Pith. sign in

REVIEW 4 major objections 5 minor 70 references

Point2Quad: Generating Quad Meshes from Point Clouds via Face Prediction

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Point2Quad claims that generating quad-only meshes from point clouds reduces to classifying candidate faces, with experiments on 1,641 models backing the claim.

desk verdict A solid first learning-based point-to-quad face-prediction pipeline with a real open question about whether its candidate generation actually covers the true faces. read the letter →

arxiv 2504.19545 v1 pith:VTTWYO3B submitted 2025-04-28 cs.CV cs.AI

Zezeng Li , Zhihui Qi , Weimin Wang , Ziliang Wang , Junyi Duan , Na Lei This is my paper · ORCID
classification cs.CVcs.AI
keywords quadmeshgenerationpointcloudfacepredictionlearning-basedcandidatequadrilateralclassificationqualitywatertightnessquad-only
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

Point2Quad sets out to prove that generating quadrilateral-only meshes from raw point clouds can be treated as a per-face classification problem, and that a network which scores candidate quads outperforms prior remeshing and edge-prediction baselines on both surface fit and mesh quality. The motivation is practical: quads are preferred in simulation, CAD, and isogeometric analysis, yet their coplanarity, convexity, and squareness constraints make direct generation harder than triangle meshing. The method builds a small candidate pool per point using $k$-nearest neighbors, filters it with geometric thresholds, and trains a classifier on fused pointwise and facewise features. If the claim holds, quad-only meshes from noisy scans become available without parameterization, Voronoi tessellation, or an intermediate triangle mesh.

What carries the argument

The load-bearing mechanism is the candidate-face classifier. For each point, the 12 nearest Euclidean neighbors are sampled in random triples, ordered counter-clockwise, and filtered by thresholds on min-max edge ratio (0.25), sine of interior angles (0.3), and normal-dot-product coplanarity (0.5), leaving 12 candidate quads per point. Geometric features come from a point-convolution encoder, and facewise features come from a second encoder operating on a 29-dimensional per-candidate vector that includes coordinates, scaled Jacobian, four sine values, and four vertex normals; the concatenated features are scored by a shared-weight MLP. This framing converts the non-differentiable selection of faces into differentiable scoring, and the compound loss (weighted cross-entropy plus a face loss that penalizes confident mistakes) together with the score-based post-processing is what turns classifier outputs into watertight, manifold quad meshes.

What would settle it

Take a high-curvature or anisotropically sampled object (for example a thin tube or sharp ridge), build the candidate pool exactly as described with $k=12$, and compute the fraction of ground-truth quadrilateral faces that appear among the candidates. If that candidate recall is materially below the recall the classifier reports, the bottleneck is candidate generation rather than the classifier, and no training change within this pipeline can recover those faces; this directly tests the paper's stated degradation on sparse and non-uniform inputs.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that Point2Quad is the first learning-based method for quad-only mesh generation directly from point clouds. The combinatorial problem of choosing which vertices connect into quads is bypassed by generating candidate quadrilateral faces and asking a binary classifier which candidates lie on the ground-truth surface. The authors report that a simple shared-MLP classifier, fed by concatenated pointwise geometric features and facewise features (vertex coordinates, scaled Jacobian, sine of interior angles, and vertex normals), achieves precision/recall near 0.94/0.99 on data with 10 percent noise and produces meshes with scaled Jacobian 0.986, watertightness 0.99998, and Chamfer distance 0.02064, beating the Instant-Meshes and IER baselines on almost every metric. They further show the contribution of each component through ablations.

Load-bearing premise

The method depends on the assumption that, for every vertex, the true neighboring vertices of the target quad mesh are among its 12 nearest Euclidean neighbors and that the random triple sampling places the true face among the 12 candidate faces; on sparse, non-uniform, or high-curvature regions this coverage can fail, and then the correct face cannot be predicted at all.

Editorial extensions

If this is right

  • Quad-only output is guaranteed by construction, in contrast to quad-dominant baselines that may mix triangles or leave holes.
  • Noise robustness: at 10 percent normal-direction noise the reported quality metrics degrade only slightly, while both baselines degrade markedly.
  • The method is stable across point-cloud resolutions from roughly 2,000 to over 10,000 points, with scaled Jacobian staying near 0.98 in every bin.
  • Ablations indicate the face encoder, face loss, and post-processing are each load-bearing: removing the face encoder raises angle distortion and Chamfer distance, removing the face loss can produce non-finite (NaN) loss in training, and removing post-processing drops watertightness from 0.99998 to 0.83587.
  • The authors state the output is not guaranteed 100 percent watertight or manifold on low-quality inputs, and generalization to sparse or highly non-uniform point clouds remains a limitation.

