Pith. sign in

REVIEW 4 major objections 5 minor 50 references

Point-Selection Fine-Tuning Framework for Robust Point Cloud Classification

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Pith's one-line read PSFT, a point-selection fine-tuning framework, claims that retaining minimally influential points, prompting a frozen backbone, and Beta-gated feature filtering reduces corruption error on ModelNet-C and ModelNet40-C across four pre-trained

desk verdict A useful, honestly reported robustness recipe for 3D PEFT, but the point-selection mechanism is under-verified and the stats need tightening. read the letter →

arxiv 2607.19711 v1 pith:VX4GQBI3 submitted 2026-07-22 cs.CV

classification cs.CV
keywords pointcloudclassificationcorruptionrobustnessparameter-efficientfine-tuningprompttuningselectionfeaturefilteringfrozenbackbone3Dpre-trainedmodels
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 argues that full fine-tuning of 3D pre-trained models can harm corruption robustness by overwriting robust priors, while naive parameter-efficient tuning remains sensitive to corrupted points. It proposes PSFT, which ranks points by their influence on the pooled representation, discards the most influential ones (assumed to be outliers), adapts a frozen transformer via layer-wise prompts generated from the selected subset, and refines patch tokens with a bottleneck MLP blended through a Beta-gated residual connection. The central claim is that this coupling yields lower corruption error than full fine-tuning on ModelNet-C and ModelNet40-C for all four tested backbones (Point-BERT, Point-MAE, ULIP-2, Uni3D-B), and the best ScanObjectNN-C result among compared tuning strategies for the two strongest backbones, while using only 10.9% of full fine-tuning's trainable parameters. A sympathetic reader would care because robust, parameter-efficient adaptation could make large 3D foundation models practical in noisy perception settings, and because the paper's own ScanObjectNN-C numbers show the benefit is not universal: for Point-BERT and Point-MAE, PSFT worsens mCE relative to full fine-tuning.

What carries the argument

The load-bearing object is the point-wise influence score I_F(j) = Σ_k 1{ j = argmax_n X_f(n,k) }, counted over the pre-pooling feature map's channels, together with the entropy-guided retention rule S = round((1 - H_n(p))·N) that keeps the minimally influential points. The prompt generation branch, an EdgeConv-based network, turns the selected subset into per-layer prompt tokens injected as (T_cls, P_i, T_i) into each frozen transformer layer. The feature filter module is a bottleneck MLP (down-projection, GELU, dropout, up-projection) whose output is blended with the original token by x_final = (1-α)x_f + αx_o, with α ~ Beta(λ,λ) during training — a bounded, stochastic residual mixing that

What would settle it

On ModelNet-C with additive-Gaussian noise, measure the retained ratio S/N under increasing noise severity; if S/N stays near 1 (i.e., the entropy rule does not respond to corruption) while PSFT still lowers mCE relative to full fine-tuning, then the robustness gain comes from the prompt/filter modules rather than point selection, and the selection premise is not the active mechanism.

Watch

Extended reading notes

Core claim

PSFT's central discovery is that robustness during adaptation can be improved by not fine-tuning the backbone at all, instead feeding a frozen pre-trained model a deliberately pruned point cloud. The pruning is driven by a per-point influence score: for each feature channel before global pooling, the point with the maximum activation gets a vote, and the votes are aggregated; the entropy of the normalized vote distribution sets the retention ratio S/N = 1 - H_n, and the least influential S points are kept. On top of this, an EdgeConv-based branch predicts one learnable prompt token per transformer layer from the selected points, and a funnel MLP with Beta-sampled residual blending (x_final =

Load-bearing premise

The load-bearing premise, taken from prior refocusing work and adopted in Sec. 3.1, is that the points winning the most max-pooling channels are usually corrupted outliers, so that discarding the most influential points removes noise instead of informative structure; the paper tests this only indirectly through end-to-end benchmarks, and its ScanObjectNN-C results for Point-BERT and Point-MAE show the premise can fail on real-world scans.

