Pith. sign in

REVIEW 4 major objections 6 minor 29 references

Point-GR: Graph Residual Point Cloud Network for 3D Object Classification and Segmentation

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read A residual graph network for raw point clouds reaches 73.47% mean IoU on indoor scene segmentation while using fewer parameters than graph-based baselines.

desk verdict The residual graph idea is plausible, but the architecture as written doesn't close dimensionally and the S3DIS claim is unverifiable. read the letter →

arxiv 2412.03052 v1 pith:L43HFMYL submitted 2024-12-04 cs.CV eess.IV

classification cs.CVeess.IV
keywords pointcloud3Dobjectclassificationpartsegmentationscenegraphneuralnetworkresiduallearningdynamic
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes Point-GR, a graph-based deep network that takes unordered point clouds and first transforms them through a residual embedding block before learning hierarchical features. Its central claim is that this residual transformation preserves local geometry while shrinking the parameter count relative to comparable graph networks, and that the resulting model achieves the best reported mean IoU (73.47%) on the S3DIS indoor scene segmentation benchmark. On ModelNet-40 it reports 92.71% classification accuracy and on ShapeNetPart 85.2% mIoU, enough to match the strongest existing graph methods. A sympathetic reader would take the contribution as evidence that residual graph embeddings offer a lightweight backbone that transfers across classification, part segmentation, and scene segmentation.

What carries the argument

The central object is the Point-GR Transformation block: graph construction via k-nearest neighbors, edge features $d_{pe} = [d_p, d_e]$ with $d_p = P_i$ and $d_e = P_i - P_j$, followed by a Point Residual Embedding that interleaves convolution, batch normalization, ReLU, a second convolution, and a residual skip connection before max selection over the local k-neighborhood. This block is the reusable front end for all three tasks, and it is paired with a Feature Learning Network that stacks Multi-Scale Graph Construction and Point Embedding modules at channel depths $\{6,64,128,256\}$ and concatenates per-stage skip features. The residual path is what the authors credit for stable training and parameter savings.

What would settle it

Implement the Point-GR Transformation exactly as written and run a forward pass on ModelNet-40: if the residual addition has no specified channel projection, the shapes will not align and the network will not train; alternatively, run the same pipeline with and without an inserted 1x1 projection and check whether the 73.47% S3DIS and 92.71% ModelNet-40 numbers reproduce under the stated hyperparameters.

Watch

Extended reading notes

Core claim

Point-GR is built from a Point-GR Transformation module that constructs a k-NN graph over the input points, forms edge features by concatenating each point's own coordinates with the difference to its neighbors ($d_p$ and $d_e$), and then runs two convolutions with a residual skip connection between the previous layer output and the convolution output. The resulting point residual features are passed through repeated Feature Learning blocks that build multi-scale graphs at progressive channel dimensions {6,64,128,256}, and skip connections from each stage are concatenated before a final convolutional and fully connected head. The paper claims this design yields 92.71% overall accuracy on ModelNet-40, 85.2% mean IoU on ShapeNetPart, and 73.47% mean IoU with 82.79% overall IoU on S3DIS, together with fewer trainable parameters than the graph baseline on classification (1.80M vs 2.61M) and part segmentation (1.04M vs 1.45M). The residual transformation is the load-bearing novelty that the authors say mitigates point-permutation sensitivity and vanishing gradients while keeping the network lightweight.

Load-bearing premise

The load-bearing assumption is that the described residual block is actually constructible: the skip connection must add feature maps whose channel dimensions are not stated to match, since the final transformation output is [N x 6], while the layer feeding the residual is the raw input or a 64-channel feature map.

