Pith. sign in

REVIEW 3 major objections 5 minor 43 references

AdaCo: Overcoming Visual Foundation Model Noise in 3D Semantic Segmentation via Adaptive Label Correction

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read AdaCo claims that noisy pseudo-labels from visual foundation models can be corrected adaptively, letting label-free 3D semantic segmentation reach 25.7 mIoU on SemanticKITTI and 31.2 mIoU on nuScenes.

desk verdict Large label-free segmentation gains undercut by an inverted confidence formula in the core correction module; deserves review but needs clarification. read the letter →

arxiv 2412.18255 v1 pith:J6MOARMP submitted 2024-12-24 cs.CV

classification cs.CV
keywords 3Dsemanticsegmentationlabel-freelearningpseudo-labelrefinementvisualfoundationmodelsnoisylabelpointcloudadaptivelosscross-modalsupervision
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 claims that noisy pseudo-labels extracted from visual foundation models can be systematically refurbished, so that a LiDAR point cloud segmenter can be trained with zero 3D annotations and still reach 25.7 mIoU on SemanticKITTI and 31.2 mIoU on nuScenes. The proposed AdaCo pipeline generates point-wise labels by lifting image-level mask descriptions into 3D, then iteratively corrects the noisy labels using each sample's own learning curve, and finally trains with a loss that becomes robust only after the correction point. The stated results beat previous label-free and unsupervised 3D segmenters by 11.5–12.5 mIoU on SemanticKITTI and 4.4 mIoU on nuScenes. If the claims hold, this would make foundation-model supervision a practical substitute for manual point-wise annotation in outdoor driving scenes.

What carries the argument

The machine that carries the argument is the adaptive correction scheduler. For each sample, the training mIoU against the noisy labels is recorded every epoch and fit by an exponential curve $f_s(t) = a(1-\exp(-t^{b_s}/c_s))$; when the normalized change in the derivative exceeds a threshold $r=0.9$, the sample is declared to have finished its clean-learning phase, and the network's historical predictions over the last $q$ iterations (with $q$ capped at 5) are turned into a reliability score $F(x_s;q)$ for each point. Points whose scores pass a threshold supply replacement labels via the mode of their prediction history, and those labels are spread to neighboring points inside DBSCAN clusters. A second mechanism, the Adaptive Robust Loss, keeps cross-entropy loss during the warm-up phase and then adds normalized cross-entropy and mean-absolute-error terms with a negative weight on the original cross-entropy, so the loss changes character exactly when the label corrector runs.

What would settle it

Compute $F(x_s;q)$ for a point whose last five predictions are all the same class and for a point whose last five predictions are uniformly spread across the $K$ classes; under Eq. (2) the uniform point scores 1 and the consistent point scores 0, opposite to the paper's reliable-point criterion. Replacing the reliable-point filter with the mode of historical predictions alone on SemanticKITTI would directly test whether the reported 25.7 mIoU depends on the mechanism as described.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that label noise from visual foundation models is not a fixed ceiling: it can be corrected adaptively during training. The authors report that AdaCo, built from the Cross-modal Label Generation Module (CLGM), the Adaptive Noise Corrector (ANC), and the Adaptive Robust Loss (ARL), improves the MaskCLIP-based label-free baseline from 8.1 to 25.7 mIoU on SemanticKITTI and from 16.5 to 31.2 mIoU on nuScenes, surpassing earlier label-free methods by 11.5–12.5 and 4.4 mIoU respectively. CLGM generates point-wise pseudo labels by combining class-agnostic masks with semantic descriptions and voting over adjacent frames; ANC detects a per-sample correction time from the training mIoU curve, marks points with consistent historical predictions as reliable, and propagates their labels within density-based clusters; ARL switches the loss from cross-entropy to normalized robust losses at the correction time to keep the network from memorizing residual noise. The paper concludes that label-free learning for 3D semantic segmentation can be made substantially more robust by treating the VFM-generated labels as a noisy signal to be cleaned rather than as final supervision.

