Pith. sign in

REVIEW 3 major objections 5 minor 95 references

LogoSP: Local-global Grouping of Superpoints for Unsupervised Semantic Segmentation of 3D Point Clouds

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Grouping superpoints by global graph-frequency patterns, after distilling self-supervised 2D features into 3D, produces semantic pseudo-labels that train an unsupervised 3D segmentation network to the best reported accuracy among…

desk verdict A useful spectral-grouping trick that improves unsupervised 3D segmentation on indoor benchmarks, but the 'global semantics' interpretation is under-supported and the dense graph is not implemented as described. read the letter →

arxiv 2506.07857 v1 pith:QEWPEFFR submitted 2025-06-09 cs.CV cs.AIcs.LGcs.RO

Zihui Zhang , Weisheng Dai , Hongtao Wen , Bo Yang This is my paper
classification cs.CVcs.AIcs.LGcs.RO
keywords unsupervisedsemanticsegmentation3DpointcloudssuperpointsgraphFouriertransformfrequency-domaingroupingpseudo-labelsself-superviseddistillationDINOv2features
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 tries to show that unsupervised 3D semantic segmentation improves when grouping happens not only on per-point local features but on global patterns of a superpoint graph in the frequency domain. The authors propose LogoSP, which distills self-supervised 2D image features into 3D points, grows the points into superpoints (groups of similar points), builds a similarity graph over all superpoints, and uses the eigenvectors of its normalized Laplacian as global patterns. Grouping superpoints by their coefficients on those patterns yields semantic pseudo-labels that train a segmentation network without any human annotation. On ScanNet validation the method reports 35.8 mIoU against 29.6 for the strongest unsupervised baseline, and it also leads on S3DIS and nuScenes, so if the claim holds it would substantially lower the annotation cost of 3D scene understanding.

What carries the argument

The central object is the Graph Fourier Transform applied to a global superpoint graph. Superpoints are groups of 3D points sharing similar distilled features; they become nodes of a graph whose adjacency is $a_{ij}=\exp(-\|\tilde{f}_i-\tilde{f}_j\|^2)$. The normalized Laplacian $L=D^{-1/2}(D-A)D^{-1/2}$ is diagonalized as $L=U\Lambda U^T$, and each eigenvector $u_s$ is treated as a global pattern defined over all superpoints. The paper projects superpoint features into the frequency domain with $\tilde{F}_{feq}=U^T\tilde{F}$, groups the $S$ patterns into $S'$ clusters, averages each cluster into a new pattern $v_{s'}$, and finally applies K-means to the rows of $V=[v_1,\dots,v_{S'}]$ to produce $C$ semantic pseudo-labels. This machinery is what lets the method see all superpoints at once and discover global semantic priors that local feature clustering misses.

What would settle it

Re-run LogoSP's pseudo-label generation on ScanNet with the eigenvectors of the normalized Laplacian replaced by a random orthogonal matrix of the same size while keeping every other step fixed; if validation mIoU stays near 35.8, the global patterns are not doing the work, while a large drop would confirm that the eigenvectors carry the semantic signal.

Watch

Extended reading notes

Core claim

The central claim is that semantic information for a 3D scene can be read off the global structure of a superpoint feature graph rather than only from local per-point features. LogoSP builds one graph whose nodes are all superpoints across the training scenes, with edge weight $a_{ij}=\exp(-\|\tilde{f}_i-\tilde{f}_j\|^2)$; it computes the normalized Laplacian $L=D^{-1/2}(D-A)D^{-1/2}$ and takes its eigenvectors as global patterns. Projecting superpoint features onto these patterns, clustering the patterns into a smaller set, and then clustering superpoints by their coefficients in the grouped patterns produces pseudo-labels for $C$ classes. The paper reports that this pseudo-label quality translates into the best unsupervised results it compares against: 35.8 mIoU on ScanNet validation and 32.7 on the hidden test, 46.3 mIoU in the S3DIS 6-fold setting, and 20.1 mIoU on nuScenes validation. The paper also claims that the learned global patterns themselves correspond to meaningful semantics, and that their semantic content sharpens as superpoints grow during training.

Load-bearing premise

The load-bearing assumption is that the eigenvectors of the superpoint feature-similarity graph, after being grouped, align with semantic classes; the paper supports this empirically, but if the distilled features carry no semantics, the frequency-domain grouping cannot create them.

Editorial extensions

If this is right

  • Unsupervised 3D semantic segmentation can be driven entirely by pseudo-labels derived from global graph-frequency patterns, with no human annotation in training.
  • Small and minority classes that prior unsupervised methods overlook, such as toilet, books, and truck, become recoverable because their global frequency signatures survive the grouping.
  • A model trained on ScanNet transfers to unseen S3DIS areas at 43.6 mIoU, over ten points above the best unsupervised baseline, so the learned semantics are not tied to one dataset.
  • The number of semantic classes can be chosen freely at training time, allowing the same pipeline to produce coarser or finer segmentations without new labels.
  • Distilled self-supervised features are load-bearing: removing the distillation step drops ScanNet mIoU from 35.8 to 26.8 in the paper's ablation.

Reading between the lines

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

  • This suggests that the pseudo-label generation is a form of spectral clustering on the superpoint feature graph, and if so the hyperparameter $S'$ could be set from the eigenvalue gap rather than by grid search.
  • The strong dependence on DINOv2 features implies the ceiling of this pipeline is set by the 2D teacher; a purely 3D self-supervised feature extractor could replace it and remove the need for aligned RGB-D data.
  • The same graph-frequency grouping recipe could transfer to unsupervised segmentation of 2D images or video, where graph nodes would be pixels or segments instead of superpoints.
  • The smaller outdoor gains (20.1 vs 18.2 mIoU on nuScenes) suggest a testable extension: swap the 2D teacher for a LiDAR-native self-supervised model and check whether the global grouping recovers minor outdoor classes more effectively.
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

3 major / 5 minor

Summary. The paper proposes LogoSP, an unsupervised 3D semantic segmentation pipeline that combines 2D-to-3D distillation of DINOv2 features, bottom-up superpoint growing, and a top-down module that builds a global superpoint graph, applies a Graph Fourier Transform, groups the resulting eigenvectors in the frequency domain, and K-means clusters the grouped patterns to produce semantic pseudo-labels for training a SparseConv network. Experiments are reported on ScanNet, S3DIS, and nuScenes, with large improvements over unsupervised baselines on the indoor benchmarks and smaller gains on the outdoor benchmark.

