Pith. sign in

REVIEW 4 major objections 3 minor 40 references

SegVec3D: A Method for Vector Embedding of 3D Objects Oriented Towards Robot manipulation

T0 review · 4 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read SegVec3D claims a single 3D network can segment point clouds into object instances and align those instances with natural-language descriptions in a shared embedding space, using sparse or no instance labels.

desk verdict Plausible engineering proposal with honest limitations, but the evidence is qualitative at best and the 'unsupervised' claim overstates what was actually done. read the letter →

arxiv 2507.09459 v1 pith:3V3NMPLM submitted 2025-07-13 cs.CV cs.RO

Zhihan Kang , Boyu Wang This is my paper · ORCID
classification cs.CVcs.RO
keywords 3Dpointcloudinstancesegmentationcontrastiveembeddingcross-modalalignmentlanguagegroundingattentionmechanismzero-shotretrievalrobotmanipulationweaksupervision
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes SegVec3D, a framework that performs 3D point cloud instance segmentation and cross-modal alignment to natural language in one system. The authors claim that by learning a discriminative embedding space with a contrastive pull-push loss, the model can separate raw point clouds into object instances without dense instance labels, and that by aligning these instance embeddings with text embeddings in a shared space it enables zero-shot retrieval and semantic labeling of objects from language queries. The intended payoff is a practical robot perception module: a robot could look at a scene, segment it into objects, and then respond to a command like "the chair" by pointing to the right region, all trained with sparse or no instance annotations.

What carries the argument

The load-bearing object is the contrastive embedding space: per-point embeddings are shaped by the pull-push loss $L_{\mathrm{ins}}$ in Eq. (8), and instance-level descriptors are aligned to language with an InfoNCE-style cross-modal loss in Eq. (12). The segmentation network itself is an attention-based local feature extractor with $k$-NN neighborhood graphs, a global max-pooled scene vector fused into point features, and a final embedding MLP; clustering in this space at inference yields instance masks.

What would settle it

Train SegVec3D on ScanNet with all instance labels withheld, forming positive and negative pairs only from geometric over-segmentation heuristics, and measure instance segmentation AP50 on the test split; if AP50 drops to near zero or fails to beat a simple region-growing baseline, the claim of unsupervised segmentation is falsified.

Watch

Extended reading notes

Core claim

The central claim is that instance segmentation and multimodal understanding can be unified in a single contrastively trained embedding space: a point-wise attention network with a global context vector produces per-point features, a contrastive loss pulls same-object points together and pushes different-object points apart, and a CLIP-style alignment loss projects the resulting instance descriptors and text descriptors into a common space. The authors state that experimental evaluation on a real-world lab scene confirms high semantic discriminability, robust multi-modal alignment, and practical deployability, and that this combination of capabilities is not jointly provided by existing methods such as Mask3D or ULIP.

Load-bearing premise

The unsupervised instance segmentation only works if reliable positive and negative point pairs can be formed without full instance labels; the paper assumes such weak grouping signals are available but never specifies how they are obtained or how noisy they are.

Editorial extensions

If this is right

  • If the central claim is correct, robots can perform open-vocabulary object search: given a text phrase, the robot retrieves the instance whose embedding is closest to the phrase embedding, without retraining for new object categories.
  • Instance segmentation becomes trainable with only a small fraction of labeled points, since the contrastive loss needs positive and negative point pairs rather than full masks or class labels.
  • The same embedding space doubles as a zero-shot classifier for segmented objects, labeling each instance by the nearest text token among candidate names.
  • The method can be deployed on a new scene directly after training, because inference requires only clustering in the learned embedding space and a nearest-neighbor comparison with text embeddings.
  • The architecture offers a candidate building block for embodied agents that must both parse a 3D scene and follow linguistic instructions.

