Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Revisiting Point Cloud Completion: Are We Ready For The Real-World?

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Adding 0-dimensional persistent-homology priors to point cloud completion networks improves their outputs on real-world scans, and a new dataset, RealPC, shows why synthetic-trained models fail.

desk verdict RealPC is a genuinely useful real-world completion benchmark, but the paper's topological-prior method is unsupported and needs a major rework. read the letter →

arxiv 2411.17580 v4 pith:XB4VB64J submitted 2024-11-26 cs.CV

Stuti Pathak , Prashant Kumar , Dheeraj Baiju , Nicholus Mboga , Gunther Steenackers , Rudi Penne This is my paper · ORCID
classification cs.CV
keywords pointcloudcompletionRealPCdatasetpersistenthomology0-dimensionalpriorsSamplertopologicaldataanalysisrailwayinfrastructurecloudsChamferdistance
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

Point cloud completion models are typically trained and evaluated on clean synthetic objects, and this paper argues that those benchmarks hide what real-world scans look like: noisy, non-uniformly dense, and full of topological structure. To make the case, the authors build RealPC, roughly 40,000 partial-to-complete pairs of industrial railway structures extracted from four scene-level LiDAR datasets, and show that strong completion networks that look good on PCN collapse on RealPC. Using persistent homology, they find that RealPC point clouds have many significant 0- and 1-dimensional topological features, unlike synthetic clouds, and they claim that feeding those features back as priors improves completion: a topology-loss variant of ODGNet reduces Chamfer distance, and BOSHNet, which samples cheap proxy backbones from the complete surface instead of computing homology, reaches the best reported numbers. The paper's central proposal is that topological consistency, not just pointwise closeness, is a missing ingredient for completing real-world shapes.

What carries the argument

The load-bearing mechanism is the 0-dim PH backbone: 0-dimensional persistent homology records, for each connected component born during a distance filtration, the radius at which it merges, and the paper interprets these birth-death pairs as outlining a spatial skeleton of the complete point cloud. The topology loss $L_{topo} = \sum_{i>k}(b_i - d_i)$ minimizes the persistence of those pairs on the sparse decoder seeds, forcing the seeds to coalesce into one connected skeleton (or $k$ skeletons for multi-component inputs). Because computing Vietoris-Rips persistence is expensive, BOSHNet replaces the homology computation with a Homology Sampler that directly draws multiple sparse proxy backbones from the surface of the ground-truth complete cloud, and trains the network to complete those backbones as well as the partial input. The dataset RealPC itself is the other load-bearing piece: HDBSCAN clustering of four scene-level railway point-cloud datasets, followed by manual inspection for complete ground truth and three sparsification/incompletion processes (viewpoint-based farthest-point removal, viewpoint-based probability sampling, and uniform subsampling), is what produces paired real-world training data with measurable 0- and 1-dimensional persistence.

What would settle it

Train TopODGNet with the topology loss replaced by a control loss that merely minimizes the sum of pairwise distances among seed points, without any persistence computation; if the Chamfer-distance improvement over ODGNet matches TopODGNet's, then the 0-dim PH skeleton prior is not the active ingredient, and the authors' topological explanation collapses.

Watch

Extended reading notes

Core claim

The paper's claim, stated on its own terms, is that 0-dimensional persistent homology priors give a completion network a global skeleton of the complete shape, and that enforcing this skeleton during training produces more topologically consistent completions. The evidence is: on RealPC, ODGNet's Chamfer distance goes from 119 to 103 (CD-L1) and from 111 to 80 (CD-L2) when the decoder seeds are regularized by the 0-dim PH persistence loss; and BOSHNet, which bypasses homology by sampling multiple sparse proxy backbones from the ground-truth surface, reaches CD-L1 69 and CD-L2 5.4, improving on all baselines. The authors also report that 0-dim PH priors improve scene completion and reconstruction on KITTI. The paper frames this as proof that existing completion methods fail on real-world data because they ignore the topological richness that real scans contain.

Load-bearing premise

Everything rests on identifying 0-dimensional persistent homology's merge-scale pairs with a spatial skeleton of the complete shape; standard 0-dim PH only records when connected components merge, and the paper assumes this defines a geometric backbone that guides generation.

Editorial extensions

If this is right

  • Benchmarking on RealPC rather than PCN or ShapeNet changes which methods look strong, and several baselines that report low error on synthetic data exceed 100 in CD-L1 on RealPC.
  • Existing seed-based completion models can be upgraded with a 0-dim PH regularization layer without architectural changes beyond accessing decoder seeds, as TopODGNet does.
  • Homology computation can be bypassed: proxy backbones sampled from the complete surface give most of the benefit at far lower cost, and they are available from the first training step rather than emerging late in training.
  • Topological priors transfer beyond objects: KITTI scene completion and reconstruction improve with the same 0-dim PH loss.
  • RealPC supplies paired partial/complete real-world data, enabling supervised training that datasets like ScanNet, Matterport3D, and KITTI cannot support directly.

