Pith. sign in

REVIEW 4 major objections 5 minor 74 references

WIR3D: Visually-Informed and Geometry-Aware 3D Shape Abstraction

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read WIR3D turns 3D shapes into sparse, meaningful 3D curves that stay faithful from every viewing angle.

desk verdict WIR3D is a solid methods paper with real contributions, but its arbitrary-view fidelity claim outruns the 0–30° elevation evaluation. read the letter →

arxiv 2505.04813 v2 pith:SWZYTMU3 submitted 2025-05-07 cs.GR cs.CV

Richard Liu , Daniel Fu , Noah Tan , Itai Lang , Rana Hanocka This is my paper · ORCID
classification cs.GRcs.CV
keywords 3DshapeabstractionBeziercurvessketchCLIPsupervisionlocalizedkeypointlossdifferentiablerenderingneuralSDFdeformationhandles
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

WIR3D claims that a 3D shape can be abstracted into a sparse set of 3D Bezier curves that stay faithful to both the geometry and the salient visual features, including texture, from every viewing angle. The authors argue that this cannot be done with surface analysis alone, because occluding contours are view-dependent and ignore texture, so they supervise the curves with the intermediate activations of CLIP, a pretrained vision-language model. The optimization is split into a geometry stage and a texture stage, with the texture stage driven by a newly introduced localized keypoint loss that lets specific features be emphasized. A neural SDF loss keeps curves on the surface, which is what allows the curves to double as intuitive deformation handles and enables interactive feature refinement. A sympathetic reading of the results is that WIR3D establishes a new capability: semantically meaningful, user-controllable, view-consistent 3D sketch abstraction for arbitrary meshes, including noisy photo-reconstructed ones.

What carries the argument

The load-bearing mechanism is the localized keypoint loss. Given a set of 3D keypoints, either user-selected or automatically detected by back-projecting CLIP features and clustering, each keypoint is projected into the current view and converted into a Gaussian weight map $I_{\mathrm{weight}}$; this map is downsampled and multiplied into the per-pixel difference between CLIP intermediate activation maps of the curve render and the target render, so that optimization effort concentrates around the keypoint. The loss also includes a mean-pooled weight applied to the global CLIP cosine distance and an LPIPS term. Supporting it are a two-stage optimization, with CLIP RN101 for geometry on Freestyle contour renders and CLIP RN50x64 for texture on surface renders, a neural SDF loss $L_{\mathrm{SDF}}$ that penalizes curve samples away from the zero level set, an NDC regularization keeping curves inside the frame, and differentiable rasterization of projected 3D Bezier control points.

What would settle it

Run the stage-II optimization on the same shape twice, once with keypoints on a salient feature and once with the same number of keypoints on a flat, featureless region, keeping all other settings fixed; if the two abstractions are indistinguishable by user ranking or by L2 distance between curve sets, then the localized loss is not doing spatial work. A stronger quantitative version is to take ground-truth keypoint positions, project them into views, and check whether the argmax of the CLIP intermediate activation difference, or the recovered center of the weight map, falls within a few pixels of the projected keypoint; chance-level localization would falsify the paper's central mechanism.

Watch

Extended reading notes

Core claim

The central claim is that a set of cubic Bezier curves optimized against spatially weighted CLIP activations yields a 3D abstraction that is simultaneously sparse, view-consistent, and semantically faithful. In the paper's own framing, WIR3D 'can abstract a myriad of shapes from different domains with various visual concepts, geometric structures, and textures,' and the abstractions 'maintain high fidelity across arbitrary views.' The novel component is the localized keypoint loss: 3D keypoints are projected into each sampled view, a Gaussian weight map is constructed around them, and this map multiplies the difference between CLIP intermediate activations of the rendered curves and of the target shape, focusing optimization on the feature at the keypoint. The method also anchors curves to the surface with a neural SDF loss, so the optimized strokes are not just decorative but can serve as deformation handles. The authors demonstrate the claim qualitatively across textured and untextured shapes, on messy in-the-wild reconstructions, and with user studies favoring WIR3D over the 3Doodle baseline in 88% of pairwise comparisons.

Load-bearing premise

The texture stage and feature-control application rest on the assumption that the internal feature maps of CLIP are spatially correlated with the input image, so that a weight map built around projected keypoints actually concentrates the optimization on the intended feature; if CLIP's activations do not localize that way, the localized loss collapses to the unweighted loss and keypoint control stops working.

Editorial extensions

If this is right

  • Because the strokes are defined as 3D Bezier curves, the abstraction is view-consistent by construction and avoids the flickering that plagues per-view occluding contours.
  • The level of abstraction is controlled simply by the number of curves; adding strokes automatically adds finer detail, as shown in the paper's Fig. 6.
  • Users can add or refine detail interactively by selecting keypoints, with refinement completing in roughly a minute.
  • The curves' adherence to the surface lets them act as deformation handles; a user study found WIR3D-based deformations preferable to ARAP-based deformation 80% of the time.
  • The method needs no clean input mesh and produces meaningful abstractions even from photo-reconstructed models with boundary and non-manifold edges.

Reading between the lines

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

  • A direct test of the localization assumption would be to measure whether CLIP intermediate activation maps predict the projected positions of known keypoints; the paper only tests this indirectly through the random-keypoint ablation, so the spatial-precision claim is the least-supported link.
  • The same spatially weighted CLIP supervision could be applied to abstractions in other modalities, such as video frames or multi-object scenes, where the keypoints would enforce temporal or spatial consistency.
  • Because the curves are bound to the surface through the SDF loss, they could serve as a general rigging primitive; skinning weights derived from curve distance might extend beyond the proof-of-concept deformation shown here.
  • The automatic keypoint detector inherits the biases of the CLIP features used for back-projection, so the method's notion of 'salient feature' may silently be CLIP's notion of salience rather than a universal one.
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 introduces WIR3D, a method that abstracts a 3D shape into a sparse set of 3D cubic Bezier curves intended to represent both geometry and salient texture from arbitrary viewpoints. The method optimizes the curves with a two-stage procedure: stage I captures coarse geometry using CLIP semantic losses supervised by Freestyle renders, and stage II adds curves for texture using a novel localized keypoint loss that weights CLIP intermediate feature differences around projected 3D keypoints. An SDF regularization encourages curve adherence to the surface, and the authors demonstrate applications in interactive feature refinement and curve-based deformation. The evaluation includes qualitative results on textured and untextured shapes, quantitative comparisons with 3Doodle and NEF using LPIPS, CLIP image similarity, user preference, and a Chamfer-based coverage metric, plus ablation studies and two user studies.