Editorial extensions

If this is right

  • PSFT lowers mean corruption error on ModelNet-C for all four evaluated backbones and on ModelNet40-C lowers mean error rate for all four, with the largest gains on strong backbones (ULIP-2 mCE 0.530, Uni3D-B mER 14.2).
  • Robustness gains are compatible with data augmentation: adding WOLFMix further improves synthetic-benchmark results (best ModelNet-C mCE 0.465 with Uni3D-B), establishing complementarity.
  • Parameter efficiency is preserved: 2.4M trainable parameters (10.9% of full fine-tuning) suffice to beat full fine-tuning on synthetic corruptions.
  • Ablations attribute gains to all three modules, with selection alone reducing mCE substantially and selection plus feature filtering being the strongest pair on ScanObjectNN-C.
  • The transfer to real-world corruptions is backbone-dependent: on ScanObjectNN-C, PSFT beats full fine-tuning only for ULIP-2 and Uni3D-B, and degrades mCE for Point-BERT and Point-MAE.

Reading between the lines

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

  • If the refocusing premise is right, PSFT-style influence-based pruning could be used as a plug-in input filter for other PEFT methods (adapters, LoRA), since it is computed from the frozen backbone's own features and requires no extra supervision.
  • The entropy rule S = (1-H_n)N means that under corruptions that spread influence evenly (e.g., rotation, shear), selection retains nearly all points; the paper's per-corruption tables show PSFT improves even those categories, so either the prompt branch or the feature filter — not point pruning — is doing the work there. A per-corruption ablation would separate these contributions.
  • The paper's own ScanObjectNN-C results (Point-BERT mCE 0.931→0.970, Point-MAE 0.895→1.014) and the supplementary's finding that WOLFMix raises ScanObjectNN-C mCE by 0.204 on average suggest a dataset–augmentation mismatch: on cluttered real scans, dropping minimally influential points can remove useful geometry. A testable fix would be density-aware or class-conditional selection.
  • The Beta-gated blending with optimal test-time α≈0.5 suggests that the filter's value is its bounded correction of the frozen feature, not the stochasticity per se; fixing α=0.5 during training and testing would isolate the interpolation from the regularization effect.
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 / 5 minor

Summary. The paper proposes PSFT, a three-module robust fine-tuning framework for 3D point cloud classification: (i) a point selection module that drops points with high max-pooling influence, following the refocusing idea of [11]; (ii) a prompt generation branch that creates layer-wise prompts from the selected subset for a frozen backbone; and (iii) a feature filter module with Beta-gated residual blending to refine patch tokens. Experiments on ModelNet-C, ModelNet40-C, and ScanObjectNN-C across Point-BERT, Point-MAE, ULIP-2, and Uni3D-B report consistent mCE/mER improvements on the two synthetic corruption benchmarks and strong ScanObjectNN-C results for ULIP-2 and Uni3D-B, while using about 10.9% of the trainable parameters of full fine-tuning. The paper also provides ablations, sensitivity analyses, and public code.

Significance. If the empirical claims are reliable, PSFT is a useful contribution: it offers a parameter-efficient alternative to full fine-tuning that improves or at least preserves corruption robustness on common benchmarks. Strengths include the public code, systematic module ablations, evaluation across four substantially different backbones, and honest reporting of the mixed ScanObjectNN-C behavior. The main limitations are methodological: the selection mechanism's premise is not directly verified, test-set information is used to choose hyperparameters, and the reported aggregate gains come from single seeds without variance estimates. These issues currently limit confidence in the central explanatory claim, although they are addressable in revision.