Significance. If the frequency-domain grouping is genuinely the source of the reported gains, the paper is a meaningful advance in unsupervised 3D semantic segmentation and would be of interest to the community. The paper reports strong empirical results on ScanNet (val mIoU 35.8 vs 29.6 for PointDC-DINOv2) and S3DIS (6-fold mIoU 46.3 vs 41.3), releases code, includes per-category tables and extensive ablations, and tests cross-dataset generalization. These are concrete strengths. However, the central mechanism needs a sharper control experiment to isolate its contribution, and the scalability of the global graph construction is not addressed. The outdoor results also do not support the broad 'large margins' claim in the abstract, as the conclusion itself acknowledges.

major comments (3)
  1. [Section 3.3, Eqs. (5)-(7)] The global superpoint graph is described as a dense graph over S = H*M0 nodes, with an adjacency matrix A in R^{S x S} containing all pairwise affinities. For ScanNet, H = 1201 and M0 is per-scene but likely on the order of hundreds to thousands, putting S in the 10^5-10^6 range. Storing A and computing the full eigendecomposition of the normalized Laplacian is O(S^2) in memory and O(S^3) in time, which is not feasible at that scale with the details given. No sparsification, kNN truncation, approximation, or hardware/runtime details are provided. This makes the central module as written not reproducible and raises the question of whether the actual implementation matches the description. The authors should specify exactly how the graph and eigendecomposition are computed (e.g., a sparse kNN graph with Lanczos iteration) or revise the description accordingly.
  2. [Section 3.3 and Table 8] The core novelty claim is that grouping superpoints via global patterns in the frequency domain discovers semantics beyond local-feature grouping, but the paper never runs the natural control: replacing Steps 2-3 of Section 3.3 with K-means (or standard spectral clustering) on the same superpoint features \tilde F. Ablation (12) removes the module and uses K-means on per-point local features, and Appendix B adds K-means on distilled per-point features and GrowSP-with-distillation, but none of these isolates the frequency-domain grouping at the superpoint level. The 4.4-point gap between (12) at 31.4 and the full model at 35.8 could come from the larger receptive field of superpoint-level clustering or from the spectral grouping itself. Because the paper's title and abstract emphasize local-global grouping in the frequency domain, this missing control is load-bearing for the stated contribution.
  3. [Abstract and Table 5] The abstract and introduction claim that LogoSP 'surpasses all existing unsupervised methods by large margins,' but on nuScenes validation the gain over PointDC-DINOv2 is only 1.9 mIoU points (20.1 vs 18.2), and per-category results in Table 21 show that LogoSP is worse than PointDC-DINOv2 on several classes, including barrier, pedestrian, drivable surface, sidewalk, and terrain. The conclusion itself states that 'our performance on the outdoor dataset is mixed.' The strong wording of the abstract and Section 1 should be moderated to claim large margins on the indoor benchmarks and competitive but more modest gains outdoors, so that the claims match the evidence.
minor comments (5)
  1. [Section 3.3, Eq. (5)] The affinity kernel uses the raw Euclidean norm on 384-dimensional features without specifying whether the features are normalized or how the scale of the kernel is chosen; this can strongly affect the Laplacian and the resulting clusters, and a note on normalization or scale selection would improve reproducibility.
  2. [Section 3.3, Eq. (5)] The notation S = H*M0 assumes every scene has the same number M0 of initial superpoints, but Section 3.2 states that M0 is different across scenes. The total should be written as S = sum_h M0^h, or the notation should be explicitly defined as the current superpoint count per scene.
  3. [Algorithm 1] Algorithm 1's Stage 1 groups initial superpoints into C classes via Section 3.3, while Section 3.3 is described as operating on the 'full set of superpoints' from all training scenes after the bottom-up growing step. The relationship between the initial Stage 1 pseudo-labels and the later rounds of Section 3.3 should be clarified.
  4. [Section 4.4] In the cross-dataset evaluation, the classifier is built by K-means on features from the training set and evaluated on test areas; it would be helpful to state explicitly whether Hungarian matching is applied here as well, since the class orders in ScanNet and S3DIS differ.
  5. [General] There are several typographical errors, including 'sueprpoints' in Algorithm 1, 'GowSP' in Section 4.2, and 'anther' in Section 3.3; these should be corrected in a final revision.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the pseudo-label pipeline is an unsupervised clustering and self-training method, and the global-pattern claims are empirical rather than definitional.

full rationale

The paper's derivation chain is not circular. The central pseudo-label generation (Section 3.3) constructs a superpoint graph from distilled features, computes Laplacian eigenvectors, clusters them, and then K-means clusters the resulting superpoint representations. This is a spectral-clustering-style procedure on the same features that are later clustered, but that does not make the method circular: the pseudo-labels are produced by the clustering itself, not by ground-truth labels or by a parameter fitted to the evaluation target. The claim that the 'global patterns' encode semantics is an empirical assertion supported by ablations (Table 8, variant 12) and by visualizations (Figure 6), and it could in principle be false if the distilled features carried no semantic structure. The paper does not fit any parameter to the validation labels and then report that fit as a prediction; the Hungarian matching used at evaluation is a standard label-alignment step, not a training input. The only self-citation, GrowSP [91], is used as a prior component for superpoint growing and training loss, and the paper explicitly compares against GrowSP as a baseline; this is normal building on prior work and is not load-bearing for the central novelty claim. No equation reduces to its own input by construction, and no result is imported from a self-citation chain. The absence of a direct K-means-on-superpoint-features control is a completeness concern, not evidence of circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The method relies on pretrained DINOv2 features, a known class count C, and a spectral grouping hypothesis. Key hyperparameters (S', M1, MT, T) are selected on the validation sets, meaning the final numbers are not fully label-free predictions.

