Pith. sign in

REVIEW 1 major objections 6 minor 63 references

STAR: A Spatial-Topology Aware Routing Framework for Generalizable 3D Scene Understanding

T0 review · 1 major / 6 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 The paper claims that routing 3D MoE experts by local spatial topology, rather than by task features alone, improves multi-domain scene understanding, with 80.1% mIoU on ScanNet and 77.2% on S3DIS.

desk verdict A plausible, incremental MoE routing paper whose central DSR mechanism is underspecified enough that the reported gains should be treated as provisional until the sparse-tensor construction and a coordinate-shuffle control are shown. read the letter →

arxiv 2608.11699 v1 pith:E3M6JK7Q submitted 2026-08-12 cs.CV

Mingwei Xing , Xinliang Wang , Yifeng Shi This is my paper · ORCID
classification cs.CV
keywords 3DsceneunderstandingMixtureofExpertsspatial-topologyroutingdomaingeneralizationpointcloudsegmentationself-supervisedpre-trainingentropy-controlleddynamicallocationmulti-domainjointtraining
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

STAR argues that in multi-domain 3D scene understanding, mixture-of-experts (MoE) routers that gate on learned task features alone systematically underrepresent local sampling topology—density, completeness, and neighborhood structure—so expert allocation goes wrong when semantically similar objects are captured by different sensors. To fix this, the paper proposes a dual-branch design: a frozen self-supervised branch that learns cross-domain structural priors through color, density, and completeness augmentations, and a trainable domain-aware branch whose routing is driven by 3D spatial context and by per-token routing uncertainty. On indoor and outdoor benchmarks the framework reports consistent gains over strong baselines, including 80.1% mIoU on ScanNet and 77.2% on S3DIS. A sympathetic reader would care because it points toward unified 3D perception that adapts to new sensors without per-domain retraining.

What carries the argument

Three components carry the argument. DSR reshapes the token feature map into a 3D sparse tensor, runs a 3D spatial convolution, and adds a learned domain embedding before the gating MLP, making routing sensitive to local geometry. EDA computes the Shannon entropy of the softmax routing probabilities per token and linearly maps it to a token-specific expert count, activating top-k experts so high-uncertainty tokens get more capacity. The Re branch is a frozen teacher-student pretrained branch with self-supervised color, density, and completeness augmentation that anchors cross-domain structural priors and initializes the whole model. Together they decouple stable cross-domain representation from adaptive, topology-aware expert allocation.

What would settle it

Look at the DSR routing input: replace the sparse-tensor positions in which the token features are arranged with random coordinate permutations while keeping every other parameter and the same input feature values. If ScanNet mIoU stays at the same level as with true positions, then local topology is not what drives the gain and the central claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that expert selection in a 3D MoE should be conditioned on local spatial topology, not only on intermediate semantic features. The Domain-Spatial-Guided Routing module reshapes token features into a 3D sparse tensor, applies a 3D spatial convolution, and adds a dataset-specific domain embedding before feeding the result to the gating network, so routing can respond to density and completeness differences. The Entropy-Controlled Dynamic Allocation module then maps the entropy of each token's routing distribution to the number of experts activated, giving uncertain tokens more capacity. The authors show that this design outperforms feature-only MoE baselines and a strong self-supervised backbone on ScanNet, S3DIS, nuScenes, and Waymo, and degrades less under simulated density and completeness perturbations.

Load-bearing premise

The whole routing advantage rests on the assumption that the backbone's token features, after downsampling and attention, can be faithfully rearranged into a 3D sparse tensor whose spatial layout still mirrors the original point cloud's density, completeness, and neighborhood structure; if that rearrangement loosens the geometry, the spatial convolution in DSR cannot actually sense the topological variations it is supposed to route on.

Editorial extensions

If this is right

  • STAR improves indoor segmentation to 80.1% mIoU on ScanNet and 77.2% on S3DIS, gains of 0.5 and 1.2 points over the strongest baseline.
  • Under controlled density dropout and regional masking, STAR shows smaller mIoU drops than feature-only routers, for example -6.0 versus -9.8 points under 0.9 dropout.
  • A scaled-up baseline with comparable parameters does not close the gap, implying the gain is from routing design rather than added capacity.
  • The framework transfers to multimodal object detection, improving F1@0.25 by 1.9 points over a strong baseline, and to zero-shot unseen scenes when a compatible source domain embedding is chosen.

