Pith. sign in

REVIEW 5 major objections 6 minor 68 references

Point Cloud Understanding via Attention-Driven Contrastive Learning

T0 review · 5 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 The paper claims that deliberately hiding the patches a point-cloud Transformer currently attends to most, and then training the masked and original views to agree, improves 3D representation learning and robustness.

desk verdict Useful training recipe; the 'attention-driven' narrative is under-supported by the evidence. read the letter →

arxiv 2411.14744 v1 pith:CAIMIZDU submitted 2024-11-22 cs.CV cs.AIcs.LG

Yi Wang , Jiaze Wang , Ziyu Guo , Renrui Zhang , Donghao Zhou , Guangyong Chen , Anfeng Liu , Pheng-Ann Heng This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords pointcloudunderstandingcontrastivelearningself-supervisedattention-drivenmaskingTransformer3Dclassificationrobustnessrepresentation
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 sets out to fix a blind spot in Transformer-based point-cloud models: they concentrate on a few high-attention patches, which the authors argue makes them fragile to noise and weak at capturing global structure. PointACL attacks this by dynamically masking the patches the model currently attends to most and training the model to keep the global features of the masked and original point clouds aligned through a contrastive loss, while retaining the original pre-training objective. The authors report that this recipe, dropped into Point-BERT, Point-MAE, or PointGPT-S, improves accuracy on ScanObjectNN, ModelNet40, and ShapeNetPart, sets new few-shot benchmarks, and sharply boosts robustness to Gaussian noise, rotation, scaling, and point dropout. If true, this is a cheap, plug-in improvement to existing 3D pre-training pipelines that adds no parameters.

What carries the argument

The load-bearing object is the per-patch significance score $S_j = A_{1,j}\|V_j\| \,/\, \sum_i A_{1,i}\|V_i\|$, which combines the global-token attention weight with the value norm to estimate how much each patch contributes to the global feature. From this score the method builds a dynamic masking probability $p_{\mathrm{dy}} = \log(\mathrm{Softmax}(S/\tau_{\mathrm{pro}})) - \log(-\log \varepsilon)$, applies a Top-K selection to mask the currently most-attended patches, and then runs a symmetric contrastive loss aligning the global features of the masked and unmasked views. This machinery forces the encoder to keep representing the whole object even when its preferred patches are removed, while the original reconstruction or generation loss preserves task-specific learning.

What would settle it

Remove the top-scoring patches and measure the drop in the model's reconstruction or classification accuracy: if the significance score $S_j$ identifies true dependency, removing top-$S$ patches must degrade performance far more than removing randomly chosen or low-$S$ patches, and this counterfactual is not run in the paper. A second direct check is to train a PointACL variant with random masking under the same epochs and budget; if its accuracy and robustness match the attention-driven version, the attention signal is not doing the claimed work.

Watch

Extended reading notes

Core claim

PointACL claims that the over-concentration of self-attention is itself a trainable weakness. The method reads the first row of the attention matrix, weights it by the norm of the value vectors to form a per-patch significance score, and uses a Gumbel-style dynamic masking probability to replace the most significant patches with mask tokens. The masked point cloud and the original are passed through the same encoder, a symmetric contrastive loss aligns their global features, and the total loss adds this to the backbone's original reconstruction or generation loss. The paper reports consistent gains over the same backbones on object classification, part segmentation, and few-shot classification, with the largest margins under heavy Gaussian noise.

Load-bearing premise

The method hinges on the assumption that the significance score computed from attention weights and value norms really pinpoints the patches the model depends on, so masking those patches forces the model to learn from neglected regions; if that score is unreliable, the whole procedure is just a stochastic masking plus contrastive loss whose robustness gains would need another explanation.

Editorial extensions

If this is right

  • PointACL can be dropped into existing Transformer-based point-cloud pre-training, such as Point-MAE and PointGPT, adding no parameters and improving downstream accuracy on classification, part segmentation, and few-shot tasks.
  • Models trained with PointACL are more robust to Gaussian noise, rotation, scaling, and point dropout than the same backbone trained alone, with the largest gains under strong noise.
  • Attention maps become more evenly distributed across patches, supporting the claimed shift from relying on a few salient regions to understanding global structure.
  • The reported training-cost analysis indicates the gains are not merely the result of longer training, since improvement persists when the same number of epochs is used.
  • The method is presented as architecture-agnostic, with consistent gains shown on Point-BERT, Point-MAE, and PointGPT-S backbones.