Editorial extensions

If this is right

  • A single residual graph backbone can serve classification, part segmentation, and indoor scene segmentation, so the method should transfer to other point-cloud tasks such as object detection and instance segmentation without redesign.
  • The reported parameter reductions (1.80M vs 2.61M for classification; 1.04M vs 1.45M for part segmentation) imply lower memory and faster inference for the same accuracy class, which matters on embedded and LiDAR platforms.
  • The S3DIS result, if taken at face value, suggests a substantial improvement over graph-based baselines on scene-level semantic segmentation, moving the graph approach closer to the top of the benchmark.
  • The ablation sensitivity to k (peak at k=20, degradation at k=40) gives a practical recipe: the neighborhood size should be tuned rather than defaulted to large values.

Reading between the lines

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

  • The paper does not specify how the residual skip connection aligns channels when the previous layer's feature count differs from the 6-channel output of the transformation; the described addition may require an implicit 1x1 projection or padding, and the architecture is only fully reproducible once that step is pinned down.
  • If the 73.47% S3DIS number holds under a standard 6-fold protocol, the largest gap to prior graph baselines may come more from training setup (block size, 9-channel input, optimizer schedule) than from the residual mechanism itself; an ablation that swaps only the residual path on and off would separate these contributions.
  • The same residual transformation could be tested on sparser and noisier LiDAR scans; the k-NN graph construction plus residual embedding is a natural candidate for outdoor point clouds where density varies sharply.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes Point-GR, a graph-based residual network for point cloud classification, part segmentation, and scene segmentation. The method consists of a Point-GR Transformation module that builds k-NN graphs, computes edge features, and applies residual convolutions, followed by a Point-GR Feature Learning Network with three multi-scale graph blocks. The authors report 92.71% accuracy on ModelNet-40, 85.2% mean IoU on ShapeNetPart, and 73.47% mean IoU on S3DIS, which they claim is state of the art, together with reduced parameter counts relative to the DG-CNN baseline. The paper also presents ablations over input point count and neighborhood size k.

Significance. If the reported results are reproducible, the contribution is potentially useful: a lightweight graph residual network that is competitive on three standard benchmarks and reduces parameters relative to a strong graph baseline would be of interest to the point cloud learning community. The ablation studies in Tables 5 and 6 and the parameter comparison in Table 4 are positive features of the manuscript. The strengths are, however, conditional: there is no code, no error bars, no statistical comparison to prior work, and the S3DIS comparison is too sparse to support the 'state-of-the-art' claim. More importantly, the central architectural description is internally inconsistent, so the reported numbers cannot currently be derived from the manuscript as written.