Reading between the lines

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

  • A testable ceiling on the method is candidate coverage: if the true quad of a vertex is not among the 12 candidates (for example in strongly anisotropic or high-curvature regions), no classifier can recover it; making $k$ adaptive to local density or curvature is the natural next step.
  • Because the post-process can create new vertices by interpolation when filling holes, the final mesh vertices are not strictly a subset of the input points; applications that require exact point fidelity should verify this.
  • The face-prediction formulation should transfer to other element types, such as hexahedra or mixed-element meshes, by replacing the candidate template and the geometric filters while keeping the two-encoder classification pipeline.
  • The reliance on ground-truth quad meshes for labels may limit the variety of trainable topologies; a self-supervised or weakly supervised variant that scores faces by geometric consistency could extend the approach to arbitrary scans.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper presents Point2Quad, a supervised method that generates quad-only meshes from point clouds by classifying candidate quadrilateral faces. Given an input point cloud, it builds a k-nearest-neighbor graph (k=12), samples candidate quads per point, filters them by hand-set geometric thresholds (edge ratio, sine, normal dot product), extracts pointwise features with a KPConv backbone and facewise features with an MLP face encoder, and trains a classifier with a weighted cross-entropy loss plus a proposed face loss. At inference, predicted faces are merged and repaired by a greedy post-process. Experiments on a new Thing10K-derived dataset (1,641 models) compare against Instant-Meshes and IER and report better scaled Jacobian, edge ratio, angle distortion, watertightness, and Chamfer distance under both clean and 10%-noise inputs, with ablations on the face encoder, face loss, and post-process.

Significance. If the reported results hold, Point2Quad is a useful first supervised baseline for direct quad-mesh generation from point clouds and a reasonable engineering contribution: it releases code and data, evaluates with several standard mesh-quality and fidelity metrics, and includes ablations that isolate the main pipeline components. The empirical evaluation is honest supervised learning on a held-out test split, so there is no circularity concern. The headline claims should, however, be read in light of two gaps: the ceiling set by candidate generation is not validated, and the printed face-loss equation appears to be index-swapped. If these are corrected and the candidate-recall analysis is added, the paper would provide a solid reference point for learning-based quad meshing.

major comments (4)
  1. [Sec. III-B and IV-A] The classifier's ceiling is candidate generation, but the paper never verifies that ground-truth faces are contained in the candidate set. With k=12, a true quad centered at a vertex is recoverable only if (i) the three other vertices of the face are among the 12 Euclidean nearest neighbors, (ii) the random triple sampling in Sec. IV-A retains that exact triple among the C(12,3)=220 possible triples, and (iii) the face survives the fixed thresholds 0.25 (edge ratio), 0.3 (sine), and 0.5 (normal dot product). None of these conditions is reported. Consequently, the precision/recall values in Tables II and III measure selection within F_P, not recovery of ground-truth faces, and the near-unity watertightness (0.99998 in Table II) may largely come from the greedy hole-filling step of Sec. III-E rather than from learned prediction. Please report candidate recall (fraction of ground-truth faces present among candidates), stratified by point density and curvature, and analyze sensitivity to k and to the three thresholds. The conclusion's own caveat about sparse/non-uniform point clouds makes this analysis necessary rather than optional.
  2. [Eq. (9), Sec. III-D] As printed, the face loss is inconsistent with its stated purpose. For a sample with y_i=1, L_F contributes -exp(yhat_0^i), which is minimized by making yhat_0^i large, i.e., by predicting class 0; for y_i=0 it contributes -exp(yhat_1^i), pushing toward class 1. This is the opposite of Eq. (8) and of the sentence defining class 1 as 'on the ground truth surface.' If the implementation matches Eq. (9), the absence of this loss should improve the classifier, not degrade it as reported in Table V; if the equation contains an index swap, the corrected form and the code must be provided. Because the face loss is a claimed contribution, this discrepancy affects the validity of the ablation study.
  3. [Sec. III-A vs Sec. III-E] The problem formulation states that the output quad mesh S_P has vertices coming from the input point cloud P, yet the post-process explicitly creates new points by bilinear interpolation of edge vectors when filling boundary angles. This is a direct contradiction: the final mesh is not a face subset of the input geometry. The discrepancy is material because the new vertices affect Chamfer distance and the claim of quad-only reconstruction from the input points. Please revise the formulation to allow created vertices, or constrain the post-process to output vertices in P, and state which convention was used in Tables II-IV.
  4. [Sec. IV-A, Tables II-IV] The experimental section reports single mean values without error bars, significance tests, or fixed randomness. Candidate generation uses random triple sampling (Sec. IV-A) but no seed is documented, so repeated runs may yield different candidate sets and therefore different mesh results. The reported differences are sometimes small (e.g., Chamfer distance 0.02064 vs 0.02092 in Table II), and without variance estimates it is unclear whether they are meaningful. Please fix a seed, report results over multiple runs, and provide standard deviations or confidence intervals for the main metrics.
minor comments (5)
  1. [Tables II and III] The captions say 'Qualitative comparison' but the tables contain quantitative numbers; they should be captioned 'Quantitative comparison'.
  2. [Table I and Eq. (12)] Table I lists the range of Angle distortion as [0,90], but Eq. (12) defines a squared error in degrees squared, whose maximum is 8100; either normalize the metric or correct the range.
  3. [Sec. IV-A] There is a typo: 'We construct our dateset Point&Quad' should be 'dataset'; similarly, Section II.A heading has 'Learing-based' instead of 'Learning-based'.
  4. [Sec. IV-A] The construction of the 'quadrilateral version of Thing10K' dataset is not described; please state how the ground-truth quad meshes were obtained and how the 1,641 training/test models were balanced across shape classes.
  5. [Sec. I and Sec. III-E] The introduction claims that Point2Quad is 'guaranteed to feature only quads,' but Sec. III-E describes the hole-filling patterns informally and does not prove that every created face is a quadrilateral; please add a brief justification or soften the guarantee.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: Point2Quad is a held-out supervised learning pipeline whose predictions are not equivalent to its training labels or fitted parameters.

