. Since these are important hyperparameters, please report them along with other training details (learning rate, batch size, number of sampled views per subject, validation selection).
  2. [Table 1] The H3DS columns report only Mean, while the other datasets report Median and Std. Please add the missing statistics or explain that they are unavailable from the quoted source. For reproducibility, also report the number of test subjects/sequences per dataset.
  3. [§4.2] The abstract and introduction state '3× inference speedup over VGGTFace,' but the paper only reports '<10s' for VGGTFace and '<3s' for the proposed method. A precise runtime table with per-component timings (VGGT, Pixel3DMM, lifting, fusion, mesh extraction) and hardware/software settings would be more informative.
  4. [§4.4, Fig. 3] The user study figure reports percentages for four criteria. It is helpful that the total votes per criterion are stated, but the exact number of participants per example and whether each participant saw all 24 groups should be stated. A confidence interval for the preference percentages would also be appropriate.
  5. [Supplementary, §D] The monocular extension is interesting but currently only qualitative. If monocular results are shown in the supplement, a brief quantitative comparison or at least a statement that this is not a supported setting would avoid confusion.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the UV-space fusion system is empirically evaluated, and the only self-citation (VGGTFace baseline) is not definitionally load-bearing.

full rationale

UVFaceFusion is an empirical system, not a derivation. The pipeline uses frozen pretrained backbones (VGGT, Pixel3DMM), lifts their outputs into a canonical UV space, and trains a neural fusion network with standard supervised losses (Eqs. 17-20) against ground-truth meshes registered to the FLAME topology. No equation is defined in terms of its target, and no fitted parameter is renamed as a prediction. The central claim is a benchmark comparison, not a derived identity. The paper's only self-referential element is that its primary competitor, VGGTFace, is the authors' prior work, and for H3DS/NeRSemble the VGGTFace numbers are quoted from that paper rather than re-run. This is a methodological self-citation, but it does not make the current results circular: the current method's Chamfer errors are independently measured, and the 'w/o Neural Fusion' ablation provides an internal control. Concerns about small Chamfer gaps, lack of error bars, and possible protocol differences are correctness/statistical risks, not circularity. The stated limitation about occluders being absorbed into the mesh is a robustness limitation, not a circular step. No specific reduction of the paper's predictions to its inputs by construction was found.

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

No new physical entities or representations are invented beyond the architectural combination. The free parameters are mostly unreported training constants, which slightly weakens reproducibility rather than circularity. The domain assumptions about VGGT and Pixel3DMM are the primary load-bearing premises.

free parameters (4)
  • UV resolution S = 512
    Choice of UV map resolution; larger S could improve detail but not explored. Not fitted to the test set.
  • splatting threshold tau_w = not reported
    Threshold on accumulated splatting weight for mask validity (Eq. 10); value not stated in paper.
  • loss weights lambda_pm, lambda_v, lambda_lap = not reported
    Weighting of point-map, vertex, and Laplacian losses in Eq. 20; values not reported.
  • number of training views sampled (4-16) = random 4-16 per subject
    Training data augmentation choice; could influence robustness claims.
assumptions (5)
  • domain assumption VGGT point maps are accurate enough, especially for in-the-wild geometry with expressions
    The entire input to fusion is VGGT point maps; if they are systematically wrong, no fusion can fix them. The paper's own limitations (bangs, masks) demonstrate this failure mode.
  • domain assumption Pixel3DMM UV correspondences are reliable for in-the-wild faces, expressions, and occlusions
    UV correspondence quality directly determines the quality of the lifted point maps. Mask/occlusion failure is admitted in §4.6.
  • domain assumption FLAME-template UV parameterization is a sufficient canonical space
    All geometry is represented in FLAME UV space; regions outside FLAME UV (hair, neck, ears) are excluded by design.
  • domain assumption Chamfer distance after similarity+ICP alignment is the right and sufficient accuracy metric
    The choice of alignment and metric can erase global-scale or expression errors; the paper uses only Chamfer, no surface normals, no per-landmark errors.
  • domain assumption Ava-256 registered meshes are a valid supervision distribution for in-the-wild expression geometry
    Training only on Ava-256 and evaluating on H3DS/NeRSemble/EmoTalk3D/RenderMe-360 assumes expressions and geometry transfer.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UVFaceFusion: Fast Multi-view Topologically Consistent Face Reconstruction in the Wild via UV-space Neural Fusion." pith.science (2026). https://pith.science/paper/5MCEPQA2

