Pith. sign in

REVIEW 3 major objections 5 minor 53 references

Exploring Modality Guidance to Enhance VFM-based Feature Fusion for UDA in 3D Semantic Segmentation

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Modality-guided fusion, with a hand-set bias toward whichever sensor is more reliable in the target environment, raises unsupervised domain adaptation for 3D semantic segmentation to a new average of 69.0 mIoU, 6.5 points above the…

desk verdict Solid incremental UDA paper with a useful three-stream fusion idea, but the +6.5 mIoU headline overstates what the proposed guidance contributes; the guidance itself adds about 1 mIoU over symmetric alignment. read the letter →

arxiv 2504.14231 v1 pith:W4WPOEM3 submitted 2025-04-19 cs.CV

Johannes Spoecklberger , Wei Lin , Pedro Hermosilla , Sivan Doveh , Horst Possegger , M. Jehanzeb Mirza This is my paper · ORCID
classification cs.CV
keywords unsuperviseddomainadaptation3DsemanticsegmentationLiDARpointcloudsvisionfoundationmodelscross-modalfusionmodalityguidancepredictivedistillationself-training
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 the usual way of fusing camera and LiDAR features for 3D segmentation fails under domain shift because the fusion network over-relies on features that worked on the labeled source domain. The proposed fix is a fusion branch that is actively pulled, by a single scalar lambda, toward imitating whichever modality is more trustworthy in the target environment: image features in daylight, LiDAR in darkness. When combined with a frozen vision foundation model and a pseudo-label self-training stage, the method reports an average of 69.0 mIoU across four standard adaptation tasks, which is 6.5 points above the previous best average. The reason a non-expert should care is that this would let a self-driving perception system adapt to a new city, weather, or even a virtual-to-real shift without any new point-cloud labels.

What carries the argument

The load-bearing mechanism is the fusion branch: a two-hidden-layer MLP with batch normalization, GeLU, and dropout that takes the concatenation of the 3D features and the frozen VFM-derived 2D features, where a mimicry head is an auxiliary segmentation head used only for aligning predictions across branches. It is trained with a guidance loss of the form $\mathcal{L}_{\mathrm{guide}} = \lambda \cdot \mathrm{KL}(p^{\mathrm{2D}}_{\mathrm{main}} \,\|\, p^{\mathrm{fuse}}_{\mathrm{mmc}}) + (1-\lambda)\cdot \mathrm{KL}(p^{\mathrm{3D}}_{\mathrm{main}} \,\|\, p^{\mathrm{fuse}}_{\mathrm{mmc}})$, where $\lambda$ implements the environmental prior and a separate alignment loss makes the 3D mimicry head imitate the fusion main head. The final prediction is the softmax average of the fusion and 3D main heads, and a self-training stage generates pseudo-labels from that same average.

What would settle it

Run the method on a target scene with mixed or unknown lighting, such as dusk, a tunnel, or rain at night, and compare lambda=1, lambda=0, and a symmetric-alignment version with no guidance. If neither endpoint beats the no-guidance version by a clear margin, or if the wrong endpoint drops accuracy well below it, the claim that modality pre-selection drives the improvement is falsified. A second check is a full lambda sweep on the Day-to-Night task: a plateau across lambda would contradict the story that the specific bias matters.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that the weakness of multi-modal 3D domain adaptation is not insufficient fusion capacity but a missing guidance signal: a fusion branch trained on source labels alone drifts on the target and chooses the wrong modality to trust. The authors therefore train a three-branch network with main heads and mimicry heads, where the fusion branch's mimicry head is regularized by a KL-divergence loss against either the 2D or the 3D main head, selected by a scalar lambda that encodes an environmental prior. With lambda=1 on daylight target domains, the fusion is biased toward the frozen vision foundation model's image features; with lambda=0 on night tasks, it is biased toward the 3D LiDAR stream. Across the four benchmark tasks, this modality-guided fusion plus two-stage pseudo-label self-training gives a reported average mIoU of 69.0, improving on the previous state of the art by 6.5 mIoU on average.