Reading between the lines

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

  • A direct testable extension would be to measure how much weak supervision the method actually needs: training with 0%, 1%, and 10% of instance labels on ScanNet and plotting segmentation AP would reveal how far the approach is from true unsupervised learning.
  • The paper's assumption of weak grouping signals could be satisfied by geometric over-segmentation heuristics, so a natural next step is to test the robustness of the contrastive loss to noisy pseudo-labels; the claimed unsupervised capability hinges on that robustness.
  • Because the language encoder is frozen during alignment training, I infer that the system would struggle with attribute-rich or novel descriptions (such as "the chair with wheels") unless the 3D encoder has already seen those attribute-word pairings, a limitation the authors themselves acknowledge.
  • The reported deployability rests on a single lab scene; a stronger corroboration would be a quantitative comparison on ScanNet with AP25/AP50 metrics, which the paper says it plans to release.
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

4 major / 3 minor

Summary. The paper proposes SegVec3D, a framework for 3D point cloud instance segmentation that combines a graph-based attention feature extractor, a contrastive embedding loss for separating instances, and a CLIP-inspired cross-modal alignment between point cloud instances and text. The authors claim that the method supports weakly supervised or unsupervised instance segmentation, enables zero-shot retrieval of 3D objects from language queries, and unifies instance segmentation and multimodal understanding in a way that Mask3D and ULIP do not. The experimental section presents a qualitative demonstration on a single self-captured lab scene, with no standard quantitative metrics, baselines, or error bars.

Significance. If the claims were established, SegVec3D would address a useful niche: joint instance segmentation and language grounding for robotic manipulation with sparse supervision. The combination of instance-level segmentation and cross-modal alignment is a plausible and worthwhile goal. However, the paper's evidence is far too thin to support the stated contributions. The abstract claims that experiments confirm 'high semantic discriminability, robust multi-modal alignment, and practical deployability,' but the reported evaluation is qualitative only, on a single scene, with no segmentation metrics, retrieval accuracy, baselines, or error bars. The 'unsupervised' aspect is not actually demonstrated, and the 'zero-shot' retrieval test uses a category token ('chair') that was part of the training vocabulary. The paper also explicitly defers quantitative benchmarks to future work. As a contribution to the literature, the main value is the conceptual framework and the explicit discussion of limitations, not the empirical validation.

major comments (4)
  1. [§3.2, §5.1] The central claim of unsupervised instance segmentation is not supported by the training protocol. Equation (8) requires a set S of positive point pairs from the same instance and a set D of negative pairs from different instances, and Section 5.1 states that training used 'a small percentage of points with instance labels to construct the contrastive embedding loss.' The fraction of labeled points, the sampling scheme for pairs, and the source of the weak grouping signals are never specified. The fully unsupervised scenario is only mentioned as a possibility ('spatial heuristics or over-segmentation methods, or even use temporal consistency in videos') and is never implemented or tested. The method as actually trained is weakly supervised, not unsupervised, and this mismatch is load-bearing because the abstract and introduction repeatedly claim unsupervised instance segmentation via contrastive clustering.
  2. [§5] The experimental evaluation does not substantiate the paper's central quantitative claims. Section 5 explicitly says that the authors 'present primarily qualitative results and case studies rather than extensive quantitative benchmarks' due to 'limited time and computational resources,' and Section 6 postpones metrics such as AP25, AP50, and mIoU to a later version. The reported results are qualitative visualizations on a single lab scene with no segmentation accuracy numbers, no retrieval accuracy, no comparison to any baseline, and no error bars. Consequently, statements in the abstract and Section 5.3 that the method achieves 'high semantic discriminability, robust multi-modal alignment, and practical deployability' are assertions rather than demonstrated findings. At minimum, the paper should report standard metrics on a public dataset such as ScanNet or S3DIS, along with comparisons to relevant methods.
  3. [§4.2, §5.2] The claimed zero-shot capability is not demonstrated. Equation (12) trains the projection matrices W_3D and W_txt to align 3D instance features with text tokens derived from ScanNet category labels, such as 'chair' and 'table.' The retrieval experiment in Section 5.2 uses the query 'the chair,' which is one of the training categories. This only shows that the model can match a test instance to a category name seen during training; it does not demonstrate generalization to unseen categories or to compositional descriptions. A proper zero-shot evaluation would use held-out categories, attribute-based descriptions that are not used in training, or a benchmark such as ScanRefer or ReferIt3D with unseen object classes.
  4. [§3.2, §4.2] Several hyperparameters and implementation details that are essential for reproducibility are not reported. These include the contrastive margin m in Eq. (8), the InfoNCE temperature τ in Eq. (12), the clustering radius or threshold used at inference, the neighborhood size k, the number of attention layers L, and the exact procedure for constructing the positive and negative pair sets S and D. Without these details, the experiments cannot be reproduced, and the sensitivity of the method to these choices is unknown. The paper should either provide these values or include an ablation study.
minor comments (3)
  1. [Abstract] The phrase 'proposeSegV ec3D' appears to have a missing space and an unusual spacing in the model name; the spelling should be consistent with 'SegVec3D' throughout.
  2. [§5.3] The claim that 'ULIP-based training' exposed the model to a broad semantic space is inconsistent with the method description in Section 4, which uses a pretrained sentence transformer and ScanNet category labels for alignment; the paper does not describe any ULIP pretraining of the 3D encoder.
  3. [§6] The statement that 'preliminary quantitative evaluations conducted in our earlier version (in Chinese) suggested promising results' is not verifiable from the manuscript and should not be used as evidence; any quantitative results should be reported directly in this paper.