Reading between the lines

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

  • Not tested in the paper: DSR's spatial tensor requires a coordinate-preserving token layout; one could ablate by replacing the sparse-tensor positions with random permutations to confirm the gain is genuinely topological.
  • Not tested in the paper: the entropy-to-k schedule could be applied to 2D MoE towers as a generic uncertainty-driven routing stabilizer, since the entropy formula is modality-agnostic.
  • Not tested in the paper: the zero-shot embedding selection suggests a practical recipe—matching the unseen sensor's acquisition metadata to a source embedding—but only three source embeddings are examined.
  • If the topology-sensitive routing claim is right, STAR should also help in mixed LiDAR-RGB-D joint training where density differences are larger than in the indoor-only setup; the paper does not report that combination directly.
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

1 major / 6 minor

Summary. The paper proposes STAR, a Mixture-of-Experts framework for multi-domain 3D scene understanding. STAR combines a frozen self-supervised 'Unified Representation' branch, pre-trained with color, density, and completeness augmentations, with a domain-aware expert branch. The domain-aware branch contains Domain-Spatial-Guided Routing (DSR), which injects a 3D sparse-convolution-based spatial context into the router, and Entropy-Controlled Dynamic Allocation (EDA), which varies the number of active experts per token based on routing entropy. The authors report gains over Sonata and other baselines on ScanNet, ScanNet200, S3DIS, nuScenes, Waymo, zero-shot transfer to SpatialLM and Matterport3D, and ARKitScenes detection, together with ablations and a parameter-matched scaling comparison aimed at showing that the gains come from the routing design rather than from added capacity.

Significance. If the central mechanism is substantiated, STAR would be a useful contribution: it addresses a real limitation of feature-only MoE routers for point clouds whose sampling topology varies across sensors, and it validates the idea with a broad benchmark suite. The paper's scaling-controlled comparison in Table 7 is a genuine strength, as is the perturbation study in Table 6 that directly targets density and completeness variation. The main significance is currently conditional: the mechanism that distinguishes STAR from feature-only routing, DSR, is described without the implementation detail needed to verify that it actually senses local topology, and the reported margins over Sonata (0.4 to 1.2 mIoU) are small enough that the absence of repeated-seed statistics is a real concern.

major comments (1)
  1. [Section 4.3, Tables 1 and 5] No error bars, confidence intervals, or multiple-seed results are reported. The headline improvements over Sonata are 0.4 to 1.2 mIoU in Table 1 and 0.5 to 0.6 mIoU in Table 2, and the DSR ablation gain in Table 5 is 0.7 mIoU. Such differences are within typical run-to-run variation for 3D segmentation. The paper should report at least three seeds, or otherwise provide variance estimates, for the key comparisons on ScanNet and S3DIS and for the component ablation; without this, the claim of consistent improvement cannot be evaluated quantitatively.
minor comments (6)
  1. [Section 3.3] The domain embedding d is used before it is formally introduced; please define d and state how it is obtained for each dataset.
  2. [Section 3.5] The Qwen2.5 citation appears as '[?]' in the sentence about the autoregressive detection loss; the reference should be completed.
  3. [Table 1] The row 'Point-MoE+Re' is not described in the method or the text; please clarify how the Re branch is added to Point-MoE and why this baseline is not discussed.
  4. [Table 6] The perturbation parameters 'masksize' and 'maskratio' are not defined; specify the mask geometry, the point-selection procedure, and the units or reference frame for the mask size.
  5. [Section 4.3, Table 5] The sentence 'Adding DSR with two experts brings another 0.7% gain' is unclear because the main configuration uses K=8 experts; please state whether DSR is restricted to two active experts in this row or whether the text means something else.
  6. [Figure 1] Figure 1 has no axis labels or units, which makes the claimed state-of-the-art comparison difficult to interpret.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: STAR's gains are measured empirically against held-out benchmarks with ablations; the central claim does not reduce to fitted inputs or self-citations.

full rationale

The paper is an empirical deep-learning submission. Its central claim—that adding spatial-topology cues (DSR) and entropy-controlled expert allocation (EDA) improves multi-domain 3D understanding—is supported by benchmark measurements on standard validation sets and by component ablations (Table 5), including a DSR decomposition and scaled-up Sonata variants in Table 7. No equation defines a predicted quantity in terms of the benchmark it is supposed to explain; EDA's entropy-to-k mapping (Eq. 2) is a mechanism definition, not a fitted constant masquerading as a prediction. The self-citations present (e.g., [19, 28, 51, 54]) appear in related-work and application contexts and are not load-bearing for the proposed routing mechanisms. The underspecified construction of the 3D sparse tensor in Section 3.3 is a reproducibility/attribution concern—one could not independently confirm that the spatial convolution senses density or completeness without coordinate and resolution details—but that is not circularity: the claimed effect is empirically testable and the paper provides comparative ablations rather than deriving the effect from its own assumptions. Therefore no circular step can be quoted, and the appropriate finding is no significant circularity.