major comments (4)
  1. [Section 2.1/2.3] The residual skip connection in the Point-GR Transformation is dimensionally inconsistent. Section 2.1 states that 'the output from the previous layer is added to the output of the convolution layer,' while Section 2.3 defines the transformation as two convolutions that map the input first to 64 channels and then to 6 channels, with final output [N x 6]. The only candidate 'previous layer' outputs are the block input (N x 3) and the first convolution's output (N x 64); both are incompatible with the N x 6 output of the second convolution. No 1x1 projection, padding, or other dimension-matching operation is specified anywhere in the manuscript. Since every downstream FLN block, concatenation, and the parameter count in Table 4 depends on the exact output of this transformation, the architecture cannot be instantiated from the written method, and the reported classification, part-segmentation, and S3DIS results cannot be reproduced or verified.
  2. [Table 3 / Section 3.3] The claim in the abstract and Section 3.3 that Point-GR achieves state-of-the-art scene segmentation is not supported by Table 3, which compares only PointNet, PointCNN, and DG-CNN. Several established methods (e.g., PointNet++, KPConv, and more recent S3DIS submissions) are absent, and no quantitative comparison to them is provided. A 'state-of-the-art' claim requires comparison with the current best published methods under the same evaluation protocol. The 73.47% mean IoU is therefore a standalone result, not evidence of state-of-the-art performance.
  3. [Section 3.3] The S3DIS input features are described inconsistently. The authors first say each point has '9 attributes (XYZ, RGB, and normalized spatial coordinates)' but two sentences later say the input contains '9 channels x, y, z, RGB color information, intensity, normal vector.' The second list has more than nine components (XYZ, RGB, intensity, and a 3D normal vector sum to ten), so the actual input dimension is unclear. Because the input feature dimension affects both the network architecture and the S3DIS result, this ambiguity is load-bearing for reproducibility.
  4. [Tables 1-3 / Section 3] All benchmark results are single-run accuracies with no error bars, standard deviations, seeds, or code release. In addition, the manuscript switches between 'DG-CNN' and 'DC-GNN' (Table 1 lists both DG-CNN [9] at 92.9% and DC-GNN [17] at 93.64%; Table 2 uses DC-GNN; Section 3.1 refers to 'the baseline DG-CNN model'). If DC-GNN is the authors' own prior model, this is not disclosed in the main text, and the comparison is not a standard independent baseline. The small accuracy gaps in Table 1 (0.19-0.93 percentage points) are within typical run-to-run variation for point cloud networks, so without repeated runs the comparative claims are not statistically supported.
minor comments (6)
  1. [Table 2] The class-wise IoU numbers in the DC-GNN row are run together (e.g., '84.53 78.95 54.28 85.02 95.8 85.38 81.0 92.2 82.9'), making the table very hard to read.
  2. [Table 1 / References] Table 1 cites P2Sequence as reference [20], but reference [20] is titled 'PCT: Point Cloud Transformer'; the citation appears to be mismatched.
  3. [Table 4] The PointNet row in Table 4 contains a single parameter value (0.69M) with no task breakdown, while DG-CNN and Ours have values for Classification, Part-Seg, and S3DIS; please clarify or align the entries.
  4. [Abstract / Section 1] The claim that residual-based learning mitigates point permutation problems is not justified: permutation invariance in this architecture comes from symmetric aggregation (max pooling) in the graph convolution, not from residual skip connections. Please rephrase or provide a concrete argument.
  5. [Section 3.1] The sentence 'Our model demonstrates average performance with an accuracy of 92.71%, surpassing most of the state-of-the-art networks' is misleading because 92.71% is below several entries in Table 1, including RS-CNN, KPConv, DG-CNN, and DC-GNN.
  6. [Throughout] There are numerous typos and grammatical errors, including 'Psuedo' in Algorithm 1, 'Classfication' in Algorithm 1, and 'eliminate the issue' in Section 2.1; the manuscript would benefit from careful proofreading.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the reported accuracies are external benchmark evaluations, and the only self-citation (DC-GNN as a comparison baseline) is not load-bearing.

full rationale

The central results—92.71% ModelNet-40 accuracy, 85.2% ShapeNetPart mIoU, and 73.47% S3DIS mIoU—are empirical outcomes on standard external benchmarks, not quantities derived from fitted parameters or from the paper's own equations. No equation defines a target result in terms of the input, and no fitted parameter is later renamed as a prediction. The only self-citation is the use of the authors' earlier DC-GNN [17] as a comparison baseline in Tables 1 and 2; those comparisons are secondary, and the headline S3DIS claim is made against DG-CNN [9] and PointNet [5], so the self-citation is not load-bearing. The dimension mismatch in the residual skip connection described in Sections 2.1 and 2.3 is a reproducibility and correctness concern, not a circular reduction: the written architecture is internally inconsistent, but that does not make the reported scores equivalent to the inputs by construction. Accordingly, no circular step meets the evidentiary bar, and the appropriate score is 1.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The empirical claims rest on standard benchmark assumptions plus two paper-specific architecture assumptions: that residual connections address permutation without proof, and that a skip connection between mismatched channel counts is implementable. No new physical or mathematical entities are introduced.