Circularity Check

1 steps flagged · score 6.0 of 10

The 'zero-shot' cross-modal evaluation is a fitted-input check: the text queries ('chair', 'table') are the same ScanNet category tokens used to train Eq. (12), so retrieval success does not demonstrate generalization to unseen language.

  1. fitted input called prediction [Section 4.2 (Eq. 12), Section 5.1, Section 5.2]
    "The cross-modal alignment (Section 4) was trained using category labels from ScanNet (e.g., object categories like chair, table, sofa) as textual tokens, plus a few additional descriptive phrases... we provided the system with a set of text labels: “chair,” “table,” “floor,” “cabinet,” “wall,” “monitor,” “other.” For each predicted 3D instance cluster, we computed its embedding vX and found which text label had the highest cosine similarity. The model accurately matched the large chair instance to the word “chair” and the desk instance to “table.”"

    Eq. (12) trains the projections W_3D and W_txt with an InfoNCE objective that pulls the 3D embedding of a chair instance toward the token embedding of the word 'chair' (and similarly for other ScanNet category labels). The evaluation then queries with those same token embeddings and reports the cosine-similarity argmax as 'zero-shot semantic labeling' and 'text-based instance retrieval'. Because the query tokens are exactly the tokens used to fit the projection, the reported success is a check of the fitted mapping rather than a prediction on genuinely unseen text. The paper's own text confirms the training set was 'simple category labels', so the demo does not establish open-vocabulary or zero-shot generalization to categories outside the training vocabulary.

full rationale

The central circularity is in the cross-modal evaluation: 'zero-shot' labelling and retrieval are demonstrated on the same category tokens ('chair', 'table', etc.) that were used as training text in Eq. (12). The matching result is therefore the training objective reapplied at test time, not independent evidence of open-vocabulary understanding. This warrants a score of 6 because the claim of zero-shot retrieval, one of the paper's main contributions, reduces to a fit-check by construction. I do not count the 'unsupervised instance segmentation' claim as a circular reduction: Section 5.1 states that 'a small percentage of points with instance labels' was used to construct Eq. (8), and Section 3.2 only assumes 'some weak grouping signals are available' without specifying how unsupervised pairs are obtained, so the unsupervised label is unsupported and the label fraction is unreported. That is an evidentiary gap and a mismatch between the abstract and the training protocol, but the lab-scene segmentation is still a transfer test on a new scene rather than a reduction to the training targets by construction. There is no self-citation chain, no imported uniqueness theorem, and no machine-checked external evidence to weigh. The paper is also candid that results are primarily qualitative due to 'limited time and computational resources', which reinforces that the quantitative claims are not established but is not itself circularity.