major comments (4)
  1. [Sec. 3.1 (Eq. 1) and Sec. 3.2 (Eq. 6)] The influence score I_F(j) is computed from X_f, the final pre-pooling feature map, which requires a complete forward pass through the frozen backbone with all N points. The subsequent prompt-injected forward pass in Eq. (6) uses the selected subset, implying a second forward pass. The manuscript never states this explicitly, nor does it report the added inference cost. If the second pass indeed runs, the influence scores are stale with respect to the selected subset; if there is no second pass, then selection only affects the prompt/FFM branches and cannot 'suppress outliers' in the backbone's pooled representation as claimed. Please specify the exact data flow, including whether the patch encoder and transformer layers operate on the selected subset, and give the computational overhead.
  2. [Sec. 3.1 / Table 5] The central premise—that the most influential points are corrupted outliers—is adopted from [11] and is not tested on any corruption type or backbone used here. Table 5 shows that point selection alone accounts for a large part of the ULIP-2 ModelNet-C gain (mCE from 0.863 to 0.656), so this premise is load-bearing. Table 3 shows that on ScanObjectNN-C the same selection degrades clean OA and mCE for Point-BERT and Point-MAE, which is consistent with the possibility that informative structure is being discarded. Please provide a direct test, e.g., per-corruption overlap between the discarded points and known corrupted points, or a comparison of influence-based selection with random and inverse-influence selection, reported per backbone. This would substantiate or correct the explanatory claim.
  3. [Sec. 4.3 / Fig. 4] Test-time α is selected by reading mCE on ModelNet-C, and λ=0.5 is chosen as the best training-time Beta shape on the same benchmark. Since Fig. 4 shows mCE varies substantially with α (roughly 0.53–0.66), this use of the evaluation benchmark to set hyperparameters can inflate the reported improvements over baselines whose own hyperparameters are not similarly tuned. Please fix α=0.5 a priori (it is the Beta mean and a natural default) or select α and λ on a separate validation split, and report the corresponding test results. Also state which α/λ values were used to produce Tables 1–3.
  4. [Tables 1–3] All results appear to be single runs with no error bars or multi-seed statistics. The claimed 'consistent' improvements are aggregate means over corruption types, and some per-corruption cells move in the opposite direction (e.g., Table 1 Rotate for Point-BERT: 1.316→1.591; Table 2 Rotation/Shear for Point-BERT and Point-MAE). A single seed is insufficient to establish the consistency claim. Please report mean±std over at least three seeds, or provide a statistical test for the headline mCE/mER differences.
minor comments (5)
  1. [Fig. 2 / Fig. 3 captions] Typos: 'Feature Fiter Module' and 'Feature Fiter' should be 'Feature Filter Module' and 'Feature Filter'.
  2. [Tables 1–3] The baseline row (e.g., 'Point-BERT') is implicitly full fine-tuning, but this is not stated in the captions. Since the comparison to full fine-tuning is central to the parameter-efficiency claim, make the baseline explicit.
  3. [Sec. 4.2, ScanObjectNN-C paragraph] Per-corruption tradeoffs are not discussed, e.g., PSFT improves Add-G but hurts Jitter for ULIP-2. A sentence acknowledging that mCE gains are not uniform across corruptions would improve transparency.
  4. [Table 5] On ScanObjectNN-C, the full model (PS+PG+FFM) has mCE 0.685, which is worse than PS+FFM (0.654) and lower OA (86.1 vs 87.4). The text notes this, but the choice of the full model over PS+FFM as the reported method needs a justification, e.g., average performance across all three benchmarks.
  5. [Sec. 4.1] Implementation details are sparse. Please provide batch size, learning rate schedule, number of input points, patch size (for transformer backbones), bottleneck dimension d', and EdgeConv hyperparameters to allow reproduction.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the selection module is explicitly adopted from external prior work [11], and no prediction reduces to a fitted parameter or self-citation.

full rationale

The paper's derivation chain is not circular. The point selection rule (Eqs. 1–5) defines an influence score and a retention ratio; it does not define 'corrupted outlier' as 'point dropped by the rule.' The load-bearing premise that highly influential points under max-pooling are often corrupted outliers is explicitly attributed to external prior work, Refocusing [11] (Levi & Gilboa, 3DV 2025), not to the present authors' own prior results. The prompt generation and feature filter modules are learned on clean training data and evaluated on external corruption benchmarks; no equation defining the method uses the reported mCE/mER as a fitting target. The choices α≈0.5 and λ=0.5 come from a post-hoc sensitivity analysis, which is hyperparameter selection rather than a predicted consequence of the framework, and the paper does not present these as predictions. The only self-citations are background references (e.g., [2] in a sentence about LiDAR-camera fusion) and are not load-bearing for the central claim. The paper itself candidly reports backbone-dependent failures on ScanObjectNN-C (PSFT worsens Point-BERT and Point-MAE), which is a correctness/transfer limitation, not a sign of circularity. Overall, no step in the claimed derivation reduces to its own inputs by construction.