@misc{pith2026260718798,
  author       = {Pith},
  title        = {Pith review of: UVFaceFusion: Fast Multi-view Topologically Consistent Face Reconstruction in the Wild via UV-space Neural Fusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5MCEPQA2}},
  note         = {Machine review of arXiv:2607.18798}
}
read the original abstract

Reconstructing high-fidelity facial geometry with an assigned topology is essential for digital avatar creation and animation, yet existing automated methods often trade off geometric fidelity and in-the-wild generalization. We present UVFaceFusion, a feed-forward framework for multi-view, fixed-topology face reconstruction from daily images. Our key idea is to replace heuristic topological optimization with learnable neural fusion in a canonical UV space. Given multi-view images, we first obtain dense point maps and facial UV correspondences of each view using VGGT and Pixel3DMM, respectively. Then, the view-specific point maps are lifted into the canonical UV domain and fused with a novel mask-aware neural fusion network. The network predicts a complete UV-space point map, from which a fixed-topology mesh is directly sampled. Although trained only on Ava-256, UVFaceFusion generalizes well to multiple public benchmarks and in-the-wild captures, benefiting from its canonical UV-space geometry-to-geometry fusion that reduces dependence on dataset-specific appearance and capture conditions. Experiments on various benchmarks show that UVFaceFusion achieves state-of-the-art reconstruction accuracy while reconstructing a mesh from 16 input views in less than 3 seconds on a single RTX 4090. Code is available at https://github.com/grignarder/UVFaceFusion.

Figures

Figures reproduced from arXiv: 2607.18798 by the authors.

Figure 1
Figure 1. Given casual in-the-wild multi-view captures as input, with four of the sixteen images shown here, our method reconstructs a high-fidelity mesh with [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Pipeline overview. Given multi-view face images {𝐼𝑖 } 𝑁 𝑖=1 , we use frozen pretrained models to predict pixel-aligned point maps {𝑋𝑖 } 𝑁 𝑖=1 and dense UV correspondences {𝑈𝑖 } 𝑁 𝑖=1 . Through image-to-UV geometry lifting and canonicalization with the template mesh M0, each view is converted into a partial canonical UV-space point map {𝑃 𝑢𝑣 𝑖 } 𝑁 𝑖=1 with a valid mask {𝐴 𝑢𝑣 𝑖 } 𝑁 𝑖=1 . Our mask-aware neural fusion n… view at source ↗
Figure 3
Figure 3. User study results. Participants compared the reconstruction re [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (8 more)
Figure 5
Figure 5. Figure 5: Limitations of our method. Our method can absorb bangs into the [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 4
Figure 4. Figure 4: Qualitative evaluation of the key design choices in our method. With [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 6
Figure 6. Figure 6: Qualitative comparisons on public datasets. Our method reconstructs cleaner fixed-topology meshes and better preserves challenging expression geometry around the eyes, mouth, and cheeks. Error maps are shown beside each reconstruction. ACM Trans. Graph., Vol. 1, No. 1,…
Figure 7
Figure 7. Figure 7: Qualitative comparisons on in-the-wild captures. Our method generalizes to casual multi-view images and produces stable, faithful face meshes across diverse identities, viewpoints, lighting conditions, and expressions. Error maps are shown beside each reconstruction. A…
Figure 8
Figure 8. Figure 8: Montage of densely sampled frames from one of our self-captured [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Monocular reconstruction results. Although our method is mainly [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Qualitative comparison under the 4-view input setting. For compactness, we show one reference image from the input views in the first column, while [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 11
Figure 11. Figure 11: Screenshot of the web interface used in our user study. Participants are shown the input multi-view images and four anonymized reconstruction [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

154 extracted references · 1 canonical work pages

  1. [1]

    Abril and Robert Plant

    Patricia S. Abril and Robert Plant. The patent holder's dilemma: Buy, sell, or troll?. Communications of the ACM. 2007. doi:10.1145/1188913.1188915

    arXiv 2007

  2. [2]

    Deciding equivalances among conjunctive aggregate queries

    Sarah Cohen and Werner Nutt and Yehoshua Sagic. Deciding equivalances among conjunctive aggregate queries. 2007. doi:10.1145/1219092.1219093

    arXiv 2007

  3. [3]

    Special issue: Digital Libraries. 1996

    1996

  4. [4]

    Understanding Policy-Based Networking

    David Kosiur. Understanding Policy-Based Networking. 2001

    2001

  5. [7]

    The title of book two. 2008. doi:10.1007/3-540-09237-4

    doi:10.1007/3-540-09237-4 2008

  6. [8]

    Asad Z. Spector. Achieving application requirements. Distributed Systems. 1990. doi:10.1145/90417.90738

    arXiv 1990

  7. [9]