Load-bearing premise

The load-bearing premise is that the confidence score $F(x_s;q)$ in Eq. (2) identifies points whose historical predictions are consistent (so their labels can be trusted), and that fitting an exponential to each sample's training mIoU reveals a genuine clean-then-memorize turning point; as printed, Eq. (2) is maximized by maximally uncertain predictions, so the reliable-point selection needs either a sign flip or a reinterpretation for the correction stage to do what the paper claims.

Editorial extensions

If this is right

  • A LiDAR segmenter can be trained with no point-level annotations and still reach 25.7 mIoU on SemanticKITTI and 31.2 mIoU on nuScenes, roughly tripling the label-free baseline on the former.
  • The correction timing is per-sample and automatic, replacing manually preset warm-up rounds in noisy-label training.
  • Because CLGM maps semantic descriptions to an arbitrary category vocabulary, the label generator can adapt to new classes by editing text prompts without retraining.
  • The ARL loss schedule gives a transferable recipe: train with standard cross-entropy first, then switch to normalized robust losses once the network has learned clean patterns.

Reading between the lines

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

  • Beyond the paper, the reliable-point selection can be viewed as a voting scheme over the network's own past predictions; adding geometric cues such as point-to-point distance or multi-view agreement would likely reduce the dependence on the curve-fit threshold.
  • Because the correction is applied only once per sample, a natural extension is to re-run the full correction loop with fresh historical predictions, testing whether repeated refinement improves or degrades label quality on other datasets.
  • The label generator's text-based category matching suggests a direct test of generalization: add unseen categories to the text prompts and check whether the corrected pseudo-labels improve on those categories without retraining the generator.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes AdaCo, a label-free 3D semantic segmentation method for outdoor LiDAR scenes. The method has three components: a Cross-modal Label Generation Module (CLGM) that produces point-wise pseudo labels from 2D VFMs (FastSAM, SSA-Engine, word2vec similarity with text prompts); an Adaptive Noise Corrector (ANC) that uses the network's historical predictions and DBSCAN clustering to refurbish noisy labels; and an Adaptive Robust Loss (ARL) that switches from cross-entropy to a combination of normalized cross-entropy and mean absolute error losses after a per-sample correction time. On SemanticKITTI and nuScenes, the method reports 25.7 and 31.2 mIoU, respectively, outperforming prior label-free baselines by large margins. The ablation study attributes most of the gain to CLGM+ANC over CLGM alone.

Significance. If the reported numbers are reproducible, the paper makes a strong empirical contribution: it demonstrates that label-free 3D semantic segmentation can approach the level of fully supervised methods on two standard outdoor benchmarks, and it provides a modular design with a thorough ablation. The paper is clearly written, the experimental comparison covers the relevant recent baselines, and the qualitative results support the claimed improvement in edge quality and label consistency. The main methodological contribution of adaptive correction timing is interesting and, in principle, a useful addition to noisy-label learning. However, the current manuscript contains a concrete, load-bearing technical error in the definition of the historical-prediction confidence, and the correction-timing signal is derived from the very noisy labels being corrected; both issues must be resolved before the empirical claims can be accepted.