Reading between the lines

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

  • Beyond the paper, the same 'mask what the model is confident in, then align views' recipe could transfer to other sparse 3D inputs such as LiDAR scans or RGB-D scenes, where the most salient regions are also often the most sensor-fragile.
  • A reader could test whether the significance score truly tracks dependency by removing top-scoring patches and measuring the drop in reconstruction or classification accuracy; the paper does not run this counterfactual, so the mechanism behind the gains remains open to a direct check.
  • The dynamic masking probability uses fixed temperature and mask ratio, so an annealed schedule that gradually broadens or sharpens masking over training is a natural untested extension.
  • The contrastive view here is a masked and an unmasked version of the same object; adding a second independently masked view or a cross-modal view would test whether the alignment benefit comes from view diversity rather than masking alone.
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

5 major / 6 minor

Summary. The paper presents PointACL, a self-supervised pretraining framework for 3D point cloud Transformers. During pretraining, PointACL computes per-patch significance scores S_j from the first row of the self-attention matrix and the norms of value vectors (Eq. 2), uses a Gumbel-perturbed score (Eq. 3) to mask the highest-scoring patches, and aligns the global features of the masked and unmasked branches with a contrastive loss (Eq. 5) combined with the original reconstruction/generation loss (Eq. 6). The authors evaluate PointACL on top of Point-BERT, Point-MAE, and PointGPT-S on ScanObjectNN, ModelNet40, ShapeNetPart, and few-shot classification, reporting improved accuracy and robustness, and they provide component ablations, hyperparameter studies, and a training-cost analysis.

Significance. If the reported gains are reproducible, PointACL would be a simple, backbone-agnostic improvement to existing point cloud self-supervised pipelines, with a plausible mechanism (reducing over-reliance on a few patches). The manuscript deserves credit for including component ablations (Table 4), an equal-training-time comparison (Table 6), and hyperparameter sensitivity analysis (Table 8). However, the central mechanism is under-validated, and several headline claims rest on small or single-run differences. The omission of a direct comparison with Contrast with Reconstruct (ref [35]) further weakens the state-of-the-art claim. The paper is a plausible candidate for publication after substantial revision, but the load-bearing assumptions need to be tested and the evidence base strengthened.

major comments (5)
  1. [§3.1, Eq. (2)] The validity of S_j as a measure of patch importance is the load-bearing assumption of the paper, but it is never tested. The first row of one attention matrix times the value norm is not an established causal importance measure, and the authors do not specify which transformer block or layer is used. The paper's own ablation (Table 4) shows that the attention-driven choice adds only +0.2 and +0.7 points over Random Mask when both use Lcontra, and the Low-Attention Mask is only slightly worse. To support the mechanism, please (i) report the correlation between S_j and the change in the global feature (or downstream accuracy) when patch j is removed, and (ii) provide repeated-run means and standard deviations for the mask-strategy comparisons in Table 4, so the reader can judge whether the attention-specific component is significant.
  2. [§4.3, Table 5 and Figure 3] The robustness claims are based on single runs. Gains under Gaussian noise are large and inconsistent across backbones (e.g., +13.4 for Point-MAE on OBJ-BG at sigma=0.03 but +1.9 for PointGPT-S on OBJ-ONLY), and Figure 3 shows averaged curves without error bands. Moreover, no robustness ablation compares Random Mask + Lcontra against High-Attention Mask + Lcontra under noise, so the specific contribution of the attention-driven mask to robustness is not established. Please add repeated trials and the corresponding mask-strategy ablation under noise.
  3. [Tables 1-3 and Section 2] Contrast with Reconstruct (Qi et al., 2023, ref [35]) is discussed in the related work but never included in the comparisons. CwR is the closest existing method in spirit (combining a generative pretraining loss with a contrastive objective for 3D representation learning), and its omission makes the state-of-the-art claim in the abstract and Section 4.2 difficult to evaluate. Please add CwR to the main comparison tables (where its backbone is compatible) and discuss the differences.
  4. [Table 2] The few-shot gains are within one standard deviation of the corresponding baselines for nearly all settings (e.g., PointGPT-S 5-way 10-shot: 96.8±2.0 vs 97.1±2.3; Point-MAE 5-way 10-shot: 96.3±2.5 vs 96.7±2.7; 10-way 10-shot: 92.6±4.1 vs 92.8±4.0). Therefore the statement in the abstract that PointACL 'sets new benchmarks across all evaluation tasks' is not supported. Please report paired significance tests or modestly rephrase the claim to reflect the observed overlap.
  5. [§6.2/Table 6] The equal-training-time analysis is not fully clean. The columns labeled 'Finetune Epoch' introduce PointACL during fine-tuning, which is a different protocol from the main results (where PointACL is only applied during pretraining and the backbone is fine-tuned normally). It is unclear whether the 300/600 fine-tuning numbers use the same loss and masking as the pretraining pipeline. Please specify the exact protocol used for these rows and, if the fine-tuning variant differs, report the baseline under the same protocol.
minor comments (6)
  1. [Eq. (3)] Equation (3) uses ε without defining it; the text says a perturbation probability from U[0,1], but the expression -log(-log ε) is a Gumbel noise, and pdy is not a probability. Please define ε and rename the quantity (e.g., 'perturbation score').
  2. [Eq. (1)] The attention matrix A in Eq. (1) is not specified: which transformer block/layer and which head(s) are used? The text says 'for each head separately... sum over all heads', but not the layer(s).
  3. [Eq. (5)] The projection that produces H_m and H_s from F_m and F_s is not described. Please specify the projection head architecture and whether it is discarded after pretraining.
  4. [Figure 2 caption] Figure 2 caption says the masking module selects 'less activated patches', while the text and Section 3.1 say it masks high-attention patches; the caption appears to be the opposite of the method.
  5. [Throughout] Typos and formatting errors: 'decmidrules' in Figure 3/5 captions, 'decmidrule' in Section 6.4, 'Robutness Analysis' and 'Limatation Analysis' headings, 'Unversity' in the affiliation line, and inconsistent use of 'dec midrule' in the appendix.
  6. [Table 4] Table 4 does not state whether the High-Attention and Low-Attention masks use the fixed or dynamic probability; please clarify, since Table 7 shows the distinction matters.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: PointACL's components are defined from model outputs and evaluated on external benchmarks; the unvalidated significance-score assumption is a correctness risk, not a circular reduction.

