Pith. sign in

REVIEW 4 major objections 7 minor 45 references

3D Reconstruction of Deformable Revolving Object under Heavy Hand Interaction

T0 review · 4 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read The paper claims that an unknown, deformable pottery object can be reconstructed in 3D from one or more depth sensors by accumulating radial density around the detected turntable axis, extracting a Catmull-Rom profile with a particle…

desk verdict A competent pipeline for a genuinely new problem, but the headline accuracy numbers rest on a ground-truth protocol that shares the method's own smooth radial prior. read the letter →

arxiv 1908.01523 v1 pith:ZLDWO6YQ submitted 2019-08-05 cs.CV cs.HC

classification cs.CVcs.HC
keywords 3DreconstructiondeformableobjectsparticlefilterCatmull-Romradialsymmetrydepthsensorspotteryocclusion
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tackles a problem that existing 3D reconstruction methods do not address: recovering the changing shape of an unknown, deformable object while a person's hands are occluding and remolding it, as in live pottery making. The authors claim that by detecting the turntable and its axis in depth point clouds, registering one or more sensors, and accumulating points into radial bins around the axis, they can isolate the object profile and track it over time with a particle filter. On a new 6030-frame pottery dataset they report an average symmetric reconstruction error of 7.60 mm and a Hausdorff error of 19.84 mm, about half the error of a B-spline baseline. If correct, this means a craft like wheel-throwing can be digitized without pausing the wheel or asking the potter to change how they work.

What carries the argument

The load-bearing mechanism is the radial accumulator coupled with a particle filter. The accumulator clusters the registered point cloud in polar coordinates into 3D annulii around the turntable axis, encoding the density of surface points in each radius-height cell and weighting that density by the angular spread of the points so that radially inconsistent hand points count less. A bootstrap particle filter then scores 5-knot Catmull-Rom profiles against this accumulator, averaging the top ten Gaussian-mixture probabilities along each sampled curve and normalizing by curve length; systematic resampling carries promising profiles through time. The result is a temporal profile estimate whose 3D mesh is generated by revolving the profile around the axis.

What would settle it

Run the pipeline on a synthetic sequence built from a known radially symmetric ground-truth mesh with a hand-shaped occluder whose size is controlled; if the symmetric average error on heavily occluded frames climbs well above the reported 7.60 mm, or if deliberately pinching the object into a non-radially-symmetric shape does not raise the error, the claimed mechanism is not what is driving the result.

Watch

Extended reading notes

Core claim

The central claim is that the 3D shape of a radially symmetric deforming object is fully determined, at each instant, by a 2D profile revolved around the axis shared with the turntable, and that this profile can be recovered from noisy, partially occluded depth data by radial accumulation plus temporal filtering. The paper builds a radial accumulator: the registered point cloud is binned into cylindrical annuli, each cell's density is weighted by the circular spread of its points so that non-radial clutter like hands is suppressed, and the profile is found as the 5-knot Catmull-Rom curve that best explains the accumulator under a bootstrap particle filter. The best-scoring curve is revolved around the turntable axis to produce the mesh. On the released dataset, the average symmetric profile error is 7.60 mm and the maximum symmetric Hausdorff error is 19.84 mm, both roughly half of the closest B-spline baseline, and two-sensor input improves the average error by 17.2% over the best single sensor.

Load-bearing premise

The method assumes the clay object stays radially symmetric around the turntable axis at all times, so that a single 2D profile rotated about the axis reproduces the full surface; pinching or cupping by the potter, or a poorly recovered axis, would bias the radial accumulator and the estimated profile.

Editorial extensions

If this is right

  • A live pottery session can be turned into a per-frame 3D model using two small depth sensors, without stopping the wheel.
  • With a single sensor the method still works, and adding a second sensor improves mean reconstruction error by 17.2% over the better single sensor.
  • Because the profile is tracked through time with particle resampling, the model follows unpredictable shape changes and recovers from tracking failures when hands briefly cover the object.
  • The average error of 7.60 mm comes out below the 10 mm radial cell of the default accumulator, which the paper attributes to continuous Gaussian-mixture scoring over discrete bins.