Load-bearing premise

The load-bearing premise is that the target domain's lighting condition is known in advance and that the single coefficient lambda is set correctly for it; if the environment is ambiguous or the prior is wrong, the reported gains are not shown to hold.

Editorial extensions

If this is right

  • If the claim holds, a frozen vision foundation model can replace a fully trained 2D encoder in cross-modal 3D UDA, eliminating the cost of training a separate image network.
  • A practitioner only needs a coarse environmental prior, such as day or night, to set the single coefficient lambda; no per-point or per-scene labels are required.
  • The reported gains span geographic, day-night, virtual-to-real, and sensor-difference shifts, with the largest margin on the virtual-to-real task, suggesting the guidance helps most where source and target sensors disagree most.
  • Because swapping the vision foundation model changes results by only one to two mIoU points, the method is positioned to inherit future improvements in foundation models without architectural change.

Reading between the lines

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

  • If the global day/night prior is the true driver, then estimating modality reliability per point or per local region, which the paper names as its own future direction, should outperform the single global lambda, and this is directly testable.
  • The lambda selection is currently an oracle on the environment; one testable extension is a lightweight classifier on target features that predicts lighting or modality confidence and sets lambda automatically, which would remove the main practical limitation.
  • The guidance loss on target data acts as a consistency regularizer, so a natural ablation is to corrupt or remove the source supervised labels to see whether the alignment terms alone carry the adaptation signal.
  • A sensitivity curve of lambda across [0,1] on each task would reveal whether the method is forgiving or knife-edged; the paper reports only the two endpoint choices.
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 a three-branch architecture for unsupervised domain adaptation in 3D semantic segmentation: a frozen 2D vision foundation model (AM-RADIO), a trainable 3D SparseConvNet, and a fusion branch that concatenates 2D and 3D features. The fusion branch is supervised on the source domain and regularized by KL-divergence losses to mimic either the 2D or the 3D main head, with a hand-set coefficient lambda that biases toward the 2D modality on daylight target domains and toward the 3D modality on night-time target domains. Training also includes a second stage with pseudo-label self-training on the target domain. The paper reports state-of-the-art results on four UDA benchmarks, including an average improvement of 6.5 mIoU over the previous state-of-the-art (UniDSeg).

Significance. If the central claim is supported, the paper would make a useful empirical contribution by showing that a frozen VFM can be integrated into a multi-modal UDA pipeline through a guided fusion branch, and the comparison across two VFMs (AM-RADIO and DINOv2) is a strength. The ablations in Table 2 and Table 3 are clearly presented and isolate several design choices (vanilla vs. MLP fusion, symmetric alignment vs. modality guidance, and xMUDA-style fusion with the same VFM). However, the manuscript currently overstates the role of the proposed guidance: the headline gain is not attributed to the modality-guidance mechanism, and the hand-set lambda makes the method task-specific rather than adaptive. These issues need to be resolved before the SOTA claim is credible.