Reading between the lines

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

  • If the skeleton interpretation is right, then completion models without explicit decoder seeds could be adapted by adding a lightweight auxiliary head that outputs a sparse skeleton and applying the same persistence loss, extending the recipe beyond ODGNet.
  • The BOSH proxy suggests a general training strategy: sample multi-scale skeleton-like subsets from ground truth as auxiliary completion targets, which is testable on any paired dataset and may act as a curriculum where the model first learns coarse structure.
  • The dataset-creation pipeline (clustering, manual ground-truth selection, and multiple corruption processes) could be rerun on other scene-level LiDAR collections, producing topological completion benchmarks for urban, agricultural, or indoor environments without new sensors.
  • A discriminating test: train BOSHNet on PCN or ShapeNet; if its advantage shrinks or vanishes on synthetic shapes, that would confirm topological richness, not the sampling scheme itself, is what drives the RealPC gains.
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 / 6 minor

Summary. The manuscript has two main goals. First, it introduces RealPC, a paired real-world industrial point cloud completion dataset with roughly 40,000 partial/complete pairs across 21 railway-related object categories, built from four open scene-level LiDAR datasets. The paper characterizes RealPC against PCN and ShapeNet using noise, non-uniformity, and 0- and 1-dimensional persistent homology (H0/H1) statistics, and benchmarks several completion, reconstruction, and generation baselines on it. Second, it proposes topological-prior methods: TopODGNet, which adds a 0-dim PH regularizer to ODGNet, and BOSHNet, which replaces homology computation with a 'Homology Sampler' that samples proxy backbones from ground-truth surfaces. The paper claims to prove that integrating these topological priors improves completion performance, citing Table 5 as evidence.

Significance. If the claims were correct, the paper would make a significant contribution: a real-world paired object completion benchmark, a systematic failure analysis of synthetic-trained models on real data, and a principled topological regularizer. The dataset itself is potentially valuable, and the comparative noise/non-uniformity/PH analysis is suggestive and useful. However, the central methodological claim is not currently supported. The topological loss in Eq. (1) is mis-specified relative to its stated objective, the claimed relationship between H0 persistence and a geometric skeleton is not established, BOSHNet's training loss leaks ground-truth samples, and the headline experimental table is internally inconsistent. These are load-bearing issues for the paper's main claim, not presentation problems.

major comments (4)
  1. [Section 5.1.2, Eq. (1)] The topological loss as written, sum over i>k of (b_i - d_i), is inconsistent with the stated goal. In a Vietoris–Rips filtration, every 0-dimensional homology class is born at b_i = 0, so the summand equals -d_i and minimizing the loss maximizes death times, i.e., maximizes persistence. This is the opposite of the text's claim that the loss minimizes persistence so that components merge early. The paper also switches conventions: Section 3 defines persistence as death minus birth, while Section 5.1.2 calls (b-d) the persistence. As a result, the mechanism claimed for TopODGNet is not the mechanism implemented by the printed loss, and the abstract's 'proof' claim is not supported by the actual equation.
  2. [Section 5.1.2 and Figure 7] The paper asserts that '0-dim PH ensures the extraction of a complete PC skeleton' and that minimizing H0 persistence guides point generation along that skeleton. H0 birth-death pairs only record the scales at which connected components merge; they carry no spatial coordinates, no edges, and no branch or backbone structure. The manuscript provides no algorithm or derivation that converts the H0 persistence diagram into the rendered 'skeleton' of Figure 7. At best, an H0 loss encourages the seed cloud to become connected or compact; it does not place generated points on a particular geometric skeleton. The explanatory mechanism for TopODGNet is therefore unsupported, and BOSHNet's sampled subsets are not demonstrated to be proxies for H0 persistence in any formal sense.
  3. [Section 5.2, Eq. (2)] The first term of Eq. (2) is M(Net(BOSH(c_i,j)), c_i), where c_i is the complete ground-truth cloud and BOSH(c_i,j) are sparse subsets sampled from that same complete cloud. This trains the network to reconstruct a complete cloud from ground-truth-derived samples, which is a much easier task than completion from a real partial scan and is unavailable at test time. The large CD-L2 improvement reported for BOSHNet in Table 5 could therefore be an artifact of this ground-truth supervision shortcut rather than evidence for a topological prior. The paper needs ablations that (a) train the same backbone with random or partial-input samples in place of BOSH samples, and (b) compare against an oracle that simply upsamples the BOSH samples. Without these, the superiority claim for BOSHNet is not established.
  4. [Table 5 and Section 6] The reported quantitative evidence is internally inconsistent. Under the metric definitions in Section 9, CD-L2 is a sum of squared nearest-neighbor distances, so for point clouds with nearest-neighbor distances below 1 it must be smaller than CD-L1. Yet Table 5 reports SnowflakeNet with CD-L1=60 and CD-L2=72, while BOSHNet is reported with CD-L1=69 and CD-L2=5.4. The drastically different relationship between CD-L1 and CD-L2 across rows is not explained, and no error bars or repeated runs are provided for any method. This undermines the quantitative basis for the central claim that topological priors improve completion.
minor comments (6)
  1. [Abstract and Contributions] The words 'prove' and 'proof' are used in the abstract and contributions, but the paper contains no formal theorem or rigorous mathematical proof; recommend replacing these with 'demonstrate' or 'show'.
  2. [Section 4.2] Please specify exactly how the 'average of persistence averages' is computed across classes, and how the essential 0-dimensional class of infinite persistence is treated when averaging H0 persistence values.
  3. [Table 4] The SnowflakeNet row contains missing entries represented by '-' without a note; please state how the class mean was computed from the available classes, or report complete results.
  4. [Supplementary Section 12.1] The statement that setting k=1 'works fairly well' appears to conflict with the main text's discussion that k>=2 is needed for disconnected partial inputs; please clarify the criterion and sensitivity for selecting k.
  5. [Figure 7] Please specify what is plotted as the 'skeleton' and how it is obtained from the persistence pairs; without this, the figure cannot be reproduced.
  6. [Section 4.1] The dataset release statement says the data and code will be made open-source, but no URL or license is provided; please include the actual artifact links.