Significance. If the central claims hold, WIR3D would be a useful contribution to 3D sketch abstraction, addressing a gap left by occlusion-contour methods and by dataset-dependent reconstruction approaches. The two-stage optimization, the localized keypoint weighting, and the SDF-based surface anchoring are reasonable and potentially transferable ideas. The paper's strengths include the breadth of qualitative results across varied shapes, the inclusion of two user studies that provide partially independent evidence, and the use of established baselines (3Doodle, NEF). The automatic keypoint detection via back-projection and clustering is also a sensible mechanism. However, the quantitative evidence is weaker than the narrative suggests: the headline metrics overlap in family with the optimization losses, no variance information is reported for stochastic optimization, and the 'arbitrary views' claim is not tested outside the training elevation band. These issues need to be addressed before the contribution can be fully assessed.

major comments (4)
  1. [Sec. 3.2 and Sec. 4.2, Table 1] The quantitative comparison is partly circular. The optimization losses in Eq. (2) and Eq. (4) use cosine similarity on global CLIP embeddings plus L2 distances on intermediate CLIP features, with ResNet CLIP variants (RN101 and RN50x16), and the LPIPS term uses the VGG variant (supplemental Sec. C). The headline evaluation in Table 1 uses CLIP ViT/B-32 image similarity and AlexNet LPIPS. The architecture split reduces but does not eliminate the concern: Table 1 still measures the same family of perceptual losses that the method is explicitly trained against, so the reported superiority over 3Doodle on LPIPS and CLIP image similarity is partly an artifact of optimizing objectives of the same type. I do not regard this as fatal because the user study (N=96) is independent and is the strongest evidence in the paper, but the claim that WIR3D 'outperforms' baselines on the first two metrics is not independent confirmation. Please report per-shape results, metrics obtained from unrelated evaluators (e.g., human judgment already available), or at least discuss this limitation quantitatively.
  2. [Sec. 4, paragraphs 1-2; Table 1 and Table 2] No variance or confidence information is reported for any quantitative result. The optimization is stochastic: one random view is sampled per iteration, random initializations are used, and the CLIP supervision is augmented (supplemental Sec. G). Tables 1 and 2 report only point estimates. Consequently, small differences such as LPIPS 0.227 vs 0.229 (No SDF) or CLIP image similarity 0.909 vs 0.904 cannot be interpreted as meaningful, and even the larger apparent gains over 3Doodle lack error bars. Please report means with standard deviations or confidence intervals across multiple optimization runs and across views, and state the number of runs used.
  3. [Sec. 4, first paragraph; Sec. 4.1; Fig. 16] The central claim that abstractions 'maintain high fidelity across arbitrary views' is not supported by the evaluation. All optimization views are sampled from elevations 0-30 degrees and azimuth 0-360 degrees, and no held-out or high-elevation evaluation is reported. Table 1 averages over novel views without a per-view breakdown, and the qualitative multi-view evidence (Fig. 16 and the supplemental) appears to demonstrate azimuthal rotation rather than a full sphere of viewpoints. Since there is no explicit multi-view consistency loss beyond the shared 3D curve parameters, it remains possible that the curves satisfy the training view distribution while misrepresenting the shape at high elevations. Please either add quantitative evaluation on held-out elevation bands (e.g., 30-60 and 60-90 degrees) and per-view error statistics, or soften the 'arbitrary views' claim to match the evaluated range.
  4. [Sec. 3.2, Eq. (4) and Sec. 3.3] The localized keypoint loss rests on the assumption that CLIP intermediate activations carry spatial information that can be meaningfully localized through a Gaussian weight map. The paper cites Shomron and Weiser [51] for this, but that reference studies spatial correlation in CNN activations generally for value prediction, not CLIP feature localization specifically, and no experiment in the paper directly validates that CLIP layers 3 and 4 provide the required spatial resolution for the proposed weighting. This is load-bearing because the texture stage and the user-control application both depend on the localization. The 'No Local' ablation and the noisy-keypoint experiment in Fig. 9 show robustness and overall utility, but they do not establish that the keypoint projections correlate with the features being emphasized. Please provide a direct spatial-localization check, for example by visualizing CLIP activation maps at keypoint locations or by comparing optimization outcomes with keypoints placed on different features against ground-truth feature regions.
minor comments (5)
  1. [Fig. 6 caption] Typo: 'progressvely' should be 'progressively'.
  2. [Sec. 4.2 and Sec. H] The paper uses inconsistent capitalization: 'Wir3D' appears in the user-study passages and in Fig. 24, while the method name is 'WIR3D' elsewhere. Please standardize.
  3. [Supplemental Sec. A] The term 'Janusing artifacts' should likely be 'Janus artifacts', and reference [54] on generative AI is an unusual citation for this graphics-specific phenomenon; please use a more standard reference.
  4. [Sec. 4.2, Table 1] The metric named 'Coverage' is a one-direction Chamfer distance from surface samples to curves. This measures proximity rather than coverage in the sense of how much of the surface is represented; please either rename the metric or add an explicit justification that this quantity captures coverage.
  5. [Sec. 1 and Data Availability] The paper states that code will be released 'in the near future' but no code is provided. Given that the method is an optimization with several tuned hyperparameters, releasing the code would substantially aid reproducibility; please state in the final version where and when the code will be available.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: WIR3D is an optimization method validated by independent user studies and by evaluation metrics that are architecture-split from the training losses.