major comments (3)
  1. [Section 4.3, Table 1] The headline '+6.5 mIoU' compares OursPL (two-stage pseudo-label self-training) against UniDSeg, which is a single-stage method. The proposed modality guidance is not the main contributor to this gap: Table 2 shows that MLP+MG improves over MLP+SymAl by only 0.8 mIoU on USA→Singapore and 1.4 mIoU on VK→SK, and the remaining margin to UniDSeg comes largely from the frozen AM-RADIO VFM and the second-stage pseudo-label training. The paper should report an ablation of Ours without the PL stage and an OursPL variant with symmetric alignment, so that the contribution of the guidance can be separated from the other components.
  2. [Section 4.2, Eq. (3)] The core mechanism of the paper, the modality-guidance coefficient lambda, is set to 1 for daylight target tasks and 0 for night tasks based on the known target domain. This makes the method task-specific rather than adaptive, and the paper's own Limitations section admits that the approach 'may be less effective in ambiguous conditions.' As written, the evidence only shows that two hand-picked lambda values work on four benchmarks; a sensitivity analysis over lambda (e.g., 0, 0.25, 0.5, 0.75, 1) and an experiment that estimates modality reliability from data would be needed to support the claim that the guidance mechanism is robust and generally applicable.
  3. [Section 4.3, Table 1 caption] The caption states that for the '2D' column of Ours, the reported result is the output of the fusion network, not a pure 2D branch. Therefore, the three columns (2D, 3D, 2D3D) are not independent, and the average mIoU computed across all three columns double-counts the fusion branch. This inflates the reported aggregate improvement (e.g., OursPL 69.0 vs. UniDSeg 62.5). The average should be recomputed using a single representative column, such as the 2D3D softmax average, or the aggregation method should be justified explicitly.
minor comments (5)
  1. [Section 4.2 and throughout] The model name is spelled inconsistently as 'AM-Radio' in the text and 'AM-RADIO' in the related work; please unify the spelling.
  2. [Section 4.2, paragraph on VFM generalization] The sentence 'we observe minimal performance fluctuation' is vague; the actual numbers are given in Figure 4, so the text should refer to the figure explicitly and state the direction of the change for each task.
  3. [Table 2 and Table 3] No standard deviations or number of seeds are reported for the main results or ablations. Given that the proposed guidance adds only 0.8–1.4 mIoU in the ablation, variance estimates are important for assessing whether the improvement is significant.
  4. [Equation (6)] The pseudo-label formula uses a simple average of softmax scores; the paper should clarify whether these are temperature-scaled and whether any confidence threshold is applied before using the pseudo-labels as supervision.
  5. [Figure 4 caption] The caption says 'We report the mIoU % for our main heads including the VFM head utilized for the fusion regularization,' but it is unclear which heads correspond to which bars; please label the figure more explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported gains are evaluated against external baselines, and the hand-set λ is a disclosed design choice rather than a fitted prediction.

full rationale

The paper's central comparisons are against external methods (xMUDA, MM2D3D, FtD++, UniDSeg, etc.) on four standard benchmarks, so the headline +6.5 mIoU is not produced by re-inserting the paper's own outputs as inputs. The modality-guidance coefficient λ is explicitly set in Sec. 4.2 ('The fusion modality guidance λ is set to 1 for adaptation in daylight target domain tasks and 0 for night tasks') and its dependence on a known environmental prior is acknowledged in the Limitations ('The fusion guidance relies on a predefined modality preference... may be less effective in ambiguous conditions'); it is a disclosed hyperparameter, not a hidden fit renamed as a prediction. The ablation in Table 2 isolates the guidance component against symmetric alignment, and although the gain is small, this is a standard component analysis rather than a circular derivation. The only overlapping-author reference ([32], Mirza et al.) appears in a list of normalization-based adaptation methods and is not load-bearing for the core approach. Table 1's note that the '2D' column for Ours reports the fusion head is a reporting and attribution caveat that can inflate the aggregate margin, but it is not an equation-level reduction of a claimed result to its own inputs. No load-bearing self-citation chain or imported uniqueness theorem is present.

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

The method's contributions rest mostly on the choice of lambda and the pretrained VFM, not on new theoretical constructs. No new physical entities are introduced.

