Pith. sign in

REVIEW 4 major objections 5 minor 60 references

Curve-Aware Gaussian Splatting for 3D Parametric Curve Reconstruction

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A one-stage pipeline reconstructs 3D parametric curves directly from 2D edge maps via curve-aware Gaussian splatting, beating two-stage methods in accuracy, speed, and compactness.

desk verdict Genuinely new one-stage curve reconstruction via curve-coupled Gaussians, with solid ABC results, but overclaimed baselines and real-world tests that depend on unstated SfM initialization and custom edge maps. read the letter →

arxiv 2506.21401 v3 pith:NVRNNUA2 submitted 2025-06-26 cs.CV

Zhirui Gao , Renjiao Yi , Yaqiao Dai , Xuening Zhu , Wei Chen , Chenyang Zhu , Kai Xu This is my paper · ORCID
classification cs.CV
keywords parametriccurvereconstructionGaussiansplattingBéziercurvesmulti-viewedgemapsdifferentiablerenderingone-stageoptimizationtopology3D
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 aims to show that 3D parametric curves—Bézier curves and line segments—can be optimized directly from multi-view 2D edge maps, bypassing the usual two-stage pipeline of first reconstructing an edge point cloud and then fitting curves. To make this possible, the paper couples each curve to a set of edge-oriented Gaussians whose positions, orientations, and scales are analytic functions of the curve's control points, so ordinary differentiable Gaussian splatting renders the curves and backpropagates loss gradients to the curve parameters. A training-time topology controller linearizes, merges, splits, and prunes curves to keep the reconstruction compact. On the ABC-NEF benchmark the method reports roughly 14.5% better reconstruction accuracy, 33% faster runtime, and 70.5% fewer output curves than a strong two-stage baseline. If correct, this shows that the intermediate 3D edge cloud is unnecessary and that its errors can be avoided, making curve reconstruction faster and cleaner for CAD-style objects.

What carries the argument

The load-bearing object is the CurveGaussian representation: a Bézier curve $c_j(t)$ spawns $N=12$ edge-oriented Gaussians through uniform sampling, with the $i$-th Gaussian's position $p_j(t_i)$, orientation (principal axis = curve tangent), and scale (axial length from neighboring sample distance, plus curve-specific thickness) all computed analytically from the control points. It carries the argument because it makes parametric curves renderable through standard splatting while keeping the curve itself the only optimizable entity, so photometric edge consistency and geometric curve regularization act on the same parameters. Around this core sit four adaptive operators—curve linearization (cubic to line when nearly straight), merging (lines and cubics), splitting (de Casteljau at abrupt tangent changes or low mask values), and pruning (low opacity or masked-out Gaussians)—that adjust topology during training.

What would settle it

Take a CAD model with known ground-truth Bézier curves, render views, and corrupt the edge maps by deleting a controlled fraction of edge pixels and adding spurious edges; if the one-stage method's reconstruction error grows roughly with the corruption rate and never recovers the occluded portions (edges invisible in every view), that confirms the central premise. A sharper test: an edge that is present in only one view yet consistent with a wrong 3D line, which the multi-view photometric loss would not correct, showing that 2D edge quality, not reconstruction, is the limiting factor.

Watch

Extended reading notes

Core claim

The central discovery is a bi-directional coupling between parametric curves and Gaussian primitives, called CurveGaussian. Each Bézier curve $c_j$ is sampled at $N$ evenly spaced parameters $t_i$, and the $i$-th Gaussian is anchored at the curve point $p(t_i)$, with its principal axis aligned to the curve tangent, its axial scale set to the inter-sample distance along with a curve thickness, and its opacity inherited from the curve. Because every Gaussian attribute is an analytic function of the control points, the rendering function $f$ in the optimization objective is exactly curve-aware Gaussian splatting, and multi-view rendering losses backpropagate straight to the control points, thicknesses, and opacities. The paper argues this eliminates the error accumulation of separate point-cloud reconstruction and RANSAC-style fitting, and empirically supports it with accuracy, completeness, precision, recall, runtime, and curve-count comparisons on ABC-NEF, MV2Cyl real objects, and Replica scenes.

Load-bearing premise

The whole optimization trusts that the input 2D edge maps correctly mark the target curves, so if edge detection is noisy, incomplete, or inconsistent across views, the one-stage optimizer has no independent signal to pull the curves back to the true geometry, and the paper's own conclusion and MV2Cyl experiments acknowledge this dependence.

Editorial extensions

If this is right

  • Because the intermediate edge point cloud is eliminated, the method's output curve count drops (28.9 average versus 97.9 for the two-stage Gaussian baseline on ABC-NEF), so downstream CAD-style editing and rendering start from a much sparser representation.
  • Optimization variables are the curve control points rather than a dense set of per-Gaussian positional parameters, which is why the paper reports 33% shorter runtime at comparable or better accuracy.
  • The adaptive topology rules (linearize, merge, split, prune) let the number of curves shrink from a large random initialization to a compact set during training, so the method does not require knowing the curve count in advance.
  • Because rendering is standard splatting, the same machinery extends to any scene where edges can be extracted as 2D maps, including real-world captures and indoor scenes (shown on MV2Cyl and Replica).