full rationale

PointACL's derivation chain is self-contained rather than circular. Equation (2) defines a significance score from the model's own attention row A_{1,j} and value norms ||V_j||; Equations (3)-(4) convert that score into a dynamic masking probability and a Top-K mask; Equations (5)-(6) add a contrastive alignment loss between standard and masked branches. None of these steps defines the claimed target (classification accuracy, part segmentation mIoU, robustness under noise) in terms of the method's own inputs or fitted constants. The reported gains are measured on ScanObjectNN, ModelNet40, and ShapeNetPart, which are external benchmarks not used to fit Equation (2) or the masking rule. Hyperparameters such as tau_pro, lambda, and mask ratio are selected on ScanObjectNN, but the central claims are also tested on ModelNet40 and ShapeNetPart, and Table 6 compares against same-training-time baselines, so no fitted quantity is renamed as a prediction. The only apparent self-citation is Point-M2AE [59] in related work, whose authors overlap with the present paper; that citation is not load-bearing because PointACL is built on Point-MAE [32] and PointGPT [5], not on Point-M2AE. Appendix 6.6 ('Limatation Analysis') asserts only generic generalizability limitations and does not claim or exhibit a circular step. The skeptic's concern that the significance score in Equation (2) is never validated as a measure of true causal patch reliance, and that Table 4 shows only a small margin over random masking, is a legitimate correctness and robustness concern, but it is not a circularity: the paper does not derive its results from the assumption that S_j is a faithful explanation, and its empirical claims are evaluated on data external to the pretraining objective. Accordingly, no step reduces by construction to its own inputs, and the appropriate circularity score is 0.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

All hyperparameters that affect the central results are tuned on the evaluation datasets themselves (ScanObjectNN), which introduces selection bias but not circularity. The two behavioral assumptions (attention equals dependency, and masking teaches robustness) are empirical hypotheses rather than derived statements.

free parameters (4)
  • tau_pro (probability temperature) = 0.5
    Chosen by ablating values 0.3 to 0.9 on ScanObjectNN OBJ-BG and OBJ-ONLY (Table 8a). The default PointACL setting uses 0.5.
  • lambda (contrastive loss weight) = 0.6
    Chosen by ablating 0.4 to 1.0 on ScanObjectNN (Table 8b); 0.6 gives the best OBJ-BG and OBJ-ONLY accuracies.
  • Mask ratio k = 0.6
    Ablated in Table 7; dynamic mask ratio 0.6 yields 92.3 and 91.6, the best results. Ratio 0.8 degrades performance.
  • Contrastive temperature tau_sim = not reported
    Used in Eq. (5) for InfoNCE; the paper does not state the value used in experiments.
assumptions (3)
  • domain assumption The significance score S_j = A_{1,j} ||V_j|| / sum_i A_{1,i} ||V_i|| reflects the importance of patch j to the global representation.
    This is assumed in Sec. 3.1 (Eq. 2) and is the basis for selecting patches to mask. If attention norms do not track true dependency, the masking is arbitrary.
  • domain assumption Masking high-attention patches during pretraining and aligning the masked branch with the unmasked branch improves robustness and global understanding at inference.
    This is the core hypothesis tested by the experiments; it is not derived from first principles and is validated only empirically on specific datasets.
  • domain assumption The original pretraining losses of Point-MAE (reconstruction) and PointGPT (generation) are retained and their backbones are used without masking at test time.
    The method assumes that adding the contrastive loss to the original loss does not degrade task-specific capabilities; Sec. 3.2. Ablations support this assumption only for the studied settings.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Point Cloud Understanding via Attention-Driven Contrastive Learning." pith.science (2026). https://pith.science/paper/CAIMIZDU