Reading between the lines

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

  • One extension the paper leaves implicit is that the same axis-symmetric accumulation idea could transfer to other revolving-workpiece processes such as woodturning or metal spinning, as long as the rotation axis can be detected or assumed.
  • A natural extension the paper does not test is allowing a low-frequency angular perturbation of the profile; measuring how much error a known asymmetric dent introduces would quantify how much the radial-symmetry assumption costs.
  • The method's dependence on a good turntable-axis estimate suggests a testable engineering trade-off: improving axis localization, for instance by calibrating the turntable once, might buy more accuracy than adding a third depth sensor.
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 / 7 minor

Summary. The manuscript presents a pipeline for reconstructing a deformable, radially symmetric object (a clay pot on a turntable) from one or more depth sensors while a potter's hands partially occlude and deform it. The proposed method detects the turntable plane and center in each view, registers the views using the turntable model, builds a radial accumulator in polar coordinates, weights the accumulator by radial spread to suppress non-radial artifacts such as hands, and fits a Catmull-Rom C5 profile with a bootstrap particle filter whose likelihood is the mean of the top ten weighted Gaussian densities of the accumulator. The recovered profile is revolved about the axis to produce a 3D mesh. Experiments on three recorded sequences (6030 frames total) report a mean symmetric average error of 7.60 mm and a Hausdorff error of 19.84 mm, roughly half the error of a B-spline baseline [41], together with ablations over sensor count, particle count, and accumulator resolution.

Significance. If the reported accuracy holds, the paper addresses a genuinely difficult and under-studied setting: unknown, deformable objects under strong occlusion and interaction. The algorithmic choices are clearly specified and the pipeline is a plausible practical contribution, particularly the use of a radial-symmetry prior combined with a particle filter over Catmull-Rom profiles, and a scalable multi-sensor turntable-based registration. The paper also provides a complete algorithm in Appendix A, ablation studies, and a commitment to release the dataset upon publication. However, the headline accuracy numbers rest on a ground-truth protocol that shares the method's symmetry and smoothness priors, which as presented prevents the central quantitative claim from being verified.