Reading between the lines

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

  • If the claimed dependence on 2D edge quality is the binding constraint, pairing the method with learned edge detectors that fuse depth or normal cues (as the supplement does for MV2Cyl) could push the practical ceiling higher than the ABC-NEF results alone suggest; the paper gestures at this by proposing to lift 2D edge foundation models into 3D.
  • The bi-directional coupling is general: any parametric primitive whose sampling yields analytic Gaussian attributes (e.g., surfaces or open B-splines) could inherit the same one-stage optimization, a direction the paper does not explore.
  • A direct comparison against a two-stage pipeline that uses the same edge maps and the same Gaussian splatting backend would isolate exactly how much of the reported gain comes from removing the fitting stage versus from the curve-aware regularization; the paper compares against existing systems rather than this matched baseline.
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. This paper proposes CurveGaussian, a hybrid representation that couples cubic and first-order Bézier curves with edge-oriented 3D Gaussian components, enabling one-stage optimization of parametric curves directly from multi-view edge maps through differentiable Gaussian splatting. The method introduces adaptive topology operations (linearization, merging, splitting, pruning) and several custom loss terms. On the ABC-NEF benchmark, the method reports improved accuracy, compactness, and training time over NEF, EMAP, and EdgeGaussians under the DexiNed edge detector; on MV2Cyl real objects it reports gains over EdgeGaussians; on Replica it shows qualitative results. The paper claims that the one-stage formulation bypasses the error accumulation and efficiency bottlenecks of two-stage pipelines.

Significance. If the central claims hold, the core idea is valuable: it demonstrates that parametric Bézier curves can be optimized directly through Gaussian splatting, removing the explicit two-stage edge-point-cloud fitting and potentially reducing error accumulation. The reported 70.5% reduction in curve count and 33% faster training on ABC-NEF are concrete, externally comparable improvements, and the authors commit to releasing code and data, which supports reproducibility. However, the strength of the real-world evidence is currently limited because the paper does not disclose in the main text that COLMAP-based scenes use SfM point clouds for initialization and that MV2Cyl uses custom SAM2-plus-normal edge maps, and the claim of outperforming all learning-based methods across all metrics is contradicted by the paper's own PiDiNet results. These issues are fixable but require additional experiments and corrections.

major comments (4)
  1. [Section 4.2, Table 1] The claim that "the proposed method outperforms all learning-based methods across all metrics" is not supported by Table 1 when the PiDiNet edge detector is used. For example, EMAP (PiDiNet) has Accuracy 9.2 vs. Ours 10.8 (lower is better), and EdgeGaussians (PiDiNet) has Completeness 10.3 vs. Ours 12.3. The statement should be qualified to the DexiNed condition or to the specific metrics where the improvement holds, and the text should acknowledge that the relative performance depends on the edge detector.
  2. [Supplementary Section 9; main-text Section 3.3] The main text states in Section 3.3 that the optimization is initialized with "random Bézier parameterized curves," but Supplementary Section 9 states that for COLMAP-based datasets the midpoints of the Bézier curves are initialized using the SfM point cloud. This is a direct contradiction and is load-bearing for the claimed one-stage, edge-map-only behavior. If SfM initialization places the initial curves near the true 3D edges, the reported real-world gains (e.g., Table 3) may be partly due to this extra 3D input. The paper must disclose this initialization in the main text, incorporate it into the problem formulation in Eq. (1), or justify its absence, and provide an ablation on at least one COLMAP scene with purely random initialization to quantify the contribution.
  3. [Supplementary Section 6; Section 4.2 (MV2Cyl)] On MV2Cyl, the proposed method uses edge maps generated by SAM2 segmentation plus a monocular normal estimator, which the paper says are superior to standard detectors. The main-text comparison in Table 3 is only against EdgeGaussians, and the paper never states whether EdgeGaussians received the same custom edge maps. If EdgeGaussians used standard detector edge maps (e.g., DexiNed or PiDiNet), the comparison is confounded: the reported gains in accuracy and F-score could reflect the higher quality of the input edge maps rather than the one-stage optimization. The paper must specify the exact edge maps supplied to each method and, if feasible, run EdgeGaussians with the same SAM2-plus-normal edge maps.
  4. [Section 3.3; Supplementary Section 9] The adaptive topology operations depend on multiple hand-set thresholds (τ_l, τ_la, τ_ld, τ_b, θ_s, τ_m, τ_d, η, τ), but only a few values are reported in the supplementary (opacity below 0.05, bending angle above 20°, mask below 0.01, and the merge threshold taken from EMAP). For reproducibility and to assess sensitivity, the paper should list all threshold values in one place and include a brief analysis of how the results vary with a reasonable perturbation of these thresholds.
minor comments (5)
  1. [Section 4.1] The metrics Accuracy, Completeness, Recall, Precision, and F-score are used in Table 1 without being defined in the main text; they are only defined in Supplementary Section 8. A one-sentence definition or a reference should be added in Section 4.1.
  2. [Section 4.2 (MV2Cyl)] The statement that "NEF and EMAP fail to produce reasonable 3D edges in our experiments" is made without supporting evidence. A qualitative figure or a brief explanation of the failure mode would strengthen the claim and allow readers to judge the comparison.
  3. [Equation (8)] The weighting in Eq. (8) is easy to misread: the first term over non-edge pixels N_I is weighted by |M_I|/|E_I|, and the second term over edge pixels M_I is weighted by |N_I|/|E_I|. A short explanatory sentence after the equation would clarify the intended balancing of the two classes.
  4. [Supplementary Section 8] The so-called "new metric, Curve Count" is just the number of output curves and is not a novel metric in the conventional sense. Consider renaming it to "output curve count" to avoid overclaiming.
  5. [Throughout] The notation for Gaussian indices is inconsistent (e.g., G_i^j in Section 3.2 versus v_j,i in Eq. (10)); please standardize the subscripts and superscripts.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is an empirical method comparison, and the curve-to-Gaussian coupling is a rendering construction rather than a fitted input renamed as a prediction.