full rationale

WIR3D is an optimization method rather than a fitted predictor, so the standard circularity patterns do not apply. The Bezier curves are optimized from furthest-point initialization against CLIP, LPIPS, SDF, and NDC losses; no quantity is fitted to a subset of data and then reported as a prediction of the same quantity. The closest candidate, using perceptual metrics in Table 1 that resemble the semantic losses in Eq. (2) and Eq. (4), is mitigated by the paper's explicit architecture split (ResNet CLIP for optimization vs ViT/B-32 for CLIP img; VGG LPIPS for optimization vs AlexNet for evaluation) and by independent user studies (88% preference over 3Doodle and 80% preference over ARAP). The coverage metric is not identical to the SDF loss: the SDF loss only requires curve samples to lie near the surface, while coverage measures whether surface points are near the curve set, so the metric is not forced by the loss. The automatic keypoint detection uses CLIP features, but this is a design choice rather than a circular derivation, since the claims about visual salience are validated by human raters rather than by CLIP alone. The discussion of arbitrary views is an evaluation-scope concern, not a circularity. No load-bearing argument is justified by a self-citation; the GeoCode and PointGMM citations are related-work only.

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

The method rests on pretrained model priors (CLIP, LPIPS) and hand-tuned optimization settings. No new physical entities are introduced. The main free choices are the CLIP architecture split, loss weights, curve counts, and keypoint count.

free parameters (4)
  • CLIP architecture split = RN101 (stage 1), RN50x64 (stage 2; supplementary says RN50x16)
    The two-stage pipeline depends on the empirical claim that RN101 captures geometry and RN50x64 captures high-level texture semantics. This split is chosen by hand and no systematic comparison is provided.
  • Stage 2 loss weights and Gaussian scale = lambda_fc=75, lambda_lpips=0.1, sigma=0.1
    The localized keypoint loss (Eq. 4) weights CLIP global, LPIPS, and spatial dropoff with these hand-tuned values; no sensitivity analysis is reported.
  • Curve count = 20 geometry curves and 20 texture curves
    The abstraction level is set by the number of Bezier curves; all experiments use the same count. The paper does not study how this interacts with shape complexity.
  • Keypoint count k = Set to the stage 2 curve count (20)
    Automatic keypoint detection uses KMeans clustering over backprojected CLIP features; k is a free choice, and the paper ties it to the number of texture curves.
assumptions (4)
  • domain assumption Theorem 1 from 3Doodle: with a sufficiently distant camera, optimizing 3D Bezier curves and projecting them is equivalent to optimizing 2D rational Bezier curves.
    The camera is assumed far from the shape such that orthographic and perspective projection nearly match (Sec. 3.1). This is cited, not re-derived.
  • domain assumption CLIP intermediate activations are spatially correlated with input image locations.
    Sec. 3.2 bases the localized keypoint loss on this, citing Shomron and Weiser (2019) for CNNs generally; the paper does not validate it for CLIP.
  • ad hoc to paper Freestyle renders isolate geometric structure and serve as an appropriate target for stage 1.
    Sec. A (supplemental) introduces Freestyle rendering as a styling choice; the assumption that this stylization captures the right geometry is design-dependent and not independently validated.
  • domain assumption The fitted neural SDF MLP accurately approximates the true distance field of the shape.
    Sec. 3.2 uses an MLP fit to the SDF for the SDF loss; the accuracy of this fit is assumed adequate for curve adherence, with no error analysis.

how reviews work

0 comments
Cite this review

Pith. "Pith review of WIR3D: Visually-Informed and Geometry-Aware 3D Shape Abstraction." pith.science (2026). https://pith.science/paper/SWZYTMU3

@misc{pith2026250504813,
  author       = {Pith},
  title        = {Pith review of: WIR3D: Visually-Informed and Geometry-Aware 3D Shape Abstraction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SWZYTMU3}},
  note         = {Machine review of arXiv:2505.04813}
}
read the original abstract

In this work we present WIR3D, a technique for abstracting 3D shapes through a sparse set of visually meaningful curves in 3D. We optimize the parameters of Bezier curves such that they faithfully represent both the geometry and salient visual features (e.g. texture) of the shape from arbitrary viewpoints. We leverage the intermediate activations of a pre-trained foundation model (CLIP) to guide our optimization process. We divide our optimization into two phases: one for capturing the coarse geometry of the shape, and the other for representing fine-grained features. Our second phase supervision is spatially guided by a novel localized keypoint loss. This spatial guidance enables user control over abstracted features. We ensure fidelity to the original surface through a neural SDF loss, which allows the curves to be used as intuitive deformation handles. We successfully apply our method for shape abstraction over a broad dataset of shapes with varying complexity, geometric structure, and texture, and demonstrate downstream applications for feature control and shape deformation.

Figures

Figures reproduced from arXiv: 2505.04813 by the authors.