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

The method's theoretical content is thin: it borrows the influence-score premise from [11], inherits PEFT assumptions from VPT/IDPT, and introduces a hand-designed stochastic filter. No new entities are postulated; the free parameters are benchmark-tuned scalars and an unspecified bottleneck width.

free parameters (3)
  • test-time blending coefficient α = ≈0.5
    Eq. 8 introduces α; Fig. 4 sweeps it on ModelNet-C and selects the optimum near 0.5, so reported mCE partly reflects tuning on the evaluation benchmark.
  • Beta shape λ = 0.5
    Fig. 4 compares Beta(0.5,0.5), Beta(1.0,1.0), and Beta(1.5,1.5), selecting λ=0.5 as most robust; this is a benchmark-driven hyperparameter.
  • FFM bottleneck dimension d' = not stated
    The bottleneck width d'<d in Eq. 7 is never specified; it controls filter capacity and residual influence, so the method as printed is not fully reproducible without code.
assumptions (3)
  • domain assumption Highly influential points under pooling are likely corrupted outliers, so discarding them helps robustness
    Sec. 3.1 cites [11] for this observation and builds the entire selection stage on it; the premise is not re-derived in this paper.
  • domain assumption Frozen pre-trained backbones retain robustness priors that full fine-tuning overwrites
    Motivated by Fig. 1's Point-MAE preliminary study, but there is no controlled proof that freezing is the causal reason; it is the design rationale for prompt tuning.
  • domain assumption Prompt tokens and feature filters provide enough adaptation capacity without updating the backbone
    Standard parameter-efficient fine-tuning assumption; supported indirectly by clean accuracy, not independently verified here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Point-Selection Fine-Tuning Framework for Robust Point Cloud Classification." pith.science (2026). https://pith.science/paper/VX4GQBI3

@misc{pith2026260719711,
  author       = {Pith},
  title        = {Pith review of: Point-Selection Fine-Tuning Framework for Robust Point Cloud Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VX4GQBI3}},
  note         = {Machine review of arXiv:2607.19711}
}
read the original abstract

Noisy and corrupted points can substantially degrade point cloud recognition performance, especially under challenging corruption settings. In particular, full fine-tuning of 3D pre-trained models may amplify the influence of outliers and overwrite robustness priors learned during pre-training, while naive parameter-efficient adaptation remains sensitive to corrupted tokens. To address this issue, we propose PSFT, a point-selection fine-tuning framework that improves robustness while remaining parameter-efficient. PSFT first estimates point-wise influence from pre-pooling features and adaptively retains minimally influential points to suppress outliers. Based on the selected subset, a prompt generation branch predicts layer-wise prompt tokens and injects them into a frozen backbone for lightweight downstream adaptation. To further mitigate residual noise after selection, we append a lightweight feature filter with bottleneck MLP transformation and Beta-gated residual blending to refine patch-token representations before prediction. Extensive experiments show that PSFT consistently reduces corruption error on ModelNet-C and ModelNet40-C across all tested 3D pre-trained backbones, while achieving the strongest ScanObjectNN-C results with ULIP-2 and Uni3D-B among the evaluated tuning strategies. Our implementation can be found at https://github.com/CVChMA/PSFT/tree/master.

Figures

Figures reproduced from arXiv: 2607.19711 by the authors.

Figure 1
Figure 1. Effect of Fine-Tuning Strategy on Robustness. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of PSFT. The point selection module ranks points by contribution and retains a robust subset. The prompt generation module uses an EdgeConv-based branch to produce prompt tokens for a frozen transformer backbone. The feature filter module refines patch features by suppressing residual noise while preserving informative signals through residual blending. the adaptation process itself. Prior work has shown th… view at source ↗
Figure 3
Figure 3. PSFT Workflow. PSFT first selects minimally influential points to suppress outliers, then generates layer-wise prompt tokens from the selected subset, and finally refines encoder features with a lightweight filter module before prediction. Formally, the contribution score IF (j) for point j ∈ {1, . . . , N} is defined as: IF (j) = Xκ k=1 I [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Effect of Test-Time Blending Coefficient [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

50 extracted references · 2 canonical work pages

  1. [11]

    Levi, M.Y., Gilboa, G.: Robustifying point cloud networks by refocusing. In: 3DV. pp. 1144–1159 (2025). https://doi.org/10.1109/3DV66043.2025.00110

  2. [1]

    In: ICML

    Ai, Z., Liu, Z., Lei, Y., Cui, Z., Zou, X., Zhou, J.: Gaprompt: Geometry-aware point cloud prompt for 3d vision model. In: ICML. pp. 796–808 (2025)

  3. [2]

    In: WACV

    Cai, H., Yin, D., Yu, F.R., Xiong, S.: DSTR: Dual scenes transformer for cross-modal fusion in 3d object detection. In: WACV. pp. 3064–3073 (2025). https://doi.org/10.1109/WACV61041.2025.00303

  4. [3]