free parameters (3)
  • Modality guidance coefficient lambda = 1 for daylight target tasks (USA to Singapore, VK to SK, A2D2 to SK); 0 for night task (Day to Night)
    Chosen by hand per task based on known target lighting conditions; it selects whether the fusion branch imitates the 2D VFM or the 3D branch. This is the defining hyperparameter of the method.
  • Source and target alignment coefficients lambda_S, lambda_T = lambda_S=1, lambda_T=0.1 for nuScenes tasks; lambda_S=0.5, lambda_T=0.5 for A2D2 to SK and VK to SK
    Tuned per benchmark without reported sensitivity analysis.
  • Pseudo-label loss weight lambda_PL = 1
    Set for all datasets; standard self-training coefficient.
assumptions (4)
  • domain assumption Features from frozen AM-RADIO VFM generalize across domains and are informative for LiDAR segmentation
    Stated in Section 3.1 and used throughout; the entire guidance toward the 2D branch for daylight tasks relies on this.
  • domain assumption Target-domain lighting condition is known a priori
    The modality guidance coefficient lambda is set to 1 or 0 depending on whether the target is nighttime (Day to Night) or daylight. Invoked in Section 4.2.
  • domain assumption KL divergence between main-head and mimicry-head predictions is an effective alignment for cross-modal UDA
    Adopted from xMUDA [18] and used in Eqs. (1) to (3); not re-derived.
  • standard math Standard probability and optimization identities (KL, softmax, cross-entropy) hold
    Used in loss definitions, Eqs. (1) to (6).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exploring Modality Guidance to Enhance VFM-based Feature Fusion for UDA in 3D Semantic Segmentation." pith.science (2026). https://pith.science/paper/W4WPOEM3

@misc{pith2026250414231,
  author       = {Pith},
  title        = {Pith review of: Exploring Modality Guidance to Enhance VFM-based Feature Fusion for UDA in 3D Semantic Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W4WPOEM3}},
  note         = {Machine review of arXiv:2504.14231}
}
read the original abstract

Vision Foundation Models (VFMs) have become a de facto choice for many downstream vision tasks, like image classification, image segmentation, and object localization. However, they can also provide significant utility for downstream 3D tasks that can leverage the cross-modal information (e.g., from paired image data). In our work, we further explore the utility of VFMs for adapting from a labeled source to unlabeled target data for the task of LiDAR-based 3D semantic segmentation. Our method consumes paired 2D-3D (image and point cloud) data and relies on the robust (cross-domain) features from a VFM to train a 3D backbone on a mix of labeled source and unlabeled target data. At the heart of our method lies a fusion network that is guided by both the image and point cloud streams, with their relative contributions adjusted based on the target domain. We extensively compare our proposed methodology with different state-of-the-art methods in several settings and achieve strong performance gains. For example, achieving an average improvement of 6.5 mIoU (over all tasks), when compared with the previous state-of-the-art.

Figures

Figures reproduced from arXiv: 2504.14231 by the authors.

Figure 1
Figure 1. (a) Cross modal learning with frozen 2D (VFM) back￾bone features using a learned fusion representation. Fusion net￾works can lead to a suboptimal feature utilization and unwanted modality bias on the target domain. Therefore, we propose reg￾ularizing the fusion by the most effective modality in a certain environment (e.g., based on lighting conditions). (b) mIoU Com￾parison of xMUDA with different fusion variants an… view at source ↗
Figure 2
Figure 2. Our architecture for the cross-modal learning consists of a Vision Foundation Model (VFM) as the 2D encoder and a 3D [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison of our method on an example from each dataset. We show the softmax average of our fusion and [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Comparison of current SOTA VFMs on USA → Sing. and VK → SK. We report the mIoU % for our main heads includ￾ing the VFM head utilized for the fusion regularization. ered fusion. The MLP fusion is improved when aligning the fusion from both the VFM and the 3D teacher, su…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 45 canonical work pages

  1. [1]

    Beit: Bert pre-training of image transformers

    Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei. Beit: Bert pre-training of image transformers. In ICLR, 2022. 2

    work page 2022

  2. [2]

    Cycle and seman- tic consistent adversarial domain adaptation for reducing simulation-to-real domain shift in lidar bird’s eye view

    Alejandro Barrera, Jorge Beltr ´an, Carlos Guindel, Jose An- tonio Iglesias, and Fernando Garc ´ıa. Cycle and seman- tic consistent adversarial domain adaptation for reducing simulation-to-real domain shift in lidar bird’s eye view. In ITSC, pages 3081–3086, 2021. 2

    work page 2021

  3. [3]