free parameters (3)
  • k-nearest neighbor count (k) = 20 (classification), 40 (part segmentation), unspecified for S3DIS
    Selected by ablation in Section 3.4, Table 6; accuracy varies from 29.53% (k=5) to 92.7% (k=20), so the choice directly affects the reported classification result.
  • Sampled points per cloud = 1024 (classification), 2048 (part segmentation), 4096 (S3DIS)
    Input size chosen by hand and ablation (Table 5); classification accuracy degrades from 92.7% at 1024 points to 88.08% at 512 points, so this is a tunable quantity affecting the headline numbers.
  • Point-GR Transformation output dimension = 6
    Hand-designed embedding size introduced in Section 2.3. It is also the dimension that conflicts with the unstated residual projection, making it load-bearing for the architecture's implementability.
assumptions (5)
  • standard math k-NN graph construction with Euclidean distance gives a valid local neighborhood representation.
    Used in Section 2.1, Eq. 1, as the basis for all edge features. This is a standard tool, though the paper does not justify the specific distance choice.
  • ad hoc to paper Residual (skip) connections mitigate point permutation problems.
    Asserted in the Abstract and Section 2.1 without proof or analysis. Permutation invariance in the described pipeline actually comes from symmetric max pooling and aggregation, not from residual connections.
  • ad hoc to paper A skip connection can be added between feature maps of different channel dimensions without a specified projection layer.
    Section 2.1 and Section 2.3 imply adding 3- or 64-channel features to a 6-channel output; no 1x1 convolution or channel alignment is described.
  • domain assumption Standard benchmark splits and preprocessing protocols are adequate and equivalent across compared methods.
    Invoked for ModelNet-40, ShapeNetPart, and S3DIS in Sections 3.1-3.3. No code or preprocessing scripts are provided to verify that the comparisons are apples-to-apples.
  • domain assumption Reported training hyperparameters are sufficient to reproduce the claimed accuracies.
    Section 3 gives learning rates, momentum, and batch size, but omits random seeds, epoch counts, and detailed scheduling, so the training setup is not fully specified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Point-GR: Graph Residual Point Cloud Network for 3D Object Classification and Segmentation." pith.science (2026). https://pith.science/paper/L43HFMYL

@misc{pith2026241203052,
  author       = {Pith},
  title        = {Pith review of: Point-GR: Graph Residual Point Cloud Network for 3D Object Classification and Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L43HFMYL}},
  note         = {Machine review of arXiv:2412.03052}
}
read the original abstract

In recent years, the challenge of 3D shape analysis within point cloud data has gathered significant attention in computer vision. Addressing the complexities of effective 3D information representation and meaningful feature extraction for classification tasks remains crucial. This paper presents Point-GR, a novel deep learning architecture designed explicitly to transform unordered raw point clouds into higher dimensions while preserving local geometric features. It introduces residual-based learning within the network to mitigate the point permutation issues in point cloud data. The proposed Point-GR network significantly reduced the number of network parameters in Classification and Part-Segmentation compared to baseline graph-based networks. Notably, the Point-GR model achieves a state-of-the-art scene segmentation mean IoU of 73.47% on the S3DIS benchmark dataset, showcasing its effectiveness. Furthermore, the model shows competitive results in Classification and Part-Segmentation tasks.

Figures

Figures reproduced from arXiv: 2412.03052 by the authors.

Figure 1
Figure 1. The figure shows the Point-GR Transformation network. It consists of [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Shows the Point-GR Features Learning Network(F.L.N). It consists of [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Deep learning based architecture of (a) The extended Point-GR architec [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: The visual representation shows how the network segments each shape [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 25 canonical work pages

  1. [9]

    Dynamic graph cnn for learning on point clouds

    Yue Wang, Yongbin Sun, Ziwei Liu, Sanjay E Sarma, Michael M Bronstein, and Justin M Solomon. Dynamic graph cnn for learning on point clouds. Acm Transactions On Graphics (tog) , 38(5):1–12, 2019

  2. [17]

    Md Meraz, Md Afzal Ansari, Mohammed Javed, and Pavan Chakraborty. Dc-gnn: drop channel graph neural network for object classification and part segmentation in the point cloud.International Journal of Multimedia Information Retrieval, 11(2):123–133, 2022