full rationale

Point2Quad's chain is an empirical learning pipeline, not a derivation that reduces to its own inputs. The model is trained with Eq. (7) against ground-truth candidate labels and evaluated on a disjoint test split (Sec. IV-A: 'We randomly split the dataset into 1,312 models (80%) for training and 312 models for testing. All quantitative results reported in this section represent average scores across the entire test set.'). No parameter is fitted to the test set, and the hand-set thresholds (k=12, aspect ratio 0.25, sine 0.3, normal dot product 0.5) are fixed design choices, not fitted values renamed as predictions. The near-watertight outputs are partly produced by the deterministic post-process of Sec. III-E, but the paper explicitly attributes this to post-processing ('The post-process primarily enhances the watertightness and manifold properties of the output meshes') and the ablation 'w/o PP' quantifies the drop; this is an honest decomposition, not a circular step. The only overlapping-author citations ([33], [35]) appear in related work and are not load-bearing; no uniqueness theorem or prior result is invoked to force the architecture. The conclusion's admission that the method 'cannot guarantee to generate a 100% watertight and manifold quad mesh' for low-quality inputs further undercuts any reading that the headline numbers are assumed by construction. Candidate coverage with k=12 is an empirical limitation that could weaken generalization, but it is not a reduction of the reported predictions to the paper's inputs.

Assumptions & free parameters 6 free parameters · 3 assumptions · 0 invented entities

The method's performance depends on several hand-chosen parameters (k, three geometric thresholds, the loss weight, and post-process weights) and on an unverified coverage assumption about the k-NN candidate set. These are not fitted to the test set in a circular way, but they are free choices that could affect the reported results. No new physical entities are introduced.

free parameters (6)
  • k (number of nearest neighbors) = 12
    Chosen in Section IV-A; controls the candidate face set and computational cost.
  • Edge ratio filter threshold = 0.25
    Section IV-A; filters out elongated candidate faces.
  • Sine value filter threshold = 0.3
    Section IV-A; filters faces with non-orthogonal corners.
  • Normal dot product threshold = 0.5
    Section IV-A; filters non-coplanar candidate faces.
  • Cross-entropy class weight w = Not specified; adjusted based on label ratio
    Eq (8); tuned on the training set based on preliminary experiments.
  • Post-process score weights = 10 and 1
    Eq (10); hand-chosen to prioritize removal of non-manifold faces.
assumptions (3)
  • domain assumption Training and test point clouds are drawn from the same distribution of quad-meshable shapes.
    Standard supervised learning assumption; required for the reported test performance to be meaningful.
  • domain assumption Ground-truth meshes in the dataset are watertight and manifold, and the point cloud is a sampling of these surfaces plus small normal noise.
    Stated in Section IV-A; required for label construction and candidate face scoring.
  • domain assumption For each vertex, the true quad-mesh neighbors are within the 12 nearest neighbors in Euclidean distance, and the random triple selection includes the true face among the 12 candidates.
    Required for the classifier to have the correct faces in its candidate set; not proven or checked.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Point2Quad: Generating Quad Meshes from Point Clouds via Face Prediction." pith.science (2026). https://pith.science/paper/VTTWYO3B

@misc{pith2026250419545,
  author       = {Pith},
  title        = {Pith review of: Point2Quad: Generating Quad Meshes from Point Clouds via Face Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VTTWYO3B}},
  note         = {Machine review of arXiv:2504.19545}
}
read the original abstract

Quad meshes are essential in geometric modeling and computational mechanics. Although learning-based methods for triangle mesh demonstrate considerable advancements, quad mesh generation remains less explored due to the challenge of ensuring coplanarity, convexity, and quad-only meshes. In this paper, we present Point2Quad, the first learning-based method for quad-only mesh generation from point clouds. The key idea is learning to identify quad mesh with fused pointwise and facewise features. Specifically, Point2Quad begins with a k-NN-based candidate generation considering the coplanarity and squareness. Then, two encoders are followed to extract geometric and topological features that address the challenge of quad-related constraints, especially by combining in-depth quadrilaterals-specific characteristics. Subsequently, the extracted features are fused to train the classifier with a designed compound loss. The final results are derived after the refinement by a quad-specific post-processing. Extensive experiments on both clear and noise data demonstrate the effectiveness and superiority of Point2Quad, compared to baseline methods under comprehensive metrics.