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

The central claim is empirical, so the ledger lists the manually chosen design constants and the domain assumptions that must hold for the proposed mechanisms to work. No new physical or abstract entities are postulated; the domain embedding is a learned conditioning vector, not an invented entity with independent falsifiable evidence.

free parameters (4)
  • Number of experts K = 8
    Maximum number of activated experts per token, used in the MoE and in EDA. Set by hand.
  • Load-balancing loss weight λ = 0.001
    Manual hyperparameter in equations (6) and (9); controls the strength of the expert balance loss.
  • Entropy-to-expert mapping bounds (k_min, k_max) = k_min=1, k_max=K
    Defines equation (2); a design choice, not derived from data.
  • SSL augmentation ratios = Not reported
    Color dropout, density dropout, and patch masking ratios shape the pretraining distribution; values said to be in the supplementary.
assumptions (3)
  • domain assumption Token features f can be reshaped into a 3D sparse tensor that preserves local point-cloud topology.
    Section 3.3, DSR: the routing input z is built from a sparse tensor of f; no construction details are given.
  • domain assumption The selected self-supervised augmentations (color dropout, density dropout, patch masking) simulate real cross-domain topological variations.
    Section 3.2 uses only these augmentations to anchor cross-domain priors; if real sensor shifts differ, the frozen Re branch may anchor the wrong structure.
  • domain assumption A single per-dataset domain embedding captures dataset-level structural priors and can be manually selected for unseen datasets.
    Section 3.3 and Table 3 rely on this; the method is not fully automatic and depends on metadata.

how reviews work

0 comments
Cite this review

Pith. "Pith review of STAR: A Spatial-Topology Aware Routing Framework for Generalizable 3D Scene Understanding." pith.science (2026). https://pith.science/paper/E3M6JK7Q

@misc{pith2026260811699,
  author       = {Pith},
  title        = {Pith review of: STAR: A Spatial-Topology Aware Routing Framework for Generalizable 3D Scene Understanding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E3M6JK7Q}},
  note         = {Machine review of arXiv:2608.11699}
}
read the original abstract

Constructing a unified 3D scene understanding model has long been hindered by the topological discrepancies across sensor modalities. While applying the Mixture-of-Experts (MoE) architecture is a flexible approach for multi-domain 3D understanding, we observe that conventional feature-only MoE routers may underrepresent local sampling topology under semantic supervision, making expert allocation difficult when semantic consistency coexists with geometric heterogeneity. To overcome this challenge, we propose STAR (Spatial-Topology Aware Routing Framework). Specifically, we introduce a multi-attribute self-supervised pre-training branch, covering topological and textural variations, to anchor cross-domain structural priors. Building upon this, we design a domain-aware expert branch with two mechanisms: Domain-Spatial-Guided Routing (DSR), which captures local topological variations from spatial context, and Entropy-controlled Dynamic Allocation (EDA), which adjusts the number of activated experts according to routing uncertainty. Together, these branches combine stable cross-domain representation learning with adaptive expert allocation. Extensive experiments across various tasks, encompassing both indoor and outdoor scenes, demonstrate the effectiveness of STAR. It achieves 80.1% mIoU on the ScanNet validation set and 77.2% mIoU on S3DIS, consistently improving over strong baselines. Code is available at our project page (https://xmw666.github.io/STAR/).

Figures

Figures reproduced from arXiv: 2608.11699 by the authors.

Figure 1
Figure 1. Performance of STAR. STAR achieves state-of-the-art performance across multiple 3D scene understanding benchmarks. eling [14, 15, 19, 27, 45, 54] has diversified the sensing modalities, acquisition pipelines, and geometric represen￾tations used to capture and model 3D scenes. While these developments provide increasingly rich and complementary sources of 3D data, they create a growing need for general￾izable represe… view at source ↗
Figure 2
Figure 2. Routing behavior under topology variations. Vanilla MoE denotes our standard feature-only MoE baseline. We compare Vanilla MoE, Point-MoE [6], and STAR under controlled density and completeness perturbations. For samples from the same semantic category (e.g., “bed” or “floor”), feature-only routers tend to keep similar expert assignments despite local geometric changes, causing larger performance drops (e.g., a 19.8… view at source ↗
Figure 3
Figure 3. STAR architecture. It combines two branches: a domain-aware branch (Do) and a unified-representation branch (Re). STAR starts with a pretrained model generated through self-supervised learning on multi-attribute data. Re freezes this pretrained knowledge to extract cross-domain geometric and structural patterns. Do uses Domain-Spatial-Guided Routing (DSR) and Entropy-Controlled Dynamic Allocation (EDA) to capture do… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative analysis. Visualization of different methods on ScanNet [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Expert activation across real cross-domain sampling dif [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 40 canonical work pages