Circularity Check

2 steps flagged · score 6.0 of 10

BOSHNet's 'topological prior' is sampled from the ground-truth target itself, so its gain partly reduces to target-copying; Eq. 1's persistence loss is internally signed backwards.

  1. self definitional [Section 5.2, Eq. 2 (BOSHNet loss)]
    "BOSH directly samples numerous backbones from the surface of the ground truth PC. ... The new total loss function is given by: nX i=1 kX j=1 M (N et(BOSH (ci, j)), ci) + nX i=1 M (N et(pi), ci)"

    The '0-dim PH prior' is constructed by taking sparse samples of the very cloud the network is asked to complete. Eq. 2 trains Net to map BOSH(ci,j), a subset of the ground-truth complete cloud, back to the same complete cloud ci. No persistence pair or Vietoris-Rips complex is computed; 'Homology' in BOSH is a renaming of target subsampling. The low CD-L2 (5.4) in Table 5 is therefore partly a measure of learning to copy/expand target-derived inputs, not evidence for an independent topological prior. The proof claim that integrating topological priors improves completion reduces, by construction, to training on privileged samples of the prediction target.

  2. other [Section 5.1.2, Eq. 1 and Section 4.2]
    "Points further from the diagonal represent features with larger persistence (death − birth). ... The persistence of each pair is (b-d) (Section 3). ... Topo Loss = nX i=0 1{i > k}(bi −di) = nX i=k+1 (bi −di) (1)"

    The paper's own definition of persistence is death − birth, but Eq. 1 minimizes b_i − d_i, the negative of persistence. In a Vietoris-Rips filtration all H0 components are born at b_i = 0, so Eq. 1 equals −Σ d_i; minimizing it maximizes death times and hence maximizes persistence, the opposite of the stated goal of merging components early. The claimed mechanism 'minimize the sum of persistence ... ensures that at the end of the filtration, there is one connected component left' is the reverse of what Eq. 1 does by the paper's own definitions. The skeleton prior is therefore not derived from the loss as written.

full rationale

The RealPC dataset, the three-metric comparison, and the non-neural/neural benchmarking are independent contributions and do not contain circular reasoning; the cited baselines are external. The circularity is confined to the paper's central 'proof' that topological priors improve completion. BOSHNet's Eq. 2 defines the topological prior by sampling the ground-truth target itself, so its reported advantage (especially CD-L2) is built into the training objective rather than being an independent demonstration of topological regularization. TopODGNet's Eq. 1 is internally inconsistent with the paper's own persistence definition, and Section 5.1.2 asserts without derivation that 0-dim PH produces a spatial skeleton; no standard H0 persistence computation outputs such a skeleton. These issues mean the claimed proof is not supported by the equations as written, although the dataset contribution remains valuable and the empirical comparison on RealPC is a legitimate benchmark.

Assumptions & free parameters 4 free parameters · 4 assumptions · 2 invented entities

The central method rests on two unverified equivalences: 0-dim PH features as a skeleton, and random sparse surface sampling as a proxy for PH backbones. The only explicit hand-tuned parameter is the truncation index k in the topological loss; BOSHNet's sampling configuration is unspecified. No independent evidence exists for the invented 'homology backbone' concept.