@misc{pith2026241114744,
  author       = {Pith},
  title        = {Pith review of: Point Cloud Understanding via Attention-Driven Contrastive Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CAIMIZDU}},
  note         = {Machine review of arXiv:2411.14744}
}
read the original abstract

Recently Transformer-based models have advanced point cloud understanding by leveraging self-attention mechanisms, however, these methods often overlook latent information in less prominent regions, leading to increased sensitivity to perturbations and limited global comprehension. To solve this issue, we introduce PointACL, an attention-driven contrastive learning framework designed to address these limitations. Our method employs an attention-driven dynamic masking strategy that guides the model to focus on under-attended regions, enhancing the understanding of global structures within the point cloud. Then we combine the original pre-training loss with a contrastive learning loss, improving feature discrimination and generalization. Extensive experiments validate the effectiveness of PointACL, as it achieves state-of-the-art performance across a variety of 3D understanding tasks, including object classification, part segmentation, and few-shot learning. Specifically, when integrated with different Transformer backbones like Point-MAE and PointGPT, PointACL demonstrates improved performance on datasets such as ScanObjectNN, ModelNet40, and ShapeNetPart. This highlights its superior capability in capturing both global and local features, as well as its enhanced robustness against perturbations and incomplete data.

Figures

Figures reproduced from arXiv: 2411.14744 by the authors.

Figure 1
Figure 1. Illustration of PointACL’s Advantages. Point-MAE is employed as the backbone of our proposed PointACL. Left: PointACL emphasizes extracting global information from a greater number of patches. Right: PointACL demonstrates greater robustness than previous methods. to focus on under-attended regions. Specifically, we con￾struct a dynamic masking probability based on the latest self-attention significance scores, prior… view at source ↗
Figure 2
Figure 2. Overview of the PointACL Framework. PointACL consists of two branches that share the same weights: a standard mode branch and a masked mode branch. An attention-driven dynamic masking module generates a masked point cloud by selecting less activated patches from the output of the standard mode branch. Both branches process their respective inputs through the shared Transformer blocks to obtain latent representations… view at source ↗
Figure 4
Figure 4. Attention visualization of PointACL with Point-MAE and PointGPT. Patches with high attention are closer to red, while patches with low attention are closer to blue. Point-MAE is em￾ployed as the backbone of our proposed PointACL. icantly enhances the understanding of global structures in point clouds, leading to notable improvements across various tasks, including object classification, part seg￾mentation, and few-s… view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Gaussian noise analysis on ScanObjectNN. While the performance of existing methods decmidrules sharply with in￾creasing Gaussian noise, this issue is mitigated by incorporating PointACL. Notably, when Point-MAE is used as the backbone net￾work, our PointACL significant…
Figure 6
Figure 6. Figure 6: Feature distribution visualization on ModelNet40. Top: An overview of the evolution of feature distributions across all 40 classes. Bottom: Detailed depiction of the evolution of feature distributions for selected typical classes. valuable. Its strong performance under…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 39 canonical work pages

  1. [35]

    Contrast with reconstruct: Contrastive 3d representation learning guided by generative pretraining

    Zekun Qi, Runpei Dong, Guofan Fan, Zheng Ge, Xiangyu Zhang, Kaisheng Ma, and Li Yi. Contrast with reconstruct: Contrastive 3d representation learning guided by generative pretraining. In International Conference on Machine Learn- ing, pages 28223–28243. PMLR, 2023. 3

    work page 2023

  2. [1]

    Maskclr: Attention-guided contrastive learning for robust action representation learning

    Mohamed Abdelfattah, Mariam Hassan, and Alexandre Alahi. Maskclr: Attention-guided contrastive learning for robust action representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18678–18687, 2024. 2

    work page 2024

  3. [2]

    Learning representations and generative models for 3d point clouds

    Panos Achlioptas, Olga Diamanti, Ioannis Mitliagkas, and Leonidas Guibas. Learning representations and generative models for 3d point clouds. In International conference on machine learning, pages 40–49. PMLR, 2018. 3

    work page 2018

  4. [3]

    An overview of augmented reality

    Fabio Arena, Mario Collotta, Giovanni Pau, and Francesco Termine. An overview of augmented reality. Computers, 11 (2):28, 2022. 1

    2022

  5. [4]

    Beit: Bert pre-training of image transformers

    Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei. Beit: Bert pre-training of image transformers. arXiv preprint arXiv:2106.08254, 2021. 2

    arXiv 2021

  6. [5]

    Pointgpt: Auto-regressively generative pre- training from point clouds

    Guangyan Chen, Meiling Wang, Yi Yang, Kai Yu, Li Yuan, and Yufeng Yue. Pointgpt: Auto-regressively generative pre- training from point clouds. Advances in Neural Information Processing Systems, 36, 2024. 1, 3, 5, 6, 7, 11

    work page 2024

  7. [6]

    3d point cloud processing and learning for autonomous driving: Impacting map cre- ation, localization, and perception