full rationale

The paper's central claim is an empirical method claim, not a derived prediction. Equation (1) defines an optimization over Bezier curve parameters against multi-view edge maps, and Section 3.2 constructs the rendering function f by coupling each curve to N edge-oriented Gaussians whose positions, orientations, and scales are analytic functions of the curve geometry. The Gaussians are not fitted first and then renamed as curves; they are deterministically generated from the curve parameters, so backpropagation to control points is a construction rather than a hidden fit. Quantitative comparisons are made against external CAD ground truth on ABC-NEF and MV2Cyl using standard accuracy, completeness, precision, recall, and F-score metrics, and the reported gains are benchmark measurements rather than consequences of the definitions. The only self-citation appearing in a design choice is the edge-aware rendering loss attributed to NEF [52], but it is a training-loss component, not a load-bearing premise, and its use does not force the reported reconstruction result. The compactness metric (curve count) and the threshold choices are aligned with the method's design goal, but that is metric selection and hyperparameter tuning, not circular derivation. Real-world evaluation does depend on custom SAM2+normal edge maps and SfM-based midpoint initialization (Supplementary Sections 6 and 9), which is a legitimate experimental-support concern about whether the one-stage, edge-map-only claim is cleanly instantiated, but it is not a case of the output being equivalent to the input by construction. No equation-level reduction, renamed known result, or imported uniqueness theorem was found.

Assumptions & free parameters 5 free parameters · 6 assumptions · 1 invented entities

The empirical claim depends on the Gaussian proxy faithfully rendering curves, on accurate multi-view edge maps, and on the generalization of many hand-set thresholds. These are listed above; none are derived from first principles.

free parameters (5)
  • Adaptive topology thresholds τ_l, τ_la, τ_ld, τ_b, θ_s, τ_m, τ_d = partial: opacity prune 0.05, split angle 20°, mask threshold 0.01; remaining values deferred to code
    Section 3.3 and Supplementary Section 9 set thresholds for linearization, merging, splitting, and pruning without sensitivity analysis; these directly control compactness and completeness.
  • Gaussians per curve N = 12
    Supplementary Section 9 fixes N=12; this determines how well the Gaussian proxy can represent curve geometry in rendering.
  • Loss weights λ1, λ2, λ3, λ4 = 0.01, 0.01, 0.01, 0.0005
    Supplementary Section 9 reports the weights; no ablation of these weights is shown.
  • Edge/non-edge threshold η and endpoint connection threshold τ = η=0.1; τ unspecified
    Used in Eq. 8 and Eq. 9 to define edge pixels and connection regularization; values are chosen manually.
  • Curve initialization scheme = 15×15×15 uniform grid for ABC; SfM point cloud for COLMAP-based datasets
    Supplementary Section 9; initialization affects convergence and final topology, and differs by dataset.
assumptions (6)
  • standard math Bernstein basis functions and de Casteljau splitting correctly describe cubic Bézier curves.
    Used in Eq. 2 and Eqs. 5-6 for curve parameterization and splitting.
  • domain assumption The 3D Gaussian splatting alpha-compositing model provides a differentiable rendering function f(·) whose gradients are usable for curve optimization.
    Section 3.2 adopts the standard 3DGS pipeline but removes color; no proof is given that this proxy preserves enough gradient information for Bézier control points, especially at high curvature.
  • domain assumption Scene edges can be represented by cubic and first-order Bézier segments with N=12 uniformly sampled Gaussians per curve.
    Section 3.1 states these curve types are sufficient; Supplementary Section 9 fixes N=12. Complex or highly curved edges may violate this.
  • domain assumption The input 2D edge maps are reliable enough to guide multi-view optimization.
    The conclusion explicitly says the method is contingent on 2D edge quality, and Supplementary Section 6 admits standard edge detectors fail on MV2Cyl, requiring SAM2 plus normal estimation.
  • ad hoc to paper The hand-set adaptive topology rules (linearize, merge, split, prune) with fixed thresholds improve rather than degrade the optimization.
    Section 3.3 introduces these heuristics without theoretical or statistical justification; ablations show most help, but no sensitivity analysis is reported.
  • domain assumption The edge-aware loss in Eq. 8 avoids gradient collapse on sparse edge maps.
    Adopted in Section 3.4 from NEF [52]; effectiveness is asserted and only indirectly ablated.
invented entities (1)
  • CurveGaussian (curve-aware Gaussian representation) independent evidence
    purpose: Acts as a differentiable rendering proxy for parametric Bézier curves, with each curve generating N Gaussians whose spatial attributes are analytic functions of control points.
    CurveGaussian is a new modeling construct, not a physical entity. It has falsifiable handles through reconstructed edge maps and quantitative comparisons against CAD ground truth, and the authors promise code at the stated repository.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Curve-Aware Gaussian Splatting for 3D Parametric Curve Reconstruction." pith.science (2026). https://pith.science/paper/NVRNNUA2