free parameters (4)
  • Topological loss truncation index k = k=1 (or manual per-class)
    Eq. 1/3: determines how many 0-dim persistence pairs are ignored in the loss. Requires manual inspection of partial point clouds; k=1 'works fairly well' (Supplementary 12.1).
  • Number of homology backbones k in BOSHNet = Not specified
    Eq. 2: k is the number of sampled backbones per complete cloud, but no value is reported.
  • BOSH sparsity levels = Not specified
    Section 5.2: backbones are sampled at multiple sparsity levels, but the sampling densities and algorithm are not described.
  • Number of points N in sparsification/incompletion = Not specified
    Section 4.1: N values for uniform/non-uniform sparsification and incompletion are not reported, affecting dataset statistics.
assumptions (4)
  • ad hoc to paper 0-dim PH persistence pairs define a geometric skeleton of a point cloud
    Section 5.1.2: '0-dim PH ensures the extraction of a complete PC skeleton...' This is asserted without derivation; standard 0-dim PH only records component merge scales.
  • ad hoc to paper Minimizing persistence of 0-dim PH features forces generation along the skeleton
    Section 5.1.2: the topological loss is defined as sum of (bi-di) for i>k, but no proof links this to point placement along a skeleton.
  • ad hoc to paper BOSH-sampled sparse subsets are proxies for 0-dim PH backbones
    Section 5.2: 'We hypothesize that these can be directly sampled from the complete shapes surfaces' and calls them 'proxy PH backbones', without evidence that random/sparse surface sampling mimics 0-dim PH.
  • domain assumption Sparse seed point clouds at intermediate decoder layers are sufficient for topological priors
    Section 5.1: the integration criterion assumes models with sparse seeds at intermediate layers allow compute-efficient PH; not universally true.
invented entities (2)
  • Homology Sampler (BOSH) and proxy backbones
    purpose: Bypass expensive persistence computation by sampling sparse subsets of the complete cloud and using them as training seeds
    Introduced in Section 5.2; no falsifiable prediction outside the paper, and its equivalence to PH features is assumed.
  • 0-dim PH skeleton
    purpose: A geometric prior claimed to guide point generation
    Section 5.1.2 and Figure 7; the existence of a skeleton as a spatial object is not established by persistence pairs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Revisiting Point Cloud Completion: Are We Ready For The Real-World?." pith.science (2026). https://pith.science/paper/XB4VB64J

@misc{pith2026241117580,
  author       = {Pith},
  title        = {Pith review of: Revisiting Point Cloud Completion: Are We Ready For The Real-World?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XB4VB64J}},
  note         = {Machine review of arXiv:2411.17580}
}
read the original abstract

Point clouds acquired in constrained, challenging, uncontrolled, and multi-sensor real-world settings are noisy, incomplete, and non-uniformly sparse. This presents acute challenges for the vital task of point cloud completion. Using tools from Algebraic Topology and Persistent Homology (PH), we demonstrate that current benchmark object point clouds lack rich topological features that are integral part of point clouds captured in realistic environments. To facilitate research in this direction, we contribute the first real-world industrial dataset for point cloud completion, RealPC - a diverse, rich and varied set of point clouds. It consists of ~ 40,000 pairs across 21 categories of industrial structures in railway establishments. Benchmark results on several strong baselines reveal that existing methods fail in real-world scenarios. We discover a striking observation - unlike current datasets, RealPC consists of multiple 0- and 1-dimensional PH-based topological features. We prove that integrating these topological priors into existing works helps improve completion. We present how 0-dimensional PH priors extract the global topology of a complete shape in the form of a 3D skeleton and assist a model in generating topologically consistent complete shapes. Since computing Homology is expensive, we present a simple, yet effective Homology Sampler guided network, BOSHNet that bypasses the Homology computation by sampling proxy backbones akin to 0-dim PH. These backbones provide similar benefits of 0-dim PH right from the start of the training, unlike similar methods where accurate backbones are obtained only during later phases of the training.

Figures

Figures reproduced from arXiv: 2411.17580 by the authors.

Figure 1
Figure 1. Qualitative comparison of current existing and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. (a) to (c) Progression of filtration on a PC over dif [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Scene-level PCs from different acquisition tech [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Object-level training dataset creation methodol [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Sub-optimal completion results on a RealPC PC using the best-performing completion baseline Snowflak￾eNet ( [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: TopODGNet. We calculate 0-dim PH based topological priors over sparse seeds and integrate it into the loss function. It enables completion along a topologically consistent skeleton. topology-based backbone we use 0-dim PH priors on the seed PCs. 0-dim PH ensures the ex…
Figure 8
Figure 8. Figure 8: BOSHNet. Our compute-efficient Homology Sampler samples proxy PH backbones from the surface of the complete scan. These guide the completion process from the start of training. levels. We circumvent the costly PH computations by in￾troducing a simple and novel Homology…
Figure 9
Figure 9. Figure 9: We find that TopODGNet is able to follow the global topology of the ground truth PC better than ODGNet, as indicated by the tapering upper half of the output and overall consistency. TopODGNet shows visible benefits of introducing topological priors. Incomplete input B…
Figure 10
Figure 10. Figure 10: BOSHNet, on account of multiple 0-dim PH priors, is able to reconstruct the incomplete PC reasonably. in a slight reduction of Chamfer distance as shown in Ta￾ble 5. While both the outputs are still noisy, TopODGNet maintains a better topological consistency w.r.t. th…
Figure 11
Figure 11. Figure 11: (a) to (c) Progression of filtration on a point cloud over different spatial resolutions as the distance threshold [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]
Figure 12
Figure 12. Figure 12: Surface reconstruction of some instances from different categories of ShapeNet and RealPC using alpha shapes. [PITH_FULL_IMAGE:figures/full_fig_p020_12.png]
Figure 13
Figure 13. Figure 13: Left: Complete GT Middle: Partial Input to Homology Sampler based Model Right: Output. Our Ho￾mology Sampler model, on account of multiple 0-dim PH priors is able to accurately reconstruct the incomplete PC. 9 [PITH_FULL_IMAGE:figures/full_fig_p021_13.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. A Multi-View High-Resolution Foot-Ankle Complex Point Cloud Dataset During Gait for Occlusion-Robust 3D Completion

    cs.CV 2025-07 conditional novelty 7.0 of 10

    FootGait3D provides graded partial-to-complete real-world point clouds of the foot-ankle complex during gait, enabling benchmarking of occlusion-robust 3D completion.

