DyConfidMatch: Dynamic Thresholding and Re-sampling for 3D Semi-supervised Learning

Bing Li; Zhimin Chen

REVIEW 4 major objections 5 minor 48 references

DyConfidMatch: Dynamic Thresholding and Re-sampling for 3D Semi-supervised Learning

T0 review · 4 major / 5 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 Class-level confidence rebalances 3D semi-supervised learning

desk verdict A solid incremental 3D SSL method whose core mechanism is from the authors' WACV 2023 paper; the journal version's SOTA claim is weakened by test-set-tuned constants, but the empirical work is substantial enough for peer review. read the letter →

arxiv 2411.08340 v1 pith:46C2RKZL submitted 2024-11-13 cs.CV

Zhimin Chen , Bing Li This is my paper · ORCID

classification cs.CV

keywords Semi-supervisedlearning3DobjectdetectionpointcloudclassificationClassimbalancePseudo-labelingDynamicthresholdingRe-samplingClass-levelconfidence

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

DyConfidMatch claims that the average confidence of a model's own predictions on unlabeled data, taken per class, is a faithful indicator of how well that class is being learned. The paper uses this class-level confidence to do two things: dynamically lower the pseudo-label threshold for classes that lag behind, and re-sample unlabeled data so struggling classes contribute more often. On ModelNet40, ScanObjectNN, SUN RGB-D, ScanNet, and KITTI, the method reports gains over fixed-threshold baselines, with the largest improvements at very low label fractions, for example roughly +7 mAP@0.25 over 3DIoUMatch at 2% labels on ScanNet. The payoff is that semi-supervised 3D pipelines can adapt to data imbalance without manual threshold tuning.

What carries the argument

The load-bearing object is class-level confidence, $P_c = \frac{1}{|C_c|} \sum_{u_j \in C_c} \max(p_m(y|\alpha(u_j)))$, the mean of the model's maximum predicted probability over unlabeled points whose argmax class is $c$. It feeds a concave mapping $M(P_c)=P_c/(2-P_c)$ to produce a per-class dynamic threshold $\tau_e(c)$, clamped by a global threshold $\tau=\exp(-2P_{\text{ave}}^2)$, and it also sets per-instance sampling weights that boost classes with low learning status. Together these two mechanisms carry the argument: they convert a scalar confidence signal into both a pseudo-label gate and a data-rebalancing schedule, and the ablation study shows each component contributes to the reported gains.

What would settle it

Compute the per-class rank correlation between class-level confidence $P_c$ and per-class test accuracy on, say, ModelNet40 with 10% labeled data, sampled every 50 epochs; if the correlation is weak (for instance below roughly 0.5) or turns negative for minority classes at any stage, the steering signal is unreliable and the reported gains should shrink or reverse when the correlation is adversarially broken.

Watch

Extended reading notes

Core claim

The central claim is that a per-class statistic—the average maximum softmax probability over unlabeled points assigned to that class—tracks the class's true learning status (test accuracy) throughout training, in both classification and detection settings. On this basis the paper constructs DyConfidMatch, which sets each class's pseudo-label threshold at a concave function of that class-level confidence, clips it with a global threshold that also adapts, and assigns each unlabeled instance a sampling weight that up-weights classes with low learning status. The paper shows this combination outperforms existing 3D SSL methods across classification and detection benchmarks, with the strongest margins in low-label regimes. The method is presented as a drop-in replacement for the threshold and sampling components of existing pseudo-label learners such as FixMatch and 3DIoUMatch.

Load-bearing premise

The load-bearing premise is that the average confidence the model assigns to unlabeled points in a class accurately tracks how well that class is actually learned, across training and even for rare classes; if that correlation fails, the dynamic thresholds and re-sampling weights would push training toward the wrong classes.

Editorial extensions

If this is right

Low-label 3D detection and classification no longer need a manually chosen fixed pseudo-label threshold; the threshold adapts as the model learns.
Underrepresented or harder classes get more unlabeled data earlier, which is where the reported gains concentrate, such as +7 mAP@0.25 over 3DIoUMatch at 2% labels on ScanNet.
The dynamic threshold and re-sampling components can be dropped into existing pseudo-label SSL methods, so the same mechanism could lift other 3D learners without changing the backbone.
Re-sampling by learning status rather than class count avoids over-sampling easy minority classes, a behavior the paper demonstrates by outperforming CReST and BiS with only the sampling component swapped.
The KITTI results indicate the mechanism transfers from indoor to outdoor 3D detection, where class imbalance and confidence distributions differ.