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

SegVec3D introduces no new physical or mathematical entities; it recombines existing modules: attention-based point feature extraction, contrastive embedding loss, and a CLIP-style projection to a shared text-3D space. All free parameters are unspecified hyperparameters or thresholds. The load-bearing assumptions are the availability of weak grouping signals for the contrastive loss, the sufficiency of ScanNet category labels for language alignment, the reliability of a pretrained sentence transformer, and the transferability of the segmentation model to a new sensor domain.

free parameters (4)
  • contrastive loss margin m = not specified
    Appears in Eq. (8) and controls how far apart different instances must be in embedding space. The value is hand-chosen and no sensitivity analysis is reported.
  • InfoNCE temperature tau = not specified
    Appears in Eq. (12) and controls the sharpness of the cross-modal contrastive distribution. No value or ablation is reported.
  • clustering radius or threshold at inference = not specified
    Determines final instance masks from the learned embeddings. The authors note in Section 5.3 that tuning this threshold changes fragmentation, so results depend on this hand-picked parameter.
  • k-NN neighborhood size k and number of attention layers L = not specified
    These control the receptive field and local context in Sections 3.1 and 3.2. Their values are not given, yet they directly affect segmentation quality.
assumptions (5)
  • ad hoc to paper Weak grouping signals (spatial heuristics, over-segmentation, or sparse instance IDs) are available to form positive and negative pairs for contrastive training.
    Section 3.2 states 'we assume some weak grouping signals are available to form positive and negative pairs for training'. Without this, Eq. (8) is undefined and the unsupervised instance segmentation claim collapses.
  • domain assumption ScanNet category labels plus a few descriptive phrases provide sufficient text supervision to align 3D instances with language.
    Sections 4.2 and 5.1 describe training the cross-modal alignment on ScanNet category labels. There is no evidence that this generalizes to open-vocabulary descriptors or attribute combinations.
  • domain assumption A pretrained sentence transformer encodes object names and attributes in a semantically useful way for the target classes.
    Section 4.1 relies on a BERT-based sentence transformer to provide text embeddings. The quality of the shared space is inherited from that pretrained model, and no measurement of that transfer is provided.
  • domain assumption k-NN spatial adjacency plus stacked attention yields point features that separate object instances in embedding space.
    Sections 3.1 and 3.2 assume that the learned embeddings cluster by object instance. Only qualitative visualization supports this, with no clustering accuracy or instance IoU reported.
  • domain assumption A model trained on ScanNet transfers to a different LiDAR and PolyCam lab scene.
    Sections 5.1 and 5.2 use the lab scene as a test set. The authors assert strong generalization, but domain shift is not quantified and no failure analysis is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SegVec3D: A Method for Vector Embedding of 3D Objects Oriented Towards Robot manipulation." pith.science (2026). https://pith.science/paper/3V3NMPLM

@misc{pith2026250709459,
  author       = {Pith},
  title        = {Pith review of: SegVec3D: A Method for Vector Embedding of 3D Objects Oriented Towards Robot manipulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3V3NMPLM}},
  note         = {Machine review of arXiv:2507.09459}
}
read the original abstract

We propose SegVec3D, a novel framework for 3D point cloud instance segmentation that integrates attention mechanisms, embedding learning, and cross-modal alignment. The approach builds a hierarchical feature extractor to enhance geometric structure modeling and enables unsupervised instance segmentation via contrastive clustering. It further aligns 3D data with natural language queries in a shared semantic space, supporting zero-shot retrieval. Compared to recent methods like Mask3D and ULIP, our method uniquely unifies instance segmentation and multimodal understanding with minimal supervision and practical deployability.

Figures

Figures reproduced from arXiv: 2507.09459 by the authors.