    Siheng Chen, Baoan Liu, Chen Feng, Carlos Vallespi- Gonzalez, and Carl Wellington. 3d point cloud processing and learning for autonomous driving: Impacting map cre- ation, localization, and perception. IEEE Signal Processing Magazine, 38(1):68–86, 2020. 1

    2020

  8. [7]

    Multi-view 3d object detection network for autonomous driving

    Xiaozhi Chen, Huimin Ma, Ji Wan, Bo Li, and Tian Xia. Multi-view 3d object detection network for autonomous driving. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 1907–1915, 2017. 1 Table 5. Robustness analysis. We report the classification accuracy (%) with four noisy environments: Gaussian noise, rotation, scali...

    work page 1907

Show all 68 references
  1. [8]

    Pra-net: Point relation-aware network for 3d point cloud analysis

    Silin Cheng, Xiwu Chen, Xinwei He, Zhe Liu, and Xiang Bai. Pra-net: Point relation-aware network for 3d point cloud analysis. IEEE Transactions on Image Processing, 30:4436– 4448, 2021. 6

    2021

  2. [9]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018. 1, 2

    2018 arXiv

  3. [10]

    Autoen- coders as cross-modal teachers: Can pretrained 2d image transformers help 3d representation learning? arXiv preprint arXiv:2212.08320, 2022

    Runpei Dong, Zekun Qi, Linfeng Zhang, Junbo Zhang, Jian- jian Sun, Zheng Ge, Li Yi, and Kaisheng Ma. Autoen- coders as cross-modal teachers: Can pretrained 2d image transformers help 3d representation learning? arXiv preprint arXiv:2212.08320, 2022. 3

    2022 arXiv

  4. [11]

    Why does unsupervised pre-training help deep learning? In Proceedings of the thirteenth international con- ference on artificial intelligence and statistics , pages 201–

    Dumitru Erhan, Aaron Courville, Yoshua Bengio, and Pas- cal Vincent. Why does unsupervised pre-training help deep learning? In Proceedings of the thirteenth international con- ference on artificial intelligence and statistics , pages 201–

  5. [12]

    Shape2scene: 3d scene representation learning through pre- training on shape data

    Tuo Feng, Wenguan Wang, Ruijie Quan, and Yi Yang. Shape2scene: 3d scene representation learning through pre- training on shape data. arXiv preprint arXiv:2407.10200 ,

    arXiv

  6. [13]

    Point cloud interaction and ma- nipulation in virtual reality

    Daniel Garrido, Rui Rodrigues, A Augusto Sousa, Joao Ja- cob, and Daniel Castro Silva. Point cloud interaction and ma- nipulation in virtual reality. In 2021 5th International Con- ference on Artificial Intelligence and Virtual Reality (AIVR), pages 15–20, 2021. 1

    2021

  7. [14]

    Generative adversarial networks

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial networks. Commu- nications of the ACM, 63(11):139–144, 2020. 2

    2020

  8. [15]

    Mvtn: Multi-view transformation network for 3d shape recognition

    Abdullah Hamdi, Silvio Giancola, and Bernard Ghanem. Mvtn: Multi-view transformation network for 3d shape recognition. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 1–11, 2021. 6

    2021

  9. [16]

    Mamba3d: Enhancing local features for 3d point cloud anal- ysis via state space model.arXiv preprint arXiv:2404.14966,

    Xu Han, Yuan Tang, Zhaoxuan Wang, and Xianzhi Li. Mamba3d: Enhancing local features for 3d point cloud anal- ysis via state space model.arXiv preprint arXiv:2404.14966,

    arXiv

  10. [17]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll´ar, and Ross Girshick. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 16000– 16009, 2022. 1, 2

    2022

  11. [18]

    Atten- tion discriminant sampling for point clouds

    Cheng-Yao Hong, Yu-Ying Chou, and Tyng-Luh Liu. Atten- tion discriminant sampling for point clouds. In Proceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 14429–14440, 2023. 6

    2023

  12. [19]

    Clip2point: Transfer clip to point cloud classifica- tion with image-depth pre-training

    Tianyu Huang, Bowen Dong, Yunhan Yang, Xiaoshui Huang, Rynson WH Lau, Wanli Ouyang, and Wangmeng Zuo. Clip2point: Transfer clip to point cloud classifica- tion with image-depth pre-training. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 221...