free parameters (6)
  • S' (number of grouped global patterns) = 50 (ScanNet), 10 (S3DIS)
    Selected via ablation on validation sets; controls granularity of the grouped patterns.
  • M1 (superpoints after first growing) = 80
    Chosen on ScanNet validation; larger values reduce early grouping errors.
  • MT (final superpoint count) = 40
    Chosen on ScanNet validation; controls final superpoint size.
  • T (growing rounds) = 5
    Number of growth rounds, fixed across datasets.
  • C (semantic class count) = 20 (ScanNet), 12 (S3DIS), 16 (nuScenes)
    Assumed known from the dataset; used as K in the final clustering step.
  • RANSAC/Euclidean thresholds (nuScenes) = 0.2m
    Manual thresholds for initial superpoint generation in outdoor scenes (Appendix A).
assumptions (5)
  • domain assumption DINOv2 self-supervised features, when projected to 3D, provide semantic priors.
    Section 3.1 relies on DINOv2 features as the initial source of semantics without verification on 3D data.
  • domain assumption The number of semantic classes C is known.
    Section 3 states this assumption; it is required for K-means to produce C clusters.
  • ad hoc to paper Grouped Laplacian eigenvectors of the superpoint graph correspond to semantic structure.
    Core novelty; asserted via Eqs. 8-10 and supported only by downstream segmentation results.
  • domain assumption Initial superpoints are sufficiently pure and semantically coherent.
    Appendix A and C discuss purity; the pipeline cannot recover if initial superpoints mix many classes.
  • ad hoc to paper The Gaussian affinity kernel in Eq. 5, without scale selection, is an appropriate similarity measure.
    No analysis of kernel bandwidth or normalization effects is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LogoSP: Local-global Grouping of Superpoints for Unsupervised Semantic Segmentation of 3D Point Clouds." pith.science (2026). https://pith.science/paper/QEWPEFFR

@misc{pith2026250607857,
  author       = {Pith},
  title        = {Pith review of: LogoSP: Local-global Grouping of Superpoints for Unsupervised Semantic Segmentation of 3D Point Clouds},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QEWPEFFR}},
  note         = {Machine review of arXiv:2506.07857}
}
read the original abstract

We study the problem of unsupervised 3D semantic segmentation on raw point clouds without needing human labels in training. Existing methods usually formulate this problem into learning per-point local features followed by a simple grouping strategy, lacking the ability to discover additional and possibly richer semantic priors beyond local features. In this paper, we introduce LogoSP to learn 3D semantics from both local and global point features. The key to our approach is to discover 3D semantic information by grouping superpoints according to their global patterns in the frequency domain, thus generating highly accurate semantic pseudo-labels for training a segmentation network. Extensive experiments on two indoor and an outdoor datasets show that our LogoSP surpasses all existing unsupervised methods by large margins, achieving the state-of-the-art performance for unsupervised 3D semantic segmentation. Notably, our investigation into the learned global patterns reveals that they truly represent meaningful 3D semantics in the absence of human labels during training.

Figures

Figures reproduced from arXiv: 2506.07857 by the authors.

Figure 1
Figure 1. Given an input point cloud with complex structures from ScanNet dataset, our LogoSP automatically discovers accurate semantic [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The framework of LogoSP. The leftmost block shows the 2D-to-3D distillation module, the middle block illustrates the bottom-up [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. An example of high and low frequency features in 2D. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Qualitative results on ScanNet dataset (the first row) and S3DIS dataset (the second row). Red circles highlight the differences. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results on the nuScenes dataset. Red circles highlight the differences. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Examples of a salient global pattern in each scene after grouping superpoints in frequency domain, and per-point local features. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Examples of initial superpoints. B. Additional Ablation Study We conduct additional ablation studies to examine the im￾pact of the semantic category number, distillation process, and segmentation component. (1) K-means on distilled features: The distilled point feature…
Figure 8
Figure 8. Figure 8: Qualitative results of our method and baselines on the validation set of ScanNet dataset. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: Qualitative results of our method and baselines on the S3DIS dataset. [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Qualitative results of our method and baselines on the nuScenes dataset. [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

95 extracted references · 74 canonical work pages

  1. [1]

    Seeded region growing

    Rolf Adams and Leanne Bischof. Seeded region growing. TPAMI, 1994. 12

    1994

  2. [2]

    Zamir, and Silvio Savarese

    Iro Armeni, Sasha Sax, Amir R. Zamir, and Silvio Savarese. Joint 2D-3D-Semantic Data for Indoor Scene Understand- ing.arXiv:1702.01105, 2017. 5, 6, 7

    arXiv 2017

  3. [3]

    nuScenes: A multimodal dataset for autonomous driving.CVPR, 2020

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh V ora, Venice Erin Liong, and Qiang Xu. nuScenes: A multimodal dataset for autonomous driving.CVPR, 2020. 5

    2020

  4. [4]

    Deep Clustering for Unsupervised Learn- ing of Visual Features.ECCV, 2018

    Mathilde Caron, Piotr Bojanowski, Armand Joulin, and Matthijs Douze. Deep Clustering for Unsupervised Learn- ing of Visual Features.ECCV, 2018. 2

    2018

  5. [5]

    Emerg- ing Properties in Self-Supervised Vision Transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing Properties in Self-Supervised Vision Transformers. ICCV, 2021. 2, 3, 5

    2021

  6. [6]

    CLIP2Scene: Towards Label-efficient 3D Scene Un- derstanding by CLIP.CVPR, 2023

    Runnan Chen, Youquan Liu, Lingdong Kong, Xinge Zhu, Yuexin Ma, Yikang Li, Yuenan Hou, Yu Qiao, and Wenping Wang. CLIP2Scene: Towards Label-efficient 3D Scene Un- derstanding by CLIP.CVPR, 2023. 2

    2023

  7. [7]