Reading between the lines

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

A concrete way to stress-test the paper's premise would be to compute a rank correlation between class-level confidence and per-class test accuracy across training; the paper only shows visual plots, so the confidence proxy remains quantitatively unvalidated.
If the confidence proxy survives that test, the same learning-status re-sampling idea could be transferred to 2D semi-supervised learning or long-tail recognition, likely after recalibrating confidence for those settings.
An oracle experiment—using true per-class test accuracy instead of $P_c$ to set thresholds and sampling—would bound how much of the reported gain comes from the rebalancing scheme itself rather than from the accuracy of the confidence estimate.
Because pseudo-class assignments come from the model's own argmax, very early training or poorly calibrated backbones could poison the class-level statistics; a warm start or uncertainty-weighted averaging might be needed in practice.

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.

Desk Editor's Note

read the letter

Colleague,

Quick take: this is a journal extension of the authors' WACV 2023 paper (Class-level confidence based 3D semi-supervised learning). The core mechanism—class-level confidence as a proxy for learning status, used to set per-class pseudo-label thresholds and re-sampling weights—is unchanged. What's new is a global threshold formula (Eq. 7), KITTI detection results, and a correlation analysis for detection. The paper is honest about its relationship to the conference version.

The method does several things well. The idea is simple and plausible: classes with low average prediction confidence on unlabeled data are likely under-learned, so lower their threshold and sample them more. The experiments are broad: four classification/detection benchmarks plus KITTI, and the gains are consistent—up to ~7 mAP@0.25 over 3DIoUMatch at 2% labels on ScanNet. The ablation studies show both components (dynamic threshold and re-sampling) contribute, and the comparison against CReST/BiS re-sampling suggests the learning-status-based sampling is not just the usual quantity-based rebalancing. The math is simple and coherent; no obvious issues there.

The soft spots are real. Most importantly, the two key constants—the '2' in the concave mapping M(P_c)=P_c/(2−P_c) and the '2' in the comprehensive threshold τ=exp(−2·P_ave²)—are selected by trying values 1, 2, 3 on the exact datasets and label fractions used in the headline comparisons (ModelNet40 at 10%, ScanNet at 5%, §4.6, Tables 8 and 9). No held-out validation is mentioned. The differences between constant choices are modest (roughly 1–3 points), so this may not be catastrophic, but it does mean the reported numbers are the best of several runs on the test set, and the paper's claim that the comprehensive threshold 'reduces the dependency on the manner of setting threshold value' sits awkwardly next to the tuned exponent.

Two more issues, in decreasing order of importance. Classification results have no error bars, so we can't tell if the 1–3 point mean-accuracy gains are within run-to-run variation. And several recent baselines cited in the text (Ssda3d, ProposalContrast, Dqs3d, Diffusion-SS3D) are missing from the comparisons; the baselines used are primarily from 2021 and the authors' own 2023 paper. No code is released.

On the central hypothesis: the paper shows only visual correlation plots (Figs. 2b and 4) between class-level confidence and test accuracy; no correlation coefficient or significance test. That's a weak spot, but the hypothesis is reasonable and the method works even if the correlation is approximate.

Who should read this? People working on 3D semi-supervised classification or detection, particularly on imbalanced data. It's a useful incremental method, not a paradigm shift. I would send it to peer review—the experiments are substantial and the idea is worth scrutiny—but I'd insist on a validation-based constant selection protocol and error bars for classification.

For your own work, I wouldn't cite the SOTA claim as a given, but the class-level confidence idea might be worth citing as related work once the protocol issues are cleared up.

Best.

Referee Report

4 major / 5 minor

Summary. The paper proposes DyConfidMatch, a semi-supervised learning method for 3D classification and detection. The key idea is to use class-level confidence computed from unlabeled data as a proxy for each class's learning status, then dynamically adjust the pseudo-label threshold per class and re-weight the sampling of unlabeled instances so that underperforming classes receive more training signal. The method is built on top of Pseudo-Labeling/FixMatch for classification and 3DIoUMatch for detection. Experiments on ModelNet40, ScanObjectNN, SUN RGB-D, ScanNet, and KITTI report consistent gains over the compared baselines, especially at low label fractions, and ablations are provided for the dynamic threshold, re-sampling, mapping function, and the two constants that define the thresholds.