major comments (4)
  1. [Section 4, 'Data labeling' and 'Metrics for evaluation'] The ground-truth profiles are produced by operators labeling the object profile in the same manually registered point clouds used by the method, and the labels are then fitted to a C5 Catmull-Rom spline. This is the same curve family and smoothness prior that the pipeline outputs, and the metric compares 2D profiles in polar space rather than an independent 3D mesh. In frames with heavy hand occlusion—the regime the paper targets—much of the object surface is unobserved, so the natural interpolation made by the labelers is exactly the smooth, radially symmetric profile the pipeline enforces. The reported average error of 7.60 mm and Hausdorff error of 19.84 mm may therefore measure agreement between two curves sharing the same prior rather than fidelity to the true deformed surface. Please provide an independent evaluation, for example synthetic point clouds with known deformation fields, or a laser-scanned or structured-light mesh of the final object, or at minimum release the raw data and annotations during review so the labeling protocol can be audited.
  2. [Section 3.3.2 and Fig. 7] The model assumes that the object remains radially symmetric about the turntable axis at all times, but the authors acknowledge near Fig. 7 that pinching the object or fully cupping the clay 'affects the radial estimation.' Those are precisely the heavy hand interactions named in the title and contribution, and the dataset appears to include such frames. As written, the claimed robustness to heavy hand interaction is not supported for asymmetric deformations. Please quantify performance separately for frames with symmetric versus asymmetric interaction, or validate the method on synthetic sequences with controlled asymmetric deformations, and state clearly which interaction regimes the method is and is not claimed to handle.
  3. [Section 4.1, Table 1] The text states that the proposed method 'is significantly better' than [41], but no significance test, confidence interval, or paired comparison is reported. Given the large per-frame standard deviations (e.g., 8.64 mm against a mean of 7.60 mm for the temporal variant), a statistical test is needed to support the claim. Furthermore, several hyperparameters—particle count, resampling ratio (0.8), motion variance σm, and accumulator resolution—appear to be selected on the evaluation dataset without a separate validation split. Please report a defined validation protocol and provide confidence intervals or significance tests for the main accuracy comparisons.
  4. [Section 4.1, comparison with [41]] The comparison to [41] is presented as a state-of-the-art baseline, but [41] is a B-spline fitting method that the authors themselves describe as assuming noise-free data and no occlusion. The authors also selected its parameters with a grid search. This makes the comparison favorable to the proposed method but not necessarily informative about state-of-the-art performance under occlusion and deformation. Please either add a baseline that is designed for occluded or deformable point-cloud fitting, or explicitly frame [41] as a standard fitting upper bound rather than a deformable reconstruction baseline.
minor comments (7)
  1. [Section 3.3.1] In the definition of polar coordinates, 'ρ ∈ N' should presumably be 'ρ ∈ ℝ' (or 'ρ ≥ 0'), since distances are continuous real values rather than natural numbers.
  2. [Section 3.3.2, Eq. (6)] The quantity P(x|Γ) is described as a probability, but it is an unnormalized average of ten Gaussian density values; please either normalize it or refer to it as a likelihood score rather than a probability.
  3. [Section 4, 'Data labeling'] The 'greedy spline fitting algorithm' used to convert manual label points into C5 Catmull-Rom curves is not described; please specify the fitting procedure for reproducibility.
  4. [Section 4.1] The phrase 'cherry picking their parameters' is informal for a journal article; please replace it with a neutral description such as 'tuning via grid search'.
  5. [Fig. 9 and Fig. 10] The text refers to 'tab. 9(a)' and 'tab. 10(b)' for results that are in figures; please fix these cross-references.
  6. [Algorithm 1, Appendix A] Line 7 uses a threshold σc that is never defined, and the control flow for 'Next data frame' inside the detection loop is unclear; please clarify the termination and update logic.
  7. [Section 4.1] The statement that an average error smaller than the accumulator cell size 'proves the benefit of the continuous Gaussian Mixture representation' is too strong, since sub-cell error can also arise from interpolation between cells. Please soften the claim or provide a direct comparison against a non-Gaussian scoring function.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the pipeline converts raw point clouds to a profile via independent steps (turntable detection, registration, radial accumulation, particle filtering), and the only self-citation is a non-load-bearing sensor-placement choice.

full rationale

I walked the claimed derivation chain from raw depth sensors to the final 3D mesh. The method detects the turntable (Section 3.1), registers point clouds from the turntable model (Section 3.2), builds a radial accumulator from the registered point cloud (Eqs. 1–3), and extracts a Catmull-Rom profile via a particle filter whose score is a Gaussian-mixture likelihood over the accumulator (Eqs. 6–7). The output profile is then revolved around the axis to form the mesh. No equation in the paper defines the output profile in terms of the error metric, and no fitted parameter is renamed as a prediction: the particle filter is initialized randomly and evaluated against the accumulator, not against the manual labels or the error metric. The ground-truth protocol does rely on manual labels in the same point clouds, later fitted to C5 Catmull-Rom curves, which shares a representation family with the method's output; this is a legitimate concern about evaluation independence, especially for asymmetric deformations, but it is not a construction-level equivalence between input and output. The one self-citation, [22], is used only for choosing sensor mounting height at shoulder level, which is a recording setup choice and does not carry the paper's mathematical or experimental conclusions. Accordingly, the central reconstruction claim has independent content, and the paper does not exhibit circular derivation.

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

The central claim rests on the radial-symmetry domain assumption and on accurate turntable-axis recovery; the evaluation relies on manual C5 Catmull-Rom ground truth and multiple hand-set hyperparameters. No new physical entity is introduced.