Figure 1
Figure 1. WIR3D produces 3D shape abstractions in the form of 3D strokes. The abstractions retain the overall shape structure and capture [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Motivating example. Abstracting even a simple shape like a cylinder is nontrivial. Rendering the contours of the cylinder (inset) results in a static image from every view, removing any sense of 3D volume. Backprojecting these contours into 3D results in a dense cluster of lines spanning the body of the cylinder, which is both non-sparse and unsatisfying aesthetically. In contrast, our method effectively abstracts t… view at source ↗
Figure 3
Figure 3. Method overview. In the first stage, our WIR3D learns to abstract the underlying geometry of the shape. In the second stage, we freeze the curves from the first stage and add new curves that are optimized to represent the shape’s texture. methods. Furthermore, these methods cannot account for texture features which do not induce a prominent edge map. Curve-based abstraction. Our work is in the domain of sketch abstr… view at source ↗
Figures from the paper (19 more)
Figure 4
Figure 4. Figure 4: Localized keypoint loss. Our localized keypoints weight the loss between the intermediate feature maps of the en￾coded curve render Icurve and the target shape render Itarget. This weight is obtained through projecting 3D keypoints (red), followed by a Gaussian filter …
Figure 5
Figure 5. Figure 5: Qualitative results for textured objects. We show WIR3D’s the result on a collection of textured meshes. that RN101 tends to be more sensitive to geometric struc￾tures, whereas RN50x64 is more sensitive to higher-level visual concepts. 4. Experiments We evaluate WIR3D …
Figure 7
Figure 7. Figure 7: In-The-Wild Reconstructions. We abstract two shapes from the Real-World-Textured Things dataset [36], which pro￾vides textured models reconstructed using modern off-the-shelf photo-reconstruction tools on in-the-wild images. The shoe (top) contains 360 boundary edges, …
Figure 8
Figure 8. Figure 8: Qualitative comparison. We compare a subset of our textured and untextured results against 3Doodle. 3Doodle pro￾duces reasonable abstractions that capture the overall structure but lack precision when it comes to the feature details. truth render. Thus, the numerical d…
Figure 9
Figure 9. Figure 9: Keypoint Robustness. Our localized keypoint loss is robust. When keypoints are random (“Random Keypoints”; posi￾tions in insets shown), the result is similar to when the localized keypoint loss is not used at all (“No Keypoints”) [PITH_FULL_IMAGE:figures/full_fig_p007…
Figure 10
Figure 10. Figure 10: Deformation application. The curve abstractions make for intuitive control handles for shape deformation. The curves wrap salient features such that deformations are transferred smoothly through a simple L2 distance skinning procedure [PITH_FULL_IMAGE:figures/full_fi…
Figure 11
Figure 11. Figure 11: Keypoint control. Our spatial weighting framework enables user-interactive detail refinement. 4.4. Applications We demonstrate two applications enabled by our 3D curve representation. The first is user-interactive feature control. By selecting additional keypoints, ne…
Figure 12
Figure 12. Figure 12: NEF qualitative comparison. We show NEF results on the same models we compare to 3Doodle in the main paper. NEF is specialized for simple manufactured CAD shapes, so it struggles to fit edges to more complex surfaces. This limitation was similarly observed in 3Doodle.…
Figure 13
Figure 13. Figure 13: Freestyle render ablation. Running our method with￾out freestyle renders still produces a reasonable abstraction, but key geometric features, such as the wheels of the car, may be missed due to the lack of visual signal from the surface renders [PITH_FULL_IMAGE:figur…
Figure 14
Figure 14. Figure 14: Perceptual metrics reliability. We show the unreli￾ability of CLIPimg in evaluating semantic similarity of curve ab￾straction to a target. We show for a given view, our stroke abstrac￾tion, 3Doodle’s, the edge map for the view extracted using Canny edge detection [3],…
Figure 16
Figure 16. Figure 16: Multi-view fidelity. WIR3D adheres to the abstracted object in a 3D-consistent manner such that its properties can be perceived from every viewing angle. Curve-Based Shape Deformation. Our deformation ap￾plication exploits the close correspondence between the op￾timiz…
Figure 19
Figure 19. Figure 19: SDF ablation. The SDF loss helps to ensure abstracted visual features will stay anchored to the surface implied by the strokes. Without it, some features may hover outside the surface, such as the smaller spots on Bob [PITH_FULL_IMAGE:figures/full_fig_p015_19.png]
Figure 20
Figure 20. Figure 20: Texture keypoint control. We expand on the keypoint control example shown in the main paper with a textured example. We show how by selecting keypoints on the texture on the plane, we are able to refine the abstraction by incorporating those texture elements. velop a …
Figure 18
Figure 18. Figure 18: CLIP layers ablation. Supervising with the intermedi￾ate activations of CLIP is critical for maintaining coherent geom￾etry. Using only the fully-connected CLIP output results in rough semantic abstraction, but the input shape geometric features are not well-preserved…
Figure 21
Figure 21. Figure 21: Additional textured results. We show additional tex￾tured results from the Meta DTC dataset [41]. the procedure can be performed at interactive speeds. We implement this deformation system as a proof-of-concept script, and show videos of the working system in the supp…
Figure 22
Figure 22. Figure 22: Scene abstraction. Our method extends to scene abstraction. Note how our method reproduces the global scene layout and captures all the objects in the scene despite the large scale differences [PITH_FULL_IMAGE:figures/full_fig_p017_22.png]
Figure 23
Figure 23. Figure 23: Perceptual Study Screenshots. Screenshots from our perceptual study. The question order and the order of Wir3D versus 3Doodle assignment to “Sketch1/Sketch2” are randomized for each respondent [PITH_FULL_IMAGE:figures/full_fig_p017_23.png]
Figure 24
Figure 24. Figure 24: Deformation Application User Study. Screenshots from our user study comparing our deformation application using WIR3D curves as handles against ARAP [53] The question order and the order of Wir3D versus 3Doodle assignment to “Option 1/Option 2” are randomized for each…
Figure 25
Figure 25. Figure 25: Comprehensive Localized Keypoint Loss. We show a comprehensive visualization of the localized keypoint loss [PITH_FULL_IMAGE:figures/full_fig_p018_25.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 69 canonical work pages

  1. [51]

    Spatial correlation and value prediction in convolutional neural networks

    Gil Shomron and Uri Weiser. Spatial correlation and value prediction in convolutional neural networks. IEEE Comput. Archit. Lett., 18(1):10–13, 2019. 4

    work page 2019

  2. [1]

    3D Surface Reconstruction Using Occluding Contours.Interna- tional Journal of Computer Vision, 22(3):219–233,

    Edmond Boyer and Marie-Odile Berger. 3D Surface Reconstruction Using Occluding Contours.Interna- tional Journal of Computer Vision, 22(3):219–233,

    work page

  3. [2]

    Line Draw- ings from 3D Models: A Tutorial.Foundations and Trends® in Computer Graphics and Vision, 11(1-2): 1–159, 2019