  3. [1]

    A concise and provably informative multi-scale signature based on heat diffusion

    Jian Sun, Maks Ovsjanikov, and Leonidas Guibas. A concise and provably informative multi-scale signature based on heat diffusion. In Computer graphics forum, volume 28, pages 1383–1392. Wiley Online Library, 2009

  4. [2]

    Aligning point cloud views using persistent feature histograms

    Radu Bogdan Rusu, Nico Blodow, Zoltan Csaba Marton, and Michael Beetz. Aligning point cloud views using persistent feature histograms. In 2008 IEEE/RSJ international conference on intelligent robots and systems , pages 3384–3391. IEEE, 2008

  5. [3]

    Using spin images for efficient object recognition in cluttered 3d scenes

    Andrew E Johnson and Martial Hebert. Using spin images for efficient object recognition in cluttered 3d scenes. IEEE Transactions on pattern analysis and machine intelligence , 21(5):433–449, 1999

  6. [4]

    3d shapenets: A deep representation for volumetric shapes

    Zhirong Wu, Shuran Song, Aditya Khosla, Fisher Yu, Linguang Zhang, Xi- aoou Tang, and Jianxiong Xiao. 3d shapenets: A deep representation for volumetric shapes. In Proceedings of the IEEE conference on computer vi- sion and pattern recognition, pages 1912–1920, 2015

  7. [5]

    Pointnet: Deep learning on point sets for 3d classification and segmentation

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. InProceedings of the IEEE conference on computer vision and pattern recognition , pages 652–660, 2017

  8. [6]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space.arXiv preprint arXiv:1706.02413, 2017

    Charles R Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space.arXiv preprint arXiv:1706.02413, 2017

Show all 29 references
  1. [7]

    Dynamicedge-conditionedfilters in convolutional neural networks on graphs

    MartinSimonovskyandNikosKomodakis. Dynamicedge-conditionedfilters in convolutional neural networks on graphs. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , July 2017

  2. [8]

    Grid-gcn for fast and scalable point cloud learning

    Qiangeng Xu, Xudong Sun, Cho-Ying Wu, Panqu Wang, and Ulrich Neu- mann. Grid-gcn for fast and scalable point cloud learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 5661–5670, 2020

  3. [10]

    Voxnet: A 3d convolutional neu- ral network for real-time object recognition

    Daniel Maturana and Sebastian Scherer. Voxnet: A 3d convolutional neu- ral network for real-time object recognition. In2015 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS) ,pages922–928. IEEE, 2015

  4. [11]

    Escape from cells: Deep kd-networks for the recognition of 3d point cloud models

    Roman Klokov and Victor Lempitsky. Escape from cells: Deep kd-networks for the recognition of 3d point cloud models. InProceedings of the IEEE International Conference on Computer Vision (ICCV) , Oct 2017

  5. [12]

    Octnet: Learning deep 3d representations at high resolutions

    Gernot Riegler, Ali Osman Ulusoy, and Andreas Geiger. Octnet: Learning deep 3d representations at high resolutions. In Proceedings of the IEEE Title Suppressed Due to Excessive Length 15 conference on computer vision and pattern recognition , pages 3577–3586, 2017

  6. [13]

    Densenet: Implementing efficient convnet de- scriptor pyramids

    Forrest Iandola, Matt Moskewicz, Sergey Karayev, Ross Girshick, Trevor Darrell, and Kurt Keutzer. Densenet: Implementing efficient convnet de- scriptor pyramids. arXiv preprint arXiv:1404.1869 , 2014

  7. [14]

    Interpolated convolu- tional networks for 3d point cloud understanding

    Jiageng Mao, Xiaogang Wang, and Hongsheng Li. Interpolated convolu- tional networks for 3d point cloud understanding. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 1578– 1587, 2019