  1. [1]

    3d semantic parsing of large-scale indoor spaces

    Iro Armeni, Ozan Sener, Amir R Zamir, Helen Jiang, Ioannis Brilakis, Martin Fischer, and Silvio Savarese. 3d semantic parsing of large-scale indoor spaces. InCVPR, 2016. 5

    2016

  2. [2]

    Arkitscenes: A diverse real-world dataset for 3d indoor scene understanding using mobile rgb-d data.arXiv preprint arXiv:2111.08897, 2021

    Gilad Baruch, Zhuoyuan Chen, Afshin Dehghan, Tal Dimry, Yuri Feigin, Peter Fu, Thomas Gebauer, Brandon Joffe, Daniel Kurz, Arik Schwartz, et al. Arkitscenes: A diverse real-world dataset for 3d indoor scene understanding using mobile rgb-d data.arXiv preprint arXiv:2111.08897, 2021. 5, 7

    arXiv 2021

  3. [3]

    nuscenes: A multi- modal dataset for autonomous driving

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Gi- ancarlo Baldan, and Oscar Beijbom. nuscenes: A multi- modal dataset for autonomous driving. InCVPR, 2020. 6

    2020

  4. [4]

    Unsupervised learn- ing of visual features by contrasting cluster assignments

    Mathilde Caron, Ishan Misra, Julien Mairal, Priya Goyal, Pi- otr Bojanowski, and Armand Joulin. Unsupervised learn- ing of visual features by contrasting cluster assignments. NeurIPS, 2020. 4

    2020

  5. [5]

    Matterport3d: Learning from rgb-d data in indoor environments.arXiv preprint arXiv:1709.06158, 2017

    Angel Chang, Angela Dai, Thomas Funkhouser, Maciej Halber, Matthias Niessner, Manolis Savva, Shuran Song, Andy Zeng, and Yinda Zhang. Matterport3d: Learning from rgb-d data in indoor environments.arXiv preprint arXiv:1709.06158, 2017. 7

    arXiv 2017

  6. [6]

    Point-moe: Towards cross-domain general- ization in 3d semantic segmentation via mixture-of-experts

    Xuweiyi Chen, Wentao Zhou, Aruni RoyChowdhury, and Zezhou Cheng. Point-moe: Towards cross-domain general- ization in 3d semantic segmentation via mixture-of-experts. arXiv preprint arXiv:2505.23926, 2025. 1, 2, 3, 6, 7, 8

    work page arXiv 2025

  7. [7]

    Transiff: An instance-level feature fusion framework for vehicle- infrastructure cooperative 3d detection with transformers

    Ziming Chen, Yifeng Shi, and Jinrang Jia. Transiff: An instance-level feature fusion framework for vehicle- infrastructure cooperative 3d detection with transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 18205–18214, 2023. 1

    work page 2023

  8. [8]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In CVPR, 2017. 5

    work page 2017

Show all 63 references
  1. [9]

    Generalizable person re-identification with relevance-aware mixture of experts

    Yongxing Dai, Xiaotong Li, Jun Liu, Zekun Tong, and Ling-Yu Duan. Generalizable person re-identification with relevance-aware mixture of experts. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16145–16154, 2021. 3

    2021

  2. [10]

    Switch transformers: Scaling to trillion parameter models with sim- ple and efficient sparsity.JMLR, 2022

    William Fedus, Barret Zoph, and Noam Shazeer. Switch transformers: Scaling to trillion parameter models with sim- ple and efficient sparsity.JMLR, 2022. 5

    2022

  3. [11]

    3d-front: 3d furnished rooms with layouts and semantics

    Huan Fu, Bowen Cai, Lin Gao, Ling-Xiao Zhang, Jiaming Wang, Cao Li, Qixun Zeng, Chengyue Sun, Rongfei Jia, Bin- qiang Zhao, et al. 3d-front: 3d furnished rooms with layouts and semantics. InICCV, 2021. 5

    2021

  4. [12]

    Pointinst3d: Segmenting 3d instances by points

    Tong He, Wei Yin, Chunhua Shen, and Anton Van den Hen- gel. Pointinst3d: Segmenting 3d instances by points. In ECCV, 2022. 3