    Douglass and David Harel and Mark B

    Bruce P. Douglass and David Harel and Mark B. Trakhtenbrot. Statecarts in use: structured analysis and object-orientation. Lectures on Embedded Systems. 1998. doi:10.1007/3-540-65193-4_29

    doi:10.1007/3-540-65193-4_29 1998

  8. [10]

    Donald E. Knuth. The Art of Computer Programming, Vol. 1: Fundamental Algorithms (3rd. ed.). 1997

    1997

Show all 154 references
  1. [11]

    Donald E. Knuth. The Art of Computer Programming. 1998

    1998

  2. [12]

    Structured Variational Inference Procedures and their Realizations (as incol)

    Dan Geiger and Christopher Meek. Structured Variational Inference Procedures and their Realizations (as incol). Proceedings of Tenth International Workshop on Artificial Intelligence and Statistics, The Barbados

  3. [13]

    Stan W. Smith. An experiment in bibliographic mark-up: Parsing metadata for XML export. Proceedings of the 3rd. annual workshop on Librarians and Computers. 2010. doi:99.9999/woot07-S422

    2010

  4. [14]

    Catch me, if you can: Evading network signatures with web-based polymorphic worms

    Matthew Van Gundy and Davide Balzarotti and Giovanni Vigna. Catch me, if you can: Evading network signatures with web-based polymorphic worms. Proceedings of the first USENIX workshop on Offensive Technologies. 2007

    2007

  5. [15]

    Catch me, if you can: Evading network signatures with web-based polymorphic worms

    Matthew Van Gundy and Davide Balzarotti and Giovanni Vigna. Catch me, if you can: Evading network signatures with web-based polymorphic worms. Proceedings of the first USENIX workshop on Offensive Technologies. 2008

    2008

  6. [16]

    Catch me, if you can: Evading network signatures with web-based polymorphic worms

    Matthew Van Gundy and Davide Balzarotti and Giovanni Vigna. Catch me, if you can: Evading network signatures with web-based polymorphic worms. Proceedings of the first USENIX workshop on Offensive Technologies. 2009

    2009

  7. [17]

    Predicate Path expressions

    Sten Andler. Predicate Path expressions. Proceedings of the 6th. ACM SIGACT-SIGPLAN symposium on Principles of Programming Languages. 1979. doi:10.1145/567752.567774

    1979

  8. [18]

    LOGICS of Programs: AXIOMATICS and DESCRIPTIVE POWER

    David Harel. LOGICS of Programs: AXIOMATICS and DESCRIPTIVE POWER. 1978

    1978

  9. [19]

    Anisi , title =

    David A. Anisi , title =

  10. [20]

    Clarkson

    Kenneth L. Clarkson. Algorithms for Closest-Point Problems (Computational Geometry). 1985

    1985

  11. [21]

    Introduction to Bayesian Statistics

    Harry Thornburg. Introduction to Bayesian Statistics. 2001

    2001

  12. [22]

    CLIFFORD: a Maple 11 Package for Clifford Algebra Computations, version 11

    Rafal Ablamowicz and Bertfried Fauser. CLIFFORD: a Maple 11 Package for Clifford Algebra Computations, version 11. 2007

    2007

  13. [23]

    Stats and Analysis

    Poker-Edge.Com. Stats and Analysis. 2006

    2006

  14. [24]

    A more perfect union

    Barack Obama. A more perfect union. 2008

    2008

  15. [25]

    The fountain of youth

    Joseph Scientist. The fountain of youth. 2009

    2009

  16. [26]

    Solder man

    Dave Novak. Solder man. ACM SIGGRAPH 2003 Video Review on Animation theater Program: Part I - Vol. 145 (July 27--27, 2003). 2003. doi:99.9999/woot07-S422

    2003

  17. [27]

    Interview with Bill Kinder: January 13, 2005

    Newton Lee. Interview with Bill Kinder: January 13, 2005. Comput. Entertain. 2005. doi:10.1145/1057270.1057278

    2005

  18. [28]

    The Enabling of Digital Libraries

    Bernard Rous. The Enabling of Digital Libraries. Digital Libraries. 2008

    2008

  19. [30]

    (new) Finding minimum congestion spanning trees , journal =

    Werneck, Renato and Setubal, Jo\. (new) Finding minimum congestion spanning trees , journal =. 2000 , issn =. doi:10.1145/351827.384253 , acmid =