Significance. If the reported results are reproducible and the evaluation protocol is sound, DyConfidMatch is a simple, general add-on to pseudo-label SSL that could be useful for imbalanced 3D data. The paper includes extensive experiments across classification and detection, including outdoor detection, and compares against several class-imbalanced SSL re-sampling strategies. The central premise—that class-level confidence tracks test accuracy—is intuitive and the visual evidence (Figs. 2b and 4) is suggestive. However, the empirical claim of surpassing state-of-the-art is currently weakened by test-set hyperparameter selection, missing recent baselines, and an unexplained inconsistency in the core threshold formula. The paper does not provide code or model releases, which limits reproducibility checks.

major comments (4)

[§4.6, Tables 8 and 9] The two constants that drive the method—the denominator 2 in M(P_c)=P_c/(2−P_c) and the exponent 2 in τ=exp(−2·P_ave²)—are selected by ablations performed on ModelNet40 at 10% labels and ScanNet at 5% labels, which are exactly the datasets and label fractions used in the headline comparisons (Tables 1 and 4). No held-out validation split is mentioned. This constitutes test-set selection bias and can inflate the reported margins over baselines that were not tuned in the same way. Please either perform the ablations on a separate validation split and only then evaluate on the test set, or demonstrate that the reported gains are robust across a range of constants (e.g., report performance for constants 1, 2, 3 on the final test sets and show that the conclusion is unchanged).
[§3.3, Eqs. (6)-(7)] The clamping in Eq. (6) is internally inconsistent whenever the comprehensive threshold satisfies τ < 0.5, which occurs when P_ave > sqrt(ln 2 / 2) ≈ 0.589. In that regime, 1−τ > τ, so the conditions 'if M(P_c) < 1−τ' and 'elif M(P_c) > τ' can both be true or both be false for a given M(P_c). Concretely, for P_ave = 0.8 (τ = 0.278, 1−τ = 0.722), a low-confidence class with P_c = 0.3 (M = 0.176) is assigned threshold 0.722, while a high-confidence class with P_c = 0.9 (M = 0.818) is assigned threshold 0.278—the opposite of the intended dynamic. Since many reported final accuracies imply P_ave > 0.589, this is not a corner case. Please clarify the intended clamping (e.g., the bounds should likely be [τ, 1−τ] or the inequalities reversed) and re-run the experiments with the corrected formula, or show that an alternative formulation yields the same results.
[§3.3 and §4.7, Figs. 2b and 4] The load-bearing premise of the method is that class-level confidence P_c from unlabeled data accurately reflects per-class test accuracy throughout training. The only support offered is two visual correlation plots and the qualitative phrase 'strong correlation'; no correlation coefficient, confidence interval, or statistical test is reported for either classification (Fig. 2b) or detection (Fig. 4). If this premise fails for underrepresented or hard classes, both the dynamic thresholds (Eq. 6) and the re-sampling weights (Eq. 10) will systematically re-weight data toward the wrong classes. Please provide quantitative correlation measures (e.g., Pearson and Spearman coefficients with bootstrap confidence intervals) computed on a validation split at several training epochs for both tasks.
[§2.3 and §4.4] The related work section cites several recent 3D SSL detection methods—Ssda3d [33], ProposalContrast [34], Dqs3d [35], and Diffusion-SS3D [36]—but the detection experiments compare only VoteNet, SESS, 3DIoUMatch, and the authors' own prior work. To support the abstract's claim of 'surpasses state-of-the-art counterparts,' at least the most relevant of these should be included in the comparison table, or the authors should explicitly state why they are excluded (e.g., different task settings, unavailability of code). Similarly, the classification comparison omits more recent SSL methods such as DMT, which is cited in the introduction.

minor comments (5)

[Eq. (5)] The definition of the per-class set C_c is malformed: the set-builder notation should specify that u_b ranges over the unlabeled set N_U, and the summation index in P_c should run from 1 to |C_c| rather than from 0 to |C_c|.
[§3.4, Eq. (10)] The quantity in Eq. (10) is called a 'sampling probability,' but the two cases produce values in different ranges (the second case can exceed 1), and no normalization is described. Please clarify whether these are unnormalized sampling weights and how they are used by the data loader.
[§4.9, Fig. 5] The text states that Fig. 5a shows FlexMatch's thresholds and that Fig. 5b shows our method's thresholds, but the figure caption assigns the opposite: (a) is 'Thresholds for each class in the last epoch of our method' and (b) is 'Each class's threshold of FlexMatch.' Please correct the text or the caption so the figure supports the argument.
[§4.2, detection] The sentence 'For scenes with multiple objects, we select the lowest confidence object' is ambiguous; it is unclear whether only one pseudo-labeled object per scene is used and how the selected object is chosen. Please specify the exact handling of multi-object scenes.
[Tables 1 and 2] The tables list both 'Flex-PL [4]' and 'FlexMatch [4]' as separate rows, but the difference between these two baselines is not explained in the text or in the table caption. Please clarify whether Flex-PL denotes FlexMatch applied to Pseudo-Labeling and FlexMatch denotes the original FixMatch-based method, and state which backbone each uses.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported SOTA margins are partly an artifact of selecting the dynamic-threshold constants on the exact evaluation sets, so the headline results are tuned rather than predicted.