free parameters (5)
  • sigma_m = 2mm
    Standard deviation of the 2D Gaussian used to perturb particle knots between frames (Eq. 5); set 'Experimentally' in Section 3.3.2.
  • particle resampling ratio = 0.8
    Fraction of particles replaced with random samples each frame; chosen in Section 4.1.
  • number of particles = 1000
    Selected as accuracy/speed trade-off in Section 4.1; 5000 particles improve error by only 4.60%.
  • accumulator resolution = 16x16 (10mm cells)
    Default in Tab. 1; ablation shows 64x64 reduces mean error from 9.39mm to 8.43mm at 5.35x cost.
  • Gaussian variance in mixture scoring = accumulator cell size
    Set to cell size to keep scoring invariant to accumulator resolution (Section 3.3.2).
assumptions (5)
  • domain assumption The turntable is a rigid flat disk of known radius r and is visible enough in each depth view for mSAC and weighted MeanShift to recover its plane and center.
    Section 3.1 takes r as input and estimates normal/center; inaccurate axis estimation would corrupt registration and the radial accumulator.
  • domain assumption The pottery object is radially symmetric around the turntable axis at all times, so a 2D profile revolved around the axis describes its 3D surface.
    Section 3.3 builds the accumulator and profile model on this property; the paper acknowledges pinching and cupping break it.
  • domain assumption Manual annotations, fitted to C5 Catmull-Rom curves, provide an unbiased ground-truth profile.
    Section 4, Data labeling: operators labeled profiles in manually registered point clouds; spline fitting smooths the annotation.
  • domain assumption Depth sensors are radially distributed and provide sufficient angular coverage for the radial-spread weighting to be meaningful.
    Section 3 setup uses radial sensors; the method is tested with one or two sensors, and angular coverage limits the radial spread estimate.
  • domain assumption The 2006 B-spline fitting method of Wang et al. [41] is the appropriate state-of-the-art baseline for this task.
    Section 4.1 uses [41] as baseline with grid-searched parameters, but the method assumes noise-free, occlusion-free input.

how reviews work

0 comments
Cite this review

Pith. "Pith review of 3D Reconstruction of Deformable Revolving Object under Heavy Hand Interaction." pith.science (2026). https://pith.science/paper/ZLDWO6YQ

@misc{pith2026190801523,
  author       = {Pith},
  title        = {Pith review of: 3D Reconstruction of Deformable Revolving Object under Heavy Hand Interaction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZLDWO6YQ}},
  note         = {Machine review of arXiv:1908.01523}
}
read the original abstract

We reconstruct 3D deformable object through time, in the context of a live pottery making process where the crafter molds the object. Because the object suffers from heavy hand interaction, and is being deformed, classical techniques cannot be applied. We use particle energy optimization to estimate the object profile and benefit of the object radial symmetry to increase the robustness of the reconstruction to both occlusion and noise. Our method works with an unconstrained scalable setup with one or more depth sensors. We evaluate on our database (released upon publication) on a per-frame and temporal basis and shows it significantly outperforms state-of-the-art achieving 7.60mm average object reconstruction error. Further ablation studies demonstrate the effectiveness of our method.

Figures

Figures reproduced from arXiv: 1908.01523 by the authors.

Figure 1
Figure 1. (a) Reconstruction of unknown 3D objects in the context of wheel [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Estimation of the turntable normal 2(a) and location 2(b). 2(a) is the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Radial accumulator 64×64 (normalized for display), for chronolog￾ically ordered frames. Top shows original accumulator (eq. 1) and bottom shows radially enhanced accumulator (eq. 3). Non-radial artifacts (red circle) such as the potter’s hands are less visible in the enhanced version. Pictures are better seen on a screen. For all, r = 160mm and ∆h = ∆ρ = r 64 . where |.| is the cardinality of A, the set of points in… view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: Radial accumulator represented as weighted Gaussian Mixture. [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 8
Figure 8. Figure 8: Frame-wise errors (10-runs average) on a sample full bowl making [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]
Figure 7
Figure 7. Figure 7: Sample reconstruction with our method (top) and the corresponding [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 9
Figure 9. Figure 9: Errors (mm) while varying sensor setups 9(a) or number of particles [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 10
Figure 10. Figure 10: Study of the accumulator sizes. (a) Sample reconstruction show [PITH_FULL_IMAGE:figures/full_fig_p007_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages

  1. [41]

    Fitting B-spline curves to point clouds by curvature-based squared distance minimization