    2000

  20. [32]

    and Mei, Alessandro , title =

    Conti, Mauro and Di Pietro, Roberto and Mancini, Luigi V. and Mei, Alessandro , title =. Inf. Fusion , volume =. 2009 , issn =. doi:10.1016/j.inffus.2009.01.002 , acmid =

    2009 doi

  21. [33]

    and Hutchful, David K

    Li, Cheng-Lun and Buyuktur, Ayse G. and Hutchful, David K. and Sant, Natasha B. and Nainwal, Satyendra K. , title =. CHI '08 extended abstracts on Human factors in computing systems , year =. doi:10.1145/1358628.1358946 , acmid =

  22. [34]

    , title =

    Hollis, Billy S. , title =. 1999 , isbn =

    1999

  23. [35]

    Goossens, Michel and Rahtz, S. P. and Moore, Ross and Sutor, Robert S. , title =. 1999 , isbn =

    1999

  24. [36]

    and Rosenberg, Arnold L

    Buss, Jonathan F. and Rosenberg, Arnold L. and Knott, Judson D. , title =. 1987 , source =

    1987

  25. [37]

    CHI '08: CHI '08 extended abstracts on Human factors in computing systems , year =

    , note =. CHI '08: CHI '08 extended abstracts on Human factors in computing systems , year =

  26. [38]

    Algorithms for Closest-Point Problems (Computational Geometry) , year =

    Clarkson, Kenneth Lee , advisor =. Algorithms for Closest-Point Problems (Computational Geometry) , year =

  27. [39]

    SIGCOMM Comput. Commun. Rev. , year =

  28. [40]

    2004 , isbn =

    IEEE TCSC Executive Committee , booktitle =. 2004 , isbn =. doi:http://dx.doi.org/10.1109/ICWS.2004.64 , acmid =

    2004 doi

  29. [41]

    Distributed systems (2nd Ed.) , year =

  30. [42]

    , title =

    Petrie, Charles J. , title =. 1986 , source =

    1986

  31. [43]

    Donald E. Knuth. Seminumerical Algorithms. 1981

    1981

  32. [44]

    E-commerce and cultural values , year =

    Kong, Wei-Chang , Title =. E-commerce and cultural values , year =

  33. [45]

    E-commerce and cultural values , year =

    Kong, Wei-Chang , type =. E-commerce and cultural values , year =

  34. [46]

    Chapter 9 , booktitle =

    Kong, Wei-Chang , editor =. Chapter 9 , booktitle =. 2002 , address =

    2002

  35. [47]

    E-commerce and cultural values , editor =

    Kong, Wei-Chang , title =. E-commerce and cultural values , editor =. 2003 , isbn =

    2003

  36. [48]

    E-commerce and cultural values - (InBook-num-in-chap) , chapter =

    Kong, Wei-Chang , editor =. E-commerce and cultural values - (InBook-num-in-chap) , chapter =. 2004 , address =

    2004

  37. [49]

    E-commerce and cultural values (Inbook-text-in-chap) , chapter =

    Kong, Wei-Chang , editor =. E-commerce and cultural values (Inbook-text-in-chap) , chapter =. 2005 , address =

    2005

  38. [50]

    E-commerce and cultural values (Inbook-num chap) , chapter =

    Kong, Wei-Chang , editor =. E-commerce and cultural values (Inbook-num chap) , chapter =. 2006 , address =

    2006

  39. [51]

    Microelectron

    Mehdi Saeedi and Morteza Saheb Zamani and Mehdi Sedighi , title =. Microelectron. J. , volume =. 2010 , pages =

    2010

  40. [52]

    Mehdi Saeedi and Morteza Saheb Zamani and Mehdi Sedighi and Zahra Sasanian , title =. J. Emerg. Technol. Comput. Syst. , volume =

  41. [53]

    Kirschmer, Markus and Voight, John , title =. SIAM J. Comput. , issue_date =. 2010 , issn =. doi:https://doi.org/10.1137/080734467 , acmid =

    2010 doi

  42. [54]

    Hoare, C. A. R. , title =. Structured programming (incoll) , editor =. 1972 , isbn =

    1972

  43. [55]

    History of programming languages I (incoll) , editor =

    Lee, Jan , title =. History of programming languages I (incoll) , editor =. 1981 , isbn =. doi:http://doi.acm.org/10.1145/800025.1198348 , acmid =

    1981

  44. [56]

    , title =

    Dijkstra, E. , title =. Classics in software engineering (incoll) , year =

  45. [57]

    , title =

    Wenzel, Elizabeth M. , title =. Multimedia interface design (incoll) , year =. doi:10.1145/146022.146089 , acmid =

  46. [58]