    2023

  13. [20]

    Self-supervised modal and view invariant fea- ture learning

    Longlong Jing, Yucheng Chen, Ling Zhang, Mingyi He, and Yingli Tian. Self-supervised modal and view invariant fea- ture learning. arXiv preprint arXiv:2005.14169, 2020. 3

    2005 arXiv

  14. [21]

    So-net: Self- organizing network for point cloud analysis

    Jiaxin Li, Ben M Chen, and Gim Hee Lee. So-net: Self- organizing network for point cloud analysis. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 9397–9406, 2018. 3

    2018

  15. [22]

    Pointcnn: Convolution on x-transformed points

    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. 6

    2018

  16. [23]

    General point model pretrain- ing with autoencoding and autoregressive

    Zhe Li, Zhangyang Gao, Cheng Tan, Bocheng Ren, Lau- rence T Yang, and Stan Z Li. General point model pretrain- ing with autoencoding and autoregressive. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20954–20964, 2024. 6, 7

    2024

  17. [24]

    Pointmamba: A simple state space model for point cloud analysis

    Dingkang Liang, Xin Zhou, Wei Xu, Xingkui Zhu, Zhikang Zou, Xiaoqing Ye, Xiao Tan, and Xiang Bai. Pointmamba: A simple state space model for point cloud analysis. In Ad- vances in Neural Information Processing Systems , 2024. 2, 6, 7

    2024

  18. [25]

    Masked dis- crimination for self-supervised learning on point clouds

    Haotian Liu, Mu Cai, and Yong Jae Lee. Masked dis- crimination for self-supervised learning on point clouds. In European Conference on Computer Vision, pages 657–675. Springer, 2022. 6, 7

    2022

  19. [26]

    Point mamba: A novel point cloud backbone based on state space model with octree-based ordering strategy

    Jiuming Liu, Ruiji Yu, Yian Wang, Yu Zheng, Tianchen Deng, Weicai Ye, and Hesheng Wang. Point mamba: A novel point cloud backbone based on state space model with octree-based ordering strategy. arXiv preprint arXiv:2403.06467, 2024. 3

    2024 arXiv

  20. [27]

    Deep image translation with an affinity-based change prior for un- supervised multimodal change detection

    Luigi Tommaso Luppino, Michael Kampffmeyer, Fil- ippo Maria Bianchi, Gabriele Moser, Sebastiano Bruno Ser- pico, Robert Jenssen, and Stian Normann Anfinsen. Deep image translation with an affinity-based change prior for un- supervised multimodal change detection. IEEE Transact...

    2021

  21. [28]

    Rethinking network design and local geometry in point cloud: A simple residual mlp framework

    Xu Ma, Can Qin, Haoxuan You, Haoxi Ran, and Yun Fu. Rethinking network design and local geometry in point cloud: A simple residual mlp framework. arXiv preprint arXiv:2202.07123, 2022. 6

    2022 arXiv

  22. [29]

    V oxel-mae: Masked autoencoders for pre-training large-scale point clouds

    Chen Min, Dawei Zhao, Liang Xiao, Yiming Nie, and Bin Dai. V oxel-mae: Masked autoencoders for pre-training large-scale point clouds. arXiv preprint arXiv:2206.09900, 3, 2022. 3

    2022 arXiv

  23. [30]

    Self-supervised learning of pretext-invariant representations

    Ishan Misra and Laurens van der Maaten. Self-supervised learning of pretext-invariant representations. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6707–6717, 2020. 2

    2020

  24. [31]

    From image collections to point clouds with self-supervised shape and pose networks

    KL Navaneet, Ansu Mathew, Shashank Kashyap, Wei-Chih Hung, Varun Jampani, and R Venkatesh Babu. From image collections to point clouds with self-supervised shape and pose networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 113...

    2020

  25. [32]

    Masked autoencoders for point cloud self-supervised learning

    Yatian Pang, Wenxiao Wang, Francis EH Tay, Wei Liu, Yonghong Tian, and Li Yuan. Masked autoencoders for point cloud self-supervised learning. In European conference on computer vision, pages 604–621. Springer, 2022. 1, 3, 5, 6, 7, 11

    2022

  26. [33]

    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. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 652–660,

  27. [34]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. Advances in neural information processing systems, 30, 2017. 6, 7

    2017

  28. [36]

    Shapellm: Universal 3d object understanding for embodied interaction

    Zekun Qi, Runpei Dong, Shaochen Zhang, Haoran Geng, Chunrui Han, Zheng Ge, Li Yi, and Kaisheng Ma. Shapellm: Universal 3d object understanding for embodied interaction. In European Conference on Computer Vision , pages 214–

  29. [37]

    Pointnext: Revisiting pointnet++ with improved training and scaling strategies

    Guocheng Qian, Yuchen Li, Houwen Peng, Jinjie Mai, Hasan Hammoud, Mohamed Elhoseiny, and Bernard Ghanem. Pointnext: Revisiting pointnet++ with improved training and scaling strategies. Advances in neural informa- tion processing systems, 35:23192–23204, 2022. 3, 6

    2022

  30. [38]

    Spatiotempo- ral contrastive video representation learning

    Rui Qian, Tianjian Meng, Boqing Gong, Ming-Hsuan Yang, Huisheng Wang, Serge Belongie, and Yin Cui. Spatiotempo- ral contrastive video representation learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6964–6974, 2021. 2

    2021

  31. [39]