@misc{pith2026250621401,
  author       = {Pith},
  title        = {Pith review of: Curve-Aware Gaussian Splatting for 3D Parametric Curve Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NVRNNUA2}},
  note         = {Machine review of arXiv:2506.21401}
}
read the original abstract

This paper presents an end-to-end framework for reconstructing 3D parametric curves directly from multi-view edge maps. Contrasting with existing two-stage methods that follow a sequential ``edge point cloud reconstruction and parametric curve fitting'' pipeline, our one-stage approach optimizes 3D parametric curves directly from 2D edge maps, eliminating error accumulation caused by the inherent optimization gap between disconnected stages. However, parametric curves inherently lack suitability for rendering-based multi-view optimization, necessitating a complementary representation that preserves their geometric properties while enabling differentiable rendering. We propose a novel bi-directional coupling mechanism between parametric curves and edge-oriented Gaussian components. This tight correspondence formulates a curve-aware Gaussian representation, \textbf{CurveGaussian}, that enables differentiable rendering of 3D curves, allowing direct optimization guided by multi-view evidence. Furthermore, we introduce a dynamically adaptive topology optimization framework during training to refine curve structures through linearization, merging, splitting, and pruning operations. Comprehensive evaluations on the ABC dataset and real-world benchmarks demonstrate our one-stage method's superiority over two-stage alternatives, particularly in producing cleaner and more robust reconstructions. Additionally, by directly optimizing parametric curves, our method significantly reduces the parameter count during training, achieving both higher efficiency and superior performance compared to existing approaches.

Figures

Figures reproduced from arXiv: 2506.21401 by the authors.

Figure 1
Figure 1. Our method achieves clean and accurate parametric [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of CurveGaussian. We propose a curve-aware Gaussian representation that optimizes parametric curves through a one-stage optimization in a self-supervised manner by re-rendering losses. The method employs many adaptive strategies, including curve linearization, merging, splitting, and pruning, to dynamically adjust the curves during training. tion using Gaussian splatting [15], it still fails to overcome the… view at source ↗
Figure 3
Figure 3. Visualization of splitting strategy. The first row demon [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparisons on ABC-NEF [52]. Distinct colors represent different curves/lines. Our method achieves more complete and accurate edge reconstruction of objects while maintaining parametric compactness. In summary, the final loss function for training the para￾…
Figure 5
Figure 5. Figure 5: Qualitative comparisons on Replica [34]. Distinct colors represent different parametric edges, with the numbers in the images indicating the count of parametric edges. Our method achieves accurate edge reconstruction, balancing compactness and completeness. GT Ours w/o…
Figure 6
Figure 6. Figure 6: Qualitative ablations on key components of our method. Excluding critical designs from the full version introduces additional noise, incompleteness, or redundancy in the qualitative reconstruction results. Method Acc.↓ Comp.↓ R5 ↑ P5 ↑ F5 ↑ Ncurve↓ Nline↓ Nsum↓ Ours 8.…
Figure 7
Figure 7. Figure 7: Qualitative comparisons on ABC-NEF [52]. Distinct colors represent different curves. Our method achieves more complete and accurate edge reconstruction of objects while maintaining parametric compactness. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 55 canonical work pages

  1. [1]

    L2d2: Learnable line detector and descriptor

    Hichem Abdellali, Robert Frohlich, Viktor Vilagos, and Zoltan Kato. L2d2: Learnable line detector and descriptor. In2021 International Conference on 3D Vision (3DV), pages 442–452. IEEE, 2021. 2

    work page 2021

  2. [2]

    Automatic line matching and 3d recon- struction of buildings from multiple views

    Caroline Baillard, Cordelia Schmid, Andrew Zisserman, and Andrew Fitzgibbon. Automatic line matching and 3d recon- struction of buildings from multiple views. InISPRS Con- ference on Automatic Extraction of GIS Objects from Digital Imagery, pages 69–80, 1999. 2

    work page 1999

  3. [3]

    Structure-from-motion us- ing lines: Representation, triangulation, and bundle adjust- ment.Computer vision and image understanding, 100(3): 416–441, 2005

    Adrien Bartoli and Peter Sturm. Structure-from-motion us- ing lines: Representation, triangulation, and bundle adjust- ment.Computer vision and image understanding, 100(3): 416–441, 2005. 2

    work page 2005

  4. [4]

    A variational approach to edge detection

    John F Canny. A variational approach to edge detection. In AAAI, pages 54–58, 1983. 3

    work page 1983

  5. [5]