    Wang, W., Pottmann, H., Liu, Y ., 2006. Fitting B-spline curves to point clouds by curvature-based squared distance minimization. ACM Trans- actions on Graphics

  2. [1]

    As-rigid-as-possible shape in- terpolation, in: Conference on Computer Graphics and Interactive Tech- niques, pp

    Alexa, M., Cohen-Or, D., Levin, D., 2000. As-rigid-as-possible shape in- terpolation, in: Conference on Computer Graphics and Interactive Tech- niques, pp. 157–164

  3. [2]

    Shape-from-template

    Bartoli, A., G ´erard, Y ., Chadebecq, F., Collins, T., Pizarro, D., 2015. Shape-from-template. IEEE Transactions on Pattern Analysis and Ma- chine Intelligence 37, 2099–2118

  4. [3]

    Bootstrap particle filtering

    Candy, J., 2007. Bootstrap particle filtering. IEEE Signal Processing Magazine

  5. [4]

    A class of local interpolating splines

    Catmull, E., Rom, R., 1974. A class of local interpolating splines. Com- puter aided geometric design

  6. [5]

    Robust reconstruction of indoor scenes, in: IEEE Conference on Computer Vision and Pattern Recogni- tion, pp

    Choi, S., Zhou, Q.Y ., Koltun, V ., 2015. Robust reconstruction of indoor scenes, in: IEEE Conference on Computer Vision and Pattern Recogni- tion, pp. 5556–5565

  7. [6]

    Locally optimized RANSAC

    Chum, O., Matas, J., Kittler, J., 2003. Locally optimized RANSAC. Pat- tern Recognition

  8. [7]

    Mean shift: a robust approach toward fea- ture space analysis

    Comaniciu, D., Meer, P., 2002. Mean shift: a robust approach toward fea- ture space analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence 24, 603–619

Show all 45 references
  1. [8]

    Real- time markerless tracking for augmented reality: the virtual visual servo- ing framework

    Comport, A.I., Marchand, E., Pressigout, M., Chaumette, F., 2006. Real- time markerless tracking for augmented reality: the virtual visual servo- ing framework. IEEE Transactions on visualization and computer graph- ics 12, 615–628

  2. [9]

    Robust 3d tracking with descriptor fields, in: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Crivellaro, A., Lepetit, V ., 2014. Robust 3d tracking with descriptor fields, in: IEEE Conference on Computer Vision and Pattern Recognition, pp. 3414–3421

  3. [10]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and au- tomated cartography

    Fischler, M.A., Bolles, R.C., 1981. Random sample consensus: a paradigm for model fitting with applications to image analysis and au- tomated cartography. Communications of the ACM 24, 381–395

  4. [11]

    Comparing and combining lighting in- sensitive approaches for face recognition

    Gopalan, R., Jacobs, D., 2010. Comparing and combining lighting in- sensitive approaches for face recognition. Computer Vision and Image Understanding 114, 135–145

  5. [12]

    Hilsmann, A., Eisert, P., 2008. Tracking deformable surfaces with opti- cal flow in the presence of self occlusion in monocular image sequences, in: IEEE Conference on Computer Vision and Pattern Recognition Work- shops, pp. 1–6

  6. [13]

    Deformable template models: A review

    Jain, A.K., Zhong, Y ., Dubuisson-Jolly, M.P., 1998. Deformable template models: A review. Signal processing 71, 109–129

  7. [14]

    Topics in circular statistics

    Jammalamadaka, S.R., Sengupta, A., 2001. Topics in circular statistics. volume 5. world scientific

  8. [15]

    Potential of automatic 3d object recon- struction from multiple images for applications in architecture, cultural heritage and archaeology

    Kersten, T., Lindstaedt, M., 2012. Potential of automatic 3d object recon- struction from multiple images for applications in architecture, cultural heritage and archaeology. International Journal of Heritage in the Digital Era 1, 399–420

  9. [16]

    Physically plausible 3d scene tracking: The single actor hypothesis, in: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Kyriazis, N., Argyros, A., 2013. Physically plausible 3d scene tracking: The single actor hypothesis, in: IEEE Conference on Computer Vision and Pattern Recognition, pp. 9–16