Figure 1
Figure 1. Illustration of the local perception structure and point [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the global feature fusion mechanism. A [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Overall architecture of the attention-based point cloud [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Multimodal alignment via contrastive learning. Dur [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Overall cross-modal semantic alignment pipeline. The [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: 3D reconstruction of the lab room point cloud (interior view). This is the input point cloud for our experiment, obtained [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Semantic instance segmentation result on the lab room point cloud. Each object instance is highlighted in a different color [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 38 canonical work pages

  1. [1]

    B., & Cousins, S

    Rusu, R. B., & Cousins, S. (2011). 3D is here: Point Cloud Library (PCL).Proceedings of the IEEE Interna- tional Conference on Robotics and Automation (ICRA). IEEE, 2011: 1–4

    work page 2011

  2. [2]

    R., Su, H., Mo, K., & Guibas, L

    Qi, C. R., Su, H., Mo, K., & Guibas, L. J. (2017). Point- Net: Deep Learning on Point Sets for 3D Classification and Segmentation.Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2017: 652–660

    work page 2017

  3. [3]

    Zhang, J., & Singh, S. (2014). LOAM: Lidar Odometry 10 and Mapping in Real-time.Robotics: Science and Systems (RSS), 2014

    work page 2014

  4. [4]

    SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmen- tation.Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR)

    Chen, L., Qi, X.,et al.(2018). SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmen- tation.Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR). IEEE, 2018: 2569–2578

    work page 2018

  5. [5]

    W.,et al.(2021)

    Radford, A., Kim, J. W.,et al.(2021). Learning Transfer- able Visual Models From Natural Language Supervision. Proceedings of the International Conference on Machine Learning (ICML). PMLR, 2021: 8748–8763

    work page 2021

  6. [6]

    LanguageRefer: 3D Visual Grounding on Point Clouds with Language.Proceedings of the European Conference on Computer Vision (ECCV)

    Xue, F., Chen, Z.,et al.(2022). LanguageRefer: 3D Visual Grounding on Point Clouds with Language.Proceedings of the European Conference on Computer Vision (ECCV). Springer, 2022: 468–484

    work page 2022

  7. [7]

    Tombari, F., Salti, S., & Di Stefano, L. (2010). Unique Sig- natures of Histograms for Local Surface Description.Pro- ceedings of the European Conference on Computer Vision (ECCV). Springer, 2010: 356–369

    work page 2010

  8. [8]

    Dynamic Graph CNN for Learning on Point Clouds.ACM Transactions on Graphics (TOG), 38(5): 1–12

    Wang, Y ., Sun, Y ., Liu, Z.,et al.(2019). Dynamic Graph CNN for Learning on Point Clouds.ACM Transactions on Graphics (TOG), 38(5): 1–12

    work page 2019

Show all 40 references
  1. [9]

    W.,et al.(2021)

    Zhao, H., Jiang, L., Fu, C. W.,et al.(2021). Point Transformer.Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). IEEE, 2021: 16259–16268

    2021

  2. [10]

    ULIP: Unifying Language and Image Pre-training for Vision-Language Understand- ing.Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR)

    Huang, Y ., Gu, J.,et al.(2023). ULIP: Unifying Language and Image Pre-training for Vision-Language Understand- ing.Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR). IEEE, 2023: 250–260

    2023

  3. [11]

    3D-ULIP: Uni- fied Language and 3D Pretraining for Open-V ocabulary 3D Understanding.Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR)

    Luo, C., Zhao, Y ., Yang, Q.,et al.(2023). 3D-ULIP: Uni- fied Language and 3D Pretraining for Open-V ocabulary 3D Understanding.Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR). IEEE, 2023: 245–255

    2023

  4. [12]

    Ni, B., Li, X., Nishino, K., & Yang, M.-H. (2022). Point- CLIP: Point Cloud Understanding by CLIP.Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR). IEEE, 2022: 8557–8567

    2022

  5. [13]

    Rabbani, T., van den Heuvel, F., & V osselman, G. (2006). Segmentation of Point Clouds Using Smoothness Con- straint.International Archives of the Photogrammetry, Re- mote Sensing and Spatial Information Sciences (ISPRS), 36(5): 248–253

    2006

  6. [14]

    R., Yi, L., Su, H., & Guibas, L

    Qi, C. R., Yi, L., Su, H., & Guibas, L. J. (2017). Point- Net++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space.Advances in Neural Information Pro- cessing Systems (NeurIPS), 30: 5099–5108