    Moving in stereo: Efficient structure and motion us- ing lines

    Manmohan Chandraker, Jongwoo Lim, and David Krieg- man. Moving in stereo: Efficient structure and motion us- ing lines. In2009 IEEE 12th International Conference on Computer Vision, pages 1741–1748. IEEE, 2009. 2

    work page 2009

  6. [6]

    EdgeGaussians -- 3D Edge Mapping via Gaussian Splatting

    Kunal Chelani, Assia Benbihi, Torsten Sattler, and Fredrik Kahl. Edgegaussians–3d edge mapping via gaussian splat- ting.arXiv preprint arXiv:2409.12886, 2024. 1, 3, 6, 7, 12

    work page Pith review arXiv 2024

  7. [7]

    3doodle: Compact abstraction of objects with 3d strokes.ACM Trans

    Changwoon Choi, Jaeah Lee, Jaesik Park, and Young Min Kim. 3doodle: Compact abstraction of objects with 3d strokes.ACM Trans. Graph., 43(4), 2024. 3

    work page 2024

  8. [8]

    Fischler and Robert C

    Martin A. Fischler and Robert C. Bolles. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Commun. ACM, 24:381–395, 1981. 1, 2, 3

    work page 1981

Show all 60 references
  1. [9]

    Fdc-nerf: learning pose-free neural radiance fields with flow-depth consistency

    Huachen Gao, Shihe Shen, Zhe Zhang, Kaiqiang Xiong, Rui Peng, Zhirui Gao, Qi Wang, Yugui Xie, and Ronggang Wang. Fdc-nerf: learning pose-free neural radiance fields with flow-depth consistency. InICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Pr...

    2024

  2. [10]

    Learning accurate template match- ing with differentiable coarse-to-fine correspondence refine- ment.Computational Visual Media, 10(2):309–330, 2024

    Zhirui Gao, Renjiao Yi, Zheng Qin, Yunfan Ye, Chenyang Zhu, and Kai Xu. Learning accurate template match- ing with differentiable coarse-to-fine correspondence refine- ment.Computational Visual Media, 10(2):309–330, 2024. 1

    2024

  3. [11]

    Self-supervised learning of hy- brid part-aware 3d representations of 2d gaussians and su- perquadrics, 2025

    Zhirui Gao, Renjiao Yi, Yuhang Huang, Wei Chen, Chenyang Zhu, and Kai Xu. Self-supervised learning of hy- brid part-aware 3d representations of 2d gaussians and su- perquadrics, 2025. 3

    2025

  4. [12]

    Generic objects as pose probes for few- shot view synthesis.IEEE Transactions on Circuits and Sys- tems for Video Technology, 2025

    Zhirui Gao, Renjiao Yi, Chenyang Zhu, Ke Zhuang, Wei Chen, and Kai Xu. Generic objects as pose probes for few- shot view synthesis.IEEE Transactions on Circuits and Sys- tems for Video Technology, 2025. 3

    2025

  5. [13]

    Efficient 3d scene abstraction using line segments.Computer Vision and Image Understanding, 157:167–178, 2017

    Manuel Hofer, Michael Maurer, and Horst Bischof. Efficient 3d scene abstraction using line segments.Computer Vision and Image Understanding, 157:167–178, 2017. 3

    2017

  6. [14]

    Mv2cyl: Reconstructing 3d extru- sion cylinders from multi-view images.arXiv preprint arXiv:2406.10853, 2024

    Eunji Hong, Minh Hieu Nguyen, Mikaela Angelina Uy, and Minhyuk Sung. Mv2cyl: Reconstructing 3d extru- sion cylinders from multi-view images.arXiv preprint arXiv:2406.10853, 2024. 6, 7, 12

    2024 arXiv

  7. [15]

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

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

    2023

  8. [16]

    Abc: A big cad model dataset for geometric deep learning

    Sebastian Koch, Albert Matveev, Zhongshi Jiang, Francis Williams, Alexey Artemov, Evgeny Burnaev, Marc Alexa, Denis Zorin, and Daniele Panozzo. Abc: A big cad model dataset for geometric deep learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Rec...

    2019

  9. [17]

    Arm3d: Attention-based re- lation module for indoor 3d object detection.Computational Visual Media, pages 1–20, 2022

    Yuqing Lan, Yao Duan, Chenyi Liu, Chenyang Zhu, Yueshan Xiong, Hui Huang, and Kai Xu. Arm3d: Attention-based re- lation module for indoor 3d object detection.Computational Visual Media, pages 1–20, 2022. 1

    2022

  10. [18]

    Compact 3d gaussian representation for radiance field

    Joo Chan Lee, Daniel Rho, Xiangyu Sun, Jong Hwan Ko, and Eunbyung Park. Compact 3d gaussian representation for radiance field. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 21719–21728, 2024. 4, 5

    2024

  11. [19]

    3d neural edge recon- struction

    Lei Li, Songyou Peng, Zehao Yu, Shaohui Liu, R´emi Pautrat, Xiaochuan Yin, and Marc Pollefeys. 3d neural edge recon- struction. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 1, 3, 6, 7, 12, 13

    2024

  12. [20]