Reference graph

Works this paper leans on

72 extracted references · 66 canonical work pages · cited by 1 Pith paper

  1. [1]

    https:// ressources.data.sncf.com/explore/ dataset/nuage-points-3d/table/, 2017

    SNCF R ´eseau railway dataset. https:// ressources.data.sncf.com/explore/ dataset/nuage-points-3d/table/, 2017. 5, 2

    work page 2017

  2. [2]

    Alliegro, D

    A. Alliegro, D. Valsesia, G. Fracastoro, E. Magli, and T. Tommasi. Denoise and contrast for category agnostic shape completion. In CVPR, 2021. 3

    work page 2021

  3. [3]

    Bingchen, N

    G. Bingchen, N. Yinyu, L. Yiqun, H. Xiaoguang, and Y . Yizhou. Me-pcn: Point completion conditioned on mask emptiness. In ICCV, 2021. 3

    work page 2021

  4. [4]

    Br ¨uel-Gabrielsson, V

    R. Br ¨uel-Gabrielsson, V . Ganapathi-Subramanian, P. Skraba, and L. J. Guibas. Topology-aware surface reconstruction for point clouds. In Computer Graphics Forum, volume 39, pages 197–207. Wiley Online Library, 2020. 3

    work page 2020

  5. [5]

    Topology-Aware Surface Reconstruction for Point Clouds

    R. Br ¨uel-Gabrielsson, V . Ganapathi-Subramanian, P. Skraba, and L. Guibas. Topology-aware surface reconstruction for point clouds. arXiv preprint arXiv:1811.12543, 2019. 3

    work page Pith review arXiv 2019

  6. [6]

    P. Cai, D. Scott, X. Li, and S. Wang. Orthogonal dictionary guided shape completion network for point cloud. In AAAI,

    work page

  7. [7]

    R. J. Campello, D. Moulavi, and J. Sander. Density-based clustering based on hierarchical density estimates. InPacific- Asia conference on knowledge discovery and data mining , pages 160–172. Springer, 2013. 4, 3

    work page 2013

  8. [8]

    Z. Cang, L. Mu, and G.-W. Wei. Representability of alge- braic topology for biomolecules in machine learning based scoring and virtual screening. PLoS computational biology, 14(1):e1005929, 2018. 2

    work page 2018

Show all 72 references
  1. [9]

    Carlsson, A

    G. Carlsson, A. Zomorodian, A. Collins, and L. Guibas. Per- sistence barcodes for shapes. In Proceedings of the 2004 Eu- rographics/ACM SIGGRAPH symposium on Geometry pro- cessing, pages 124–135, 2004. 3

    2004

  2. [10]

    Chang, A

    A. Chang, A. Dai, T. Funkhouser, M. Halber, M. Niessner, M. Savva, S. Song, A. Zeng, and Y . Zhang. Matterport3d: Learning from rgb-d data in indoor environments. arXiv preprint arXiv:1709.06158, 2017. 3

    2017 arXiv

  3. [11]

    A. X. Chang, T. Funkhouser, L. Guibas, P. Hanrahan, Q. Huang, Z. Li, S. Savarese, M. Savva, S. Song, H. Su, et al. Shapenet: An information-rich 3d model repository. arXiv preprint arXiv:1512.03012, 2015. 3

    2015 arXiv

  4. [12]

    C. Chen, X. Ni, Q. Bai, and Y . Wang. A topological regular- izer for classifiers via persistent homology. In The 22nd In- ternational Conference on Artificial Intelligence and Statis- tics, pages 2573–2582. PMLR, 2019. 3

    2019

  5. [13]

    Z. Chen, F. Long, Z. Qiu, T. Yao, W. Zhou, J. Luo, and T. Mei. Anchorformer: Point cloud completion from dis- criminative nodes. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 13581–13590, 2023. 2, 6

    2023

  6. [14]

    J. R. Clough, I. Oksuz, N. Byrne, J. A. Schnabel, and A. P. King. Explicit topological priors for deep-learning based image segmentation using persistent homology. In Inter- national Conference on Information Processing in Medical Imaging, pages 16–28. Springer, 2019. 3

    2019

  7. [15]

    M. Cserep. Hungarian mls point clouds of railroad environ- ment and annotated ground truth data. Mendeley Data. DOI: https://doi. org/10.17632/ccxpzhx9dj, 1, 2022. 5, 2

    2022 doi

  8. [16]