    2017

  7. [15]

    R., Deschaud, J

    Thomas, H., Qi, C. R., Deschaud, J. E.,et al.(2019). KPConv: Flexible and Deformable Convolution for Point Clouds.Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). IEEE, 2019: 6410–6419

    2019

  8. [16]

    RandLA-Net: Effi- cient Semantic Segmentation of Large-Scale Point Clouds

    Hu, Q., Yang, B., Xie, L.,et al.(2020). RandLA-Net: Effi- cient Semantic Segmentation of Large-Scale Point Clouds. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2020: 11105–11114

    2020

  9. [17]

    H., V o, X., Tran, N.-K.,et al.(2022)

    Vu, T. H., V o, X., Tran, N.-K.,et al.(2022). SoftGroup: Learning an Adaptive Grouping for Point Cloud Instance Segmentation.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2022: 12589–12598

    2022

  10. [18]

    Hamilton, W., Ying, R., & Leskovec, J. (2017). Induc- tive Representation Learning on Large Graphs.Advances in Neural Information Processing Systems (NeurIPS), 30: 1024–1034

    2017

  11. [19]

    Graph Attention Networks.Proceedings of the Interna- tional Conference on Learning Representations (ICLR), 2018

    Veli ˇckovi´c, P., Cucurull, G., Casanova, A.,et al.(2018). Graph Attention Networks.Proceedings of the Interna- tional Conference on Learning Representations (ICLR), 2018

    2018

  12. [20]

    PAConv: Position Adaptive Convolution with Dynamic Kernel As- sembling on Point Clouds.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Xu, Q., Zhang, X., Huang, X.,et al.(2021). PAConv: Position Adaptive Convolution with Dynamic Kernel As- sembling on Point Clouds.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2021: 11171–11180

    2021

  13. [21]

    CLIP2Point: Trans- fer CLIP to Point Cloud Understanding with Contextual Prompting.arXiv preprint arXiv:2211.11688, 1–13

    Zeng, Z., Li, Q., Xu, Z.,et al.(2022). CLIP2Point: Trans- fer CLIP to Point Cloud Understanding with Contextual Prompting.arXiv preprint arXiv:2211.11688, 1–13

    2022 arXiv

  14. [22]

    Neural Fields in Vi- sual Computing and Beyond: A Survey.Computer Graph- ics Forum, 41(6): 641–666

    Xie, Y ., Yao, H., Sun, X.,et al.(2022). Neural Fields in Vi- sual Computing and Beyond: A Survey.Computer Graph- ics Forum, 41(6): 641–666

    2022

  15. [23]

    X., Funkhouser, T., Guibas, L.,et al.(2015)

    Chang, A. X., Funkhouser, T., Guibas, L.,et al.(2015). ShapeNet: An Information-Rich 3D Model Repository. arXiv preprint arXiv:1512.03012, 1–12

    2015 arXiv

  16. [24]

    Vector Neu- rons: A General Framework for SO(3)-Equivariant Net- works.Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR)

    Deng, B., Yang, M., Liu, J.,et al.(2021). Vector Neu- rons: A General Framework for SO(3)-Equivariant Net- works.Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR). IEEE, 2021: 14321–14332

    2021

  17. [25]

    AdaFit: Learn- ing to Fit Multiple Geometric Primitives for Robust Shape Segmentation.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Wang, Y ., Zhao, S., Gou, R.,et al.(2021). AdaFit: Learn- ing to Fit Multiple Geometric Primitives for Robust Shape Segmentation.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2021: 7445–7455

    2021

  18. [26]

    Segmentation-based Urban Localization Using 3D Point Clouds.Robotics and Autonomous Systems, 59(10): 738–748

    Douillard, B., Fox, D., Ramos, F.,et al.(2011). Segmentation-based Urban Localization Using 3D Point Clouds.Robotics and Autonomous Systems, 59(10): 738–748

    2011

  19. [27]

    A.,et al.(2017)