    , title =

    Mumford, E. , title =. Critical issues in information systems research (incoll) , year =

  47. [59]

    and Golden, Donald G

    McCracken, Daniel D. and Golden, Donald G. , title =. 1990 , isbn =

    1990

  48. [60]

    The analysis of linear partial differential operators

    H. The analysis of linear partial differential operators. 1985 , PAGES =

    1985

  49. [61]

    IEEE", address =

    A. Adya and P. Bahl and J. Padhye and A.Wolman and L. Zhou , title =. Proceedings of the IEEE 1st International Conference on Broadnets Networks (BroadNets'04) , publisher = "IEEE", address = "Los Alamitos, CA", year =

  50. [62]

    I. F. Akyildiz and W. Su and Y. Sankarasubramaniam and E. Cayirci , title =. Comm. ACM , volume = 38, number = "4", year =

  51. [63]

    I. F. Akyildiz and T. Melodia and K. R. Chowdhury , title =. Computer Netw. , volume = 51, number = "4", year =

  52. [64]

    ACM", address =

    P. Bahl and R. Chancre and J. Dungeon , title =. Proceeding of the 10th International Conference on Mobile Computing and Networking (MobiCom'04) , publisher = "ACM", address = "New York, NY", year =

  53. [65]

    8 (Special Issue on Sensor Networks)

    D. Culler and D. Estrin and M. Srivastava , title =. IEEE Comput. , volume = 37, number = "8 (Special Issue on Sensor Networks)", publisher = "IEEE", address = "Los Alamitos, CA", year =

  54. [66]

    Natarajan and M

    A. Natarajan and M. Motani and B. de Silva and K. Yap and K. C. Chua , title =. Network Architectures , editor =. 960935712

  55. [67]

    Tzamaloukas and J

    A. Tzamaloukas and J. J. Garcia-Luna-Aceves , title =

  56. [68]

    Zhou and J

    G. Zhou and J. Lu and C.-Y. Wan and M. D. Yarvis and J. A. Stankovic , title =

  57. [69]

    Mapping Powerlists onto Hypercubes

    Jacob Kornerup. Mapping Powerlists onto Hypercubes. 1994

    1994

  58. [70]

    Automatic Parallelization for Distributed-Memory Multiprocessing Systems

    Michael Gerndt. Automatic Parallelization for Distributed-Memory Multiprocessing Systems

  59. [71]

    J. E. Archer, Jr. and R. Conway and F. B. Schneider. User recovery and reversal in interactive systems. ACM Trans. Program. Lang. Syst

  60. [72]

    D. D. Dunlop and V. R. Basili. Generalizing specifications for uniformly implemented loops. ACM Trans. Program. Lang. Syst

  61. [73]

    Heering and P

    J. Heering and P. Klint. Towards monolingual programming environments. ACM Trans. Program. Lang. Syst

  62. [74]

    Donald E. Knuth. The book

  63. [75]

    Korach and D

    E. Korach and D. Rotem and N. Santoro. Distributed algorithms for finding centers and medians in networks. ACM Trans. Program. Lang. Syst

  64. [76]

    : A Document Preparation System

    Leslie Lamport. : A Document Preparation System

  65. [77]

    F. Nielson. Program transformations in a denotational setting. ACM Trans. Program. Lang. Syst

  66. [78]

    Brian K. Reid. A high-level approach to computer document formatting. Proceedings of the 7th Annual Symposium on Principles of Programming Languages

  67. [79]

    and Abdelzaher, Tarek F

    Zhou, Gang and Wu, Yafeng and Yan, Ting and He, Tian and Huang, Chengdu and Stankovic, John A. and Abdelzaher, Tarek F. , title =. ACM Trans. Embed. Comput. Syst. , issue_date =. doi:10.1145/1721695.1721705 , acmid = 1721705, publisher =

  68. [80]

    Institutional members of the Users Group

  69. [81]

    Boris Veytsman , title =

  70. [82]

    and Peterson, Larry L

    Bowman, Mic and Debray, Saumya K. and Peterson, Larry L. , title =. ACM Trans. Program. Lang. Syst. , volume =. 1993 , doi =

    1993

  71. [83]

    TUGboat , volume =

    Braams, Johannes , title =. TUGboat , volume =

  72. [84]

    Post Congress Tristesse

    Malcolm Clark. Post Congress Tristesse. TeX90 Conference Proceedings

  73. [85]

    ACM Trans

    Herlihy, Maurice , title =. ACM Trans. Program. Lang. Syst. , volume =. 1993 , doi =

    1993

  74. [86]