Figures

Figures reproduced from arXiv: 2504.19545 by the authors.

Figure 1
Figure 1. Full pipeline of our Point2Quad: given a point cloud as input, we first propose a set of candidate quadrilateral. During training (dashed arrow), the network is trained to classify the candidates with the supervision of face labels. During inference, the predicted label is used to filter out quadrilaterals, which are then merged into the output mesh with a post-process [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Illustration of notions in quadrilaterals. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Scoring example: The blue and red edges are non￾manifold and boundary edges respectively [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (5 more)
Figure 3
Figure 3. Figure 3: post-process: ① remove non-manifold parts with scores in Eq. (10); ② fill holes automatically. Non-manifold and boundary edges are highlighted in blue and red. In the post process, we first remove the overlapped elements that cause some small holes in the surface, then…
Figure 6
Figure 6. Figure 6: Visualization results and close-up views of quad meshes generated from noisy point clouds. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Distribution of the number of points for each 3D object [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Visualization results and close-up views of generated quad meshes in high-genus scenarios. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 10
Figure 10. Figure 10: Quad mesh results: before and after the proposed post [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 67 canonical work pages

  1. [1]

    Quadrilateral mesh gen- eration ii: Meromorphic quartic differentials and abel–jacobi condition,

    N. Lei, X. Zheng, Z. Luo, F. Luo, and X. Gu, “Quadrilateral mesh gen- eration ii: Meromorphic quartic differentials and abel–jacobi condition,” CMAME, vol. 366, p. 112980, 2020. 1

    work page 2020

  2. [2]

    Hexahedral mesh generation con- straints,

    J. F. Shepherd and C. R. Johnson, “Hexahedral mesh generation con- straints,” Eng. Comput. , vol. 24, no. 3, pp. 195–213, 2008. 1

    work page 2008

  3. [3]

    Quad-mesh generation and processing: A survey,

    D. Bommes, B. L ´evy, N. Pietroni, E. Puppo, C. Silva, M. Tarini, and D. Zorin, “Quad-mesh generation and processing: A survey,” in Comput Graph F orum., vol. 32, no. 6, 2013, pp. 51–76. 1, 2

    work page 2013

  4. [4]

    Are bilinear quadrilaterals better than linear triangles?

    E. F. D’azevedo, “Are bilinear quadrilaterals better than linear triangles?” Siam J Sci Comput , vol. 22, no. 1, pp. 198–217, 2000. 1

    work page 2000

  5. [5]

    Quadrilateral mesh generation i: Metric based method,

    W. Chen, X. Zheng, J. Ke, N. Lei, Z. Luo, and X. Gu, “Quadrilateral mesh generation i: Metric based method,” CMAME, vol. 356, pp. 652– 668, 2019. 1

    work page 2019

  6. [6]

    Global parametrization of range image sets,

    N. Pietroni, M. Tarini, O. Sorkine, and D. Zorin, “Global parametrization of range image sets,” in SIGGRAPH Asia , 2011, pp. 1–10. 1, 2

    work page 2011

  7. [7]

    Meshless quadrangulation by global parameterization,

    E. Li, B. L ´evy, X. Zhang, W. Che, W. Dong, and J.-C. Paul, “Meshless quadrangulation by global parameterization,” Computers & Graphics , vol. 35, no. 5, pp. 992–1000, 2011. 1, 2

    work page 2011

  8. [8]

    Direct quad- dominant meshing of point cloud via global parameterization,

    E. Li, W. Che, X. Zhang, Y .-K. Zhang, and B. Xu, “Direct quad- dominant meshing of point cloud via global parameterization,” Com- puters & Graphics , vol. 35, no. 3, pp. 452–460, 2011. 1, 2, 3

    work page 2011

Show all 70 references
  1. [9]

    Instant field-aligned meshes,

    W. Jakob, M. Tarini, D. Panozzo, O. Sorkine-Hornung et al. , “Instant field-aligned meshes,” TOG, vol. 34, no. 6, p. 189, 2015. 1, 2, 3, 8, 9

    2015

  2. [10]

    Field-aligned online surface reconstruction,

    N. Schertler, M. Tarini, W. Jakob, M. Kazhdan, S. Gumhold, and D. Panozzo, “Field-aligned online surface reconstruction,” TOG, vol. 36, no. 4, pp. 1–13, 2017. 1, 2, 3