major comments (3)
  1. [Methodology, Eq. (2) and Algorithm 1, line 8] The confidence score F(x_s;q) in Eq. (2) is defined as -entropy(P(k|x_s;q)) / log(1/K). With base-e logarithms, log(1/K) = -ln K, so F(x_s;q) = entropy(P)/ln K, which ranges from 0 for a deterministic prediction history to 1 for a uniform distribution. The text in 'Reliable Labels Prediction' states that samples with consistent historical predictions should be considered reliable, but the formula as written assigns the highest confidence to maximally inconsistent histories. Algorithm 1 line 8 then selects points with F(x_s;q) > threshold, so the reliable set would be composed of high-entropy points, and Eq. (5) would take the argmax of an almost uniform frequency vector, yielding effectively arbitrary corrections. This issue is central: the ablation in Table 2 attributes a 5.8 mIoU gain to CLGM+ANC (19.5 to 25.3), so the label-correction stage is the main source of the reported improvement. Please correct the formula (e.g., F = 1 - entropy(P)/ln K) and ensure that Algorithm 1 uses the same convention, or explicitly state if the implemented form differs from the text.
  2. [Adaptive Correction Timing Search, Eq. (1), and Algorithm 1, line 5] The correction timing is determined by fitting the exponential function f_s(t) to the per-sample training mIoU, but Algorithm 1 line 5 computes Is.append(mIoU(\hat{y}_s, L_s)) using the noisy labels L_s. The fitted curve therefore measures agreement with the noisy supervision, not with clean labels. The text in the Additional Implementation Details explicitly states that 'training IoU is calculated from the prediction results and noisy labels'. The claim that the derivative-change threshold r identifies the moment when the model has finished learning clean labels is not justified by the manuscript: under standard noisy-label training, accuracy measured against the noisy labels often increases monotonically, and no clean-label signal is available to validate the turning point. This is a correctness risk for the ANC, since the correction time directly controls when historical predictions are used to overwrite labels. Please provide empirical evidence that the fitted t_c correlates with the early-learning/memorization transition; for example, on SemanticKITTI, where clean training labels are available, the paper could compare the fitted derivative-change point with the peak of the early-learning IoU computed against the clean labels.
  3. [Experimental protocol, Tables 1–3] All results are reported as single numbers with no error bars or multiple seeds. Given that many hyperparameters (r, q, ω, λ, β, σ, FastSAM thresholds, DBSCAN eps and min_points, block size and stride) are tuned on the validation set, the large reported margins over prior methods need to be supported by a few independent training runs with mean and standard deviation. This is particularly important for the main SemanticKITTI result (25.7 mIoU) and for the ablation in Table 3, where the difference between the best and second-best settings is only 0.1–0.3 mIoU in some rows.
minor comments (5)
  1. [Experiments section heading] The section heading reads 'Experiements'; please correct the spelling to 'Experiments'.
  2. [Figure 4 caption] The caption states that the early-learning IoU curve is computed with the correct ground truth, while the method itself does not use any clean labels during training. Please clarify that this figure is an analysis diagnostic and is not part of the training procedure, to avoid confusion with the circularity concern raised in the major comments.
  3. [Eq. (7)] The notation in the normalized cross-entropy loss is unclear: q(k|x) is defined as the distribution over labels, but the denominator uses both q(y=j|x) and p(k|x); please define the terms precisely and specify over which set the outer sum runs.
  4. [Tables 6–7] The label mapping tables list 'Our pre-defined labels' as a manual dictionary, but the text in the Additional Implementation Details says a manual dictionary was maintained following OpenScene. Please provide the exact reference to OpenScene and state whether these mappings are used only for the manual-dictionary comparison or also in the final word2vec-based pipeline.
  5. [Eq. (10)] The use of a negative weight σ = -0.99 for the cross-entropy term L_CE in L_ARL is unusual and is only briefly motivated. Please provide a short explanation of how this negative weighting is intended to behave, and ideally show a curve of the loss value to rule out unintended optimization behavior.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported gains are benchmarked against external ground truth, and the self-citations are not load-bearing.

full rationale

The paper's derivation chain is empirical rather than formal: CLGM generates noisy pseudo-labels from VFM outputs, ANC refurbishes them using historical predictions plus DBSCAN clustering, and ARL reweights losses. None of these steps defines its output in terms of the quantity it claims to predict; the headline results (25.7 mIoU on SemanticKITTI, 31.2 on nuScenes) are measured against external ground truth, not against the fitted mIoU curves or the correction schedule. The ANC correction timing is obtained by fitting Eq. (1) to training mIoU computed against noisy labels, but this is a training heuristic to locate the early-learning/memorization transition, not a fitted parameter renamed as a benchmark prediction. The self-citations present in the paper (OPOCA, COIN, HINTED) appear only as related-work and dataset context and are not load-bearing for the central claim. The manuscript does contain internal-consistency concerns: Eq. (2) as written assigns maximal confidence to maximally uncertain prediction histories, and Eq. (10) uses a negative cross-entropy weight; hyperparameters are also tuned on the validation set used for reporting. These are reproducibility and correctness risks, but they are not examples of a result reducing to its inputs by construction. The acknowledged limitations (calibration dependence, VFM semantic limits) further confirm that the method is presented as an empirical engineering contribution rather than as a derivation from assumptions. No circular step meeting the quoted-evidence standard was found.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The method rests on several domain assumptions about VFM reliability, early-learning dynamics, and clustering semantics, plus a set of tuned hyperparameters. No new physical or ontological entities are introduced.