    2022

  5. [13]

    Ropebev: A multi-camera roadside perception network in bird’s-eye- view.arXiv preprint arXiv:2409.11706, 2024

    Jinrang Jia, Guangqi Yi, and Yifeng Shi. Ropebev: A multi-camera roadside perception network in bird’s-eye- view.arXiv preprint arXiv:2409.11706, 2024. 1

    2024 arXiv

  6. [14]

    Panoworld: A generative spatial world model for con- sistent whole-house panorama synthesis.arXiv preprint arXiv:2605.17916, 2026

    Jinrang Jia, Zhenjia Li, Yijiang Hu, and Yifeng Shi. Panoworld: A generative spatial world model for con- sistent whole-house panorama synthesis.arXiv preprint arXiv:2605.17916, 2026. 1

    2026 arXiv

  7. [15]

    You only gaussian once: Controllable 3d gaussian splatting for ultra-densely sampled scenes.arXiv preprint arXiv:2604.21400, 2026

    Jinrang Jia, Zhenjia Li, and Yifeng Shi. You only gaussian once: Controllable 3d gaussian splatting for ultra-densely sampled scenes.arXiv preprint arXiv:2604.21400, 2026. 1

    2026 arXiv

  8. [16]

    Pointgroup: Dual-set point grouping for 3d instance segmentation

    Li Jiang, Hengshuang Zhao, Shaoshuai Shi, Shu Liu, Chi- Wing Fu, and Jiaya Jia. Pointgroup: Dual-set point grouping for 3d instance segmentation. InCVPR, 2020. 3

    2020

  9. [17]

    Jia Jinrang, Zhenjia Li, and Yifeng Shi. Monouni: A unified vehicle and infrastructure-side monocular 3d object detec- tion network with sufficient depth clues.Advances in Neural Information Processing Systems, 36:11703–11715, 2023. 1

    2023

  10. [18]

    Danet: Dimen- sion apart network for radar object detection

    Bo Ju, Wei Yang, Jinrang Jia, Xiaoqing Ye, Qu Chen, Xiao Tan, Hao Sun, Yifeng Shi, and Errui Ding. Danet: Dimen- sion apart network for radar object detection. InProceedings of the 2021 international conference on multimedia retrieval, pages 533–539, 2021. 1

    2021

  11. [19]

    Sat2realcity: Geometry-aware and appearance-controllable 3d urban generation from satellite imagery.arXiv preprint arXiv:2511.11470, 2025

    Yijie Kang, Xinliang Wang, Zhenyu Wu, Yifeng Shi, and Hailong Zhu. Sat2realcity: Geometry-aware and appearance-controllable 3d urban generation from satellite imagery.arXiv preprint arXiv:2511.11470, 2025. 1

    2025 arXiv

  12. [20]

    Unidet3d: Multi- dataset indoor 3d object detection

    Maksim Kolodiazhnyi, Anna V orontsova, Matvey Skripkin, Danila Rukhovich, and Anton Konushin. Unidet3d: Multi- dataset indoor 3d object detection. InProceedings of the AAAI Conference on Artificial Intelligence, pages 4365– 4373, 2025. 3

    2025

  13. [21]

    Dusa: Decoupled unsupervised sim2real adaptation for vehicle-to-everything collaborative perception

    Xianghao Kong, Wentao Jiang, Jinrang Jia, Yifeng Shi, Run- sheng Xu, and Si Liu. Dusa: Decoupled unsupervised sim2real adaptation for vehicle-to-everything collaborative perception. InProceedings of the 31st ACM International Conference on Multimedia, pages 1943–1954, 2023. 1

    1943

  14. [22]

    Pointpillars: Fast encoders for object detection from point clouds

    Alex H Lang, Sourabh V ora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom. Pointpillars: Fast encoders for object detection from point clouds. InCVPR, pages 12697–12705, 2019. 3

    2019

  15. [23]

    Sparse mixture- of-experts are domain generalizable learners.arXiv preprint arXiv:2206.04046, 2022

    Bo Li, Yifei Shen, Jingkang Yang, Yezhen Wang, Jiawei Ren, Tong Che, Jun Zhang, and Ziwei Liu. Sparse mixture- of-experts are domain generalizable learners.arXiv preprint arXiv:2206.04046, 2022. 3

    2022 arXiv

  16. [24]

    Ragtrack: Language-aware rgbt tracking with retrieval-augmented generation.arXiv preprint arXiv:2603.03617, 2026

