Pith. sign in

REVIEW 3 major objections 5 minor 50 references

LSVG: Language-Guided Scene Graphs with 2D-Assisted Multi-Modal Encoding for 3D Visual Grounding

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that 3D visual grounding improves when the relational scene graph is built from the words in the description rather than from all nearby objects, and that a frozen 2D vision-language model can supply both the features…

desk verdict Solid within-subfield improvement: language-guided scene graphs plus CLIP-supervised 3D encoding beat prior work on Nr3D/Sr3D, but the abstract overclaims on ScanRefer and the graph's vocabulary dependence is under-analyzed. read the letter →

arxiv 2505.04058 v3 pith:AIVOO2YG submitted 2025-05-07 cs.CV

Feng Xiao , Hongbin Xu , Guocan Zhao , Wenxiong Kang This is my paper · ORCID
classification cs.CV
keywords 3Dvisualgroundingscenegraphlanguage-guided2D-3Dmulti-modalencodingreferredobjectrelationalreasoningvision-languagealignmentdistractordiscrimination
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 attacks a specific failure mode in 3D visual grounding, the task of locating the one object in a 3D scene that a natural-language sentence describes: when several objects of the same class are present, models tend to spread relational attention over everything instead of over the objects the sentence actually mentions. Its proposal is to let the sentence decide which objects matter, by constructing a language-guided scene graph whose nodes are only the candidates and referred objects found through a coarse 3D-language matching step. On the ReferIt3D benchmarks the framework reaches 67.2% overall accuracy on Nr3D and 76.0% on Sr3D, ahead of previous state-of-the-art methods, and the biggest gains are on hard scenes with many same-category distractors. A second claim is that a frozen 2D vision-language model can supply both extra visual features and contrastive supervision, and that this supervision is what makes the 3D-language matching reliable enough to build the graph.

What carries the argument

The central object is the language-guided scene graph: a graph over object proposals whose edge set $A_{ij}\in\{0,1\}$ is not spatial proximity but semantic match between lemmatized (base-form) description words and class-aware prompts of the form “The object is [instance name]”, with instance names drawn from the training utterances. Nodes are then updated by graph attention with that masked adjacency. The other load-bearing component is the dual-branch 3D encoder: a point-cloud branch produces geometric features $F_P$, while a second branch fed by a frozen 2D vision-language model produces 2D-enhanced features $F_M$, and contrastive losses $L_{ot}$ align $F_P$, $F_M$, and textual features so the language-guided graph has a trustworthy matching signal. A cross-modal interaction block alternates object-text cross-attention with graph attention, injecting rotated-box spatial embeddings before each round.

What would settle it

Build a validation subset whose referring landmarks are designated only by attributes rather than class names (for instance, “the chair with the torn cushion”), compare accuracy with sentences whose landmarks appear in the prompt list, and check whether accuracy collapses; if it does, lexical prompt-matching is doing the work, and if it does not, the graph construction is not the active mechanism.

Watch

Extended reading notes

Core claim

The central claim is that relational reasoning in 3D grounding should be confined to the objects the referring expression names or implies, rather than spread over every object in the scene. LSVG therefore replaces fully connected or KNN-based scene graphs with a sparse graph whose adjacency is set by semantic consistency: description words are lemmatized and matched against class-aware prompts, and only objects identified as potential targets or referred objects become nodes. Graph attention then aggregates features over exactly those nodes, alternating with vision-language cross-attention. The paper reports that this produces the strongest results on Nr3D and Sr3D, with particular gains on hard samples, and attributes them to the target node being strongly activated through its contextual connection to correctly identified referred objects.

Load-bearing premise

The load-bearing premise is that a sentence's referred object can be found by matching its words against a fixed list of class-aware prompts built from training-set instance names; if a description refers to a landmark by an attribute, color, or word outside that list, the graph simply lacks the node the entire relational argument depends on.

Editorial extensions

If this is right

  • On scenes with many same-category distractors, accuracy should improve most; the paper reports a 3.8-point gain on Nr3D Hard and a 4.0-point gain on ScanRefer's Multiple subset over its closest comparison.
  • Because the graph is sparse and built once from a single 2D view per object, the approach should be cheaper than methods that aggregate all views or keep fully connected graphs.
  • If the language-guided edge selection is the active mechanism, the graph should handle long-range relations too, since a referred object does not need to be a spatial neighbor of the target.
  • The two-layer optimum for graph attention suggests that these sparse graphs need only limited aggregation, and that more layers can smear the target node's activation.