    A. Dai, A. X. Chang, M. Savva, M. Halber, T. Funkhouser, and M. Nießner. Scannet: Richly-annotated 3d reconstruc- tions of indoor scenes. In Proceedings of the IEEE con- ference on computer vision and pattern recognition , pages 5828–5839, 2017. 3

    2017

  9. [17]

    T. K. Dey, K. Li, C. Luo, P. Ranjan, I. Safa, and Y . Wang. Persistent heat signature for pose-oblivious matching of in- complete models. In Computer Graphics Forum, volume 29, pages 1545–1554. Wiley Online Library, 2010. 3

    2010

  10. [18]

    Edelsbrunner

    H. Edelsbrunner. Computational topology an introduction,

  11. [19]

    Edelsbrunner and J

    H. Edelsbrunner and J. Harer. Computational Topology - an Introduction. American Mathematical Society, 2010. 3

    2010

  12. [20]

    Edelsbrunner, D

    H. Edelsbrunner, D. Kirkpatrick, and R. Seidel. On the shape of a set of points in the plane. IEEE Transactions on infor- mation theory, 29(4):551–559, 1983. 5, 3

    1983

  13. [21]

    B. Fei, W. Yang, W.-M. Chen, Z. Li, Y . Li, T. Ma, X. Hu, and L. Ma. Comprehensive review of deep learning- based 3d point cloud completion processing and analysis. IEEE Transactions on Intelligent Transportation Systems , 23(12):22862–22883, 2022. 2

    2022

  14. [22]

    R. B. Gabrielsson and G. Carlsson. Exposition and interpre- tation of the topology of neural networks. In 2019 18th ieee international conference on machine learning and applica- tions (icmla), pages 1069–1076. IEEE, 2019. 3

    2019

  15. [23]

    R. B. Gabrielsson, B. J. Nelson, A. Dwaraknath, and P. Skraba. A topology layer for machine learning. In Inter- national Conference on Artificial Intelligence and Statistics, pages 1553–1563. PMLR, 2020. 3

    2020

  16. [24]

    Gebhart and P

    T. Gebhart and P. Schrater. Adversarial examples tar- get topological holes in deep networks. arXiv preprint arXiv:1901.09496, 2019. 3

    1901 arXiv

  17. [25]

    Geiger, P

    A. Geiger, P. Lenz, and R. Urtasun. Are we ready for au- tonomous driving? the kitti vision benchmark suite. In 2012 IEEE conference on computer vision and pattern recogni- tion, pages 3354–3361. IEEE, 2012. 3

    2012

  18. [26]

    H. Geng, Z. Li, Y . Geng, J. Chen, H. Dong, and H. Wang. Partmanip: Learning cross-category generalizable part ma- nipulation policy from point cloud observations. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2978–2988, 2023. 2

    2023

  19. [27]

    Giansiracusa, R

    N. Giansiracusa, R. Giansiracusa, and C. Moon. Persis- tent homology machine learning for fingerprint classifica- tion. arXiv preprint arXiv:1711.09158, 2017. 3

    2017 arXiv

  20. [28]

    Z. Guo, H. Liu, H. Shi, F. Li, X. Guo, and B. Cheng. Kd-tree- based euclidean clustering for tomographic sar point cloud extraction and segmentation. IEEE Geoscience and Remote Sensing Letters, 20:1–5, 2023. 2

    2023

  21. [29]

    Hofer, R

    C. Hofer, R. Kwitt, M. Niethammer, and A. Uhl. Deep learn- ing with topological signatures. Advances in neural informa- tion processing systems, 30, 2017. 3

    2017

  22. [30]

    T. Hu, Z. Han, and M. Zwicker. 3d shape completion with multi-view consistent inference. In AAAI, 2020. 3

    2020

  23. [31]

    Huang, D

    H. Huang, D. Li, H. Zhang, U. Ascher, and D. Cohen-Or. Consolidation of unorganized point clouds for surface recon- struction. ACM transactions on graphics (TOG), 28(5):1–7,

  24. [32]

    Huang, S

    H. Huang, S. Wu, M. Gong, D. Cohen-Or, U. Ascher, and H. Zhang. Edge-aware point set resampling. ACM transac- tions on graphics (TOG), 32(1):1–12, 2013. 6, 3

    2013

  25. [33]

    S. Li, P. Gao, X. Tan, and M. Wei. Proxyformer: Proxy alignment assisted point cloud completion with missing part sensitive transformer. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 9466–9475, 2023. 2

    2023

  26. [34]

    Lipman, D

    Y . Lipman, D. Cohen-Or, D. Levin, and H. Tal-Ezer. Parameterization-free projection for geometry reconstruc- tion. ACM Transactions on Graphics (ToG) , 26(3):22–es,

  27. [35]

    Liu, S.-K

    J.-Y . Liu, S.-K. Jeng, and Y .-H. Yang. Applying topological persistence in convolutional neural network for music audio signals. arXiv preprint arXiv:1608.07373, 2016. 3

    2016 arXiv

  28. [36]

    M. Moor, M. Horn, B. Rieck, and K. Borgwardt. Topolog- ical autoencoders. In International conference on machine learning, pages 7045–7054. PMLR, 2020. 4, 2

    2020

  29. [37]

    Nguyen, M

    M. Nguyen, M. Aktas, and E. Akbas. Bot detection on so- cial networks using persistent homology. Mathematical and Computational Applications, 25(3):58, 2020. 2

    2020

  30. [38]

    Nguyen, S

    M. Nguyen, S. Vats, S. Van Damme, J. Van Der Hooft, M. T. Vega, T. Wauters, C. Timmerer, and H. Hellwagner. Impact of quality and distance on the perception of point clouds in mixed reality. In 2023 15th International Conference on Quality of Multimedia Experience (QoMEX) , pa...