    Se- mantickITTI: A dataset for semantic scene understanding of LiDAR sequences

    Jens Behley, Martin Garbade, Andres Milioto, Jan Quen- zel, Sven Behnke, Cyrill Stachniss, and J ¨urgen Gall. Se- mantickITTI: A dataset for semantic scene understanding of LiDAR sequences. In CVPR, pages 9297–9307, 2019. 5

    work page 2019

  4. [4]

    CAFuser: Condition-Aware Multimodal Fusion for Robust Semantic Perception of Driving Scenes

    Tim Br ¨odermann, Christos Sakaridis, Yuqian Fu, and Luc Van Gool. Condition-aware multimodal fusion for ro- bust semantic perception of driving scenes. arXiv preprint arXiv:2410.10791, 2024. 2, 3

    work page Pith review arXiv 2024

  5. [5]

    nuscenes: A mul- timodal dataset for autonomous driving

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuscenes: A mul- timodal dataset for autonomous driving. In CVPR, pages 11621–11631, 2020. 5

    2020

  6. [6]

    Mopa: Multi-modal prior aided domain adaptation for 3d semantic segmentation

    Haozhi Cao, Yuecong Xu, Jianfei Yang, Pengyu Yin, Sheng- hai Yuan, and Lihua Xie. Mopa: Multi-modal prior aided domain adaptation for 3d semantic segmentation. In ICRA, pages 9463–9470, 2024. 3

    work page 2024

  7. [7]

    Exploiting the complementarity of 2d and 3d networks to address domain-shift in 3d semantic segmentation

    Adriano Cardace, Pierluigi Zama Ramirez, Samuele Salti, and Luigi Di Stefano. Exploiting the complementarity of 2d and 3d networks to address domain-shift in 3d semantic segmentation. In CVPR Workshop, pages 98–109, 2023. 2, 4, 6

    work page 2023

  8. [8]

    Emerg- ing properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing properties in self-supervised vision transformers. In CVPR, pages 9650–9660, 2021. 2

    work page 2021

Show all 53 references
  1. [9]

    Self-training avoids using spurious features under domain shift

    Yining Chen, Colin Wei, Ananya Kumar, and Tengyu Ma. Self-training avoids using spurious features under domain shift. In NeurIPS, pages 21061–21071, 2020. 5

    2020

  2. [10]

    Stargan: Unified genera- tive adversarial networks for multi-domain image-to-image translation

    Yunjey Choi, Minje Choi, Munyoung Kim, Jung-Woo Ha, Sunghun Kim, and Jaegul Choo. Stargan: Unified genera- tive adversarial networks for multi-domain image-to-image translation. In CVPR, pages 8789–8797, 2018. 2

    2018

  3. [11]

    An image is worth 16x16 words: Trans- formers for image recognition at scale

    Alexey Dosovitskiy. An image is worth 16x16 words: Trans- formers for image recognition at scale. In ICLR, 2021. 2

    2021

  4. [12]

    Virtual worlds as proxy for multi-object tracking anal- ysis

    Adrien Gaidon, Qiao Wang, Yohann Cabon, and Eleonora Vig. Virtual worlds as proxy for multi-object tracking anal- ysis. In CVPR, pages 4340–4349, 2016. 5

    2016

  5. [13]

    Domain-adversarial training of neural networks

    Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pas- cal Germain, Hugo Larochelle, Franc ¸ois Laviolette, Mario March, and Victor Lempitsky. Domain-adversarial training of neural networks. Journal of Machine Learning Research, 17(59):1–35, 2016. 2

    2016

  6. [14]

    A2d2: Audi autonomous driving dataset

    Jakob Geyer, Yohannes Kassahun, Mentar Mahmudi, Xavier Ricou, Rupesh Durgesh, Andrew S Chung, Lorenz Hauswald, Viet Hoang Pham, Maximilian M ¨uhlegg, Sebas- tian Dorn, et al. A2d2: Audi autonomous driving dataset. arXiv preprint arXiv:2004.06320, 2020. 5