free parameters (8)
  • Correction derivative threshold r = 0.9
    Determines when ANC starts correcting labels via fitted learning curve; tuned on SemanticKITTI validation (Tab. 5).
  • Historical prediction length q (max tm) = 5
    Length of prediction history for mode voting and confidence; set to 5 due to memory (Implementation Details, Tab. 3 History Seq.).
  • Reliable-label confidence threshold gamma = not stated
    Threshold in F(x_s;q) >= gamma used to select reliable points; no value is reported in the paper, and Eq. (2) makes its meaning ambiguous.
  • Cluster winner ratio omega = 3
    Selects winner labels in DBSCAN intersection voting; tuned on validation (Tab. 5).
  • ARL weights lambda, beta, sigma = lambda=100, beta=1, sigma=-0.99
    Weights for normalized CE, MAE, and cross-entropy in adaptive robust loss; lambda and beta taken from Ma et al. 2020, sigma tuned in Tab. 3.
  • FastSAM segmentation threshold and mask confidence = 0.7 and 0.4
    Control mask generation quality in CLGM; set by hand (Implementation Details).
  • DBSCAN eps and min_points = eps=0.6, min_points=5
    Instance clustering parameters for label propagation; set in Additional Implementation Details.
  • Voxel size, block size, stride = 0.05/0.1 m, block [10,10,z], stride 10
    Spatial resolution choices for voxel voting, clustering, and training; follow prior work or hand-set.
assumptions (6)
  • domain assumption VFMs (FastSAM + SSA-Engine) produce semantically accurate masks and descriptions in outdoor driving scenes.
    CLGM's pseudo labels inherit VFM accuracy; this is the entire supervision source (Sec. Cross-modal Label Generation Module).
  • ad hoc to paper The exponential learning-curve model f_s(t) = a(1 - exp(-t^b/c)) captures when a sample finishes learning clean labels.
    Used to locate correction round t_c in ANC; no justification that mIoU on noisy labels follows this form (Eq. (1)).
  • domain assumption Deep networks learn clean samples before noisy ones in this 3D setting.
    ANC relies on the early-learning phenomenon (Sec. Adaptive Noise Corrector, inspired by Liu et al. 2022).
  • domain assumption DBSCAN clusters of non-ground points correspond to object instances suitable for label propagation.
    Clustering label propagation assumes each cluster is semantically homogeneous (Sec. Clustering Label Propagation).
  • domain assumption word2vec semantic similarity between SSA-Engine text descriptions and class prompt words is a valid label assignment signal.
    CLGM chooses pixel labels by highest word2vec similarity; validated only by final mIoU (Additional Implementation Details, Tab. 4).
  • domain assumption Pixel-to-point calibration and inter-frame alignment are accurate enough after voxel voting.
    Point pseudo labels are obtained by back-projecting 2D labels through calibration matrices (Sec. Point-wise Label Unprojection and Voxel Refinement).