    Salas and Einar Hille

    S.L. Salas and Einar Hille. Calculus: One and Several Variable. 1978

    1978

  75. [87]

    Publication quality tables in

    Simon Fear , month =. Publication quality tables in

  76. [88]

    Using the amsthm Package , organization =

  77. [89]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Topologically Consistent Multi-View Face Inference Using Volumetric Sampling , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  78. [90]

    Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Instant Multi-View Head Capture through Learnable Registration , author=. Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  79. [91]

    European Conference on Computer Vision , pages=

    GRAPE: Generalizable and Robust Multi-view Facial Capture , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  80. [92]

    2023 , eprint=

    A Hierarchical Representation Network for Accurate and Detailed Face Reconstruction from In-The-Wild Images , author=. 2023 , eprint=

    2023

  81. [93]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Deep Facial Non-Rigid Multi-View Stereo , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  82. [94]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Fml: Face model learning from videos , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  83. [95]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Vggt: Visual geometry grounded transformer , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  84. [96]

    2025 , url=

    Pixel3DMM: Versatile Screen-Space Priors for Single-Image 3D Face Reconstruction , author=. 2025 , url=

    2025

  85. [97]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Mvf-net: Multi-view 3d face morphable model regression , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  86. [98]

    IEEE Computer Vision and Pattern Recognition Workshops , year=

    Accurate 3D Face Reconstruction with Weakly-Supervised Learning: From Single Image to Image Set , author=. IEEE Computer Vision and Pattern Recognition Workshops , year=

  87. [99]

    and Bolkart, Timo , journal =

    Feng, Yao and Feng, Haiwen and Black, Michael J. and Bolkart, Timo , journal =. Learning an Animatable Detailed. 2021 , url =

    2021

  88. [100]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    3D Face Reconstruction with the Geometric Guidance of Facial Part Segmentation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  89. [101]

    Acm siggraph 2009 courses , pages=

    The digital emily project: photoreal facial modeling and animation , author=. Acm siggraph 2009 courses , pages=

    2009

  90. [102]

    ACM SIGGRAPH 2013 Posters , pages=

    Digital ira: Creating a real-time photoreal digital actor , author=. ACM SIGGRAPH 2013 Posters , pages=

    2013

  91. [103]

    , author=

    Single-shot high-quality facial geometry and skin appearance capture. , author=. ACM Trans. Graph. , volume=

  92. [104]

    Structure-from-Motion Revisited , booktitle=

    Sch\". Structure-from-Motion Revisited , booktitle=

  93. [105]

    Pixelwise View Selection for Unstructured Multi-View Stereo , booktitle=

    Sch\". Pixelwise View Selection for Unstructured Multi-View Stereo , booktitle=

  94. [106]

    Large Steps in Inverse Rendering of Geometry

    Nicolet, Baptiste and Jacobson, Alec and Jakob, Wenzel. Large Steps in Inverse Rendering of Geometry. ACM Transactions on Graphics (Proceedings of SIGGRAPH Asia). 2021. doi:10.1145/3478513.3480501

    2021

  95. [107]

    International workshop on vision algorithms , pages=

    Bundle adjustment—a modern synthesis , author=. International workshop on vision algorithms , pages=. 1999 , organization=

    1999

  96. [108]

    2023 , month =

    Agarwal, Sameer and Mierle, Keir and The Ceres Solver Team , title =. 2023 , month =

    2023

  97. [109]

    Li, Tianye and Bolkart, Timo and Black, Michael. J. and Li, Hao and Romero, Javier , journal =. Learning a model of facial shape and expression from. 2017 , url =

    2017

  98. [110]

    CVPR , year=

    Han, Yuxuan and Wang, Zhibo and Xu, Feng , title =. CVPR , year=

  99. [111]

    Self-Supervised Multi-Level Face Model Learning for Monocular Reconstruction at Over 250 Hz , booktitle =

    Ayush Tewari and Michael Zollh. Self-Supervised Multi-Level Face Model Learning for Monocular Reconstruction at Over 250 Hz , booktitle =. 2018 , url =. doi:10.1109/CVPR.2018.00270 , timestamp =

    2018

  100. [112]

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Yang, Haotian and Zhu, Hao and Wang, Yanru and Huang, Mingkai and Shen, Qiu and Yang, Ruigang and Cao, Xun , title =. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2020 , page =

    2020

  101. [113]

    A Morphable Model for the Synthesis of 3D Faces , booktitle =

    Volker Blanz and Thomas Vetter , editor =. A Morphable Model for the Synthesis of 3D Faces , booktitle =. 1999 , url =