fitted input called prediction [Sec. 4.6 (Ablation Study on Other Design Choices), Eq. (6) M(x)=x/(2-x), Eq. (7) tau=exp(-2 P_ave^2), Tables 8 and 9]
"We chose a constant value of 2 because we wanted the threshold to approach 1 as the confidence neared 1 ... As illustrated in Table 8, setting the constant to 2 achieved the best results. ... As illustrated in Table 9, setting the constant to 2 achieved the best results across all evaluated metrics."

The constants used in the dynamic threshold—the denominator 2 in M(P_c)=P_c/(2-P_c) (Eq. 6) and the exponent 2 in tau=exp(-2*P_ave^2) (Eq. 7)—are selected by ablation on ModelNet40 with 10% labeled data and ScanNet with 5% labeled data, which are exactly the settings where Tables 1 and 4 report the headline SOTA improvements. The selection procedure compares constant values 1, 2, and 3 on the same evaluation metrics and datasets that the paper later presents as evidence of superiority. Thus the reported gains are not out-of-sample predictions of a fixed method; they are the output of a hyperparameter-selection process on the evaluation data, making the comparison statistically forced rather than derived from the proposed learning-status principle. No held-out validation split is mentioned.

full rationale

The paper's central derivation—using class-level confidence (Eq. 5) to modulate per-class thresholds and re-sampling weights—is not circular by construction: confidence is an empirical proxy for learning status, and the authors provide only visual evidence, not an equation, for that correlation. The self-citation to the authors' conference version [15] is disclosed and used primarily as a comparative baseline, not as load-bearing evidence for the method's correctness. However, the empirical claim that DyConfidMatch 'surpasses state-of-the-art counterparts' is undermined by the test-set hyperparameter selection in Sec. 4.6: the very constants that define the dynamic threshold are chosen on the datasets and label fractions used for the final reported results. This is a fitted input presented as a prediction, since the constants were tuned against the same metrics later reported as SOTA. The core idea retains independent content, but the headline quantitative claim is partially an artifact of this selection procedure, giving a moderate circularity score.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central method is an empirical heuristic that depends on four tuned constants and on several unverified domain assumptions about confidence calibration and pseudo-label quality. No code or data are released, and the key mapping and global threshold formulas are justified only by test-set ablations.

free parameters (5)

concave mapping constant = 2
In Eq. (6), M(x) = x/(2-x); the ablation in Table 8 selects 2 as best on ModelNet40 10% and ScanNet 5%.
comprehensive threshold exponent constant = 2
In Eq. (7), tau = exp(-2*(Pave)^2); the ablation in Table 9 selects 2 by best test performance.
warm-up decay constant = 5
In Eq. (10), W(e) = exp(-5*(1-e/Emax)^2); borrowed from prior work [40,41] without ablation on 3D tasks.
pseudo-label base threshold = 0.8
Set to 0.8 for both classification and detection in Section 4.2; not tuned in the paper.
re-sampling loader update period = 50 epochs
Data loader is updated every 50 epochs in both tasks (Section 4.2); no ablation on this period.

assumptions (4)

domain assumption Class-level confidence P_c from unlabeled data accurately reflects class learning status (test accuracy).
Introduced in Section 3.3 Eq. (5) and the core hypothesis behind both thresholding and re-sampling. Only support is scatter-plot inspection in Figs. 2 and 4; no correlation coefficient or statistical test is reported.
domain assumption The model's argmax prediction on unlabeled data is a valid pseudo-label source.
Required for Eq. (5), Eq. (8), and the pseudo-label loss in Eq. (9). The paper does not measure confirmation bias or pseudo-label error rates.
domain assumption Softmax confidence is comparable across classes and can be used to derive thresholds.
Underlies Eqs. (6) and (7). No calibration analysis is provided, and the paper itself notes that 3D confidence is generally lower than in 2D, so cross-class comparability is not established.
domain assumption Re-sampling weights in Eq. (10) do not distort the learned distribution in a harmful way.
Eq. (10) defines per-sample weights from W(e), P_c, and p_i; the formula can exceed 1 for low-confidence cases, so the paper implicitly assumes the weights act as unnormalized sampling probabilities without harmful distribution shift.