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdaCo: Overcoming Visual Foundation Model Noise in 3D Semantic Segmentation via Adaptive Label Correction." pith.science (2026). https://pith.science/paper/J6MOARMP

@misc{pith2026241218255,
  author       = {Pith},
  title        = {Pith review of: AdaCo: Overcoming Visual Foundation Model Noise in 3D Semantic Segmentation via Adaptive Label Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/J6MOARMP}},
  note         = {Machine review of arXiv:2412.18255}
}
read the original abstract

Recently, Visual Foundation Models (VFMs) have shown a remarkable generalization performance in 3D perception tasks. However, their effectiveness in large-scale outdoor datasets remains constrained by the scarcity of accurate supervision signals, the extensive noise caused by variable outdoor conditions, and the abundance of unknown objects. In this work, we propose a novel label-free learning method, Adaptive Label Correction (AdaCo), for 3D semantic segmentation. AdaCo first introduces the Cross-modal Label Generation Module (CLGM), providing cross-modal supervision with the formidable interpretive capabilities of the VFMs. Subsequently, AdaCo incorporates the Adaptive Noise Corrector (ANC), updating and adjusting the noisy samples within this supervision iteratively during training. Moreover, we develop an Adaptive Robust Loss (ARL) function to modulate each sample's sensitivity to noisy supervision, preventing potential underfitting issues associated with robust loss. Our proposed AdaCo can effectively mitigate the performance limitations of label-free learning networks in 3D semantic segmentation tasks. Extensive experiments on two outdoor benchmark datasets highlight the superior performance of our method.

Figures

Figures reproduced from arXiv: 2412.18255 by the authors.

Figure 1
Figure 1. In outdoor scenes, variable environmental condi [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The illustration of our label-free 3D semantic segmentation method AdaCo. (a) The 3D noisy pseudo labels are [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. CLGM pipeline. We utilize the composition of SAM and SSA-Engine as our 2D-PLGE to segment masks with [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The learning curve of different mIoU in the Se [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results in SemanticKITTI (a)-(d) and nuScenes (e)-(h), the noisy ground truth is generated from CLGM. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 7
Figure 7. Figure 7: Our CLGM demonstrates better instance segmen [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 6
Figure 6. Figure 6: Qualitative results of 2D semantic labels generated in SemanticKITTI. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Qualitative results of 2D semantic labels generated in nuScenes. [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Qualitative results in SemanticKITTI (Row 1-2) and nuScenes (Row 3-5). [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 31 canonical work pages

  1. [1]

    Behley, J.; Garbade, M.; Milioto, A.; Quenzel, J.; Behnke, S.; Stachniss, C.; and Gall, J. 2019. SemanticKITTI: A Dataset for Semantic Scene Understanding of LiDAR Sequences . In Proc. of the IEEE/CVF International Conf. on Computer Vision (ICCV)

  2. [2]

    J.; and Goldberger, J

    Bekker, A. J.; and Goldberger, J. 2016. Training deep neural-networks based on unreliable labels. In 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2682--2686. IEEE

  3. [3]

    R.; and Blaschko, M

    Berman, M.; Triki, A. R.; and Blaschko, M. B. 2018. The lov \'a sz-softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, 4413--4421

  4. [4]

    H.; Vora, S.; Liong, V

    Caesar, H.; Bankiti, V.; Lang, A. H.; Vora, S.; Liong, V. E.; Xu, Q.; Krishnan, A.; Pan, Y.; Baldan, G.; and Beijbom, O. 2019. nuScenes: A multimodal dataset for autonomous driving. arXiv preprint arXiv:1903.11027

  5. [5]

    Cen, J.; Zhang, S.; Pei, Y.; Li, K.; Zheng, H.; Luo, M.; Zhang, Y.; and Chen, Q. 2023. CMDFusion: Bidirectional Fusion Network With Cross-Modality Knowledge Distillation for LiDAR Semantic Segmentation. IEEE Robotics and Automation Letters, 9(1): 771--778

  6. [6]

    Chen, J.; Yang, Z.; and Zhang, L. 2023. Semantic Segment Anything. https://github.com/fudan-zvg/Semantic-Segment-Anything

  7. [7]

    Chen, R.; Liu, Y.; Kong, L.; Chen, N.; Xinge, Z.; Ma, Y.; Liu, T.; and Wang, W. 2023 a . Towards label-free scene understanding by vision foundation models. In Thirty-seventh Conference on Neural Information Processing Systems

  8. [8]