    Hao Li, Yuhao Wang, Wenning Hao, Pingping Zhang, Dong Wang, and Huchuan Lu. Ragtrack: Language-aware rgbt tracking with retrieval-augmented generation.arXiv preprint arXiv:2603.03617, 2026. 1

    2026

  17. [25]

    Cadtrack: Learning contextual aggregation with deformable alignment for robust rgbt tracking

    Hao Li, Yuhao Wang, Xiantao Hu, Wenning Hao, Pingping Zhang, Dong Wang, and Huchuan Lu. Cadtrack: Learning contextual aggregation with deformable alignment for robust rgbt tracking. InProceedings of the AAAI Conference on Artificial Intelligence, pages 6109–6117, 2026. 1

    2026

  18. [26]

    Monolss: Learnable sample selection for monocular 3d detection

    Zhenjia Li, Jinrang Jia, and Yifeng Shi. Monolss: Learnable sample selection for monocular 3d detection. In2024 Inter- national Conference on 3D Vision (3DV), pages 1125–1135. IEEE, 2024. 1

    2024

  19. [27]

    Pano2world: End- to-end 3d generation via unified multi-view sequences.arXiv preprint arXiv:2607.00832, 2026

    Zhenjia Li, Jinrang Jia, and Yifeng Shi. Pano2world: End- to-end 3d generation via unified multi-view sequences.arXiv preprint arXiv:2607.00832, 2026. 1

    2026 arXiv

  20. [28]

    Llava-sp: Enhancing visual representa- tion with visual spatial tokens for mllms.arXiv preprint arXiv:2507.00505, 2025

    Haoran Lou, Chunxiao Fan, Ziyan Liu, Yuexin Wu, and Xinliang Wang. Llava-sp: Enhancing visual representa- tion with visual spatial tokens for mllms.arXiv preprint arXiv:2507.00505, 2025. 1

    2025 arXiv

  21. [29]

    Spatiallm: Train- ing large language models for structured indoor modeling

    Yongsen Mao, Junhao Zhong, Chuan Fang, Jia Zheng, Rui Tang, Hao Zhu, Ping Tan, and Zihan Zhou. Spatiallm: Train- ing large language models for structured indoor modeling. In NeurIPS, 2025. 5, 7

    2025

  22. [30]

    Repre- sentation learning with contrastive predictive coding.arXiv preprint arXiv:1807.03748, 2018

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Repre- sentation learning with contrastive predictive coding.arXiv preprint arXiv:1807.03748, 2018. 5

    2018 arXiv

  23. [31]

    Maxime Oquab, Timoth ´ee Darcet, Theo Moutakanni, Huy V . V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, Rus- sell Howes, Po-Yao Huang, Hu Xu, Vasu Sharma, Shang- Wen Li, Wojciech Galuba, Mike Rabbat, Mido Assran, Ni...

    2023

  24. [32]

    Masked autoencoders for point cloud self-supervised learning

    Yatian Pang, Wenxiao Wang, Francis EH Tay, Wei Liu, Yonghong Tian, and Li Yuan. Masked autoencoders for point cloud self-supervised learning. InECCV, 2022. 2, 3

    2022

  25. [33]

    Pointnet: Deep learning on point sets for 3d classification and segmentation

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. InCVPR, 2017. 3

    2017

  26. [34]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space.NeurIPS, 2017

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space.NeurIPS, 2017. 3

    2017

  27. [35]

    Pointnext: Revisiting pointnet++ with improved training and scaling strategies.NeurIPS, 2022

    Guocheng Qian, Yuchen Li, Houwen Peng, Jinjie Mai, Hasan Hammoud, Mohamed Elhoseiny, and Bernard Ghanem. Pointnext: Revisiting pointnet++ with improved training and scaling strategies.NeurIPS, 2022. 3

    2022

  28. [36]

    Qwen2.5 technical report, 2025

    Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Jun- yang Lin, Kai Dang, Keming Lu, Keqin Bao, Kexin Yang...

    2025

  29. [37]

    Spreading vectors for similarity search

    Alexandre Sablayrolles, Matthijs Douze, Cordelia Schmid, and Herv ´e J ´egou. Spreading vectors for similarity search. arXiv preprint arXiv:1806.03198, 2018. 4

    2018 arXiv

  30. [38]

    Pv-rcnn: Point- voxel feature set abstraction for 3d object detection

    Shaoshuai Shi, Chaoxu Guo, Li Jiang, Zhe Wang, Jianping Shi, Xiaogang Wang, and Hongsheng Li. Pv-rcnn: Point- voxel feature set abstraction for 3d object detection. In CVPR, 2020. 3

    2020

  31. [39]

    Open-transmind: A new baseline and benchmark for 1st foundation model challenge of intelligent transportation

    Yifeng Shi, Feng Lv, Xinliang Wang, Chunlong Xia, Shaojie Li, Shujie Yang, Teng Xi, and Gang Zhang. Open-transmind: A new baseline and benchmark for 1st foundation model challenge of intelligent transportation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pa...