  10. [17]

    Scalable 3d tracking of multiple in- teracting objects, in: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Kyriazis, N., Argyros, A., 2014. Scalable 3d tracking of multiple in- teracting objects, in: IEEE Conference on Computer Vision and Pattern Recognition, pp. 3430–3437

  11. [18]

    Fixing the Locally Optimized RANSAC, in: British Machine Vision Conference, pp

    Lebeda, K., Matas, J., Chum, O., 2012. Fixing the Locally Optimized RANSAC, in: British Machine Vision Conference, pp. 95.1–95.11

  12. [19]

    Resampling methods for particle filtering: classification, implementation, and strategies

    Li, T., Bolic, M., Djuric, P.M., 2015. Resampling methods for particle filtering: classification, implementation, and strategies. IEEE Signal Pro- cessing Magazine 32, 70–86

  13. [20]

    Deep marching cubes: Learn- ing explicit surface representations, in: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Liao, Y ., Donn ´e, S., Geiger, A., 2018. Deep marching cubes: Learn- ing explicit surface representations, in: IEEE Conference on Computer Vision and Pattern Recognition, pp. 2916–2925

  14. [21]

    Distinctive image features from scale-invariant key- points

    Lowe, D.G., 2004. Distinctive image features from scale-invariant key- points. International Journal of Computer Vision 60, 91–110

  15. [22]

    Cap- ture, modeling, and recognition of expert technical gestures in wheel- throwing art of pottery

    Manitsaris, S., Glushkova, A., Bevilacqua, F., Moutarde, F., 2014. Cap- ture, modeling, and recognition of expert technical gestures in wheel- throwing art of pottery. Journal Computing and Cultural Heritage 7, 10

  16. [23]

    Fast segmentation, tracking, and analysis of deformable objects, in: IEEE International Conference on Computer Vision, pp

    Nastar, C., Ayache, N., 1993. Fast segmentation, tracking, and analysis of deformable objects, in: IEEE International Conference on Computer Vision, pp. 275–279

  17. [24]

    Kinect- fusion: Real-time dense surface mapping and tracking, in: IEEE interna- tional symposium on Mixed and augmented reality, pp

    Newcombe, R.A., Izadi, S., Hilliges, O., Molyneaux, D., Kim, D., Davi- son, A.J., Kohi, P., Shotton, J., Hodges, S., Fitzgibbon, A., 2011. Kinect- fusion: Real-time dense surface mapping and tracking, in: IEEE interna- tional symposium on Mixed and augmented reality, pp. 127–136

  18. [25]

    Full dof tracking of a hand interacting with an object by modeling occlusions and physical 8 constraints, in: IEEE International Conference on Computer Vision, pp

    Oikonomidis, I., Kyriazis, N., Argyros, A.A., 2011. Full dof tracking of a hand interacting with an object by modeling occlusions and physical 8 constraints, in: IEEE International Conference on Computer Vision, pp. 2088–2095

  19. [26]

    Laplacian meshes for monocular 3d shape recovery, in: European Conference on Computer Vi- sion, Springer

    ¨Ostlund, J., Varol, A., Ngo, D.T., Fua, P., 2012. Laplacian meshes for monocular 3d shape recovery, in: European Conference on Computer Vi- sion, Springer. pp. 412–425

  20. [27]

    Fast and globally optimal single view reconstruction of curved objects, in: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Oswald, M.R., T ¨oppe, E., Cremers, D., 2012. Fast and globally optimal single view reconstruction of curved objects, in: IEEE Conference on Computer Vision and Pattern Recognition, pp. 534–541

  21. [28]

    3d tracking of human hands in interaction with unknown objects., in: British Machine Vision Conference, pp

    Panteleris, P., Kyriazis, N., Argyros, A.A., 2015. 3d tracking of human hands in interaction with unknown objects., in: British Machine Vision Conference, pp. 123–1

  22. [29]

    As-rigid-as- possible volumetric shape-from-template, in: IEEE International Confer- ence on Computer Vision, pp