    Chen, R.; Liu, Y.; Kong, L.; Zhu, X.; Ma, Y.; Li, Y.; Hou, Y.; Qiao, Y.; and Wang, W. 2023 b . Clip2scene: Towards label-efficient 3d scene understanding by clip. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 7020--7030

Show all 43 references
  1. [9]

    Church, K. W. 2017. Word2Vec. Natural Language Engineering, 23(1): 155--162

  2. [10]

    Contributors, S. 2022. Spconv: Spatially Sparse Convolution Library. https://github.com/traveller59/spconv

  3. [11]

    Ester, M.; Kriegel, H.-P.; Sander, J.; Xu, X.; et al. 1996. A density-based algorithm for discovering clusters in large spatial databases with noise. In kdd, volume 96, 226--231

  4. [12]

    Genova, K.; Yin, X.; Kundu, A.; Pantofaru, C.; Cole, F.; Sud, A.; Brewington, B.; Shucker, B.; and Funkhouser, T. 2021. Learning 3d semantic segmentation with only 2d image supervision. In 2021 International Conference on 3D Vision (3DV), 361--372. IEEE

  5. [13]

    Hu, Q.; Yang, B.; Fang, G.; Guo, Y.; Leonardis, A.; Trigoni, N.; and Markham, A. 2022. Sqn: Weakly-supervised semantic segmentation of large-scale 3d point clouds. In Computer Vision--ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23--27, 2022, Proceedings, Par...

  6. [14]

    Hu, Q.; Yang, B.; Xie, L.; Rosa, S.; Guo, Y.; Wang, Z.; Trigoni, N.; and Markham, A. 2020. Randla-net: Efficient semantic segmentation of large-scale point clouds. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 11108--11117

  7. [15]

    Huang, L.; Zhang, C.; and Zhang, H. 2020. Self-adaptive training: beyond empirical risk minimization. Advances in neural information processing systems, 33: 19365--19376

  8. [16]

    Huang, W.; Zou, P.; Xia, Y.; Wen, C.; Zang, Y.; Wang, C.; and Zhou, G. 2024. OPOCA: One Point One Class Annotation for LiDAR Point Cloud Semantic Segmentation. IEEE Transactions on Geoscience and Remote Sensing, 62: 1--10

  9. [17]

    Kang, X.; Chu, L.; Li, J.; Chen, X.; and Lu, Y. 2024. Hierarchical Intra-modal Correlation Learning for Label-free 3D Semantic Segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 28244--28253

  10. [18]

    C.; Lo, W.-Y.; et al

    Kirillov, A.; Mintun, E.; Ravi, N.; Mao, H.; Rolland, C.; Gustafson, L.; Xiao, T.; Whitehead, S.; Berg, A. C.; Lo, W.-Y.; et al. 2023. Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 4015--4026

  11. [19]

    Lee, K.; Yun, S.; Lee, K.; Lee, H.; Li, B.; and Shin, J. 2019. Robust inference via generative classifiers for handling noisy labels. In International conference on machine learning, 3763--3772. PMLR

  12. [20]

    Lee, S.; Lim, H.; and Myung, H. 2022. Patchwork++: Fast and robust ground segmentation solving partial under-segmentation using 3D point cloud . In Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., 13276--13283

  13. [21]

    Li, H.; Sun, Z.; Wu, Y.; and Song, Y. 2021. Semi-supervised point cloud segmentation using self-training with label confidence prediction. Neurocomputing, 227–237

  14. [22]

    P.; and Shum, H

    Liu, J.; Yu, Z.; Breckon, T. P.; and Shum, H. P. 2024. U3DS3: Unsupervised 3D Semantic Scene Segmentation. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 3759--3768

  15. [23]

    Liu, S.; Liu, K.; Zhu, W.; Shen, Y.; and Fernandez-Granda, C. 2022. Adaptive early-learning correction for segmentation from noisy annotations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2606--2616

  16. [24]

    Liu, Z.; Qi, X.; and Fu, C.-W. 2021. One thing one click: A self-training approach for weakly supervised 3d semantic segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 1726--1736

  17. [25]