    2004 arXiv

  7. [15]

    3d semantic segmentation with submanifold sparse convolutional networks

    Benjamin Graham, Martin Engelcke, and Laurens Van Der Maaten. 3d semantic segmentation with submanifold sparse convolutional networks. In CVPR, pages 9224–9232,

  8. [16]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR, pages 770–778, 2016. 1

    2016

  9. [17]

    Cycada: Cycle-consistent adversarial domain adaptation

    Judy Hoffman, Eric Tzeng, Taesung Park, Jun-Yan Zhu, Phillip Isola, Kate Saenko, Alexei Efros, and Trevor Darrell. Cycada: Cycle-consistent adversarial domain adaptation. In ICML, pages 1989–1998, 2018. 2

    1989

  10. [18]

    xMUDA: Cross-modal unsuper- vised domain adaptation for 3D semantic segmentation

    Maximilian Jaritz, Tuan-Hung Vu, Raoul de Charette, Emilie Wirbel, and Patrick P ´erez. xMUDA: Cross-modal unsuper- vised domain adaptation for 3D semantic segmentation. In CVPR, 2020. 2, 3, 4, 6, 7

    2020

  11. [19]

    Cross-modal learning for domain adaptation in 3D semantic segmentation

    Maximilian Jaritz, Tuan-Hung Vu, Raoul de Charette, Emilie Wirbel, and Patrick P´erez. Cross-modal learning for domain adaptation in 3D semantic segmentation. In PAMI, 2022. 2, 4, 5, 6, 8

    2022

  12. [20]

    Scaling up visual and vision-language representation learning with noisy text supervision

    Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc Le, Yun-Hsuan Sung, Zhen Li, and Tom Duerig. Scaling up visual and vision-language representation learning with noisy text supervision. In ICML, pages 4904– 4916, 2021. 2

    2021

  13. [21]

    Segment any- thing

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment any- thing. In CVPR, pages 4015–4026, 2023. 2

    2023

  14. [22]

    Temporal ensembling for semi- supervised learning

    Samuli Laine and Timo Aila. Temporal ensembling for semi- supervised learning. ICLR, 2017. 2

    2017

  15. [23]

    Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks

    Dong-Hyun Lee et al. Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks. In ICML Workshop, page 896, 2013. 2

    2013

  16. [24]

    Mseg3d: Multi-modal 3d semantic segmentation for autonomous driv- ing

    Jiale Li, Hang Dai, Hao Han, and Yong Ding. Mseg3d: Multi-modal 3d semantic segmentation for autonomous driv- ing. In CVPR, pages 21694–21704, 2023. 2

    2023

  17. [25]

    Adaptive batch normalization for practical do- main adaptation

    Yanghao Li, Naiyan Wang, Jianping Shi, Xiaodi Hou, and Jiaying Liu. Adaptive batch normalization for practical do- main adaptation. Pattern Recognition, 80:109–117, 2018. 2

    2018

  18. [26]

    Cycle self-training for domain adaptation

    Hong Liu, Jianmin Wang, and Mingsheng Long. Cycle self-training for domain adaptation. NeurIPS, pages 22968– 22981, 2021. 5

    2021

  19. [27]

    Adversarial unsupervised domain adaptation for 3d semantic segmentation with multi-modal learning

    Wei Liu, Zhiming Luo, Yuanzheng Cai, Ying Yu, Yang Ke, Jos´e Marcato Junior, Wesley Nunes Gonc ¸alves, and Jonathan Li. Adversarial unsupervised domain adaptation for 3d semantic segmentation with multi-modal learning. ISPRS Journal of Photogrammetry and Remote Sensing , 176:2...