    Pierre B ´enard and Aaron Hertzmann. Line Draw- ings from 3D Models: A Tutorial.Foundations and Trends® in Computer Graphics and Vision, 11(1-2): 1–159, 2019. 2

    work page 2019

  4. [3]

    A Computational Approach to Edge Detection.IEEE Transactions on Pattern Analy- sis and Machine Intelligence, PAMI-8(6):679–698,

    John Canny. A Computational Approach to Edge Detection.IEEE Transactions on Pattern Analy- sis and Machine Intelligence, PAMI-8(6):679–698,

    work page

  5. [4]

    WireframeNet: A novel method for wireframe gener- ation from point cloud.Computers & Graphics, 115: 226–235, 2023

    Li Cao, Yike Xu, Jianwei Guo, and Xiaoping Liu. WireframeNet: A novel method for wireframe gener- ation from point cloud.Computers & Graphics, 115: 226–235, 2023. 3

    work page 2023

  6. [5]

    Algebraic Smooth Occluding Con- tours

    Ryan Capouellez, Jiacheng Dai, Aaron Hertzmann, and Denis Zorin. Algebraic Smooth Occluding Con- tours. InACM SIGGRAPH 2023 Conference Proceed- ings, pages 1–10, 2023. 1, 2

    work page 2023

  7. [6]

    Example- based facial sketch generation with non-parametric sampling

    Hong Chen, Ying-Qing Xu, Heung-Yeung Shum, Song-Chun Zhu, and Nan-Ning Zheng. Example- based facial sketch generation with non-parametric sampling. InProceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001, pages 433–438 vol.2, 2001. 3

    work page 2001

  8. [7]

    3Doodle: Compact Abstraction of Objects with 3D Strokes.ACM Transactions on Graphics (TOG), 43(4):1–13, 2024

    Changwoon Choi, Jaeah Lee, Jaesik Park, and Young Min Kim. 3Doodle: Compact Abstraction of Objects with 3D Strokes.ACM Transactions on Graphics (TOG), 43(4):1–13, 2024. 3, 4, 6

    work page 2024

Show all 74 references
  1. [8]

    Online 3D Edge Reconstruction of Wiry Struc- tures From Monocular Image Sequences.IEEE Robotics and Automation Letters, 8(11):7479–7486,

    Hyelim Choi, Minji Lee, Jiseock Kang, and Dongjun Lee. Online 3D Edge Reconstruction of Wiry Struc- tures From Monocular Image Sequences.IEEE Robotics and Automation Letters, 8(11):7479–7486,

  2. [9]

    Ro- bust fairing via conformal curvature flow.ACM Trans- actions on Graphics (TOG), 32(4):1–10, 2013

    Keenan Crane, Ulrich Pinkall, and Peter Schr¨oder. Ro- bust fairing via conformal curvature flow.ACM Trans- actions on Graphics (TOG), 32(4):1–10, 2013. 5

    2013

  3. [10]

    Suggestive contours for conveying shape.ACM Trans

    Doug DeCarlo, Adam Finkelstein, Szymon Rusinkiewicz, and Anthony Santella. Suggestive contours for conveying shape.ACM Trans. Graph., 22(3):848–855, 2003. 2

    2003

  4. [11]

    Suggestive Contours for Conveying Shape

    Doug DeCarlo, Adam Finkelstein, Szymon Rusinkiewicz, and Anthony Santella. Suggestive Contours for Conveying Shape. InSeminal Graphics Papers: Pushing the Boundaries, Volume 2, pages 401–408. ACM New York, NY , USA, 2023. 1

    2023

  5. [12]

    Cvxnet: Learnable Convex Decomposition

    Boyang Deng, Kyle Genova, Soroosh Yazdani, Sofien Bouaziz, Geoffrey Hinton, and Andrea Tagliasacchi. Cvxnet: Learnable Convex Decomposition. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), pages 31–44,

  6. [13]

    CLIP- Draw: Exploring Text-to-Drawing Synthesis through Language-Image Encoders.Advances in Neural In- formation Processing Systems, 35:5207–5218, 2022

    Kevin Frans, Lisa Soros, and Olaf Witkowski. CLIP- Draw: Exploring Text-to-Drawing Synthesis through Language-Image Encoders.Advances in Neural In- formation Processing Systems, 35:5207–5218, 2022. 3

    2022

  7. [14]

    iWIRES: An analyze-and-edit approach to shape manipulation.ACM Transactions on Graphics (proceedings of ACM SIGGRAPH), 28(3):33:1–33:10,

    Ran Gal, Olga Sorkine, Niloy Mitra, and Daniel Cohen-Or. iWIRES: An analyze-and-edit approach to shape manipulation.ACM Transactions on Graphics (proceedings of ACM SIGGRAPH), 28(3):33:1–33:10,

  8. [15]

    Breathing Life Into Sketches Using Text-to-Video Pri- ors

    Rinon Gal, Yael Vinker, Yuval Alaluf, Amit Bermano, Daniel Cohen-Or, Ariel Shamir, and Gal Chechik. Breathing Life Into Sketches Using Text-to-Video Pri- ors. In2024 IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR), pages 4325– 4336, Seattle, W A, USA, 2...

    2024

  9. [16]

    Semantic Abstrac- tion: Open-World 3D Scene Understanding from 2D Vision-Language Models, 2022

    Huy Ha and Shuran Song. Semantic Abstrac- tion: Open-World 3D Scene Understanding from 2D Vision-Language Models, 2022. arXiv:2207.11514 [cs]. 2

    2022 arXiv

  10. [17]

    Hennessey, Han Liu, Holger Winnem ¨oller, Mira Dontcheva, and Niloy J

    James W. Hennessey, Han Liu, Holger Winnem ¨oller, Mira Dontcheva, and Niloy J. Mitra. How2sketch: Generating easy-to-follow tutorials for sketching 3d objects.Symposium on Interactive 3D Graphics and Games, 2017. 3

    2017

  11. [18]