    2023

  32. [40]

    Mdt3d: Multi-dataset training for lidar 3d object detection generalization

    Louis Soum-Fontez, Jean-Emmanuel Deschaud, and Franc ¸ois Goulette. Mdt3d: Multi-dataset training for lidar 3d object detection generalization. In2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 5765–5772. IEEE, 2023. 3

    2023

  33. [41]

    Scalability in perception for autonomous driving: Waymo open dataset

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, et al. Scalability in perception for autonomous driving: Waymo open dataset. InCVPR,

  34. [42]

    Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results.NeurIPS, 2017

    Antti Tarvainen and Harri Valpola. Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results.NeurIPS, 2017. 3

    2017

  35. [43]

    Kpconv: Flexible and deformable convolution for point clouds

    Hugues Thomas, Charles R Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, Franc ¸ois Goulette, and Leonidas J Guibas. Kpconv: Flexible and deformable convolution for point clouds. InICCV, 2019. 3

    2019

  36. [44]

    Rt- detrv3: Real-time end-to-end object detection with hierarchi- cal dense positive supervision

    Shuo Wang, Chunlong Xia, Feng Lv, and Yifeng Shi. Rt- detrv3: Real-time end-to-end object detection with hierarchi- cal dense positive supervision. InWACV, pages 1628–1636,

  37. [45]

    Artifactworld: Scaling 3d gaussian splatting artifact restoration via video generation models.arXiv preprint arXiv:2604.12251, 2026

    Xinliang Wang, Yifeng Shi, and Zhenyu Wu. Artifactworld: Scaling 3d gaussian splatting artifact restoration via video generation models.arXiv preprint arXiv:2604.12251, 2026. 1

    2026 arXiv

  38. [46]

    One for all: Multi-domain joint training for point cloud based 3d object detection.Advances in Neural Infor- mation Processing Systems, 37:56859–56877, 2024

    Zhenyu Wang, Yali Li, Hengshuang Zhao, and Shengjin Wang. One for all: Multi-domain joint training for point cloud based 3d object detection.Advances in Neural Infor- mation Processing Systems, 37:56859–56877, 2024. 1, 2, 3

    2024

  39. [47]

    Point transformer v2: Grouped vector atten- tion and partition-based pooling.NeurIPS, 2022

    Xiaoyang Wu, Yixing Lao, Li Jiang, Xihui Liu, and Heng- shuang Zhao. Point transformer v2: Grouped vector atten- tion and partition-based pooling.NeurIPS, 2022. 3

    2022

  40. [48]

    Point transformer v3: Simpler faster stronger

    Xiaoyang Wu, Li Jiang, Peng-Shuai Wang, Zhijian Liu, Xi- hui Liu, Yu Qiao, Wanli Ouyang, Tong He, and Hengshuang Zhao. Point transformer v3: Simpler faster stronger. In CVPR, 2024. 3, 6, 8

    2024

  41. [49]

    Towards large- scale 3d representation learning with multi-dataset point prompt training

    Xiaoyang Wu, Zhuotao Tian, Xin Wen, Bohao Peng, Xihui Liu, Kaicheng Yu, and Hengshuang Zhao. Towards large- scale 3d representation learning with multi-dataset point prompt training. InCVPR, 2024. 1, 2, 3, 5, 6, 7, 8

    2024

  42. [50]

    Sonata: Self- supervised learning of reliable point representations

    Xiaoyang Wu, Daniel DeTone, Duncan Frost, Tianwei Shen, Chris Xie, Nan Yang, Jakob Engel, Richard New- combe, Hengshuang Zhao, and Julian Straub. Sonata: Self- supervised learning of reliable point representations. In CVPR, 2025. 1, 2, 3, 4, 5, 6, 7, 8

    2025

  43. [51]

    Beyondsparse: Facilitating mamba to enhance cross-domain 3d semantic segmentation in adverse weather

    Yao Wu, Mingwei Xing, Yachao Zhang, Fangyong Wang, Xiaopei Zhang, and Yanyun Qu. Beyondsparse: Facilitating mamba to enhance cross-domain 3d semantic segmentation in adverse weather. InProceedings of the AAAI Conference on Artificial Intelligence, 2026. 3