    Parashar, S., Pizarro, D., Bartoli, A., Collins, T., 2015. As-rigid-as- possible volumetric shape-from-template, in: IEEE International Confer- ence on Computer Vision, pp. 891–899

  23. [30]

    Octnetfusion: Learning depth fusion from data, in: IEEE International Conference on 3D Vision, pp

    Riegler, G., Ulusoy, A.O., Bischof, H., Geiger, A., 2017. Octnetfusion: Learning depth fusion from data, in: IEEE International Conference on 3D Vision, pp. 57–66

  24. [31]

    Reconstructing sharply folding surfaces: A convex formulation, in: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Salzmann, M., Fua, P., 2009. Reconstructing sharply folding surfaces: A convex formulation, in: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1054–1061

  25. [32]

    Tracking deformable objects with point clouds, in: IEEE International Conference on Robotics and Automation, pp

    Schulman, J., Lee, A., Ho, J., Abbeel, P., 2013. Tracking deformable objects with point clouds, in: IEEE International Conference on Robotics and Automation, pp. 1130–1137

  26. [33]

    As-rigid-as-possible surface modeling, in: Symposium on Geometry processing, pp

    Sorkine, O., Alexa, M., 2007. As-rigid-as-possible surface modeling, in: Symposium on Geometry processing, pp. 109–116

  27. [34]

    Modeling surfaces of arbitrary topology with dynamic particles, in: IEEE Computer Vision and Pattern Recognition, pp

    Szeliski, R., Tonnesen, D., Terzopoulos, D., 1993. Modeling surfaces of arbitrary topology with dynamic particles, in: IEEE Computer Vision and Pattern Recognition, pp. 82–87

  28. [35]

    Constraints on deformable models: Recovering 3d shape and nonrigid motion

    Terzopoulos, D., Witkin, A., Kass, M., 1988. Constraints on deformable models: Recovering 3d shape and nonrigid motion. Artificial intelligence 36, 91–123

  29. [36]

    Tien Ngo, D., Park, S., Jorstad, A., Crivellaro, A., Yoo, C.D., Fua, P.,

  30. [37]

    MLESAC: A new robust estimator with application to estimating image geometry

    Torr, P., Zisserman, A., 2000. MLESAC: A new robust estimator with application to estimating image geometry. Computer Vision and Image Understanding

  31. [38]

    Joint 3d tracking of a deformable object in interaction with a hand, in: European Conference on Computer Vision

    Tsoli, A., Argyros, A.A., 2018. Joint 3d tracking of a deformable object in interaction with a hand, in: European Conference on Computer Vision

  32. [39]

    Shape matching: Similarity measures and algo- rithms, in: International Conference on Shape Modeling and Applica- tions, pp

    Veltkamp, R.C., 2001. Shape matching: Similarity measures and algo- rithms, in: International Conference on Shape Modeling and Applica- tions, pp. 188–197

  33. [40]

    Balloon shapes: Reconstructing and deforming objects with volume from images, in: IEEE International conference on 3D Vision

    Vicente, S., Agapito, L., . Balloon shapes: Reconstructing and deforming objects with volume from images, in: IEEE International conference on 3D Vision

  34. [42]

    Lane detection using catmull-rom spline

    Wang, Y ., Shen, D., Teoh, E., 1998. Lane detection using catmull-rom spline. IEEE International Conference on Intelligent Vehicles

  35. [43]

    Parameterization and applica- tions of CatmullRom curves

    Yuksel, C., Schaefer, S., Keyser, J., 2011. Parameterization and applica- tions of CatmullRom curves. Computer-Aided Design

  36. [44]

    Fast B-spline curve fitting by L-BFGS

    Zheng, W., Bo, P., Liu, Y ., Wang, W., 2012. Fast B-spline curve fitting by L-BFGS. Computer Aided Geometric Design . Appendix A. Algorithm Algorithm 1 is the pseudo code of our entire method for mod- eling 3D revolving object, accounting for n input sensors pro- viding multipl...

  37. [2015]

    2273–2281

    Dense image registration and deformable surface reconstruction in presence of occlusions and minimal texture, in: IEEE International Conference on Computer Vision, pp. 2273–2281

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.