    Tulsiani, S., Zhou, T., Efros, A. A.,et al.(2017). Multi-view Supervision for Single-view Reconstruction via Differentiable Ray Consistency.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2017: 5876–5885

    2017

  20. [28]

    Grid-GCN for Fast and Scal- able Point Cloud Learning.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Xu, Q.,et al.(2020). Grid-GCN for Fast and Scal- able Point Cloud Learning.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2020: 3260–3269

    2020

  21. [29]

    Wu, W., Qi, Z., & Fuxin, L. (2019). PointConv: Deep Con- volutional Networks on 3D Point Clouds.Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR). IEEE, 2019: 9621–9630

    2019

  22. [30]

    Geiger, A., Lenz, P., & Urtasun, R. (2012). Are We Ready for Autonomous Driving? The KITTI Vision Bench- mark Suite.Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR). IEEE, 2012: 3354–3361

    2012

  23. [31]

    R.,et al.(2017)

    Armeni, I., Sax, A., Zamir, A. R.,et al.(2017). Joint 2D- 3D-Semantic Data for Indoor Scene Understanding.arXiv 11 preprint arXiv:1702.01105, 1–13

    2017 arXiv

  24. [32]

    Lahoud, J., Khoury, F., & Aubry, M. (2019). 3D Instance Segmentation via Multi-Task Metric Learning.Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2019: 9471–9480

    2019

  25. [33]

    Point Trans- former V2: Grouped Vector Attention and Partition-based Pooling.Advances in Neural Information Processing Sys- tems (NeurIPS), 36: 11234–11247

    Engel, N., Xie, Z., Zhu, H.,et al.(2023). Point Trans- former V2: Grouped Vector Attention and Partition-based Pooling.Advances in Neural Information Processing Sys- tems (NeurIPS), 36: 11234–11247

    2023

  26. [34]

    Baltru ˇsaitis, T., Ahuja, C., & Morency, L. P. (2019). Mul- timodal Machine Learning: A Survey and Taxonomy. IEEE Transactions on Pattern Analysis and Machine In- telligence, 41(2): 423–443

    2019

  27. [35]

    Tsai, Y . H. H., Bai, S., Yamada, M.,et al.(2019). Mul- timodal Transformer for Unaligned Multimodal Language Sequences.Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL), 2019: 655–668

    2019

  28. [36]

    R., Gotmare, A.,et al.(2021)

    Li, J., Selvaraju, R. R., Gotmare, A.,et al.(2021). Align Before Fuse: Vision and Language Representa- tion Learning with Momentum Distillation.Advances in Neural Information Processing Systems (NeurIPS), 34: 12387–12399

    2021

  29. [37]

    3D Con- cept Grounding on Point Clouds with Natural Language

    Chen, W., Zhang, H., Su, Y .,et al.(2022). 3D Con- cept Grounding on Point Clouds with Natural Language. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2022: 14353–14362

    2022

  30. [38]

    ReferIt3D: Neural Listeners for Fine-Grained 3D Object Identifica- tion in Real-World Scenes.Proceedings of the European Conference on Computer Vision (ECCV)

    Liu, L., Duan, Y ., Lin, W.,et al.(2020). ReferIt3D: Neural Listeners for Fine-Grained 3D Object Identifica- tion in Real-World Scenes.Proceedings of the European Conference on Computer Vision (ECCV). Springer, 2020: 164–181

    2020

  31. [39]

    Schroff, F., Kalenichenko, D., & Philbin, J. (2015). FaceNet: A Unified Embedding for Face Recognition and Clustering.Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015: 815–823

    2015

  32. [40]

    Mask3D: Mask Transformer for 3D Semantic Instance Segmentation.Proceedings of the IEEE International Con- ference on Robotics and Automation (ICRA)

    Schult, J., Engelmann, F., Hermans, A.,et al.(2023). Mask3D: Mask Transformer for 3D Semantic Instance Segmentation.Proceedings of the IEEE International Con- ference on Robotics and Automation (ICRA). IEEE, 2023. 12

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.