    In: NeurIPS

    Chen, G., Wang, M., Yang, Y., Yu, K., Yuan, L., Yue, Y.: Pointgpt: Auto- regressively generative pre-training from point clouds. In: NeurIPS. vol. 36, pp. 29667–29679 (2023)

  5. [4]

    In: NeurIPS

    Chen, S., Ge, C., Tong, Z., Wang, J., Song, Y., Wang, J., Luo, P.: Adaptformer: Adapting vision transformers for scalable visual recognition. In: NeurIPS. vol. 35, pp. 16664–16678 (2022)

  6. [5]

    In: ECCV

    Chen, Y., Hu, V.T., Gavves, E., Mensink, T., Mettes, P., Yang, P., Snoek, C.G.M.: Pointmixup: Augmentation for point clouds. In: ECCV. pp. 330–345 (2020)

  7. [6]

    In: CVPR

    Choy, C.B., Gwak, J., Savarese, S.: 4d spatio-temporal convnets: Minkowski con- volutional neural networks. In: CVPR. pp. 3075–3084 (2019) 14 D. Li et al

  8. [7]

    In: ICML

    Goyal, A., Law, H., Liu, B., Newell, A., Deng, J.: Revisiting point cloud shape classification with a simple and effective baseline. In: ICML. vol. 139, pp. 3809– 3820 (2021)

Show all 50 references
  1. [8]

    Computational Visual Media7, 187–199 (2021)

    Guo, M.H., Cai, J.X., Liu, Z.N., Mu, T.J., Martin, R.R., Hu, S.M.: Pct: Point cloud transformer. Computational Visual Media7, 187–199 (2021). https://doi.org/10.1007/s41095-021-0229-5

  2. [9]

    In: ECCV

    Jia, M., Tang, L., Chen, B.C., Cardie, C., Belongie, S., Hariharan, B., Lim, S.N.: Visual prompt tuning. In: ECCV. pp. 709–727 (2022)

  3. [10]

    In: ICCV

    Levi, M.Y., Gilboa, G.: Epic: Ensemble of partial point clouds for robust classifi- cation. In: ICCV. pp. 14475–14484 (2023)

  4. [12]

    IEEE Transactions on Pattern Analysis and Machine Intelligence47(4), 3003–3017 (2025)

    Li, J., Wang, J., Chen, J., Xu, T.: Towards robust point cloud recognition with sample-adaptive auto-augmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence47(4), 3003–3017 (2025)

  5. [13]

    In: ICCV

    Li, K., Ferrand, J.C.N., Sheatsley, R., Hoak, B., Beugin, Y., Pauley, E., McDaniel, P.: On the robustness tradeoff in fine-tuning. In: ICCV. pp. 4898–4907 (2025)

  6. [14]

    In: CVPR

    Li, R., Li, X., Heng, P.A., Fu, C.W.: Pointaugment: An auto-augmentation frame- work for point cloud classification. In: CVPR. pp. 6378–6387 (2020)

  7. [15]

    In: CVPR

    Liu, H., Jia, J., Gong, N.Z.: Pointguard: Provably robust 3d point cloud classifica- tion. In: CVPR. pp. 6186–6195 (2021)

  8. [16]

    In: CVPR

    Liu, Y., Fan, B., Xiang, S., Pan, C.: Relation-shape convolutional neu- ral network for point cloud analysis. In: CVPR. pp. 8894–8903 (2019). https://doi.org/10.1109/CVPR.2019.00910

  9. [17]

    In: ICLR (2022)

    Ma, X., Qin, C., You, H., Ran, H., Fu, Y.: Rethinking network design and local geometry in point cloud: A simple residual MLP framework. In: ICLR (2022)

  10. [18]

    In: Advances in Neural Information Processing Systems

    Miao, Y., Du, Y., Zheng, Y., Yang, H., Yuan, C., Gong, B.: Improving robustness of 3d point cloud recognition from a fourier perspective. In: Advances in Neural Information Processing Systems. vol. 37 (2024)

  11. [19]

    In: ECCV

    Pang, Y., Wang, W., Tay, F.E., Liu, W., Tian, Y., Yuan, L.: Masked autoencoders for point cloud self-supervised learning. In: ECCV. pp. 604–621 (2022)

  12. [20]

    In: CVPR

    Qi, C.R., Su, H., Mo, K., Guibas, L.J.: Pointnet: Deep learning on point sets for 3d classification and segmentation. In: CVPR. pp. 77–85 (2017)