    2017

  3. [11]

    Extracting lines of curvature from noisy point clouds,

    E. Kalogerakis, D. Nowrouzezahrai, P. Simari, and K. Singh, “Extracting lines of curvature from noisy point clouds,” Comput. Aided Des., vol. 41, no. 4, pp. 282–292, 2009. 1, 3

    2009

  4. [12]

    An effective quad-dominant meshing method for unorganized point clouds,

    X. Pang, Z. Song, and R. W. Lau, “An effective quad-dominant meshing method for unorganized point clouds,” Graphical models, vol. 76, no. 2, pp. 86–102, 2014. 1, 3

    2014

  5. [13]

    Hierarchical quad meshing of 3d scanned surfaces,

    D. R. Bukenberger and H. P. Lensch, “Hierarchical quad meshing of 3d scanned surfaces,” in Comput. Graphics F orum , vol. 37, no. 5, 2018, pp. 135–145. 1, 3

    2018

  6. [14]

    V oxel-based quadrilateral mesh generation from point cloud,

    B. Guan, S. Lin, R. Wang, F. Zhou, X. Luo, and Y . Zheng, “V oxel-based quadrilateral mesh generation from point cloud,” Multimed. Tools Appl., vol. 79, pp. 20 561–20 578, 2020. 1, 3

    2020

  7. [15]

    L p centroidal voronoi tessellation and its applications,

    B. L ´evy and Y . Liu, “L p centroidal voronoi tessellation and its applications,” TOG, vol. 29, no. 4, pp. 1–11, 2010. 1, 3

    2010

  8. [16]

    Point cloud resampling using centroidal voronoi tessellation methods,

    Z. Chen, T. Zhang, J. Cao, Y . J. Zhang, and C. Wang, “Point cloud resampling using centroidal voronoi tessellation methods,” Comput. Aided Des. , vol. 102, pp. 12–21, 2018. 1, 3

    2018

  9. [17]

    Surface reconstruction by parallel and unified particle-based resampling from point clouds,

    S. Zhong, Z. Zhong, and J. Hua, “Surface reconstruction by parallel and unified particle-based resampling from point clouds,” CAGD, vol. 71, pp. 43–62, 2019. 1, 3

    2019

  10. [18]

    Deep marching cubes: Learning explicit surface representations,

    Y . Liao, S. Donne, and A. Geiger, “Deep marching cubes: Learning explicit surface representations,” in CVPR, 2018, pp. 2916–2925. 1, 2

    2018

  11. [19]

    Learning elementary structures for 3d shape generation and matching,

    T. Deprelle, T. Groueix, M. Fisher, V . Kim, B. Russell, and M. Aubry, “Learning elementary structures for 3d shape generation and matching,” in NeurIPS, 2019, pp. 7433–7443. 1, 2

    2019

  12. [20]

    Needrop: Self- supervised shape representation from sparse point clouds using needle dropping,

    A. Boulch, P.-A. Langlois, G. Puy, and R. Marlet, “Needrop: Self- supervised shape representation from sparse point clouds using needle dropping,” in 3DV. IEEE, 2021, pp. 940–950. 1, 2

    2021

  13. [21]

    Deep marching tetra- hedra: a hybrid representation for high-resolution 3d shape synthesis,

    T. Shen, J. Gao, K. Yin, M.-Y . Liu, and S. Fidler, “Deep marching tetra- hedra: a hybrid representation for high-resolution 3d shape synthesis,” NeurIPS, vol. 34, pp. 6087–6101, 2021. 1, 2

    2021

  14. [22]

    Reconstructing surfaces for sparse point clouds with on-surface priors,

    B. Ma, Y .-S. Liu, and Z. Han, “Reconstructing surfaces for sparse point clouds with on-surface priors,” in CVPR, 2022, pp. 6315–6325. 1, 2

    2022

  15. [23]

    Neural fields as learnable kernels for 3d reconstruction,

    F. Williams, Z. Gojcic, S. Khamis, D. Zorin, J. Bruna, S. Fidler, and O. Litany, “Neural fields as learnable kernels for 3d reconstruction,” in CVPR, 2022, pp. 18 500–18 510. 1, 2

    2022

  16. [24]

    Neural Dual Contouring,

    Z. Chen, A. Tagliasacchi, T. Funkhouser, and H. Zhang, “Neural Dual Contouring,” ACM TOG, vol. 41, pp. 1–13, Jul. 2022. 1, 2

    2022

  17. [25]

    Poco: Point convolution for surface recon- struction,

    A. Boulch and R. Marlet, “Poco: Point convolution for surface recon- struction,” in CVPR, 2022, pp. 6302–6314. 1, 2

    2022

  18. [26]

    Shap-e: Generating conditional 3d implicit functions,

    H. Jun and A. Nichol, “Shap-e: Generating conditional 3d implicit functions,” arXiv preprint arXiv:2305.02463 , 2023. 1, 2