    Shape Self-Correction for Unsupervised Point Cloud Understanding.ICCV, 2021

    Ye Chen, Jinxian Liu, Bingbing Ni, Hang Wang, Jiancheng Yang, Ning Liu, Teng Li, Qi Tian, Huawei Hisilicon, Huawei Car Bu, and Huawei Cloud. Shape Self-Correction for Unsupervised Point Cloud Understanding.ICCV, 2021. 2

    2021

  8. [8]

    Point DC: Unsupervised Semantic Segmentation of 3D Point Clouds via Cross-modal Distillation and Super- V oxel Clustering.ICCV, 2023

    Zisheng Chen, Hongbin Xu, Weitao Chen, Zhipeng Zhou, Haihong Xiao, Baigui Sun, Xuansong Xie, and Wenxiong Kang. Point DC: Unsupervised Semantic Segmentation of 3D Point Clouds via Cross-modal Distillation and Super- V oxel Clustering.ICCV, 2023. 1, 2, 3, 5, 6, 7, 14, 15, 16

    2023

Show all 95 references
  1. [9]

    Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes

    Julian Chibane, Francis Engelmann, Tuan Anh Tran, and Gerard Pons-Moll. Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes. ECCV, 2022. 2

    2022

  2. [10]

    PiCIE: Unsupervised Semantic Segmentation us- ing Invariance and Equivariance in Clustering.CVPR, 2021

    Jang Hyun Cho, Utkarsh Mall, Kavita Bala, and Bharath Hariharan. PiCIE: Unsupervised Semantic Segmentation us- ing Invariance and Equivariance in Clustering.CVPR, 2021. 2, 5, 6, 7, 14, 15, 16

    2021

  3. [11]

    4D Spatio-Temporal ConvNets: Minkowski Convolutional Neu- ral Networks.CVPR, 2019

    Christopher Choy, JunYoung Gwak, and Silvio Savarese. 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neu- ral Networks.CVPR, 2019. 2

    2019

  4. [12]

    4d spatio-temporal convnets: Minkowski convolutional neural networks.CVPR, 2019

    Christopher Choy, JunYoung Gwak, and Silvio Savarese. 4d spatio-temporal convnets: Minkowski convolutional neural networks.CVPR, 2019. 6

    2019

  5. [13]

    Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner

    Angela Dai, Angel X. Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. Scan- Net: Richly-annotated 3D Reconstructions of Indoor Scenes. CVPR, 2017. 5, 6, 7

    2017

  6. [14]

    PLA: Language-Driven Open- V ocabulary 3D Scene Understanding.CVPR, 2023

    Runyu Ding, Jihan Yang, Chuhui Xue, Wenqing Zhang, Song Bai, and Xiaojuan Qi. PLA: Language-Driven Open- V ocabulary 3D Scene Understanding.CVPR, 2023. 2

    2023

  7. [15]

    Unsupervised Semantic Segmentation by Con- trasting Object Mask Proposals.ICCV, 2021

    Wouter Van Gansbeke, Simon Vandenhende, and Stamatios Georgoulis. Unsupervised Semantic Segmentation by Con- trasting Object Mask Proposals.ICCV, 2021. 2

    2021

  8. [16]

    3D Semantic Segmentation with Submanifold Sparse Convolutional Networks.CVPR, 2018

    Benjamin Graham, Martin Engelcke, and Laurens van der Maaten. 3D Semantic Segmentation with Submanifold Sparse Convolutional Networks.CVPR, 2018. 1, 2, 3, 6, 14, 15, 16

    2018

  9. [17]

    SAM-guided Graph Cut for 3D Instance Segmentation.ECCV, 2024

    Haoyu Guo, He Zhu, Sida Peng, Yuang Wang, Yujun Shen, Ruizhen Hu, and Xiaowei Zhou. SAM-guided Graph Cut for 3D Instance Segmentation.ECCV, 2024. 2

    2024

  10. [18]

    Martin, and Shi-Min Hu

    Meng-Hao Guo, Jun-Xiong Cai, Zheng-Ning Liu, Tai-Jiang Mu, Ralph R. Martin, and Shi-Min Hu. PCT: Point Cloud Transformer.Computational Visual Media, 2021. 2

    2021

  11. [19]

    Semantic Abstraction: Open- World 3D Scene Understanding from 2D Vision-Language Models.CoRL, 2022

    Huy Ha and Shuran Song. Semantic Abstraction: Open- World 3D Scene Understanding from 2D Vision-Language Models.CoRL, 2022. 2

    2022

  12. [20]

    Mark Hamilton, Zhoutong Zhang, Bharath Hariharan, Noah Snavely, and William T. Freeman. Unsupervised Semantic Segmentation by Distilling Feature Correspondences.ICLR,

  13. [21]

    Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts.CVPR, 2021

    Ji Hou, Benjamin Graham, Matthias Nießner, and Saining Xie. Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts.CVPR, 2021. 2

    2021

  14. [22]

    RandLA-Net: Efficient Semantic Segmentation of Large- Scale Point Clouds.CVPR, 2020

    Qingyong Hu, Bo Yang, Linhai Xie, Stefano Rosa, Yulan Guo, Zhihua Wang, Niki Trigoni, and Andrew Markham. RandLA-Net: Efficient Semantic Segmentation of Large- Scale Point Clouds.CVPR, 2020. 1, 2

    2020

  15. [23]

    Learning Semantic Segmentation of Large-scale Point Clouds with Random Sampling.TPAMI, 2021

    Qingyong Hu, Bo Yang, Linhai Xie, Stefano Rosa, Yulan Guo, Zhihua Wang, Niki Trigoni, and Andrew Markham. Learning Semantic Segmentation of Large-scale Point Clouds with Random Sampling.TPAMI, 2021. 2

    2021

  16. [24]

    Exploring the devil in graph spectral domain for 3d point cloud attacks.ECCV,

    Qianjiang Hu, Daizong Liu, and Wei Hu. Exploring the devil in graph spectral domain for 3d point cloud attacks.ECCV,

  17. [25]