    Improving language understanding by gener- ative pre-training

    Alec Radford. Improving language understanding by gener- ative pre-training. 2018. 2

    2018

  32. [40]

    Surface representa- tion for point clouds

    Haoxi Ran, Jun Liu, and Chengjie Wang. Surface representa- tion for point clouds. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 18942–18952, 2022. 6

    2022

  33. [41]

    Point-jepa: A joint em- bedding predictive architecture for self-supervised learning on point cloud

    Ayumu Saito and Jiju Poovvancheri. Point-jepa: A joint em- bedding predictive architecture for self-supervised learning on point cloud. arXiv preprint arXiv:2404.16432, 2024. 3

    2024 arXiv

  34. [42]

    Detecting formal thought disorder by deep contextualized word representations

    Justyna Sarzynska-Wawer, Aleksander Wawer, Aleksan- dra Pawlak, Julia Szymanowska, Izabela Stefaniak, Michal Jarkiewicz, and Lukasz Okruszek. Detecting formal thought disorder by deep contextualized word representations. Psy- chiatry Research, 304:114135, 2021. 2

    2021

  35. [43]

    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. Advances in Neural Information Processing Systems, 32, 2019. 3

    2019

  36. [44]

    Exploring 3d navigation: combining speed-coupled flying with orbiting

    Desney S Tan, George G Robertson, and Mary Czerwinski. Exploring 3d navigation: combining speed-coupled flying with orbiting. In Proceedings of the SIGCHI conference on Human factors in computing systems, pages 418–425, 2001. 1

    2001

  37. [45]

    Revisiting point cloud classification: A new benchmark dataset and classification model on real-world data

    Mikaela Angelina Uy, Quang-Hieu Pham, Binh-Son Hua, Thanh Nguyen, and Sai-Kit Yeung. Revisiting point cloud classification: A new benchmark dataset and classification model on real-world data. In Proceedings of the IEEE/CVF international conference on computer vision , pages 1...

    2019

  38. [46]

    Unsupervised point cloud pre-training via oc- clusion completion

    Hanchen Wang, Qi Liu, Xiangyu Yue, Joan Lasenby, and Matt J Kusner. Unsupervised point cloud pre-training via oc- clusion completion. In Proceedings of the IEEE/CVF inter- national conference on computer vision , pages 9782–9792,

  39. [47]

    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. 6, 7

    2019

  40. [48]

    P2p: Tuning pre-trained image models for point cloud analysis with point-to-pixel prompting

    Ziyi Wang, Xumin Yu, Yongming Rao, Jie Zhou, and Jiwen Lu. P2p: Tuning pre-trained image models for point cloud analysis with point-to-pixel prompting. Advances in neural information processing systems, 35:14388–14402, 2022. 3

    2022

  41. [49]

    Pointramba: A hybrid transformer-mamba framework for point cloud analysis

    Zicheng Wang, Zhenghao Chen, Yiming Wu, Zhen Zhao, Luping Zhou, and Dong Xu. Pointramba: A hybrid transformer-mamba framework for point cloud analysis. arXiv preprint arXiv:2405.15463, 2024. 3

    2024 arXiv

  42. [50]

    Point transformer v3: Simpler faster stronger

    Xiaoyang Wu, Li Jiang, Peng-Shuai Wang, Zhijian Liu, Xi- hui Liu, Yu Qiao, Wanli Ouyang, Tong He, and Hengshuang Zhao. Point transformer v3: Simpler faster stronger. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4840–4851, 2024. 3

    2024

  43. [51]

    3d shapenets: A deep representation for volumetric shapes

    Zhirong Wu, Shuran Song, Aditya Khosla, Fisher Yu, Lin- guang Zhang, Xiaoou Tang, and Jianxiong Xiao. 3d shapenets: A deep representation for volumetric shapes. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1912–1920, 2015. 5

    1912

  44. [52]

    Pointcontrast: Unsupervised pre- training for 3d point cloud understanding

    Saining Xie, Jiatao Gu, Demi Guo, Charles R Qi, Leonidas Guibas, and Or Litany. Pointcontrast: Unsupervised pre- training for 3d point cloud understanding. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part III 16 , pages...

    2020

  45. [53]

    Ulip: Learning a unified representation of language, images, and point clouds for 3d understanding

    Le Xue, Mingfei Gao, Chen Xing, Roberto Mart ´ın-Mart´ın, Jiajun Wu, Caiming Xiong, Ran Xu, Juan Carlos Niebles, and Silvio Savarese. Ulip: Learning a unified representation of language, images, and point clouds for 3d understanding. In Proceedings of the IEEE/CVF conference o...

    2023

  46. [54]

    Ulip-2: Towards scalable multimodal pre-training for 3d understanding

    Le Xue, Ning Yu, Shu Zhang, Artemis Panagopoulou, Jun- nan Li, Roberto Mart´ın-Mart´ın, Jiajun Wu, Caiming Xiong, Ran Xu, Juan Carlos Niebles, et al. Ulip-2: Towards scalable multimodal pre-training for 3d understanding. In Proceed- ings of the IEEE/CVF Conference on Computer ...