how reviews work

0 comments

Cite this review

Pith. "Pith review of DyConfidMatch: Dynamic Thresholding and Re-sampling for 3D Semi-supervised Learning." pith.science (2026). https://pith.science/paper/46C2RKZL

@misc{pith2026241108340,
  author       = {Pith},
  title        = {Pith review of: DyConfidMatch: Dynamic Thresholding and Re-sampling for 3D Semi-supervised Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/46C2RKZL}},
  note         = {Machine review of arXiv:2411.08340}
}

read the original abstract

Semi-supervised learning (SSL) leverages limited labeled and abundant unlabeled data but often faces challenges with data imbalance, especially in 3D contexts. This study investigates class-level confidence as an indicator of learning status in 3D SSL, proposing a novel method that utilizes dynamic thresholding to better use unlabeled data, particularly from underrepresented classes. A re-sampling strategy is also introduced to mitigate bias towards well-represented classes, ensuring equitable class representation. Through extensive experiments in 3D SSL, our method surpasses state-of-the-art counterparts in classification and detection tasks, highlighting its effectiveness in tackling data imbalance. This approach presents a significant advancement in SSL for 3D datasets, providing a robust solution for data imbalance issues.

Figures

Figures reproduced from arXiv: 2411.08340 by the authors.

**Figure 1.** The analysis of FixMatch, when trained on the ModelNet40 dataset using only 10% labeled data, reveals the following insights based on fixed threshold values of 0.9 and 0.3: (a) The number of pseudo-labels selected from the unlabeled data varies significantly. With a higher threshold of 0.9, only a small portion of the unlabeled data is incorporated, as the stricter criterion filters out many potential pseudo-labels.… view at source ↗

**Figure 2.** (a)The results comparing FixMatch with 0.3 and 0.9 thresholds, FlexMatch, and our method show that while a lower threshold (0.3) allows for greater utilization of unlabeled data, it also introduces more noise, which negatively impacts performance. In contrast, our method strikes a balance between pseudo-label quality and the effective use of unlabeled data, leading to superior overall performance. (b) There are nota… view at source ↗

**Figure 3.** The proposed method consists of three key components: (1) Leveraging class-level confidence [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Analysis of results for 3DIoUMatch in SUN-RGBD Dataset. The result demonstrates that the [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]

**Figure 5.** Figure 5: (a) Thresholds for each class in the last epoch of our method, applied to the ModelNet40 dataset with 10% labeled data. (b) Each class’s threshold of the FlexMatch in the last epoch. FlexMatch results in long-tail thresholds, introducing significant noise into pseudo …

**Figure 6.** Figure 6: Class-level confidence comparison in unlabeled data between FixMatch and our method, as applied to the ModelNet40 datasets and trained with 10% labeled data, is presented. The findings reveal that our approach not only enhances the learning status for each class but al…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 38 canonical work pages

[33]

Y . Wang, J. Yin, W. Li, P. Frossard, R. Yang, J. Shen, Ssda3d: Semi-supervised domain adaptation for 3d object detection from point cloud, in: Proceedings of the AAAI Conference on Artificial Intelligence, V ol. 37, 2023, pp. 2707–2715

work page 2023
[34]

J. Yin, D. Zhou, L. Zhang, J. Fang, C.-Z. Xu, J. Shen, W. Wang, Proposalcon- trast: Unsupervised pre-training for lidar-based 3d object detection, in: European conference on computer vision, Springer, 2022, pp. 17–33

work page 2022
[35]

H.-a. Gao, B. Tian, P. Li, H. Zhao, G. Zhou, Dqs3d: Densely-matched quantization-aware semi-supervised 3d detection, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 21905– 21915

work page 2023
[36]

Ho, C.-H

C.-J. Ho, C.-H. Tai, Y .-Y . Lin, M.-H. Yang, Y .-H. Tsai, Diffusion-ss3d: Diffusion model for semi-supervised 3d object detection, Advances in Neural Information Processing Systems 36 (2024)

work page 2024
[1]