  13. [21]

    In: NeurIPS

    Qi, C.R., Yi, L., Su, H., Guibas, L.J.: Pointnet++: Deep hierarchical feature learn- ing on point sets in a metric space. In: NeurIPS. pp. 5099–5108 (2017)

  14. [22]

    In: ICML

    Qi, Z., Dong, R., Fan, G., Ge, Z., Zhang, X., Ma, K., Yi, L.: Contrast with recon- struct: Contrastive 3d representation learning guided by generative pretraining. In: ICML. vol. 202, pp. 28223–28243 (2023)

  15. [23]

    In: NeurIPS

    Qian, G., Li, Y., Peng, H., Mai, J., Hammoud, H., Elhoseiny, M., Ghanem, B.: Pointnext: Revisiting pointnet++ with improved training and scaling strategies. In: NeurIPS. vol. 35, pp. 23192–23204 (2022)

  16. [24]

    Rakotosaona, M.J., La Barbera, V., Guerrero, P., Mitra, N.J., Ovsjanikov, M.: Pointcleannet: Learning to denoise and remove outliers from dense point clouds. Comput. Graph. Forum39(1), 185–203 (2020). https://doi.org/10.1111/cgf.13753

  17. [25]

    Preprint (2022)

    Ren, J., Kong, L., Pan, L., Liu, Z.: Pointcloud-c: Benchmarking and analyzing point cloud perception robustness under corruptions. Preprint (2022)

  18. [26]

    In: ICML

    Ren, J., Pan, L., Liu, Z.: Benchmarking and analyzing point cloud classification under corruptions. In: ICML. vol. 162, pp. 18559–18575 (2022)

  19. [27]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Srivastava, V., Nguyen, D., Le, C.T., Xiang, S., Link, H., Igel, C., Lensch, H.P.A.: Omnivec2 - a novel transformer-based network for large-scale point cloud segmen- Point-Selection Fine-Tuning for Robust Point Cloud Classification 15 tation. In: Proceedings of the IEEE/CVF Co...

  20. [28]

    In: ICRA

    Sun, H., Wang, Y., Chen, W., Deng, H., Li, D.: Parameter-efficient prompt learning for 3d point cloud understanding. In: ICRA. pp. 9478–9486 (2024). https://doi.org/10.1109/ICRA57147.2024.10610093

  21. [29]

    arXiv preprint arXiv:2201.12296 (2022)

    Sun, J., Zhang, Q., Kailkhura, B., Yu, Z., Xiao, C., Mao, Z.M.: Benchmarking ro- bustness of 3d point cloud recognition against common corruptions. arXiv preprint arXiv:2201.12296 (2022)

  22. [30]

    In: AAAI

    Tang, Y., Zhang, R., Guo, Z., Ma, X., Zhao, B., Wang, Z., Wang, D., Li, X.: Point- peft: Parameter-efficient fine-tuning for 3d pre-trained models. In: AAAI. vol. 38, pp. 5171–5179 (2024). https://doi.org/10.1609/aaai.v38i6.28323

  23. [31]

    In: ICCV

    Uy, M.A., Pham, Q., Hua, B., Nguyen, D.T., Yeung, S.: Revisiting point cloud classification: A new benchmark dataset and classification model on real-world data. In: ICCV. pp. 1588–1597 (2019)

  24. [32]

    In: ICCV

    Wang, J., Ding, L., Xu, T., Dong, S., Xu, X., Bai, L., Li, J.: Sample-adaptive augmentation for point cloud recognition against real-world corruptions. In: ICCV. pp. 14330–14339 (2023)

  25. [33]

    In: NeurIPS (2024)

    Wang,J.,Xu,T.,Ding,L.,Li,J.:Target-guidedadversarialpointcloudtransformer towards recognition against real-world corruptions. In: NeurIPS (2024)

  26. [34]

    ACM Trans

    Wang, Y., Sun, Y., Liu, Z., Sarma, S.E., Bronstein, M.M., Solomon, J.M.: Dynamic graph CNN for learning on point clouds. ACM Trans. Graph.38(5), 146:1–146:12 (2019). https://doi.org/10.1145/3326362

  27. [35]

    In: CVPR

    Wu, Z., Song, S., Khosla, A., Yu, F., Zhang, L., Tang, X., Xiao, J.: 3d shapenets: A deep representation for volumetric shapes. In: CVPR. pp. 1912–1920 (2015)