    MEASURING HUMAN-CLIP ALIGNMENT AT DIFFERENT ABSTRACTION LEVELS

    Pablo Hernandez-Camara and Jorge Vila-Tomas. MEASURING HUMAN-CLIP ALIGNMENT AT DIFFERENT ABSTRACTION LEVELS. 2024. 2

    2024

  12. [19]

    PointGMM: a Neural GMM Network for Point Clouds

    Amir Hertz, Rana Hanocka, Raja Giryes, and Daniel Cohen-Or. PointGMM: a Neural GMM Network for Point Clouds. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR), pages 12054–12063, 2020. 2

    2020

  13. [20]

    SPAGHETTI: Edit- ing Implicit Shapes Through Part Aware Generation

    Amir Hertz, Or Perel, Raja Giryes, Olga Sorkine- Hornung, and Daniel Cohen-Or. SPAGHETTI: Edit- ing Implicit Shapes Through Part Aware Generation. ACM Transactions on Graphics (TOG), 41(4):1–20,

  14. [21]

    New insights in smooth occluding contours for nonphotorealistic rendering.IEEE Com- puter Graphics and Applications, 44(1):76–85, 2024

    Aaron Hertzmann. New insights in smooth occluding contours for nonphotorealistic rendering.IEEE Com- puter Graphics and Applications, 44(1):76–85, 2024. 1, 2

    2024

  15. [22]

    Multi-view Wire Art.ACM Transactions on Graphics (TOG), 37(6):1–11, 2018

    Kai-Wen Hsiao, Jia-Bin Huang, and Hung-Kuo Chu. Multi-view Wire Art.ACM Transactions on Graphics (TOG), 37(6):1–11, 2018. 3

    2018

  16. [23]

    Mitra, and Adrien Bousseau

    Felix H ¨ahnlein, Changjian Li, Niloy J. Mitra, and Adrien Bousseau. CAD2Sketch: Generating Concept Sketches from CAD Sequences.ACM Transactions on Graphics, 41(6):1–18, 2022. 3

    2022

  17. [24]

    SHRED: 3D Shape Region Decomposition with Learned Local Operations.ACM Transactions on Graphics (TOG), 41(6):1–11, 2022

    R Kenny Jones, Aalia Habib, and Daniel Ritchie. SHRED: 3D Shape Region Decomposition with Learned Local Operations.ACM Transactions on Graphics (TOG), 41(6):1–11, 2022. 2

    2022

  18. [25]

    What does the occluding contour tell us about solid shape?Perception, 13(3):321–330,

    Jan J Koenderink. What does the occluding contour tell us about solid shape?Perception, 13(3):321–330,

  19. [26]

    3D Neu- ral Edge Reconstruction

    Lei Li, Songyou Peng, Zehao Yu, Shaohui Liu, Remi Pautrat, Xiaochuan Yin, and Marc Pollefeys. 3D Neu- ral Edge Reconstruction. In2024 IEEE/CVF Con- ference on Computer Vision and Pattern Recogni- tion (CVPR), pages 21219–21229, Seattle, W A, USA,

  20. [27]

    Photo-sketching: Inferring con- tour drawings from images.WACV, 2019

    Mengtian Li, Zhe Lin, Radom´ır Mˇ ech, , Ersin Yumer, and Deva Ramanan. Photo-sketching: Inferring con- tour drawings from images.WACV, 2019. 3

    2019

  21. [28]

    Differentiable vector graph- ics rasterization for editing and learning.ACM Trans

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

  22. [29]

    ConTesse: Accurate Occluding Contours for Subdivision Surfaces.ACM Transactions on Graphics, 42(1):1–16, 2023

    Chenxi Liu, Pierre B ´enard, Aaron Hertzmann, and Shayan Hoshyari. ConTesse: Accurate Occluding Contours for Subdivision Surfaces.ACM Transactions on Graphics, 42(1):1–16, 2023. 1

    2023

  23. [30]

    Neural Contours: Learning to Draw Lines from 3D Shapes

    Difan Liu, Mohamed Nabail, Aaron Hertzmann, and Evangelos Kalogerakis. Neural Contours: Learning to Draw Lines from 3D Shapes. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 5428–5436, 2020. 2

    2020

  24. [31]

    Neural Strokes: Stylized Line Drawing of 3d Shapes

    Difan Liu, Matthew Fisher, Aaron Hertzmann, and Evangelos Kalogerakis. Neural Strokes: Stylized Line Drawing of 3d Shapes. InProceedings of the IEEE/CVF International Conference on Computer Vi- sion (CVPR), pages 14204–14213, 2021. 1, 2

    2021

  25. [32]

    Lingjie Liu, Duygu Ceylan, Cheng Lin, Wenping Wang, and Niloy J. Mitra. Image-based reconstruc- tion of wire art.ACM SIGGRAPH 2017, 2017. 3

    2017

  26. [33]

    Lingjie Liu, Nenglun Chen, Duygu Ceylan, Christian Theobalt, Wenping Wang, and Niloy J. Mitra. Curve- fusion: Reconstructing thin structures from rgbd se- quences.ACM Trans. Graph., 37(6):218:1–218:12,

  27. [34]

    Marching-Primitives: Shape Abstraction from Signed Distance Function

    Weixiao Liu, Yuwei Wu, Sipu Ruan, and Gregory S Chirikjian. Marching-Primitives: Shape Abstraction from Signed Distance Function. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 8771–8780, 2023. 2

    2023

  28. [35]

    Least squares quantization in pcm.IEEE transactions on information theory, 28(2):129–137,

    Stuart Lloyd. Least squares quantization in pcm.IEEE transactions on information theory, 28(2):129–137,

  29. [36]

    Real-World Textured Things: A repository of textured models generated with modern photo-reconstruction tools.Computer Aided Geometric Design, 83:101943, 2020

    Andrea Maggiordomo, Federico Ponchio, Paolo Cignoni, and Marco Tarini. Real-World Textured Things: A repository of textured models generated with modern photo-reconstruction tools.Computer Aided Geometric Design, 83:101943, 2020. 6

    2020

  30. [37]