    2023 arXiv

  19. [27]

    Scan2Mesh: From Unstructured Range Scans to 3D Meshes,

    A. Dai and M. Nießner, “Scan2Mesh: From Unstructured Range Scans to 3D Meshes,” in CVPR, Jun. 2019, pp. 5569–5578. 1, 2

    2019

  20. [28]

    Meshing point clouds with predicted intrinsic-extrinsic ratio guidance,

    M. Liu, X. Zhang, and H. Su, “Meshing point clouds with predicted intrinsic-extrinsic ratio guidance,” in ECCV, 2020, pp. 68–84. 1, 2, 8, 9

    2020

  21. [29]

    REIN: Flexible Mesh Generation from Point Clouds,

    R. Daroya, R. Atienza, and R. Cajote, “REIN: Flexible Mesh Generation from Point Clouds,” in CVPRW, Jun. 2020, pp. 1444–1453. 1, 2

    2020

  22. [30]

    PointTriNet: Learned Triangulation of 3D Point Sets,

    N. Sharp and M. Ovsjanikov, “PointTriNet: Learned Triangulation of 3D Point Sets,” in ECCV. Springer, 2020, pp. 762–778. 1, 2

    2020

  23. [31]

    Learning delaunay surface elements for mesh reconstruction,

    M.-J. Rakotosaona, P. Guerrero, N. Aigerman, N. J. Mitra, and M. Ovs- janikov, “Learning delaunay surface elements for mesh reconstruction,” in CVPR, 2021, pp. 22–31. 1, 2

    2021

  24. [32]

    Meshanything: Artist-created mesh generation with autoregressive transformers,

    Y . Chen, T. He, D. Huang, W. Ye, S. Chen, J. Tang, X. Chen, Z. Cai, L. Yang, G. Yu et al. , “Meshanything: Artist-created mesh generation with autoregressive transformers,” arXiv preprint arXiv:2406.10163 ,

    arXiv

  25. [33]

    Mergenet: Explicit mesh reconstruction from sparse point clouds via edge prediction,

    W. Wang, Y . Deng, Z. Li, Y . Liu, and N. Lei, “Mergenet: Explicit mesh reconstruction from sparse point clouds via edge prediction,” in ICME. IEEE, 2024, pp. 1–6. 1, 2

    2024

  26. [34]

    Finite element mesh generation methods: a review and classification,

    K. Ho-Le, “Finite element mesh generation methods: a review and classification,” Comput. Aided Des. , vol. 20, pp. 27–38, 1988. 2

    1988

  27. [35]

    What’s the situation with intelligent mesh generation: A survey and perspectives,

    N. Lei, Z. Li, Z. Xu, Y . Li, and X. Gu, “What’s the situation with intelligent mesh generation: A survey and perspectives,” TVCG, vol. 30, no. 8, pp. 4997–5017, 2024. 2

    2024

  28. [36]

    Pixel2mesh: Generating 3d mesh models from single rgb images,

    N. Wang, Y . Zhang, Z. Li, Y . Fu, W. Liu, and Y .-G. Jiang, “Pixel2mesh: Generating 3d mesh models from single rgb images,” in ECCV, 2018, pp. 52–67. 2

    2018

  29. [37]

    Pixel2mesh++: Multi-view 3d mesh generation via deformation,

    C. Wen, Y . Zhang, Z. Li, and Y . Fu, “Pixel2mesh++: Multi-view 3d mesh generation via deformation,” in ICCV, 2019, pp. 1042–1051. 2

    2019

  30. [38]

    Pcgan: Prediction- compensation generative adversarial network for meshes,

    T. Li, Y . Shi, X. Sun, J. Wang, and B. Yin, “Pcgan: Prediction- compensation generative adversarial network for meshes,” TCSVT, vol. 32, no. 7, pp. 4667–4679, 2021. 2

    2021

  31. [39]

    A progressive quadric graph convolutional network for 3d human mesh recovery,

    L. Wang, X. Liu, X. Ma, J. Wu, J. Cheng, and M. Zhou, “A progressive quadric graph convolutional network for 3d human mesh recovery,” TCSVT, vol. 33, no. 1, pp. 104–117, 2022. 2

    2022

  32. [40]

    Transformer-based 3d face reconstruction with end-to-end shape-preserved domain transfer,

    Z. Chen, Y . Wang, T. Guan, L. Xu, and W. Liu, “Transformer-based 3d face reconstruction with end-to-end shape-preserved domain transfer,” TCSVT, vol. 32, no. 12, pp. 8383–8393, 2022. 2

    2022

  33. [41]

    Simultaneously recovering multi- person meshes and multi-view cameras with human semantics,

    B. Huang, J. Ju, Y . Shu, and Y . Wang, “Simultaneously recovering multi- person meshes and multi-view cameras with human semantics,” TCSVT, vol. 34, no. 6, pp. 4229–4242, 2024. 2