    2026

  44. [52]

    Vit-comer: Vision transformer with convolu- tional multi-scale feature interaction for dense predictions

    Chunlong Xia, Xinliang Wang, Feng Lv, Xin Hao, and Yifeng Shi. Vit-comer: Vision transformer with convolu- tional multi-scale feature interaction for dense predictions. InCVPR, 2024. 1

    2024

  45. [53]

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

    Saining Xie, Jiatao Gu, Demi Guo, Charles R Qi, Leonidas Guibas, and Or Litany. Pointcontrast: Unsupervised pre- training for 3d point cloud understanding. InECCV, 2020. 2, 3

    2020

  46. [54]

    Adapt- splat: Adapting vision foundation models for feed-forward 3d gaussian splatting.arXiv preprint arXiv:2605.10239,

    Mingwei Xing, Xinliang Wang, and Yifeng Shi. Adapt- splat: Adapting vision foundation models for feed-forward 3d gaussian splatting.arXiv preprint arXiv:2605.10239,

    arXiv

  47. [55]

    Cbdmoe: Consistent-but-diverse mixture of ex- perts for domain generalization.IEEE Transactions on Mul- timedia, 26:9814–9824, 2024

    Fangbin Xu, Dongyue Chen, Tong Jia, Shizhuo Deng, and Hao Wang. Cbdmoe: Consistent-but-diverse mixture of ex- perts for domain generalization.IEEE Transactions on Mul- timedia, 26:9814–9824, 2024. 3

    2024

  48. [56]

    Limoe: Mixture of lidar representation learners from automotive scenes

    Xiang Xu, Lingdong Kong, Hui Shuai, Liang Pan, Ziwei Liu, and Qingshan Liu. Limoe: Mixture of lidar representation learners from automotive scenes. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 27368– 27379, 2025. 2, 3

    2025

  49. [57]

    Habitat-matterport 3d semantics dataset

    Karmesh Yadav, Ram Ramrakhya, Santhosh Kumar Ramakr- ishnan, Theo Gervet, John Turner, Aaron Gokaslan, Noah Maestre, Angel Xuan Chang, Dhruv Batra, Manolis Savva, et al. Habitat-matterport 3d semantics dataset. InCVPR,

  50. [58]

    Uni3d-moe: Scalable multimodal 3d scene understanding via mixture of experts.arXiv preprint arXiv:2505.21079, 2025

    Yue Zhang, Yingzhao Jian, Hehe Fan, Yi Yang, and Roger Zimmermann. Uni3d-moe: Scalable multimodal 3d scene understanding via mixture of experts.arXiv preprint arXiv:2505.21079, 2025. 2, 3

    2025 arXiv

  51. [59]

    Con- certo: Joint 2d-3d self-supervised learning emerges spatial representations

    Yujia Zhang, Xiaoyang Wu, Yixing Lao, Chengyao Wang, Zhuotao Tian, Naiyan Wang, and Hengshuang Zhao. Con- certo: Joint 2d-3d self-supervised learning emerges spatial representations. InNeurIPS, 2025. 3

    2025

  52. [60]

    Structured3d: A large photo-realistic dataset for structured 3d modeling

    Jia Zheng, Junfei Zhang, Jing Li, Rui Tang, Shenghua Gao, and Zihan Zhou. Structured3d: A large photo-realistic dataset for structured 3d modeling. InECCV, 2020. 5

    2020

  53. [61]

    Meta-dmoe: Adapting to domain shift by meta- distillation from mixture-of-experts.Advances in Neural In- formation Processing Systems, 35:22243–22257, 2022

    Tao Zhong, Zhixiang Chi, Li Gu, Yang Wang, Yuanhao Yu, and Jin Tang. Meta-dmoe: Adapting to domain shift by meta- distillation from mixture-of-experts.Advances in Neural In- formation Processing Systems, 35:22243–22257, 2022. 3

    2022

  54. [62]

    Uni3d: Exploring unified 3d representation at scale.arXiv preprint arXiv:2310.06773,

    Junsheng Zhou, Jinsheng Wang, Baorui Ma, Yu-Shen Liu, Tiejun Huang, and Xinlong Wang. Uni3d: Exploring unified 3d representation at scale.arXiv preprint arXiv:2310.06773,

    arXiv

  55. [63]

    V oxelnet: End-to-end learning for point cloud based 3d object detection

    Yin Zhou and Oncel Tuzel. V oxelnet: End-to-end learning for point cloud based 3d object detection. InCVPR, 2018. 3

    2018

Pith tools

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