    SQN: Weakly-Supervised Semantic Segmentation of Large-Scale 3D Point Clouds.ECCV, 2022

    Qingyong Hu, Bo Yang, Guangchi Fang, Yulan Guo, Ales Leonardis, Niki Trigoni, and Andrew Markham. SQN: Weakly-Supervised Semantic Segmentation of Large-Scale 3D Point Clouds.ECCV, 2022. 1, 2

    2022

  18. [26]

    Spatio-temporal Self-Supervised Representation Learning for 3D Point Clouds.ICCV, 2021

    Siyuan Huang, Yichen Xie, Song-Chun Zhu, and Yixin Zhu. Spatio-temporal Self-Supervised Representation Learning for 3D Point Clouds.ICCV, 2021. 2

    2021

  19. [27]

    Invariant infor- mation clustering for unsupervised image classification and segmentation.ICCV, 2019

    Xu Ji, Andrea Vedaldi, and Joao Henriques. Invariant infor- mation clustering for unsupervised image classification and segmentation.ICCV, 2019. 2, 5, 6, 7, 14, 15, 16

    2019

  20. [28]

    EAGLE: Eigen Aggregation Learning for Object- Centric Unsupervised Semantic Segmentation.CVPR, 2024

    Chanyoung Kim, Woojung Han, Dayun Ju, and Seong Jae Hwang. EAGLE: Eigen Aggregation Learning for Object- Centric Unsupervised Semantic Segmentation.CVPR, 2024. 2

    2024

  21. [29]

    On-the-fly Category Discovery for LiDAR Semantic Segmentation.ECCV, 2024

    Hyeonseong Kim, Sung-Hoon Yoon, Minseok Kim, and Kuk-Jin Yoon. On-the-fly Category Discovery for LiDAR Semantic Segmentation.ECCV, 2024. 2

    2024

  22. [30]

    Berg, Wan-Yen Lo, Piotr Doll ´ar, and Ross Girshick

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C. Berg, Wan-Yen Lo, Piotr Doll ´ar, and Ross Girshick. Segment Anything.ICCV, 2023. 1, 2

    2023

  23. [31]

    OneFormer3D: One Transformer for Unified Point Cloud Segmentation.CVPR, 2024

    Maxim Kolodiazhnyi, Anna V orontsova, Anton Konushin, and Danila Rukhovich. OneFormer3D: One Transformer for Unified Point Cloud Segmentation.CVPR, 2024. 2

    2024

  24. [32]

    Virtual multi-view fusion for 3D semantic segmentation

    Abhijit Kundu, Xiaoqi Yin, Alireza Fathi, David Ross, Brian Brewington, Thomas Funkhouser, and Caroline Pantofaru. Virtual multi-view fusion for 3D semantic segmentation. ECCV, 2020. 2

    2020

  25. [33]

    Large-scale point cloud semantic segmentation with superpoint graphs.CVPR,

    Loic Landrieu and Martin Simonovsky. Large-scale point cloud semantic segmentation with superpoint graphs.CVPR,

  26. [34]

    Octree guided CNN with Spherical Kernels for 3D Point Clouds.CVPR,

    Huan Lei, Naveed Akhtar, and Ajmal Mian. Octree guided CNN with Spherical Kernels for 3D Point Clouds.CVPR,

  27. [35]

    Uni3DL: A Unified Model for 3D Vision-Language Understanding.ECCV, 2024

    Xiang Li, Jian Ding, Zhaoyang Chen, and Mohamed Elho- seiny. Uni3DL: A Unified Model for 3D Vision-Language Understanding.ECCV, 2024. 2

    2024

  28. [36]

    PointCNN: Convolution On X- Transformed Points.NIPS, 2018

    Yangyan Li, Rui Bu, Mingchao Sun, Wei Wu, Xinhan Di, and Baoquan Chen. PointCNN: Convolution On X- Transformed Points.NIPS, 2018. 2, 6, 14

    2018

  29. [37]

    Breckon, and Hubert P.H

    Jiaxu Liu, Zhengdi Yu, Toby P. Breckon, and Hubert P.H. Shum. U3DS3: Unsupervised 3D Semantic Scene Segmen- tation.WACV, 2024. 1

    2024

  30. [38]

    Kangcheng Liu, Yuzhi Zhao, Qiang Nie, Zhi Gao, and Ben M. Chen. Weakly Supervised 3D Scene Parsing with Region-Level Boundary Awareness and Instance Discrimi- nation.ECCV, 2022. 2

    2022

  31. [39]

    LESS: Label-Efficient Seman- tic Segmentation for LiDAR Point Clouds.ECCV, 2022

    Minghua Liu, Yin Zhou, Charles R Qi, Boqing Gong, Hao Su, and Dragomir Anguelov. LESS: Label-Efficient Seman- tic Segmentation for LiDAR Point Clouds.ECCV, 2022. 2

    2022

  32. [40]

    Segment Any Point Cloud Sequences by Distilling Vision Foundation Models.NeurIPS, 2023

    Youquan Liu, Lingdong Kong, Jun Cen, Runnan Chen, Wen- wei Zhang, Liang Pan, Kai Chen, and Ziwei Liu. Segment Any Point Cloud Sequences by Distilling Vision Foundation Models.NeurIPS, 2023. 2

    2023

  33. [41]

    Point- V oxel CNN for Efficient 3D Deep Learning.NeurIPS, 2019

    Zhijian Liu, Haotian Tang, Yujun Lin, and Song Han. Point- V oxel CNN for Efficient 3D Deep Learning.NeurIPS, 2019. 2

    2019

  34. [42]

    One Thing One Click: A Self-Training Approach for Weakly Super- vised 3D Semantic Segmentation.CVPR, 2021

    Zhengzhe Liu, Xiaojuan Qi, and Chi-Wing Fu. One Thing One Click: A Self-Training Approach for Weakly Super- vised 3D Semantic Segmentation.CVPR, 2021. 2

    2021

  35. [43]

    VV-Net: V oxel V AE Net with Group Convolu- tions for Point Cloud Segmentation.ICCV, 2019