    Ma, X.; Huang, H.; Wang, Y.; Romano, S.; Erfani, S.; and Bailey, J. 2020. Normalized loss functions for deep learning with noisy labels. In International conference on machine learning, 6543--6553. PMLR

  18. [26]

    Peng, S.; Genova, K.; Jiang, C.; Tagliasacchi, A.; Pollefeys, M.; Funkhouser, T.; et al. 2023. Openscene: 3d scene understanding with open vocabularies. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 815--824

  19. [27]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learning, 8748--8763. PMLR

  20. [28]

    P.; and Xu, X

    Schubert, E.; Sander, J.; Ester, M.; Kriegel, H. P.; and Xu, X. 2017. DBSCAN revisited, revisited: why and how you should (still) use DBSCAN. ACM Transactions on Database Systems (TODS), 42(3): 1--21

  21. [29]

    Sukhbaatar, S.; Bruna, J.; Paluri, M.; Bourdev, L.; and Fergus, R. 2014. Training convolutional networks with noisy labels. arXiv preprint arXiv:1406.2080

  22. [30]

    Wang, Y.; Ma, X.; Chen, Z.; Luo, Y.; Yi, J.; and Bailey, J. 2019. Symmetric cross entropy for robust learning with noisy labels. In Proceedings of the IEEE/CVF international conference on computer vision, 322--330

  23. [31]

    Wei, H.; Tao, L.; Xie, R.; and An, B. 2021. Open-set label noise can improve robustness against inherent label noise. Advances in Neural Information Processing Systems, 34: 7978--7992

  24. [32]

    Xia, Q.; Deng, J.; Wen, C.; Wu, H.; Shi, S.; Li, X.; and Wang, C. 2023. Coin: Contrastive instance feature mining for outdoor 3d object detection with very limited annotations. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 6254--6263

  25. [33]

    Xia, Q.; Ye, W.; Wu, H.; Zhao, S.; Xing, L.; Huang, X.; Deng, J.; Li, X.; Wen, C.; and Wang, C. 2024. HINTED: Hard Instance Enhanced Detector with Mixed-Density Feature Fusion for Sparsely-Supervised 3D Object Detection. In Proceedings of the IEEE/CVF Conference on Computer Vi...

  26. [34]

    Xia, X.; Liu, T.; Han, B.; Gong, C.; Wang, N.; Ge, Z.; and Chang, Y. 2020. Robust early-learning: Hindering the memorization of noisy labels. In International conference on learning representations

  27. [35]

    Ye, S.; Chen, D.; Han, S.; and Liao, J. 2021. Learning with noisy labels for robust point cloud segmentation. In Proceedings of the IEEE/CVF international conference on computer vision, 6443--6452

  28. [36]

    Zhang, Z.; and Sabuncu, M. 2018. Generalized cross entropy loss for training deep neural networks with noisy labels. Advances in neural information processing systems, 31

  29. [37]

    Zhang, Z.; Yang, B.; Wang, B.; and Li, B. 2023. Growsp: Unsupervised semantic segmentation of 3d point clouds. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 17619--17629

  30. [38]

    Zhao, X.; Ding, W.; An, Y.; Du, Y.; Yu, T.; Li, M.; Tang, M.; and Wang, J. 2023. Fast segment anything. arXiv preprint arXiv:2306.12156

  31. [39]

    Zheng, S.; Wu, P.; Goswami, A.; Goswami, M.; Metaxas, D.; and Chen, C. 2020. Error-bounded correction of noisy labels. In International Conference on Machine Learning, 11447--11457. PMLR

  32. [40]

    C.; and Dai, B

    Zhou, C.; Loy, C. C.; and Dai, B. 2022. Extract free dense labels from clip. In European Conference on Computer Vision, 696--712. Springer

  33. [41]

    Zhu, X.; Zhou, H.; Wang, T.; Hong, F.; Li, W.; Ma, Y.; Li, H.; Yang, R.; and Lin, D. 2021. Cylindrical and asymmetrical 3d convolution networks for lidar-based perception. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(10): 6807--6822

  34. [42]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  35. [43]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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