    Analysis of Occluding Contour.Proceed- ings of the Royal Society of London

    David Marr. Analysis of Occluding Contour.Proceed- ings of the Royal Society of London. Series B. Biolog- ical Sciences, 197(1129):441–475, 1977. 2

    1977

  31. [38]

    Abstraction of man-made shapes

    Ravish Mehra, Qingnan Zhou, Jeremy Long, Alla Sheffer, Amy Gooch, and Niloy J Mitra. Abstraction of man-made shapes. InACM SIGGRAPH Asia 2009 papers, pages 1–10. ACM New York, NY , USA, 2009. 3

    2009

  32. [39]

    Shadow Art.ACM Transactions on Graphics, 28(5):156–1, 2009

    Niloy J Mitra and Mark Pauly. Shadow Art.ACM Transactions on Graphics, 28(5):156–1, 2009. 3

    2009

  33. [40]

    Axis-Aligned Height-Field Block Decomposition of 3D Shapes

    Alessandro Muntoni, Marco Livesu, Riccardo Scateni, Alla Sheffer, and Daniele Panozzo. Axis-Aligned Height-Field Block Decomposition of 3D Shapes. ACM Transactions on Graphics (TOG), 37(5):1–15,

  34. [41]

    Aria dig- ital twin: A new benchmark dataset for egocentric 3d machine perception, 2023

    Xiaqing Pan, Nicholas Charron, Yongqian Yang, Scott Peters, Thomas Whelan, Chen Kong, Omkar Parkhi, Richard Newcombe, and Carl Yuheng Ren. Aria dig- ital twin: A new benchmark dataset for egocentric 3d machine perception, 2023. 5, 4

    2023

  35. [42]

    Juxtaform: interactive visual summarization for ex- ploratory shape design.ACM Trans

    Karran Pandey, Fanny Chevalier, and Karan Singh. Juxtaform: interactive visual summarization for ex- ploratory shape design.ACM Trans. Graph., 42(4): 52:1–52:14, 2023. 3

    2023

  36. [43]

    SketchGen: Generating Constrained CAD Sketches

    Wamiq Para, Shariq Bhat, Paul Guerrero, Tom Kelly, Niloy Mitra, Leonidas J Guibas, and Peter Wonka. SketchGen: Generating Constrained CAD Sketches. InAdvances in Neural Information Processing Sys- tems, pages 5077–5088. Curran Associates, Inc.,

  37. [44]

    Neural Parts: Learn- ing Expressive 3D Shape Abstractions with Invertible Neural Networks

    Despoina Paschalidou, Angelos Katharopoulos, An- dreas Geiger, and Sanja Fidler. Neural Parts: Learn- ing Expressive 3D Shape Abstractions with Invertible Neural Networks. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR), pages 3204–...

    2021

  38. [45]

    GeoCode: Interpretable Shape Programs.arXiv preprint arXiv:2212.11715, 2022

    Ofek Pearl, Itai Lang, Yuhua Hu, Raymond A Yeh, and Rana Hanocka. GeoCode: Interpretable Shape Programs.arXiv preprint arXiv:2212.11715, 2022. 2

    2022 arXiv

  39. [46]

    SketchLattice: Lat- ticed Representation for Sketch Manipulation

    Yonggang Qi, Guoyao Su, Pinaki Nath Chowdhury, Mingkang Li, and Yi-Zhe Song. SketchLattice: Lat- ticed Representation for Sketch Manipulation. In2021 IEEE/CVF International Conference on Computer Vi- sion (ICCV), pages 933–941, Montreal, QC, Canada,

  40. [47]

    Wired Perspectives: Multi-View Wire Art Embraces Generative AI

    Zhiyu Qu, Lan Yang, Honggang Zhang, Tao Xiang, Kaiyue Pang, and Yi-Zhe Song. Wired Perspectives: Multi-View Wire Art Embraces Generative AI. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 6149– 6158, 2024. 3

    2024

  41. [48]

    Learning Transferable Visual Models from Nat- ural Language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sas- try, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning Transferable Visual Models from Nat- ural Language supervision. InInternational Confer- ence on Machine Learnin...

    2021

  42. [49]

    Viewpoint from Occluding Contour.CVGIP: Image Understanding, 55(2):198–211, 1992

    W Brent Seales and Charles R Dyer. Viewpoint from Occluding Contour.CVGIP: Image Understanding, 55(2):198–211, 1992. 2

    1992

  43. [50]

    Tianjia Shao, Wilmot Li, Kun Zhou, Weiwei Xu, Baining Guo, and Niloy J. Mitra. Interpreting con- cept sketches.ACM Transactions on Graphics, 32(4),

  44. [52]

    Hospedales

    Jifei Song, Kaiyue Pang, Yi-Zhe Song, Tao Xiang, and Timothy M. Hospedales. Learning to Sketch with Shortcut Cycle Consistency. In2018 IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 801–810, Salt Lake City, UT, 2018. IEEE. 3

    2018

  45. [53]

    As-rigid-as-possible surface modeling

    Olga Sorkine and Marc Alexa. As-rigid-as-possible surface modeling. InProceedings of the fifth Euro- graphics symposium on Geometry processing, pages 109–116, Goslar, DEU, 2007. Eurographics Associa- tion. 8, 6

    2007

  46. [54]

    The Janus Effect of Genera- tive AI: Charting the Path for Responsible Conduct of Scholarly Activities in Information Systems.Infor- mation Systems Research, 34(2):399–408, 2023

    Anjana Susarla, Ram Gopal, Jason Bennett Thatcher, and Suprateek Sarker. The Janus Effect of Genera- tive AI: Charting the Path for Responsible Conduct of Scholarly Activities in Information Systems.Infor- mation Systems Research, 34(2):399–408, 2023. Pub- lisher: INFORMS. 1

    2023

  47. [55]

    Sphere-Meshes: shape approximation using spherical quadric error metrics.ACM Transactions on Graphics, 32(6):1–12, 2013

    Jean-Marc Thiery, ´Emilie Guy, and Tamy Boubekeur. Sphere-Meshes: shape approximation using spherical quadric error metrics.ACM Transactions on Graphics, 32(6):1–12, 2013. 2

    2013

  48. [56]

    Modern Evolution Strate- gies for Creativity: Fitting Concrete Images and Ab- stract Concepts

    Yingtao Tian and David Ha. Modern Evolution Strate- gies for Creativity: Fitting Concrete Images and Ab- stract Concepts. InArtificial Intelligence in Music, Sound, Art and Design: 11th International Confer- ence, EvoMUSART 2022, Held as Part of EvoStar 2022, Madrid, Spain, Ap...