    Hsien-Yu Meng, Gao Lin, Yu-Kun Lai, and Dinesh Manocha. VV-Net: V oxel V AE Net with Group Convolu- tions for Point Cloud Segmentation.ICCV, 2019. 2

    2019

  36. [44]

    RangeNet++: Fast and Accurate LiDAR Semantic Segmentation.IROS, 2019

    Andres Milioto, Ignacio Vizzo, Jens Behley, and Cyrill Stachniss. RangeNet++: Fast and Accurate LiDAR Semantic Segmentation.IROS, 2019. 2

    2019

  37. [45]

    Unsupervised semantic segmentation of urban high-density multispectral point clouds.arXiv:2410.18520, 2024

    Oona Oinonen, Lassi Ruoppa, Josef Taher, Matti Lehtom¨aki, Leena Matikainen, Kirsi Karila, Teemu Hakala, Antero Kukko, Harri Kaartinen, and Juha Hyypp ¨a. Unsupervised semantic segmentation of urban high-density multispectral point clouds.arXiv:2410.18520, 2024. 1

    arXiv

  38. [46]

    DINOv2: Learning Robust Visual Features without Supervision.TMLR, 2024

    Maxime Oquab, Timoth ´ee Darcet, Th´eo Moutakanni, Huy V V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-nouby, Mah- moud Assran, Nicolas Ballas, Wojciech Galuba, Russell Howes, Po-yao Huang, Shang-wen Li, Ishan Misra, Micha...

    2024

  39. [47]

    Better Call SAL: Towards Learning to Segment Anything in Lidar

    Aljo ˇsa O ˇsep, Tim Meinhardt, Francesco Ferroni, Neehar Peri, Deva Ramanan, and Laura Leal-Taix ´e. Better Call SAL: Towards Learning to Segment Anything in Lidar. ECCV, 2024. 2

    2024

  40. [48]

    Autore- gressive Unsupervised Image Segmentation.ECCV, 2020

    Yassine Ouali, C ´eline Hudelot, and Myriam Tami. Autore- gressive Unsupervised Image Segmentation.ECCV, 2020. 2

    2020

  41. [49]

    Yatian Pang, Wenxiao Wang, Francis E. H. Tay, Wei Liu, Yonghong Tian, and Li Yuan. Masked Autoencoders for Point Cloud Self-supervised Learning.ECCV, 2022. 2

    2022

  42. [50]

    V oxel cloud connectivity segmentation- supervoxels for point clouds.CVPR, 2013

    Jeremie Papon, Alexey Abramov, Markus Schoeler, and Flo- rentin Worgotter. V oxel cloud connectivity segmentation- supervoxels for point clouds.CVPR, 2013. 12

    2013

  43. [51]

    OA-CNNs: Omni-Adaptive Sparse CNNs for 3D Semantic Segmenta- tion.CVPR, 2024

    Bohao Peng, Xiaoyang Wu, Li Jiang, Yukang Chen, Heng- shuang Zhao, Zhuotao Tian, and Jiaya Jia. OA-CNNs: Omni-Adaptive Sparse CNNs for 3D Semantic Segmenta- tion.CVPR, 2024. 2

    2024

  44. [52]

    OpenScene: 3D Scene Understanding with Open V ocabular- ies.CVPR, 2023

    Songyou Peng, Kyle Genova, Chiyu ”Max” Jiang, An- drea Tagliasacchi, Marc Pollefeys, and Thomas Funkhouser. OpenScene: 3D Scene Understanding with Open V ocabular- ies.CVPR, 2023. 2, 3

    2023

  45. [53]

    Learning to Adapt SAM for Segmenting Cross-Domain Point Clouds.ECCV, 2024

    Xidong Peng, Runnan Chen, Feng Qiao, Lingdong Kong, Youquan Liu, Yujing Sun, Tai Wang, Xinge Zhu, and Yuexin Ma. Learning to Adapt SAM for Segmenting Cross-Domain Point Clouds.ECCV, 2024. 2

    2024

  46. [54]

    Qi, Hao Su, Kaichun Mo, and Leonidas J

    Charles R. Qi, Hao Su, Kaichun Mo, and Leonidas J. Guibas. PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation.CVPR, 2017. 1, 2, 6, 15, 16

    2017

  47. [55]

    Qi, Li Yi, Hao Su, and Leonidas J

    Charles R. Qi, Li Yi, Hao Su, and Leonidas J. Guibas. Point- Net++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space.NIPS, 2017. 2, 6, 14, 15, 16

    2017

  48. [56]

    Learning Transferable Visual Models From Natural Language Supervision.ICML, 2021

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning Transferable Visual Models From Natural Language Supervision.ICML, 2021. 1, 2

    2021

  49. [57]

    Global-Local Bidi- rectional Reasoning for Unsupervised Representation Learn- ing of 3D Point Clouds.CVPR, 2020

    Yongming Rao, Jiwen Lu, and Jie Zhou. Global-Local Bidi- rectional Reasoning for Unsupervised Representation Learn- ing of 3D Point Clouds.CVPR, 2020. 2

    2020

  50. [58]

    SAM 2: Segment Anything in Images and Videos

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

    2024 arXiv

  51. [59]

    Effi- cient 3d semantic segmentation with superpoint transformer

    Damien Robert, Hugo Raguet, and Loic Landrieu. Effi- cient 3d semantic segmentation with superpoint transformer. ICCV, 2023. 2

    2023

  52. [60]

    Language- Grounded Indoor 3D Semantic Segmentation in the Wild

    David Rozenberszki, Or Litany, and Angela Dai. Language- Grounded Indoor 3D Semantic Segmentation in the Wild. ECCV, 2022. 2

    2022

  53. [61]

    Unsupervised deep learning for semantic segmentation of multispectral LiDAR forest point clouds

    Lassi Ruoppa, Oona Oinonen, Josef Taher, Matti Lehtom¨aki, Narges Takhtkeshha, Antero Kukko, Harri Kaartinen, and Juha Hyypp ¨a. Unsupervised deep learning for semantic segmentation of multispectral LiDAR forest point clouds. arXiv:2502.06227, 2025. 1