  8. [15]

    Modelinglocalgeometric structure of 3d point clouds using geo-cnn

    ShiyiLan,RuichiYu,GangYu,andLarrySDavis. Modelinglocalgeometric structure of 3d point clouds using geo-cnn. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 998–1008, 2019

  9. [16]

    An introduction to kernel and nearest-neighbor nonpara- metric regression

    Naomi S Altman. An introduction to kernel and nearest-neighbor nonpara- metric regression. The American Statistician, 46(3):175–185, 1992

  10. [18]

    J. Li, B. Chen, and Gim Hee Lee. So-net: Self-organizing network for point cloud analysis. 2018 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 9397–9406, 2018

  11. [19]

    A-cnn: Annularly convo- lutional neural networks on point clouds

    Artem Komarichev, Zichun Zhong, and Jing Hua. A-cnn: Annularly convo- lutional neural networks on point clouds. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 7421–7430, 2019

  12. [20]

    Pct: Point cloud transformer

    Meng-Hao Guo, Jun-Xiong Cai, Zheng-Ning Liu, Tai-Jiang Mu, Ralph R Martin, and Shi-Min Hu. Pct: Point cloud transformer. arXiv preprint arXiv:2012.09688, 2020

  13. [21]

    Escape from cells: Deep kd-networks for the recognition of 3d point cloud models

    Roman Klokov and Victor Lempitsky. Escape from cells: Deep kd-networks for the recognition of 3d point cloud models. InProceedings of the IEEE International Conference on Computer Vision , pages 863–872, 2017

  14. [22]

    Point convolutional neural networks by extension operators.arXiv preprint arXiv:1803.10091 , 2018

    Matan Atzmon, Haggai Maron, and Yaron Lipman. Point convolutional neural networks by extension operators.arXiv preprint arXiv:1803.10091 , 2018

  15. [23]

    Di, and B

    Yangyan Li, Rui Bu, Mingchao Sun, Wei Wu, X. Di, and B. Chen. Pointcnn: Convolution on x-transformed points. InNeurIPS, 2018

  16. [24]

    Wenxuan Wu, Zhongang Qi, and F. Li. Pointconv: Deep convolutional networks on 3d point clouds. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 9613–9622, 2019

  17. [25]

    Relation- shape convolutional neural network for point cloud analysis

    Yongcheng Liu, Bin Fan, Shiming Xiang, and Chunhong Pan. Relation- shape convolutional neural network for point cloud analysis. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8895–8904, 2019

  18. [26]

    Kpconv: Flexible and deformable convolution for point clouds

    Hugues Thomas, Charles R Qi, Jean-Emmanuel Deschaud, and Marcotegui J. Kpconv: Flexible and deformable convolution for point clouds. InPro- 16 M. Meraz et al. ceedings of the IEEE/CVF International Conference on Computer Vision , pages 6411–6420, 2019

  19. [27]

    Pointgrid: A deep network for 3d shape under- standing

    Truc Le and Ye Duan. Pointgrid: A deep network for 3d shape under- standing. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9204–9214, 2018

  20. [28]

    A scal- able active framework for region annotation in 3d shape collections.ACM Transactions on Graphics (ToG), 35(6):1–12, 2016

    Li Yi, Vladimir G Kim, Duygu Ceylan, I-Chao Shen, Mengyan Yan, Hao Su, Cewu Lu, Qixing Huang, Alla Sheffer, and Leonidas Guibas. A scal- able active framework for region annotation in 3d shape collections.ACM Transactions on Graphics (ToG), 35(6):1–12, 2016

  21. [29]

    Pointcnn: Convolution on x-transformed points.Advances in neural information processing systems, 31, 2018

    Yangyan Li, Rui Bu, Mingchao Sun, Wei Wu, Xinhan Di, and Baoquan Chen. Pointcnn: Convolution on x-transformed points.Advances in neural information processing systems, 31, 2018

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.