    2023

  31. [39]

    L. Pan, X. Chen, Z. Cai, J. Zhang, H. Zhao, S. Yi, and Z. Liu. Variational relational point completion network. In CVPR,

  32. [40]

    Poulenard, P

    A. Poulenard, P. Skraba, and M. Ovsjanikov. Topological function optimization for continuous shape matching. In Computer Graphics Forum, volume 37, pages 13–25. Wiley Online Library, 2018. 3

    2018

  33. [41]

    C. S. Pun, K. Xia, and S. X. Lee. Persistent-homology- based machine learning and its applications–a survey. arXiv preprint arXiv:1811.00252, 2018. 3

    2018 arXiv

  34. [42]

    B. Qiu, Y . Zhou, L. Dai, B. Wang, J. Li, Z. Dong, C. Wen, Z. Ma, and B. Yang. Whu-railway3d: A diverse dataset and benchmark for railway point cloud semantic segmenta- tion. IEEE Transactions on Intelligent Transportation Sys- tems, 2024. 5, 2

    2024

  35. [43]

    Sarmad, H

    M. Sarmad, H. J. Lee, and Y . M. Kim. Rl-gan-net: A re- inforcement learning agent controlled gan network for real- time point cloud shape completion. In CVPR, 2019. 3

    2019

  36. [44]

    Singh, C

    Y . Singh, C. M. Farrelly, Q. A. Hathaway, T. Leiner, J. Jag- tap, G. E. Carlsson, and B. J. Erickson. Topological data analysis in medical imaging: current state of the art. Insights into Imaging, 14(1):1–10, 2023. 2

    2023

  37. [45]

    L. P. Tchapmi, V . Kosaraju, H. Rezatofighi, I. Reid, and S. Savarese. Topnet: Structural point cloud decoder. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 383–392, 2019. 2

    2019

  38. [46]

    L. P. Tchapmi, V . Kosaraju, H. Rezatofighi, I. Reid, and S. Savarese. Topnet: Structural point cloud decoder. In CVPR, 2019. 3, 6

    2019

  39. [47]

    K. W. Tesema, L. Hill, M. W. Jones, M. I. Ahmad, and G. K. Tam. Point cloud completion: A survey. IEEE Transactions on Visualization and Computer Graphics, 2023. 2

    2023

  40. [48]

    Tianxin, Z

    H. Tianxin, Z. Hao, C. Jinhao, Y . Xuemeng, W. Mengmeng, Z. Xiangrui, Z. Jiangning, Y . Yi, X. Yifan, and L. Yong. Rfnet: Recurrent forward network for dense point cloud completion. In ICCV, 2021. 3

    2021

  41. [49]

    B. Ton. Labelled high resolution point cloud dataset of 15 catenary arches in the netherlands. 4TU Res. data, The NetherlandsTech. Rep, 2022. 5, 2

    2022

  42. [50]

    Vahdat, F

    A. Vahdat, F. Williams, Z. Gojcic, O. Litany, S. Fidler, K. Kreis, et al. Lion: Latent point diffusion models for 3d shape generation. Advances in Neural Information Process- ing Systems, 35:10021–10039, 2022. 6

    2022

  43. [51]

    J. Wang, Y . Cui, D. Guo, J. Li, Q. Liu, and C. Shen. Pointattn: You only need attention for point cloud comple- tion. Proceedings of the AAAI Conference on Artificial Intel- ligence, 38(6):5472–5480, Mar. 2024. 2, 3

    2024

  44. [52]

    Y . Wang, Y . Sun, Z. Liu, S. Sarma, M. Bronstein, and J. Solomon. Dynamic graph cnn for learning on point clouds. corr. arXiv preprint arXiv:1801.07829, 2018. 6

    2018 arXiv

  45. [53]

    Y . Wang, Y . Sun, Z. Liu, S. E. Sarma, M. M. Bronstein, and J. M. Solomon. Dynamic graph cnn for learning on point clouds. ACM Transactions on Graphics (tog) , 38(5):1–12,

  46. [54]

    X. Wen, P. Xiang, Z. Han, Y .-P. Cao, P. Wan, W. Zheng, and Y .-S. Liu. Pmp-net: Point cloud completion by learning multi-step point moving paths. In CVPR, 2021. 3

    2021

  47. [55]

    T. Wu, J. Zhang, X. Fu, Y . Wang, J. Ren, L. Pan, W. Wu, L. Yang, J. Wang, C. Qian, et al. Omniobject3d: Large- vocabulary 3d object dataset for realistic perception, recon- struction and generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Reco...