    Differentiable vector graphics rasterization for editing and learning.ACM Trans

    Tzu-Mao Li, Michal Luk ´aˇc, Gharbi Micha ¨el, and Jonathan Ragan-Kelley. Differentiable vector graphics rasterization for editing and learning.ACM Trans. Graph. (Proc. SIG- GRAPH Asia), 39(6):193:1–193:15, 2020. 3

    2020

  13. [21]

    Uv-slam: Unconstrained line-based slam using vanishing points for structural mapping

    Hyunjun Lim, Jinwoo Jeon, and Hyun Myung. Uv-slam: Unconstrained line-based slam using vanishing points for structural mapping. 7(2):1518–1525, 2022. 3

    2022

  14. [22]

    3d line mapping revisited

    Shaohui Liu, Yifan Yu, R ´emi Pautrat, Marc Pollefeys, and Viktor Larsson. 3d line mapping revisited. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21445–21455, 2023. 1, 2, 6, 7, 12

    2023

  15. [23]

    Integration of 3d lines and points in 6dof visual slam by uncertain projective geometry

    Daniele Marzorati, Matteo Matteucci, Davide Migliore, and Domenico G Sorrenti. Integration of 3d lines and points in 6dof visual slam by uncertain projective geometry. InEMCR. Citeseer, 2007. 3

    2007

  16. [24]

    Def: Deep estimation of sharp geometric features in 3d shapes.ACM Transactions on Graphics, 41(4), 2022

    Albert Matveev, Ruslan Rakhimov, Alexey Artemov, Gleb Bobrovskikh, Vage Egiazarian, Emil Bogomolov, Daniele Panozzo, Denis Zorin, and Evgeny Burnaev. Def: Deep estimation of sharp geometric features in 3d shapes.ACM Transactions on Graphics, 41(4), 2022. 3

    2022

  17. [25]

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

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceed- ings...

    2020

  18. [26]

    Sold2: Self-supervised occlusion- aware line description and detection

    R ´emi Pautrat, Juan-Ting Lin, Viktor Larsson, Martin R Os- wald, and Marc Pollefeys. Sold2: Self-supervised occlusion- aware line description and detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11368–11378, 2021. 2

    2021

  19. [27]

    Deeplsd: Line segment detection and refinement with deep image gradients

    R ´emi Pautrat, Daniel Barath, Viktor Larsson, Martin R Os- wald, and Marc Pollefeys. Deeplsd: Line segment detection and refinement with deep image gradients. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17327–17336, 2023. 2

    2023

  20. [28]

    Gluestick: Robust image matching by sticking points and lines together

    R ´emi Pautrat, Iago Su ´arez, Yifan Yu, Marc Pollefeys, and Viktor Larsson. Gluestick: Robust image matching by sticking points and lines together. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 9706–9716, 2023. 2

    2023

  21. [29]

    Dense extreme inception network: Towards a robust cnn model for edge detection

    Xavier Soria Poma, Edgar Riba, and Angel Sappa. Dense extreme inception network: Towards a robust cnn model for edge detection. InProceedings of the IEEE/CVF winter conference on applications of computer vision, pages 1923– 1932, 2020. 7, 8, 12

    1923

  22. [30]

    Sam 2: Segment anything in images and videos

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman R¨adle, Chloe Rolland, Laura Gustafson, Eric Mintun, Junt- ing Pan, Kalyan Vasudev Alwala, Nicolas Carion, Chao- Yuan Wu, Ross Girshick, Piotr Doll´ar, and Christoph Feic...

    2024 arXiv

  23. [31]

    Line-based structure from motion for urban envi- ronments

    Grant Schindler, Panchapagesan Krishnamurthy, and Frank Dellaert. Line-based structure from motion for urban envi- ronments. InThird International Symposium on 3D Data Processing, Visualization, and Transmission (3DPVT’06), pages 846–853. IEEE, 2006. 2

    2006

  24. [32]

    Structure plp-slam: Efficient sparse mapping and localization using point, line and plane for monocular, rgb-d and stereo cameras

    Fangwen Shu, Jiaxuan Wang, Alain Pagani, and Didier Stricker. Structure plp-slam: Efficient sparse mapping and localization using point, line and plane for monocular, rgb-d and stereo cameras. 3

  25. [33]

    Multi-view recon- struction using photo-consistency and exact silhouette con- straints: A maximum-flow formulation

    Sudipta N Sinha and Marc Pollefeys. Multi-view recon- struction using photo-consistency and exact silhouette con- straints: A maximum-flow formulation. InICCV, pages 349–

  26. [34]

    Julian Straub, Thomas Whelan, Lingni Ma, Yufan Chen, Erik Wijmans, Simon Green, Jakob J. Engel, Raul Mur-Artal, Carl Ren, Shobhit Verma, Anton Clarkson, Mingfei Yan, Brian Budge, Yajie Yan, Xiaqing Pan, June Yon, Yuyang Zou, Kimberly Leon, Nigel Carter, Jesus Briales, Tyler Gi...