  28. [36]

    In: CVPR

    Xu, M., Ding, R., Zhao, H., Qi, X.: PAConv: position adaptive convolution with dynamic kernel assembling on point clouds. In: CVPR. pp. 3173–3182 (2021)

  29. [37]

    In: AAAI (2021)

    Xu, M., Zhang, J., Zhou, Z., Xu, M., Qi, X., Qiao, Y.: Learning geometry- disentangled representation for complementary understanding of 3d object point cloud. In: AAAI (2021)

  30. [38]

    In: CVPR

    Xue, L., Gao, M., Xing, C., Martín-Martín, R., Wu, J., Xiong, C., Xu, R., Niebles, J.C., Savarese, S.: ULIP: learning a unified representation of language, images, and point clouds for 3d understanding. In: CVPR. pp. 1179–1189 (2023)

  31. [39]

    In: CVPR

    Xue, L., Yu, N., Zhang, S., Panagopoulou, A., Li, J., Martín-Martín, R., Wu, J., Xiong, C., Xu, R., Niebles, J.C., Savarese, S.: ULIP-2: towards scalable multimodal pre-training for 3d understanding. In: CVPR. pp. 27081–27091 (2024)

  32. [40]

    In: CVPR

    Yan, X., Zheng, C., Li, Z., Wang, S., Cui, S.: Pointasnl: Robust point clouds pro- cessing using nonlocal neural networks with adaptive sampling. In: CVPR. pp. 5589–5598 (2020)

  33. [41]

    In: CVPR

    Yu, X., Tang, L., Rao, Y., Huang, T., Zhou, J., Lu, J.: Point-bert: Pre-training 3d point cloud transformers with masked point modeling. In: CVPR. pp. 19313–19322 (2022)

  34. [42]

    arXiv preprint arXiv:1811.11209 (2018)

    Yuan, W., Held, D., Mertz, C., Hebert, M.: Iterative transformer network for 3d point cloud. arXiv preprint arXiv:1811.11209 (2018)

  35. [43]

    In: ICCV

    Zha, Y., Wang, J., Dai, T., Chen, B., Wang, Z., Xia, S.T.: Instance-aware dynamic prompt tuning for pre-trained point cloud models. In: ICCV. pp. 14161–14170 (2023)

  36. [44]

    In: ICML

    Zha, Y., Wang, C., Yang, W., Zhang, T., Wu, F.: Exploring vision semantic prompt for efficient point cloud understanding. In: ICML. pp. 74271–74287 (2025)

  37. [45]

    IEEE Robotics Autom

    Zhang, J., Chen, W., Wang, Y., Vasudevan, R., Johnson-Roberson, M.: Point set voting for partial point cloud analysis. IEEE Robotics Autom. Lett.6(2), 596–603 (2021). https://doi.org/10.1109/LRA.2020.3048658 16 D. Li et al

  38. [46]

    IEEE Trans

    Zhang, Y., Lin, J., Li, R., Jia, K., Zhang, L.: Point-dae: Denoising autoencoders for self-supervised point cloud learning. IEEE Trans. Neural Netw. Learn. Syst. 36(9), 16279–16293 (2025). https://doi.org/10.1109/TNNLS.2025.3557055

  39. [47]

    Pattern Recognition 127, 108626 (2022)

    Zhao, C., Yang, J., Xiong, X., Zhu, A., Cao, Z., Li, X.: Rotation invariant point cloud analysis: Where local geometry meets global topology. Pattern Recognition 127, 108626 (2022). https://doi.org/10.1016/j.patcog.2022.108626

  40. [48]

    In: CVPR

    Zhao, H., Jiang, L., Jia, J., Torr, P.H., Koltun, V.: Point transformer. In: CVPR. pp. 16259–16268 (2021)

  41. [49]

    In: ICLR (2024)

    Zhou, J., Wang, J., Ma, B., Liu, Y., Huang, T., Wang, X.: Uni3d: Exploring unified 3d representation at scale. In: ICLR (2024)

  42. [50]

    In: CVPR

    Zhou, X., Liang, D., Xu, W., Zhu, X., Xu, Y., Zou, Z., Bai, X.: Dynamic adapter meets prompt tuning: Parameter-efficient transfer learning for point cloud analysis. In: CVPR. pp. 14707–14717 (2024) Point-Selection Fine-Tuning for Robust Point Cloud Classification 17 Supplement...

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.