    2024

  34. [42]

    High-quality animat- able dynamic garment reconstruction from monocular videos,

    X. Li, J. Zhang, Y .-K. Lai, J. Yang, and K. Li, “High-quality animat- able dynamic garment reconstruction from monocular videos,” TCSVT, vol. 34, no. 6, pp. 4243–4256, 2024. 2

    2024

  35. [43]

    Staf: 3d human mesh recovery from video with spatio-temporal alignment fusion,

    W. Yao, H. Zhang, Y . Sun, and J. Tang, “Staf: 3d human mesh recovery from video with spatio-temporal alignment fusion,” TCSVT, vol. 34, no. 11, pp. 10 564–10 577, 2024. 2

    2024

  36. [44]

    Incomplete points cloud data surface Reconstruction Based on Neural Network,

    G.-x. Li, X.-m. Wu, and W.-m. Zhao, “Incomplete points cloud data surface Reconstruction Based on Neural Network,” in IIH-MSP, 2008, pp. 913–916. 2

    2008

  37. [45]

    Meshlet priors for 3d mesh reconstruction,

    A. Badki, O. Gallo, J. Kautz, and P. Sen, “Meshlet priors for 3d mesh reconstruction,” in CVPR, 2020, pp. 2849–2858. 2

    2020

  38. [46]

    Iso-points: Optimizing neural implicit surfaces with hybrid representations,

    W. Yifan, S. Wu, C. Oztireli, and O. Sorkine-Hornung, “Iso-points: Optimizing neural implicit surfaces with hybrid representations,” in CVPR, 2021, pp. 374–383. 2

    2021

  39. [47]

    Neural-Pull: Learning Signed Distance Functions from Point Clouds by Learning to Pull Space onto Surfaces,

    B. Ma, Z. Han, Y .-S. Liu, and M. Zwicker, “Neural-Pull: Learning Signed Distance Functions from Point Clouds by Learning to Pull Space onto Surfaces,” Proceedings of the 38th International Conference on Machine Learning , vol. 139, Jan. 2021. 2

    2021

  40. [48]

    Multiview textured mesh recovery by differentiable rendering,

    L. Lin, J. Zhu, and Y . Zhang, “Multiview textured mesh recovery by differentiable rendering,” TCSVT, vol. 33, no. 4, pp. 684–696, 2022. 2

    2022

  41. [49]

    Differentiable surface triangulation,

    M.-J. Rakotosaona, N. Aigerman, N. J. Mitra, M. Ovsjanikov, and P. Guerrero, “Differentiable surface triangulation,” TOG, vol. 40, no. 6, pp. 1–13, 2021. 2

    2021

  42. [50]

    Neural Marching Cubes,

    Z. Chen and H. Zhang, “Neural Marching Cubes,” TOG, vol. 40, no. 6, pp. 1–15, Dec. 2021. 2 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY , VOL. XX, NO. XX, 2024 12

    2021

  43. [51]

    Learning delaunay surface elements for mesh reconstruction,

    M.-J. Rakotosaona, P. Guerrero, N. Aigerman, N. J. Mitra, and M. Ovs- janikov, “Learning delaunay surface elements for mesh reconstruction,” in CVPR, June 2021, pp. 22–31. 2

    2021

  44. [52]

    Shape as points: A differentiable poisson solver,

    S. Peng, C. Jiang, Y . Liao, M. Niemeyer, M. Pollefeys, and A. Geiger, “Shape as points: A differentiable poisson solver,” NeurIPS, vol. 34, pp. 13 032–13 044, 2021. 2

    2021

  45. [53]

    V oromesh: Learning watertight surface meshes with voronoi diagrams,

    N. Maruani, R. Klokov, M. Ovsjanikov, P. Alliez, and M. Desbrun, “V oromesh: Learning watertight surface meshes with voronoi diagrams,” in ICCV, 2023, pp. 14 565–14 574. 2

    2023

  46. [54]

    Marching cubes: A high resolution 3d surface construction algorithm,

    W. E. Lorensen and H. E. Cline, “Marching cubes: A high resolution 3d surface construction algorithm,” ACM siggraph computer graphics , vol. 21, no. 4, pp. 163–169, 1987. 2

    1987

  47. [55]

    Learning manifold patch- based representations of man-made shapes,

    D. Smirnov, M. Bessmeltsev, and J. Solomon, “Learning manifold patch- based representations of man-made shapes,” in ICLR, 2020. 2

    2020

  48. [56]

    Learning direction fields for quad mesh generation,

    A. Dielen, I. Lim, M. Lyon, and L. Kobbelt, “Learning direction fields for quad mesh generation,” in Comput Graph F orum. , vol. 40, no. 5, 2021, pp. 181–191. 2

    2021

  49. [57]