H. Wang, Y . Cong, O. Litany, Y . Gao, L. J. Guibas, 3dioumatch: Leveraging iou prediction for semi-supervised 3d object detection, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 14615–14624

work page 2021
[2]

Z. Chen, L. Jing, Y . Liang, Y . Tian, B. Li, Multimodal semi-supervised learning for 3d objects, arXiv preprint arXiv:2110.11601 (2021)

work page Pith review arXiv 2021
[3]

K. Sohn, D. Berthelot, C.-L. Li, Z. Zhang, N. Carlini, E. D. Cubuk, A. Kurakin, H. Zhang, C. Ra ffel, Fixmatch: Simplifying semi-supervised learning with con- sistency and confidence, arXiv preprint arXiv:2001.07685 (2020). 26

arXiv 2020
[4]

Zhang, Y

B. Zhang, Y . Wang, W. Hou, H. Wu, J. Wang, M. Okumura, T. Shinozaki, Flex- match: Boosting semi-supervised learning with curriculum pseudo labeling, Ad- vances in Neural Information Processing Systems 34 (2021)

work page 2021

Show all 48 references

[5]

Lee, et al., Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks, in: Workshop on challenges in representation learning, ICML, V ol

D.-H. Lee, et al., Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks, in: Workshop on challenges in representation learning, ICML, V ol. 3, 2013

2013
[6]

Zhao, T.-S

N. Zhao, T.-S. Chua, G. H. Lee, Sess: Self-ensembling semi-supervised 3d object detection, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 11079–11087

2020
[7]

Y . Yang, H. Zhen, J. J. Yang, An information gradient approach to optimizing traffic sensor placement in statewide networks., Information (2078-2489) 15 (10) (2024)

2024
[8]

Huang, Z

Z. Huang, Z. Shangguan, J. Zhang, G. Bar, M. Boyd, E. Ohn-Bar, Assister: Assis- tive navigation via conditional instruction generation, in: European Conference on Computer Vision, Springer, 2022, pp. 271–289

2022
[9]

Xiang, J

J. Xiang, J. Chen, Y . Liu, Hybrid multiscale search for dynamic planning of multi- agent drone traffic, Journal of Guidance, Control, and Dynamics 46 (10) (2023) 1963–1974

2023
[10]

Xiang, J

J. Xiang, J. Chen, Imitation learning-based convex approximations of probabilis- tic reachable sets, in: AIAA A VIATION FORUM AND ASCEND 2024, 2024, p. 4356

2024
[11]

Z. Yao, X. Li, B. Lang, M. C. Chuah, Goal-lbp: Goal-based local behavior guided trajectory prediction for autonomous driving, IEEE Transactions on Intelligent Transportation Systems (2023)

2023
[12]

T. Deng, H. Xie, J. Wang, W. Chen, Long-term visual simultaneous localization and mapping: Using a bayesian persistence filter-based global map prediction, IEEE Robotics & Automation Magazine 30 (1) (2023) 36–49. 27

2023
[13]

T. Deng, Y . Chen, L. Zhang, J. Yang, S. Yuan, D. Wang, W. Chen, Compact 3d gaussian splatting for dense visual slam, arXiv preprint arXiv:2403.11247 (2024)

2024 arXiv
[14]

Y . Zhao, Z. Shangguan, W. Fan, Z. Cao, J. Wang, U-net for satellite image seg- mentation: Improving the weather forecasting, in: 2020 5th International Confer- ence on Universal Village (UV), IEEE, 2020, pp. 1–6

2020
[15]

Z. Chen, L. Jing, L. Yang, Y . Li, B. Li, Class-level confidence based 3d semi- supervised learning, in: Proceedings of the IEEE /CVF Winter Conference on Applications of Computer Vision, 2023, pp. 633–642

2023
[16]

Geiger, P

A. Geiger, P. Lenz, C. Stiller, R. Urtasun, Vision meets robotics: The kitti dataset, The International Journal of Robotics Research 32 (11) (2013) 1231–1237

2013
[17]

Y . Xu, L. Shang, J. Ye, Q. Qian, Y .-F. Li, B. Sun, H. Li, R. Jin, Dash: Semi- supervised learning with dynamic thresholding, in: International Conference on Machine Learning, PMLR, 2021, pp. 11525–11536

2021
[18]

Z. Chen, B. Li, Bridging the domain gap: Self-supervised 3d scene understanding with foundation models, arXiv preprint arXiv:2305.08776 (2023)

2023 arXiv
[19]