    2025

  54. [62]

    Discrete signal processing on graphs: Graph fourier transform.ICASSP,

    Aliaksei Sandryhaila and Jos ´e MF Moura. Discrete signal processing on graphs: Graph fourier transform.ICASSP,

  55. [63]

    Self-Supervised Deep Learning on Point Clouds by Reconstructing Space

    Jonathan Sauder and Bjarne Sievers. Self-Supervised Deep Learning on Point Clouds by Reconstructing Space. NeurIPS, 2019. 2

    2019

  56. [64]

    Mask3D: Mask Trans- former for 3D Semantic Instance Segmentation.ICRA, 2023

    Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, and Bastian Leibe. Mask3D: Mask Trans- former for 3D Semantic Instance Segmentation.ICRA, 2023. 2

    2023

  57. [65]

    Progressive Proxy Anchor Propagation for Unsuper- vised Semantic Segmentation.ECCV, 2024

    Hyun Seok Seong, WonJun Moon, SuBeen Lee, and Jae-Pil Heo. Progressive Proxy Anchor Propagation for Unsuper- vised Semantic Segmentation.ECCV, 2024. 2

    2024

  58. [66]

    Weakly Supervised Segmentation on Outdoor 4D point clouds with Temporal Matching and Spatial Graph Propagation.CVPR, 2022

    Hanyu Shi, Jiacheng Wei, Ruibo Li, Fayao Liu, and Gu- osheng Lin. Weakly Supervised Segmentation on Outdoor 4D point clouds with Temporal Matching and Spatial Graph Propagation.CVPR, 2022. 2

    2022

  59. [67]

    Canonical Capsules: Unsupervised Capsules in Canoni- cal Pose.NeurIPS, 2021

    Weiwei Sun, Andrea Tagliasacchi, Boyang Deng, Sara Sabour, Soroosh Yazdani, Geoffrey Hinton, and Kwang Moo Yi. Canonical Capsules: Unsupervised Capsules in Canoni- cal Pose.NeurIPS, 2021. 2

    2021

  60. [68]

    Searching efficient 3d archi- tectures with sparse point-voxel convolution.ECCV, 2020

    Haotian Tang, Zhijian Liu, Shengyu Zhao, Yujun Lin, Ji Lin, Hanrui Wang, and Song Han. Searching efficient 3d archi- tectures with sparse point-voxel convolution.ECCV, 2020. 6, 16

    2020

  61. [69]

    Anh Thai, Weiyao Wang, Hao Tang, Stefan Stojanov, Matt Feiszli, and James M. Rehg. 3x2: 3D Object Part Segmenta- tion by 2D Semantic Correspondences.ECCV, 2024. 2

    2024

  62. [70]

    Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, Franc ¸ois Goulette, and Leonidas J

    Hugues Thomas, Charles R. Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, Franc ¸ois Goulette, and Leonidas J. Guibas. KPConv: Flexible and Deformable Convolution for Point Clouds.ICCV, 2019. 2

    2019

  63. [71]

    KPConvX: Modernizing Kernel Point Con- volution with Kernel Attention.CVPR, 2024

    Hugues Thomas, Yao-hung Hubert Tsai, Timothy D Barfoot, and Jian Zhang. KPConvX: Modernizing Kernel Point Con- volution with Kernel Attention.CVPR, 2024. 2

    2024

  64. [72]

    Unsupervised Point Cloud Co-part Segmentation via Co-attended Superpoint Generation and Aggregation.TMM,

    Ardian Umam, Cheng-Kun Yang, Jen-Hui Chuang, and Yen- Yu Lin. Unsupervised Point Cloud Co-part Segmentation via Co-attended Superpoint Generation and Aggregation.TMM,

  65. [73]

    Scribble- Supervised LiDAR Semantic Segmentation.CVPR, 2022

    Ozan Unal, Dengxin Dai, and Luc Van Gool. Scribble- Supervised LiDAR Semantic Segmentation.CVPR, 2022. 2

    2022

  66. [74]

    Hanchen Wang, Qi Liu, Xiangyu Yue, Joan Lasenby, and Matthew J. Kusner. Unsupervised Point Cloud Pre-Training via Occlusion Completion.ICCV, 2021. 2

    2021

  67. [75]

    Sarma, Michael M

    Yue Wang, Yongbin Sun, Ziwei Liu, Sanjay E. Sarma, Michael M. Bronstein, and Justin M. Solomon. Dynamic Graph CNN for Learning on Point Clouds.TOG, 38, 2019. 2, 14

    2019

  68. [76]

    Multi-Path Region Mining For Weakly Supervised 3D Semantic Segmentation on Point Clouds

    Jiacheng Wei, Guosheng Lin, Kim-Hui Yap, Tzu-Yi Hung, and Lihua Xie. Multi-Path Region Mining For Weakly Supervised 3D Semantic Segmentation on Point Clouds. CVPR, 2020. 2

    2020

  69. [77]

    SqueezeSeg: Convolutional Neural Nets with Recurrent CRF for Real-Time Road-Object Segmentation from 3D Li- DAR Point Cloud.ICRA, 2018

    Bichen Wu, Alvin Wan, Xiangyu Yue, and Kurt Keutzer. SqueezeSeg: Convolutional Neural Nets with Recurrent CRF for Real-Time Road-Object Segmentation from 3D Li- DAR Point Cloud.ICRA, 2018. 1, 2

    2018

  70. [78]

    PointConv: Deep Convolutional Networks on 3D Point Clouds.CVPR, 2019

    Wenxuan Wu, Zhongang Qi, and Li Fuxin. PointConv: Deep Convolutional Networks on 3D Point Clouds.CVPR, 2019. 2

    2019

  71. [79]

    Dual Adaptive Transformations for Weakly Su- pervised Point Cloud Segmentation.ECCV, 2022

    Zhonghua Wu, Yicheng Wu, Guosheng Lin, Jianfei Cai, and Chen Qian. Dual Adaptive Transformations for Weakly Su- pervised Point Cloud Segmentation.ECCV, 2022. 2