    1999

  102. [114]

    Computer Graphics Forum , volume=

    Practical face reconstruction via differentiable ray tracing , author=. Computer Graphics Forum , volume=. 2021 , organization=

    2021

  103. [115]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    MoSAR: Monocular semi-supervised model for avatar reconstruction using differentiable shading , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  104. [116]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Face2face: Real-time face capture and reenactment of rgb videos , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  105. [117]

    Computer graphics forum , volume=

    State of the art on monocular 3D face reconstruction, tracking, and applications , author=. Computer graphics forum , volume=. 2018 , organization=

    2018

  106. [118]

    Mallikarjun B. R. and Ayush Tewari and Hans. Learning Complete 3D Morphable Face Models From Images and Videos , booktitle =. 2021 , url =. doi:10.1109/CVPR46437.2021.00337 , timestamp =

    2021

  107. [119]

    Proceedings of the AAAI Conference on Artificial Intelligence , year=

    RAFaRe: Learning Robust and Accurate Non-parametric 3D Face Reconstruction from Pseudo 2D&3D Pairs , author=. Proceedings of the AAAI Conference on Artificial Intelligence , year=

  108. [120]

    arxiv , year=

    Unrestricted Facial Geometry Reconstruction Using Image-to-Image Translation , author=. arxiv , year=

  109. [121]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Zhang, Zhenyu and Ge, Yanhao and Chen, Renwang and Tai, Ying and Yan, Yan and Yang, Jian and Wang, Chengjie and Li, Jilin and Huang, Feiyue , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  110. [122]

    2019 , url =

    Fanzi Wu and Linchao Bao and Yajing Chen and Yonggen Ling and Yibing Song and Songnan Li and King Ngi Ngan and Wei Liu , title =. 2019 , url =. doi:10.1109/CVPR.2019.00105 , timestamp =

    2019

  111. [123]

    Debevec , title =

    Graham Fyffe and Koki Nagano and Loc Huynh and Shunsuke Saito and Jay Busch and Andrew Jones and Hao Li and Paul E. Debevec , title =. Comput. Graph. Forum , volume =. 2017 , url =. doi:10.1111/CGF.13127 , timestamp =

    2017 doi

  112. [124]

    CoRR , volume =

    Yunze Xiao and Hao Zhu and Haotian Yang and Zhengyu Diao and Xiangju Lu and Xun Cao , title =. CoRR , volume =. 2022 , url =. 2201.01016 , timestamp =

    2022 arXiv

  113. [125]

    Learning Neural Parametric Head Models , booktitle =

    Simon Giebenhain and Tobias Kirschstein and Markos Georgopoulos and Martin R. Learning Neural Parametric Head Models , booktitle =

  114. [126]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    ImFace: A Nonlinear 3D Morphable Face Model with Implicit Neural Representations , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  115. [127]

    arXiv preprint arXiv:2107.12512 , booktitle=

    H3D-Net: Few-Shot High-Fidelity 3D Head Reconstruction , author=. arXiv preprint arXiv:2107.12512 , booktitle=

    arXiv

  116. [128]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    General facial representation learning in a visual-linguistic manner , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  117. [129]

    The Thirteenth International Conference on Learning Representations , year=

    Scaling In-the-Wild Training for Diffusion-based Illumination Harmonization and Editing by Imposing Consistent Light Transport , author=. The Thirteenth International Conference on Learning Representations , year=

  118. [130]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Switchlight: Co-design of physics-driven architecture and pre-training framework for human portrait relighting , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  119. [131]

    ArXiv , month=

    Human Hair Reconstruction with Strand-Aligned 3D Gaussians , author=. ArXiv , month=

  120. [132]

    European Conference on Computer Vision , pages=

    High-quality mesh blendshape generation from face videos via neural inverse rendering , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  121. [133]

    CVPR , year=

    Han, Yuxuan and Lyu, Junfeng and Xu, Feng , title =. CVPR , year=

  122. [134]

    SIGGRAPH , year=

    Han, Yuxuan and Lyu, Junfeng and Sheng, Kuan and Que, Minghao and Zhang, Qixuan and Xu, Lan and Xu, Feng , title =. SIGGRAPH , year=

  123. [135]

    NeRSemble: Multi-View Radiance Field Reconstruction of Human Heads , year =

    Kirschstein, Tobias and Qian, Shenhan and Giebenhain, Simon and Walter, Tim and Nie. NeRSemble: Multi-View Radiance Field Reconstruction of Human Heads , year =. doi:10.1145/3592455 , journal =

    doi

  124. [136]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Implicit shape and appearance priors for few-shot full head reconstruction , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  125. [137]