Z. Feng, L. Jing, P. Yin, Y . Tian, B. Li, Advancing self-supervised monocular depth learning with sparse lidar, in: Conference on Robot Learning, PMLR, 2022, pp. 685–694

2022
[20]

Z. Feng, Q. Zhou, Q. Gu, X. Tan, G. Cheng, X. Lu, J. Shi, L. Ma, Dmt: Dynamic mutual training for semi-supervised learning, Pattern Recognition 130 (2022) 108777

2022
[21]

Z. Yao, M. C. Chuah, Event-guided low-light video semantic segmentation, arXiv preprint arXiv:2411.00639 (2024)

2024
[22]

W. Xin, K. Wang, Z. Fu, L. Zhou, Let community rules be reflected in online content moderation, arXiv preprint arXiv:2408.12035 (2024)

2024 arXiv
[23]

Z. Fu, K. Wang, W. Xin, L. Zhou, S. Chen, Y . Ge, D. Janies, D. Zhang, Detecting misinformation in multimedia content through cross-modal entity consistency: A 28 dual learning approach, ArXiv abs/2409.00022 (2024). URL https://api.semanticscholar.org/CorpusID:271293900

2024 arXiv
[24]

B. Fang, Y . Li, H. Zhang, J. C.-W. Chan, Collaborative learning of lightweight convolutional neural network and deep clustering for hyperspectral image semi- supervised classification with limited training samples, ISPRS Journal of Pho- togrammetry and Remote Sensing 161 (202...

2020
[25]

Y . Yang, Z. Xu, Rethinking the value of labels for improving class-imbalanced learning, arXiv preprint arXiv:2006.07529 (2020)

2020 arXiv
[26]

M. Hyun, J. Jeong, N. Kwak, Class-imbalanced semi-supervised learning, arXiv preprint arXiv:2002.06815 (2020)

2020 arXiv
[27]

L. Lai, Z. Shangguan, J. Zhang, E. Ohn-Bar, Xvo: Generalized visual odome- try via cross-modal self-training, in: Proceedings of the IEEE /CVF International Conference on Computer Vision, 2023, pp. 10094–10105

2023
[28]

C. Wei, K. Sohn, C. Mellina, A. Yuille, F. Yang, Crest: A class-rebalancing self- training framework for imbalanced semi-supervised learning, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 10857–10866

2021
[29]

J. He, A. Kortylewski, S. Yang, S. Liu, C. Yang, C. Wang, A. Yuille, Re- thinking re-sampling in imbalanced semi-supervised learning, arXiv preprint arXiv:2106.00209 (2021)

2021 arXiv
[30]

R. Li, D. Xue, S. Su, X. He, Q. Mao, Y . Zhu, J. Sun, Y . Zhang, Learning depth via leveraging semantics: Self-supervised monocular depth estimation with both implicit and explicit semantic guidance, Pattern Recognition 137 (2023) 109297

2023
[31]

Lin, C.-T

X. Lin, C.-T. Li, S. Adams, A. Z. Kouzani, R. Jiang, L. He, Y . Hu, M. Vernon, E. Doeven, L. Webb, et al., Self-supervised leaf segmentation under complex lighting conditions, Pattern Recognition 135 (2023) 109021. 29

2023
[32]

J. Yin, J. Fang, D. Zhou, L. Zhang, C.-Z. Xu, J. Shen, W. Wang, Semi-supervised 3d object detection with proficient teachers, in: European Conference on Com- puter Vision, Springer, 2022, pp. 727–743

2022
[37]

Berthelot, N

D. Berthelot, N. Carlini, I. Goodfellow, N. Papernot, A. Oliver, C. Ra ffel, Mixmatch: A holistic approach to semi-supervised learning, arXiv preprint arXiv:1905.02249 (2019)

2019 arXiv
[38]

Berthelot, N

D. Berthelot, N. Carlini, E. D. Cubuk, A. Kurakin, K. Sohn, H. Zhang, C. Ra ffel, Remixmatch: Semi-supervised learning with distribution alignment and augmen- tation anchoring, arXiv preprint arXiv:1911.09785 (2019)

2019 arXiv
[39]

Tarvainen, H

A. Tarvainen, H. Valpola, Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results, arXiv preprint arXiv:1703.01780 (2017)

2017 arXiv
[40]

T. Yu, C. Lan, W. Zeng, M. Feng, Z. Zhang, Z. Chen, Playvirtual: Augment- ing cycle-consistent virtual trajectories for reinforcement learning, Advances in Neural Information Processing Systems 34 (2021) 5276–5289. 30