    1906 arXiv

  27. [35]

    Pixel difference net- works for efficient edge detection

    Zhuo Su, Wenzhe Liu, Zitong Yu, Dewen Hu, Qing Liao, Qi Tian, Matti Pietikainen, and Li Liu. Pixel difference net- works for efficient edge detection. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 5117–5127, 2021. 7, 12

    2021

  28. [36]

    Pixel difference networks for efficient edge detection

    Zhuo Su, Wenzhe Liu, Zitong Yu, Dewen Hu, Qing Liao, Qi Tian, Matti Pietik ¨ainen, and Li Liu. Pixel difference networks for efficient edge detection. InProceedings of the IEEE/CVF international conference on computer vision, pages 5117–5127, 2021. 3

    2021

  29. [37]

    Fast continual multi-view clustering with incomplete views.IEEE Transactions on Image Processing, 33:2995–3008, 2024

    Xinhang Wan, Bin Xiao, Xinwang Liu, Jiyuan Liu, Weixuan Liang, and En Zhu. Fast continual multi-view clustering with incomplete views.IEEE Transactions on Image Processing, 33:2995–3008, 2024. 1

    2024

  30. [38]

    Omniearth-bench: Towards holistic evaluation of earth’s six spheres and cross-spheres interactions with multimodal observational earth data.arXiv preprint arXiv:2505.23522, 2025

    Fengxiang Wang, Mingshuo Chen, Xuming He, YiFan Zhang, Feng Liu, Zijie Guo, Zhenghao Hu, Jiong Wang, Jingyi Xu, Zhangrui Li, et al. Omniearth-bench: Towards holistic evaluation of earth’s six spheres and cross-spheres interactions with multimodal observational earth data.arXiv...

    2025

  31. [39]

    Harnessing massive satellite imagery with efficient masked image modeling, 2025

    Fengxiang Wang, Hongzhen Wang, Di Wang, Zonghao Guo, Zhenyu Zhong, Long Lan, Wenjing Yang, and Jing Zhang. Harnessing massive satellite imagery with efficient masked image modeling, 2025. 1

    2025

  32. [40]

    Pie-net: Parametric inference of point cloud edges.Advances in neu- ral information processing systems, 33:20167–20178, 2020

    Xiaogang Wang, Yuelang Xu, Kai Xu, Andrea Tagliasac- chi, Bin Zhou, Ali Mahdavi-Amiri, and Hao Zhang. Pie-net: Parametric inference of point cloud edges.Advances in neu- ral information processing systems, 33:20167–20178, 2020. 1, 3

    2020

  33. [41]

    Parametric primitive analysis of cad sketches with vision transformer.IEEE Transactions on Industrial Informatics, 20(10):12041–12050, 2024

    Xiaogang Wang, Liang Wang, Hongyu Wu, Guoqiang Xiao, and Kai Xu. Parametric primitive analysis of cad sketches with vision transformer.IEEE Transactions on Industrial Informatics, 20(10):12041–12050, 2024. 1

    2024

  34. [42]

    Elsr: Efficient line segment reconstruc- tion with planes and points guidance

    Dong Wei, Yi Wan, Yongjun Zhang, Xinyi Liu, Bin Zhang, and Xiqi Wang. Elsr: Efficient line segment reconstruc- tion with planes and points guidance. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15807–15815, 2022. 3

    2022

  35. [43]

    Fusion 360 gallery: A dataset and environ- ment for programmatic cad construction from human design sequences.ACM Transactions on Graphics (TOG), 40(4): 1–24, 2021

    Karl DD Willis, Yewen Pu, Jieliang Luo, Hang Chu, Tao Du, Joseph G Lambourne, Armando Solar-Lezama, and Wo- jciech Matusik. Fusion 360 gallery: A dataset and environ- ment for programmatic cad construction from human design sequences.ACM Transactions on Graphics (TOG), 40(4): ...

    2021

  36. [44]

    Differentiable rendering of parametric geometry.ACM Trans

    Markus Worchel and Marc Alexa. Differentiable rendering of parametric geometry.ACM Trans. Graph., 42(6), 2023. 3

    2023

  37. [45]

    Holistically- attracted wireframe parsing: From supervised to self- supervised learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

    Nan Xue, Tianfu Wu, Song Bai, Fu-Dong Wang, Gui-Song Xia, Liangpei Zhang, and Philip HS Torr. Holistically- attracted wireframe parsing: From supervised to self- supervised learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023. 2

    2023

  38. [46]

    Neat: Distilling 3d wire- frames from neural attraction fields

    Nan Xue, Bin Tan, Yuxi Xiao, Liang Dong, Gui-Song Xia, Tianfu Wu, and Yujun Shen. Neat: Distilling 3d wire- frames from neural attraction fields. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19968–19977, 2024. 3

    2024

  39. [47]

    Interpolation-based contrastive learning for few-label semi-supervised learning.IEEE Transactions on Neural Networks and Learning Systems, 35(2):2054–2065,

    Xihong Yang, Xiaochang Hu, Sihang Zhou, Xinwang Liu, and En Zhu. Interpolation-based contrastive learning for few-label semi-supervised learning.IEEE Transactions on Neural Networks and Learning Systems, 35(2):2054–2065,

  40. [48]