    2024

  47. [55]

    A scalable 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 Shef- fer, and Leonidas Guibas. A scalable active framework for region annotation in 3d shape collections.ACM Transactions on Graphics (ToG), 35(6):1–12, 2016. 5

    2016

  48. [56]

    Seq- gan: Sequence generative adversarial nets with policy gra- dient

    Lantao Yu, Weinan Zhang, Jun Wang, and Yong Yu. Seq- gan: Sequence generative adversarial nets with policy gra- dient. In Proceedings of the AAAI conference on artificial intelligence, 2017. 2

    2017

  49. [57]

    Point-bert: Pre-training 3d point cloud transformers with masked point modeling

    Xumin Yu, Lulu Tang, Yongming Rao, Tiejun Huang, Jie Zhou, and Jiwen Lu. Point-bert: Pre-training 3d point cloud transformers with masked point modeling. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 19313–19322, 2022. 1, 3, 6, 7

    2022

  50. [58]

    Towards compact 3d representations via point feature enhancement masked au- toencoders

    Yaohua Zha, Huizhen Ji, Jinmin Li, Rongsheng Li, Tao Dai, Bin Chen, Zhi Wang, and Shu-Tao Xia. Towards compact 3d representations via point feature enhancement masked au- toencoders. In Proceedings of the AAAI Conference on Arti- ficial Intelligence, pages 6962–6970, 2024. 3

    2024

  51. [59]

    Point-m2ae: multi-scale masked autoencoders for hierarchical point cloud pre-training

    Renrui Zhang, Ziyu Guo, Peng Gao, Rongyao Fang, Bin Zhao, Dong Wang, Yu Qiao, and Hongsheng Li. Point-m2ae: multi-scale masked autoencoders for hierarchical point cloud pre-training. Advances in neural information processing sys- tems, 35:27061–27074, 2022. 3, 6

    2022

  52. [60]

    Pointclip: Point cloud understanding by clip

    Renrui Zhang, Ziyu Guo, Wei Zhang, Kunchang Li, Xu- peng Miao, Bin Cui, Yu Qiao, Peng Gao, and Hongsheng Li. Pointclip: Point cloud understanding by clip. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8552–8562, 2022. 3

    2022

  53. [61]

    Learning 3d representations from 2d pre-trained models via image-to-point masked autoencoders

    Renrui Zhang, Liuhui Wang, Yu Qiao, Peng Gao, and Hong- sheng Li. Learning 3d representations from 2d pre-trained models via image-to-point masked autoencoders. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21769–21780, 2023. 3

    2023

  54. [62]

    Point could mamba: Point cloud learning via state space model

    Tao Zhang, Xiangtai Li, Haobo Yuan, Shunping Ji, and Shuicheng Yan. Point could mamba: Point cloud learning via state space model. arXiv preprint arXiv:2403.00762, 2024. 3

    2024 arXiv

  55. [63]

    Self-supervised pretraining of 3d features on any point-cloud

    Zaiwei Zhang, Rohit Girdhar, Armand Joulin, and Ishan Misra. Self-supervised pretraining of 3d features on any point-cloud. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 10252–10263, 2021. 3

    2021

  56. [64]

    Point cloud pre-training with diffusion models

    Xiao Zheng, Xiaoshui Huang, Guofeng Mei, Yuenan Hou, Zhaoyang Lyu, Bo Dai, Wanli Ouyang, and Yongshun Gong. Point cloud pre-training with diffusion models. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 22935–22945, 2024. 6

    2024

  57. [65]

    Point- clip v2: Prompting clip and gpt for powerful 3d open-world learning

    Xiangyang Zhu, Renrui Zhang, Bowei He, Ziyu Guo, Ziyao Zeng, Zipeng Qin, Shanghang Zhang, and Peng Gao. Point- clip v2: Prompting clip and gpt for powerful 3d open-world learning. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2639–2650, 2023. 3

    2023

  58. [66]

    3d-vista: Pre-trained transformer for 3d vision and text alignment

    Ziyu Zhu, Xiaojian Ma, Yixin Chen, Zhidong Deng, Siyuan Huang, and Qing Li. 3d-vista: Pre-trained transformer for 3d vision and text alignment. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2911– 2921, 2023. 2

    2023

  59. [68]

    Preliminary Transformer-based self-supervised learning

    Appendix 6.1. Preliminary Transformer-based self-supervised learning. Given a point cloud X ∈ RP ×3, we utilize Farthest Point Sam- pling (FPS) and K-Nearest Neighbors (KNN) algorithms to identify n center points C and their corresponding k near- est neighbors, forming n point...

  60. [208]

    JMLR Workshop and Conference Proceedings, 2010. 2

    2010

Pith tools

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