    2023 , booktitle=

    NeuS2: Fast Learning of Neural Implicit Surfaces for Multi-view Reconstruction , author=. 2023 , booktitle=

    2023

  126. [138]

    2003 , publisher=

    Multiple view geometry in computer vision , author=. 2003 , publisher=

    2003

  127. [139]

    Proceedings of the International Conference on Computer Vision (ICCV) , year=

    Emerging Properties in Self-Supervised Vision Transformers , author=. Proceedings of the International Conference on Computer Vision (ICCV) , year=

  128. [140]

    arXiv:2304.07193 , year=

    DINOv2: Learning Robust Visual Features without Supervision , author=. arXiv:2304.07193 , year=

    arXiv

  129. [141]

    2021 , eprint=

    High-Resolution Image Synthesis with Latent Diffusion Models , author=. 2021 , eprint=

    2021

  130. [142]

    2025 , eprint=

    Avat3r: Large Animatable Gaussian Reconstruction Model for High-fidelity 3D Head Avatars , author=. 2025 , eprint=

    2025

  131. [143]

    Ren, Xingyu and Deng, Jiankang and Cheng, Yuhao and Zhu, Wenhan and Yan, Yichao and Yang, Xiaokang and Zafeiriou, Stefanos and Ma, Chao , booktitle=. S\^

  132. [144]

    European conference on computer vision , pages=

    Towards metrical reconstruction of human faces , author=. European conference on computer vision , pages=. 2022 , organization=

    2022

  133. [145]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Emoca: Emotion driven monocular face capture and animation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  134. [146]

    arXiv preprint arXiv:2404.04104 , year=

    SMIRK: 3D Facial Expressions through Analysis-by-Neural-Synthesis , author=. arXiv preprint arXiv:2404.04104 , year=

    arXiv

  135. [147]

    Proceedings of the European conference on computer vision (ECCV) , pages=

    Joint 3d face reconstruction and dense alignment with position map regression network , author=. Proceedings of the European conference on computer vision (ECCV) , pages=

  136. [148]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    WarpHE4D: Dense 4D Head Map toward Full Head Reconstruction , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  137. [149]

    ACM Transactions on Graphics (ToG) , volume=

    3d morphable face models—past, present, and future , author=. ACM Transactions on Graphics (ToG) , volume=. 2020 , publisher=

    2020

  138. [150]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Learning detailed face reconstruction from a single image , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  139. [151]

    Computer Graphics Forum , volume=

    Multi-view stereo on consistent face topology , author=. Computer Graphics Forum , volume=. 2017 , organization=

    2017

  140. [152]

    ACM Transactions on Graphics (TOG) , volume=

    Rapid face asset acquisition with recurrent feature alignment , author=. ACM Transactions on Graphics (TOG) , volume=. 2022 , publisher=

    2022

  141. [153]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    VGGTFace: Topologically Consistent Facial Geometry Reconstruction in the Wild , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  142. [154]

    arXiv preprint arXiv:2605.01450 , year=

    Registration-Free Learnable Multi-View Capture of Faces in Dense Semantic Correspondence , author=. arXiv preprint arXiv:2605.01450 , year=

    arXiv

  143. [155]

    IEEE Transactions on pattern analysis and machine intelligence , volume=

    Least-squares estimation of transformation parameters between two point patterns , author=. IEEE Transactions on pattern analysis and machine intelligence , volume=. 1991 , publisher=

    1991

  144. [156]

    Jewett and Simon Venshtain and Christopher Heilman and Yueh-Tung Chen and Sidi Fu and Mohamed Ezzeldin A

    Julieta Martinez and Emily Kim and Javier Romero and Timur Bagautdinov and Shunsuke Saito and Shoou-I Yu and Stuart Anderson and Michael Zollhöfer and Te-Li Wang and Shaojie Bai and Chenghui Li and Shih-En Wei and Rohan Joshi and Wyatt Borsos and Tomas Simon and Jason Saragih ...

    2024

  145. [157]

    European Conference on Computer Vision (ECCV) , year=

    EmoTalk3D: High-Fidelity Free-View Synthesis of Emotional 3D Talking Head , author=. European Conference on Computer Vision (ECCV) , year=

  146. [158]

    Advances in Neural Information Processing Systems , volume=

    RenderMe-360: A Large Digital Asset Library and Benchmarks Towards High-fidelity Head Avatars , author=. Advances in Neural Information Processing Systems , volume=

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.