    2021

  20. [28]

    Segment any point cloud sequences by distilling vision foundation models

    Youquan Liu, Lingdong Kong, Jun Cen, Runnan Chen, Wen- wei Zhang, Liang Pan, Kai Chen, and Ziwei Liu. Segment any point cloud sequences by distilling vision foundation models. In NeurIPS, pages 37193–37229, 2023. 5

    2023

  21. [29]

    A convnet for the 2020s

    Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feicht- enhofer, Trevor Darrell, and Saining Xie. A convnet for the 2020s. In CVPR, pages 11976–11986, 2022. 1

    2022

  22. [30]

    In- stance adaptive self-training for unsupervised domain adap- tation

    Ke Mei, Chuang Zhu, Jiaqi Zou, and Shanghang Zhang. In- stance adaptive self-training for unsupervised domain adap- tation. In ECCV, pages 415–430, 2020. 5

    2020

  23. [31]

    Saluda: Surface- based automotive lidar unsupervised domain adaptation

    Bj ¨orn Michele, Alexandre Boulch, Gilles Puy, Tuan-Hung Vu, Renaud Marlet, and Nicolas Courty. Saluda: Surface- based automotive lidar unsupervised domain adaptation. In 3DV, pages 421–431, 2024. 2

    2024

  24. [32]

    The norm must go on: Dynamic unsuper- vised domain adaptation by normalization

    M Jehanzeb Mirza, Jakub Micorek, Horst Possegger, and Horst Bischof. The norm must go on: Dynamic unsuper- vised domain adaptation by normalization. In CVPR, pages 14765–14775, 2022. 2

    2022

  25. [33]

    Minimal-entropy correlation alignment for unsupervised deep domain adaptation

    Pietro Morerio, Jacopo Cavazza, and Vittorio Murino. Minimal-entropy correlation alignment for unsupervised deep domain adaptation. In ICLR, 2018. 2

    2018

  26. [34]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193, 2023. 2, 5

    2023 arXiv

  27. [35]

    Sparse-to-dense feature matching: Intra and inter do- main cross-modal learning in domain adaptation for 3d se- mantic segmentation

    Duo Peng, Yinjie Lei, Wen Li, Pingping Zhang, and Yulan Guo. Sparse-to-dense feature matching: Intra and inter do- main cross-modal learning in domain adaptation for 3d se- mantic segmentation. In CVPR, pages 7108–7117, 2021. 2, 6

    2021

  28. [36]

    Learning to adapt sam for segmenting cross-domain point clouds

    Xidong Peng, Runnan Chen, Feng Qiao, Lingdong Kong, Youquan Liu, T Wang, X Zhu, and Y Ma. Learning to adapt sam for segmenting cross-domain point clouds. In ECCV,

  29. [37]

    Three pillars improving vision foundation model distillation for lidar

    Gilles Puy, Spyros Gidaris, Alexandre Boulch, Oriane Sim´eoni, Corentin Sautier, Patrick P´erez, Andrei Bursuc, and Renaud Marlet. Three pillars improving vision foundation model distillation for lidar. In CVPR, pages 21519–21529,

  30. [38]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In ICML, pages 8748–8763, 2021. 2

    2021

  31. [39]

    Am-radio: Agglomerative vision foundation model reduce all domains into one

    Mike Ranzinger, Greg Heinrich, Jan Kautz, and Pavlo Molchanov. Am-radio: Agglomerative vision foundation model reduce all domains into one. In CVPR, pages 12490– 12500, 2024. 2, 5

    2024

  32. [40]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In MICCAI, pages 234–241, 2015. 5

    2015

  33. [41]

    Image-to-lidar self-supervised distillation for autonomous driving data

    Corentin Sautier, Gilles Puy, Spyros Gidaris, Alexandre Boulch, Andrei Bursuc, and Renaud Marlet. Image-to-lidar self-supervised distillation for autonomous driving data. In CVPR, pages 9891–9901, 2022. 5

    2022

  34. [42]

    Deep coral: Correlation alignment for deep domain adaptation

    Baochen Sun and Kate Saenko. Deep coral: Correlation alignment for deep domain adaptation. In ECCV Workshop, pages 443–450, 2016. 2

    2016

  35. [43]