2021
[41]

J. Ma, D. Yarats, On the adequacy of untuned warmup for adaptive optimization, in: Proceedings of the AAAI Conference on Artificial Intelligence, V ol. 35, 2021, pp. 8828–8836

2021
[42]

Guo, J.-X

M.-H. Guo, J.-X. Cai, Z.-N. Liu, T.-J. Mu, R. R. Martin, S.-M. Hu, Pct: Point cloud transformer, Computational Visual Media 7 (2) (2021) 187–199

2021
[43]

M. A. Uy, Q.-H. Pham, B.-S. Hua, D. T. Nguyen, S.-K. Yeung, Revisiting point cloud classification: A new benchmark dataset and classification model on real- world data, in: International Conference on Computer Vision (ICCV), 2019

2019
[44]

Z. Wu, S. Song, A. Khosla, F. Yu, L. Zhang, X. Tang, J. Xiao, 3d shapenets: A deep representation for volumetric shapes, in: Proceedings of the IEEE confer- ence on computer vision and pattern recognition, 2015, pp. 1912–1920

2015
[45]

S. Song, S. P. Lichtenberg, J. Xiao, Sun rgb-d: A rgb-d scene understanding benchmark suite, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 567–576

2015
[46]

A. Dai, A. X. Chang, M. Savva, M. Halber, T. Funkhouser, M. Nießner, Scannet: Richly-annotated 3d reconstructions of indoor scenes, in: Proc. Computer Vision and Pattern Recognition (CVPR), IEEE, 2017

2017
[47]

S. Shi, C. Guo, L. Jiang, Z. Wang, J. Shi, X. Wang, H. Li, Pv-rcnn: Point-voxel feature set abstraction for 3d object detection, in: Proceedings of the IEEE /CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 10529– 10538

2020
[48]

C. R. Qi, O. Litany, K. He, L. J. Guibas, Deep hough voting for 3d object detection in point clouds, in: Proceedings of the IEEE /CVF International Conference on Computer Vision, 2019, pp. 9277–9286. 31

2019

Pith tools

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

[1] [33]

Y . Wang, J. Yin, W. Li, P. Frossard, R. Yang, J. Shen, Ssda3d: Semi-supervised domain adaptation for 3d object detection from point cloud, in: Proceedings of the AAAI Conference on Artificial Intelligence, V ol. 37, 2023, pp. 2707–2715

work page 2023

[2] [34]

J. Yin, D. Zhou, L. Zhang, J. Fang, C.-Z. Xu, J. Shen, W. Wang, Proposalcon- trast: Unsupervised pre-training for lidar-based 3d object detection, in: European conference on computer vision, Springer, 2022, pp. 17–33

work page 2022

[3] [35]

H.-a. Gao, B. Tian, P. Li, H. Zhao, G. Zhou, Dqs3d: Densely-matched quantization-aware semi-supervised 3d detection, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 21905– 21915

work page 2023

[4] [36]

Ho, C.-H

C.-J. Ho, C.-H. Tai, Y .-Y . Lin, M.-H. Yang, Y .-H. Tsai, Diffusion-ss3d: Diffusion model for semi-supervised 3d object detection, Advances in Neural Information Processing Systems 36 (2024)

work page 2024

[5] [1]

H. Wang, Y . Cong, O. Litany, Y . Gao, L. J. Guibas, 3dioumatch: Leveraging iou prediction for semi-supervised 3d object detection, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 14615–14624

work page 2021

[6] [2]

Z. Chen, L. Jing, Y . Liang, Y . Tian, B. Li, Multimodal semi-supervised learning for 3d objects, arXiv preprint arXiv:2110.11601 (2021)

work page Pith review arXiv 2021

[7] [3]

K. Sohn, D. Berthelot, C.-L. Li, Z. Zhang, N. Carlini, E. D. Cubuk, A. Kurakin, H. Zhang, C. Ra ffel, Fixmatch: Simplifying semi-supervised learning with con- sistency and confidence, arXiv preprint arXiv:2001.07685 (2020). 26

arXiv 2020

[8] [4]

Zhang, Y

B. Zhang, Y . Wang, W. Hou, H. Wu, J. Wang, M. Okumura, T. Shinozaki, Flex- match: Boosting semi-supervised learning with curriculum pseudo labeling, Ad- vances in Neural Information Processing Systems 34 (2021)

work page 2021