    2022

  72. [80]

    3D Open-V ocabulary Panoptic Segmentation with 2D-3D Vision-Language Distil- lation.ECCV, 2024

    Zihao Xiao, Longlong Jing, Shangxuan Wu, Alex Zihao Zhu, Jingwei Ji, Chiyu Max Jiang, Wei-chih Hung, Thomas Funkhouser, and Weicheng Kuo. 3D Open-V ocabulary Panoptic Segmentation with 2D-3D Vision-Language Distil- lation.ECCV, 2024. 2

    2024

  73. [81]

    PointContrast: Unsupervised Pre- training for 3D Point Cloud Understanding.ECCV, 2020

    Saining Xie, Jiatao Gu, Demi Guo, Charles R Qi, Leonidas Guibas, and Or Litany. PointContrast: Unsupervised Pre- training for 3D Point Cloud Understanding.ECCV, 2020. 2

    2020

  74. [82]

    PDF: A Probability-Driven Framework for Open World 3D Point Cloud Semantic Seg- mentation.CVPR, 2024

    Jinfeng Xu, Siyuan Yang, Xianzhi Li, Yuan Tang, Yixue Hao, Long Hu, and Min Chen. PDF: A Probability-Driven Framework for Open World 3D Point Cloud Semantic Seg- mentation.CVPR, 2024. 2

    2024

  75. [83]

    Dual- level Adaptive Self-Labeling for Novel Class Discovery in Point Cloud Segmentation.ECCV, 2024

    Ruijie Xu, Chuyu Zhang, Hui Ren, and Xuming He. Dual- level Adaptive Self-Labeling for Novel Class Discovery in Point Cloud Segmentation.ECCV, 2024

    2024

  76. [84]

    RegionPLC: Regional Point-Language Contrastive Learning for Open-World 3D Scene Understanding.CVPR, 2024

    Jihan Yang, Runyu Ding, Zhe Wang, and Xiaojuan Qi. RegionPLC: Regional Point-Language Contrastive Learning for Open-World 3D Scene Understanding.CVPR, 2024

    2024

  77. [85]

    SAI3D: Segment Any Instance in 3D Scenes.CVPR, 2024

    Yingda Yin, Yuzheng Liu, Yang Xiao, Daniel Cohen-Or, Jingwei Huang, and Baoquan Chen. SAI3D: Segment Any Instance in 3D Scenes.CVPR, 2024. 2

    2024

  78. [86]

    TransFGU: A Top-down Approach to Fine-Grained Unsupervised Semantic Segmen- tation.ECCV, 2022

    Zhaoyuan Yin, Pichao Wang, Fan Wang, Xianzhe Xu, Han- ling Zhang, Hao Li, and Rong Jin. TransFGU: A Top-down Approach to Fine-Grained Unsupervised Semantic Segmen- tation.ECCV, 2022. 2

    2022

  79. [87]

    Unsupervised Se- mantic Segmentation with Self-supervised Object-Centric Representations.ICLR, 2023

    Andrii Zadaianchuk, Matthaeus Kleindessner, Yi Zhu, Francesco Locatello, and Thomas Brox. Unsupervised Se- mantic Segmentation with Self-supervised Object-Centric Representations.ICLR, 2023. 2

    2023

  80. [88]

    Perturbed Self-Distillation: Weakly Supervised Large-Scale Point Cloud Semantic Seg- mentation.ICCV, 2021

    Yachao Zhang, Yanyun Qu, Yuan Xie, Zonghao Li, Shan- shan Zheng, and Cuihua Li. Perturbed Self-Distillation: Weakly Supervised Large-Scale Point Cloud Semantic Seg- mentation.ICCV, 2021. 2

    2021

  81. [89]

    Self-Supervised Pretraining of 3D Features on any Point-Cloud.ICCV, 2021

    Zaiwei Zhang, Rohit Girdhar, Armand Joulin, and Ishan Misra. Self-Supervised Pretraining of 3D Features on any Point-Cloud.ICCV, 2021. 2

    2021

  82. [90]

    Self-Supervised Pre- training for Large-Scale Point Clouds.NeurIPS, 2022

    Zaiwei Zhang, Min Bai, and Erran Li. Self-Supervised Pre- training for Large-Scale Point Clouds.NeurIPS, 2022. 2

    2022

  83. [91]

    GrowSP: Unsupervised Semantic Segmentation of 3D Point Clouds

    Zihui Zhang, Bo Yang, Bing Wang, and Bo Li. GrowSP: Unsupervised Semantic Segmentation of 3D Point Clouds. CVPR, 2023. 1, 2, 3, 4, 5, 6, 7, 12, 13, 14, 15, 16

    2023

  84. [92]

    Unsupervised seepage segmen- tation pipeline based on point cloud projection with large vi- sion model.Tunnelling and Underground Space Technology,

    Zhaoxiang Zhang, Ankang Ji, Zhuan Xia, Limao Zhang, Yuelei Xu, and Qing Zhou. Unsupervised seepage segmen- tation pipeline based on point cloud projection with large vi- sion model.Tunnelling and Underground Space Technology,

  85. [93]

    Point Transformer.ICCV, 2021

    Hengshuang Zhao, Li Jiang, Jiaya Jia, Philip Torr, and Vladlen Koltun. Point Transformer.ICCV, 2021. 2

    2021

  86. [94]

    Extract Free Dense Labels from CLIP.ECCV, 2022

    Chong Zhou, Chen Change Loy, and Bo Dai. Extract Free Dense Labels from CLIP.ECCV, 2022. 2

    2022

  87. [95]

    Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR Seg- mentation.CVPR, 2021

    Xinge Zhu, Hui Zhou, Tai Wang, Fangzhou Hong, Yuexin Ma, Wei Li, Hongsheng Li, and Dahua Lin. Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR Seg- mentation.CVPR, 2021. 2, 6, 16 Appendix A. Initial Superpoint Generation In this section, we describe the detailed ...

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.