    Mgnet: a novel differential mesh generation method based on unsupervised neural networks,

    X. Chen, T. Li, Q. Wan, X. He, C. Gong, Y . Pang, and J. Liu, “Mgnet: a novel differential mesh generation method based on unsupervised neural networks,” Eng. Comput. , vol. 38, no. 5, pp. 4409–4421, 2022. 2

    2022

  50. [58]

    Sketch2pq: freeform planar quadrilateral mesh design via a single sketch,

    Z. Deng, Y . Liu, H. Pan, W. Jabi, J. Zhang, and B. Deng, “Sketch2pq: freeform planar quadrilateral mesh design via a single sketch,” TVCG,

  51. [59]

    Reinforcement learning for automatic quadrilateral mesh generation: A soft actor–critic approach,

    J. Pan, J. Huang, G. Cheng, and Y . Zeng, “Reinforcement learning for automatic quadrilateral mesh generation: A soft actor–critic approach,” Neural Networks , vol. 157, pp. 288–304, 2023. 2

    2023

  52. [60]

    Quadrilateral mesh generation method based on convolutional neural network,

    Y . Zhou, X. Cai, Q. Zhao, Z. Xiao, and G. Xu, “Quadrilateral mesh generation method based on convolutional neural network,” Information, vol. 14, no. 5, p. 273, 2023. 2

    2023

  53. [61]

    Srl-assisted afm: Generating planar unstructured quadrilateral meshes with supervised and reinforcement learning-assisted advancing front method,

    H. Tong, K. Qian, E. Halilaj, and Y . J. Zhang, “Srl-assisted afm: Generating planar unstructured quadrilateral meshes with supervised and reinforcement learning-assisted advancing front method,” Journal of Computational Science , vol. 72, p. 102109, 2023. 2

    2023

  54. [62]

    Craftsman: High-fidelity mesh generation with 3d native generation and interactive geometry refiner,

    W. Li, J. Liu, R. Chen, Y . Liang, X. Chen, P. Tan, and X. Long, “Craftsman: High-fidelity mesh generation with 3d native generation and interactive geometry refiner,” arXiv preprint arXiv:2405.14979 , 2024. 2

    2024 arXiv

  55. [63]

    Kpconv: Flexible and deformable convolution for point clouds,

    H. Thomas, C. R. Qi, J.-E. Deschaud, B. Marcotegui, F. Goulette, and L. J. Guibas, “Kpconv: Flexible and deformable convolution for point clouds,” in ICCV, 2019, pp. 6411–6420. 3, 7

    2019

  56. [64]

    Achieving finite element mesh quality via optimization of the jacobian matrix norm and associated quantities. part i—a framework for surface mesh optimization,

    P. M. Knupp, “Achieving finite element mesh quality via optimization of the jacobian matrix norm and associated quantities. part i—a framework for surface mesh optimization,” J. Numer . Methods Eng., vol. 48, no. 3, pp. 401–420, 2000. 4, 7

    2000

  57. [65]

    Netgen an advancing front 2d/3d-mesh generator based on abstract rules,

    J. Sch ¨oberl, “Netgen an advancing front 2d/3d-mesh generator based on abstract rules,” Comput. Vis. Sci. , vol. 1, no. 1, pp. 41–52, 1997. 6

    1997

  58. [66]

    Reliable feature-line driven quad-remeshing,

    N. Pietroni, S. Nuvoli, T. Alderighi, P. Cignoni, M. Tarini et al. , “Reliable feature-line driven quad-remeshing,” TOG, vol. 40, no. 4, pp. 1–17, 2021. 7

    2021

  59. [67]

    The verdict geometric quality library

    P. M. Knupp, C. Ernst, D. C. Thompson, C. Stimpson, and P. P. Pebay, “The verdict geometric quality library.” Sandia National Laboratories, Albuquerque, NM, and Livermore, Tech. Rep., 2006. 7

    2006

  60. [68]

    Point2mesh: a self-prior for deformable meshes,

    R. Hanocka, G. Metzer, R. Giryes, and D. Cohen-Or, “Point2mesh: a self-prior for deformable meshes,” TOG, vol. 39, no. 4, p. 126, 2020. 7

    2020

  61. [69]

    Pointtrinet: Learned triangulation of 3d point sets,

    N. Sharp and M. Ovsjanikov, “Pointtrinet: Learned triangulation of 3d point sets,” in ECCV. Springer, 2020, pp. 762–778. 7

    2020

  62. [70]

    The ball-pivoting algorithm for surface reconstruction,

    F. Bernardini, J. Mittleman, H. Rushmeier, C. Silva, and G. Taubin, “The ball-pivoting algorithm for surface reconstruction,” TVCG, vol. 5, no. 4, pp. 349–359, 1999. 8 Zezeng Li received a B.S. degree from Beijing University of Technology (BJUT) in 2015 and a Ph.D. degree from...

    1999

Pith tools

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