    2022

  49. [57]

    Fabricable 3D Wire Art

    Kenji Tojo, Ariel Shamir, Bernd Bickel, and Nobuyuki Umetani. Fabricable 3D Wire Art. InSpecial Interest Group on Computer Graphics and Interactive Tech- niques Conference Conference Papers ’24, pages 1– 11, Denver CO USA, 2024. ACM. 3

    2024

  50. [58]

    Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs, 2024

    Shengbang Tong, Zhuang Liu, Yuexiang Zhai, Yi Ma, Yann LeCun, and Saining Xie. Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs, 2024. arXiv:2401.06209 [cs]. 6

    2024 arXiv

  51. [59]

    Learning Shape Abstractions by Assembling V olumetric Primitives

    Shubham Tulsiani, Hao Su, Leonidas J Guibas, Alexei A Efros, and Jitendra Malik. Learning Shape Abstractions by Assembling V olumetric Primitives. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2635–2643,

  52. [60]

    Prior knowledge for part correspondence

    Oliver van Kaick, Andrea Tagliasacchi, Oana Sidi, Hao Zhang, Daniel Cohen-Or, Lior Wolf, and Ghassan Hamarneh. Prior knowledge for part correspondence. Computer Graphics Forum, 30(2):553–562, 2011. 5

    2011

  53. [61]

    CLIPasso: Semantically-Aware Object Sketching

    Yael Vinker, Ehsan Pajouheshgar, Jessica Y Bo, Roman Christian Bachmann, Amit Haim Bermano, Daniel Cohen-Or, Amir Zamir, and Ariel Shamir. CLIPasso: Semantically-Aware Object Sketching. ACM Transactions on Graphics (TOG), 41(4):1–11,

  54. [62]

    CLIPascene: Scene Sketching with Different Types and Levels of Abstraction

    Yael Vinker, Yuval Alaluf, Daniel Cohen-Or, and Ariel Shamir. CLIPascene: Scene Sketching with Different Types and Levels of Abstraction. In2023 IEEE/CVF International Conference on Computer Vi- sion (ICCV), pages 4123–4133, Paris, France, 2023. IEEE. 2, 3

    2023

  55. [63]

    Exploring clip for assessing the look and feel of im- ages

    Jianyi Wang, Kelvin CK Chan, and Chen Change Loy. Exploring clip for assessing the look and feel of im- ages. InAAAI, 2023. 6

    2023

  56. [64]

    Vid2Curve: simultaneous camera motion estimation and thin structure reconstruction from an RGB video

    Peng Wang, Lingjie Liu, Nenglun Chen, Hung- Kuo Chu, Christian Theobalt, and Wenping Wang. Vid2Curve: simultaneous camera motion estimation and thin structure reconstruction from an RGB video. ACM Trans. Graph., 39(4):132:132:1–132:132:12,

  57. [65]

    Diffusion Feedback Helps CLIP See Better, 2024

    Wenxuan Wang, Quan Sun, Fan Zhang, Yepeng Tang, Jing Liu, and Xinlong Wang. Diffusion Feedback Helps CLIP See Better, 2024. arXiv:2407.20171 [cs]. 6

    2024 arXiv

  58. [66]

    Back to 3d: Few-shot 3d keypoint detection with back-projected 2d features

    Thomas Wimmer, Peter Wonka, and Maks Ovs- janikov. Back to 3d: Few-shot 3d keypoint detection with back-projected 2d features. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, 2024. 5, 1

    2024

  59. [67]

    Primitive-based Shape Abstraction via Nonparametric Bayesian Inference

    Yuwei Wu, Weixiao Liu, Sipu Ruan, and Gregory S Chirikjian. Primitive-based Shape Abstraction via Nonparametric Bayesian Inference. InEuropean Con- ference on Computer Vision (ECCV), pages 479–495. Springer, 2022. 2

    2022

  60. [68]

    WireRoom: model-guided explorative design of ab- stract wire art.ACM Transactions on Graphics, 40(4): 1–13, 2021

    Zhijin Yang, Pengfei Xu, Hongbo Fu, and Hui Huang. WireRoom: model-guided explorative design of ab- stract wire art.ACM Transactions on Graphics, 40(4): 1–13, 2021. 3

    2021

  61. [69]

    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. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 8486–8495, 2023. 2, 6

    2023

  62. [70]

    The unreasonable ef- fectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable ef- fectiveness of deep features as a perceptual metric. In CVPR, 2018. 4, 6

    2018

  63. [71]

    flattening

    Yibo Zhang, Lihong Wang, Changqing Zou, Tieru Wu, and Rui Ma. Diff3DS: Generating View- Consistent 3D Sketch via Differentiable Curve Ren- dering.arXiv preprint arXiv:2405.15305, 2024. 3 WIR3D: Visually-Informed and Geometry-Aware 3D Shape Abstraction Supplementary Material A....

    2024 arXiv

  64. [74]

    Sketch1/Sketch2

    Transformations to each curve can then be automatically mapped to the surface through these skinning weights, and Figure 21.Additional textured results.We show additional tex- tured results from the Meta DTC dataset [41]. the procedure can be performed at interactive speeds. W...

  65. [1986]

    Conference Name: IEEE Transactions on Pat- tern Analysis and Machine Intelligence. 6, 2

  66. [2023]

    Conference Name: IEEE Robotics and Automa- tion Letters. 3

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.