    2023

  48. [56]

    Xiang, X

    P. Xiang, X. Wen, Y .-S. Liu, Y .-P. Cao, P. Wan, W. Zheng, and Z. Han. Snowflakenet: Point cloud completion by snowflake point deconvolution with skip-transformer. In Proceedings of the IEEE/CVF international conference on computer vision, pages 5499–5509, 2021. 2, 3, 6

    2021

  49. [57]

    Xiang, X

    P. Xiang, X. Wen, Y .-S. Liu, Y .-P. Cao, P. Wan, W. Zheng, and Z. Han. Snowflakenet: Point cloud completion by snowflake point deconvolution with skip-transformer. In ICCV, 2021. 3

    2021

  50. [58]

    H. Xie, H. Yao, S. Zhou, J. Mao, S. Zhang, and W. Sun. Gr- net: Gridding residual network for dense point cloud com- pletion. In European conference on computer vision, pages 365–381. Springer, 2020. 3, 6

    2020

  51. [59]

    Xuancheng, F

    Z. Xuancheng, F. Yutong, L. Siqi, Z. Changqing, W. Hai, Z. Xibin, G. Yandong, and G. Yue. View-guided point cloud completion. In CVPR, 2021. 3

    2021

  52. [60]

    Y . Yang, C. Feng, Y . Shen, and D. Tian. Foldingnet: Point cloud auto-encoder via deep grid deformation. In CVPR,

  53. [61]

    Y . Yang, C. Feng, Y . Shen, and D. Tian. Foldingnet: Point cloud auto-encoder via deep grid deformation. In Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 206–215, 2018. 6

    2018

  54. [62]

    X. Yu, Y . Rao, Z. Wang, Z. Liu, J. Lu, and J. Zhou. Pointr: Diverse point cloud completion with geometry-aware trans- formers. In ICCV, 2021. 2, 3, 6

    2021

  55. [63]

    J. Yuan, B. Zhang, X. Yan, B. Shi, T. Chen, Y . Li, and Y . Qiao. Ad-pt: Autonomous driving pre-training with large- scale point cloud dataset. Advances in Neural Information Processing Systems, 36, 2024. 2 11

    2024

  56. [64]

    W. Yuan, T. Khot, D. Held, C. Mertz, and M. Hebert. Pcn: Point completion network. In 2018 international conference on 3D vision (3DV), pages 728–737. IEEE, 2018. 1, 2, 3, 6

    2018

  57. [65]

    W. Yuan, T. Khot, D. Held, C. Mertz, and M. Hebert. Pcn: point completion network. In 3DV, 2018. 3

    2018

  58. [66]

    Zhang, X

    S. Zhang, X. Liu, H. Xie, L. Nie, H. Zhou, D. Tao, and X. Li. Learning geometric transformation for point cloud completion. International Journal of Computer Vision , 131(9):2425–2445, 2023. 2

    2023

  59. [67]

    R. Zhao, H. Wang, C. Zhang, and W. Cai. Pointneuron: 3d neuron reconstruction via geometry and topology learn- ing of point clouds. In Proceedings of the IEEE/CVF Win- ter Conference on Applications of Computer Vision , pages 5787–5797, 2023. 2

    2023

  60. [68]

    H. Zhou, Y . Cao, W. Chu, J. Zhu, T. Lu, Y . Tai, and C. Wang. Seedformer: Patch seeds based point cloud completion with upsample transformer. In European conference on computer vision, pages 416–432. Springer, 2022. 2, 3

    2022

  61. [69]

    Zhuang, Z

    Z. Zhuang, Z. Zhi, T. Han, Y . Chen, J. Chen, C. Wang, M. Cheng, X. Zhang, N. Qin, and L. Ma. A survey of point cloud completion. IEEE Journal of Selected Topics in Ap- plied Earth Observations and Remote Sensing, 2024. 2 12 Revisiting Point Cloud Completion: Are We Ready For ...

    2024

  62. [70]

    which uses an edge-aware method to compute normals away from the surface singularities, and then uses the data obtained to resample towards the said singularities with the help of a bilateral projector. 12 Topological Loss 12.1 Formulation The topological loss function assists...

  63. [71]

    It transforms each point of a point cloud using Edge Convolutions into higher dimensions and retains the number of points across each layer

    because it can easily be used for integrating topologi- cal priors as it satisfies the required criterion for integrating PH priors. It transforms each point of a point cloud using Edge Convolutions into higher dimensions and retains the number of points across each layer. The...

  64. [72]

    We also show some more examples of BOSHNet com- pletions using RealPC in Figure 13

    Almost all the 0-dimensional topological features (indi- cated by dot) are along or very close to the diagonal for the synthetic datasets (Table 10) as opposed to RealPC (Table 9). We also show some more examples of BOSHNet com- pletions using RealPC in Figure 13. BOSHNet pick...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.