    Dealmvc: Dual contrastive calibration for multi-view 10 clustering

    Xihong Yang, Jin Jiaqi, Siwei Wang, Ke Liang, Yue Liu, Yi Wen, Suyuan Liu, Sihang Zhou, Xinwang Liu, and En Zhu. Dealmvc: Dual contrastive calibration for multi-view 10 clustering. InProceedings of the 31st ACM International Conference on Multimedia, pages 337–346, 2023. 3

    2023

  41. [49]

    V ol- ume rendering of neural implicit surfaces.Advances in Neu- ral Information Processing Systems, 34:4805–4815, 2021

    Lior Yariv, Jiatao Gu, Yoni Kasten, and Yaron Lipman. V ol- ume rendering of neural implicit surfaces.Advances in Neu- ral Information Processing Systems, 34:4805–4815, 2021. 3

    2021

  42. [50]

    Stablenormal: Reducing diffusion variance for stable and sharp normal.ACM Transactions on Graphics (TOG), 2024

    Chongjie Ye, Lingteng Qiu, Xiaodong Gu, Qi Zuo, Yushuang Wu, Zilong Dong, Liefeng Bo, Yuliang Xiu, and Xiaoguang Han. Stablenormal: Reducing diffusion variance for stable and sharp normal.ACM Transactions on Graphics (TOG), 2024. 12

    2024

  43. [51]

    Delving into crispness: Guided label refinement for crisp edge detection.IEEE Transactions on Image Processing, 32: 4199–4211, 2023

    Yunfan Ye, Renjiao Yi, Zhirui Gao, Zhiping Cai, and Kai Xu. Delving into crispness: Guided label refinement for crisp edge detection.IEEE Transactions on Image Processing, 32: 4199–4211, 2023. 3

    2023

  44. [52]

    Nef: Neural edge fields for 3d parametric curve reconstruction from multi-view images

    Yunfan Ye, Renjiao Yi, Zhirui Gao, Chenyang Zhu, Zhiping Cai, and Kai Xu. Nef: Neural edge fields for 3d parametric curve reconstruction from multi-view images. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 8486–8495, 2023....

    2023

  45. [53]

    Pc2wf:3d wireframe reconstruction from raw point clouds

    Konrad Schindler Yujia Liu, Stefano D’Aronco and Jan Dirk Wegner. Pc2wf:3d wireframe reconstruction from raw point clouds. InProceedings of the International Conference on Learning Representations, 2021. 1, 3

    2021

  46. [54]

    Diff3ds: Generating view-consistent 3d sketch via differentiable curve rendering.arXiv preprint arXiv:2405.15305, 2024

    Yibo Zhang, Lihong Wang, Changqing Zou, Tieru Wu, and Rui Ma. Diff3ds: Generating view-consistent 3d sketch via differentiable curve rendering.arXiv preprint arXiv:2405.15305, 2024. 3

    2024 arXiv

  47. [55]

    Monomobility: Zero-shot 3d mobility analysis from monoc- ular videos, 2025

    Hongyi Zhou, Xiaogang Wang, Yulan Guo, and Kai Xu. Monomobility: Zero-shot 3d mobility analysis from monoc- ular videos, 2025. 3

    2025

  48. [56]

    Nerve: Neural volumetric edges for parametric curve extraction from point cloud

    Xiangyu Zhu, Dong Du, Weikai Chen, Zhiyou Zhao, Yinyu Nie, and Xiaoguang Han. Nerve: Neural volumetric edges for parametric curve extraction from point cloud. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13601–13610, 2023. 1, 3 1...

    2023

  49. [57]

    Detailed descriptions of each dataset and the experimental setups are provided below

    Datasets The proposed method is evaluated on three publicly avail- able datasets: ABC-NEF [52], Mv2Cyl’s Real Objects [14], and the Replica Dataset [34]. Detailed descriptions of each dataset and the experimental setups are provided below. ABC-NEF Dataset.The ABC-NEF dataset i...

  50. [58]

    These in- clude three learning-based methods—NEF [52], EdgeGaus- sians [6], and EMAP [19]—and one line-based Structure- from-Motion (SfM) method, LIMAP [22]

    Baselines Our method is compared against four state-of-the-art 3D line and curve reconstruction baselines. These in- clude three learning-based methods—NEF [52], EdgeGaus- sians [6], and EMAP [19]—and one line-based Structure- from-Motion (SfM) method, LIMAP [22]. For a fair c...

  51. [59]

    Points are uniformly sampled along both the reconstructed parametric curves and the cor- responding ground-truth edges, enabling a direct compari- son between them

    Evaluation Metrics To quantitatively assess the performance of our method, we adopt a set of evaluation metrics that align with estab- lished protocols in this field. Points are uniformly sampled along both the reconstructed parametric curves and the cor- responding ground-tru...

  52. [60]

    For all B ´ezier curves and straight lines, a default of 12 Gaussian points was sam- pled per curve

    Implementation Details The weight coefficientsλ 1,λ 2,λ 3, andλ 4 were set to0.01, 0.01,0.01, and0.0005, respectively. For all B ´ezier curves and straight lines, a default of 12 Gaussian points was sam- pled per curve. For the ABC dataset, the midpoints of B ´ezier curves wer...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.