    Cross-modal unsu- pervised domain adaptation for 3d semantic segmentation via bidirectional fusion-then-distillation

    Yao Wu, Mingwei Xing, Yachao Zhang, Yuan Xie, Jianping Fan, Zhongchao Shi, and Yanyun Qu. Cross-modal unsu- pervised domain adaptation for 3d semantic segmentation via bidirectional fusion-then-distillation. In ACMMM, pages 490–498, 2023. 2, 3, 6

    2023

  36. [44]

    Unidseg: Unified cross-domain 3d semantic segmentation via visual foundation models prior

    Yao Wu, Mingwei Xing, Yachao Zhang, Xiaotong Luo, Yuan Xie, and Yanyun Qu. Unidseg: Unified cross-domain 3d semantic segmentation via visual foundation models prior. In NeurIPS, pages 101223–101249, 2024. 3, 6

    2024

  37. [45]

    Fusion-then-distillation: Toward cross-modal positive distillation for domain adaptive 3d semantic seg- mentation, 2024

    Yao Wu, Mingwei Xing, Yachao Zhang, Yuan Xie, and Yanyun Qu. Fusion-then-distillation: Toward cross-modal positive distillation for domain adaptive 3d semantic seg- mentation, 2024. arXiv preprint. 2, 3, 6

    2024

  38. [46]

    Cross-modal contrastive learning for domain adaptation in 3d semantic segmentation

    Bowei Xing, Xianghua Ying, Ruibin Wang, Jinfa Yang, and Taiyan Chen. Cross-modal contrastive learning for domain adaptation in 3d semantic segmentation. In AAAI, pages 2974–2982, 2023. 2

    2023

  39. [47]

    Visual foundation models boost cross-modal unsupervised domain adaptation for 3d semantic segmentation, 2024

    Jingyi Xu, Weidong Yang, Lingdong Kong, Youquan Liu, Rui Zhang, Qingyuan Zhou, and Ben Fei. Visual foundation models boost cross-modal unsupervised domain adaptation for 3d semantic segmentation, 2024. arXiv preprint. 3, 6

    2024

  40. [48]

    Complete & label: A domain adaptation approach to semantic segmen- tation of lidar point clouds

    Li Yi, Boqing Gong, and Thomas Funkhouser. Complete & label: A domain adaptation approach to semantic segmen- tation of lidar point clouds. In CVPR, pages 15363–15373,

  41. [49]

    Prototype-guided multitask adversarial network for cross-domain lidar point clouds semantic segmentation

    Zhimin Yuan, Ming Cheng, Wankang Zeng, Yanfei Su, Wei- quan Liu, Shangshu Yu, and Cheng Wang. Prototype-guided multitask adversarial network for cross-domain lidar point clouds semantic segmentation. IEEE Transactions on Geo- science and Remote Sensing, pages 1–13, 2023. 2

    2023

  42. [50]

    Wide residual net- works, 2017

    Sergey Zagoruyko and Nikos Komodakis. Wide residual net- works, 2017. 1

    2017

  43. [51]

    Self-supervised ex- clusive learning for 3d segmentation with cross-modal unsu- pervised domain adaptation

    Yachao Zhang, Miaoyu Li, Yuan Xie, Cuihua Li, Cong Wang, Zhizhong Zhang, and Yanyun Qu. Self-supervised ex- clusive learning for 3d segmentation with cross-modal unsu- pervised domain adaptation. In ACMMM, pages 3338–3346,

  44. [52]

    Segment everything everywhere all at once

    Xueyan Zou, Jianwei Yang, Hao Zhang, Feng Li, Linjie Li, Jianfeng Wang, Lijuan Wang, Jianfeng Gao, and Yong Jae Lee. Segment everything everywhere all at once. InNeurIPS,

  45. [53]

    Confidence regularized self-training

    Yang Zou, Zhiding Yu, Xiaofeng Liu, BVK Kumar, and Jin- song Wang. Confidence regularized self-training. In CVPR, pages 5982–5991, 2019. 5

    2019

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.