Reading between the lines

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

  • A natural stress test is to replace lexical prompt matching with an open-vocabulary extractor for referred objects; if the accuracy gains persist, the mechanism is the sparse graph itself rather than the closed vocabulary of prompts.
  • The same two-part recipe, a frozen 2D language-vision model acting as both feature enhancer and contrastive teacher, transfers to other 3D-language tasks such as dense captioning or embodied navigation, where the modality gap is similar.
  • The authors' observation that larger frozen 2D teachers do not improve results suggests the bottleneck is not teacher capacity but the 3D-language alignment it supervises, so better alignment losses may pay off more than larger backbones.
  • The reported drop with three graph attention layers could be checked directly by measuring node-feature similarity across layers; over-smoothing predicts a sharp rise in similarity between the second and third layers.
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 LSVG, a two-stage 3D visual grounding framework that constructs a language-guided scene graph whose nodes are objects semantically matched to the textual description, then applies iterative cross-attention and graph attention to predict the target. A dual-branch PointNet++ encoder uses frozen CLIP features both to enhance the 3D representation and to supervise 3D-language alignment through contrastive losses. Experiments on Nr3D, Sr3D, and ScanRefer report state-of-the-art results on the ReferIt3D benchmarks, with ablations showing contributions from the 2D supervision, 2D enhancement, and graph learning components.

Significance. If the claims hold, the paper makes a useful practical contribution: confining relational reasoning to linguistically relevant objects is a sensible inductive bias for 3D visual grounding, and the single-view CLIP enhancement is computationally appealing. The ablation studies in Tables 3-7 are internally consistent and give clear evidence that each proposed component contributes to the reported Nr3D/Sr3D improvements. The central graph-construction mechanism, however, has a closed-vocabulary limitation that is not acknowledged, and the ScanRefer results are overstated relative to Table 2. These issues affect the paper's main claims and require clarification before the results can be fully assessed.

major comments (3)
  1. [§3.1.1, §3.2.1] The claimed 'open vocabulary' 3D-language alignment is contradicted by the graph construction described in §3.2.1. In §3.1.1 the instance names in the class-aware prompts are defined as 'the object types found in the training set utterances,' so the prompt vocabulary is fixed by the training set. The graph is then built by matching lemmatized description lexemes to these same class-aware prompts. Consequently, descriptions whose referent noun is a synonym, a free-form compound, an attribute-only phrase, or an unseen category will not produce the intended referred-object node, and the relational graph that is the paper's central novelty cannot operate on those examples. The paper reports no coverage statistics for the matching step and no comparison of performance on examples where a referred-object node is present versus absent. This is a load-bearing point: the claimed mechanism cannot be verified as the source of the hard-sample gains without such statistics.
  2. [Table 2, §4.3.2, Abstract] The abstract's claim of 'superior performance compared to state-of-the-art methods' is not supported on ScanRefer. Table 2 reports LSVG at 45.37% overall and 40.87% on the Multiple subset, below MA2TransVG at 45.7% and 41.4% respectively. The 4.0% improvement on Multiple in §4.3.2 is relative to M3DRef-CLIP (36.8%), not to the state of the art. The text should be revised to state that the state-of-the-art results are on Nr3D/Sr3D, and that on ScanRefer LSVG is competitive but not superior to MA2TransVG. This also affects the accuracy of the contribution summary in the abstract.
  3. [§3.2.1, Fig. 1] Even when the referent noun is in the prompt vocabulary, the graph connects all objects of the same semantic category (e.g., all chairs), not the specific referred instance. The matching in §3.2.1 is category-level: 'we use lexemes from the same semantic category in the modified description text to identify related objects.' Thus the graph performs category-conditioned pruning, while the later graph attention is what distinguishes the particular referred object. The paper's phrasing in Fig. 1 and the abstract, such as 'connects objects that are identified as referred objects or targets,' is therefore stronger than the actual mechanism. This distinction should be stated explicitly, and the contribution should be framed as category-conditioned graph construction plus learned instance discrimination rather than direct 'referred object discrimination' at graph-construction time.
minor comments (5)
  1. [Eq. (4)] In the graph attention update, the summation over heads k combined with the 1/sqrt(dk) scaling is notationally unclear; the scaling normally applies per-head before softmax, and the role of A_{ij} as a binary mask multiplying the attention weight should be clarified.
  2. [Eq. (1)-(2)] The contrastive loss notation Lc uses C1 and C2 without precise definitions; in particular, the relationship between the batch size, the number of classes, and the negative set for the class-aware prompts should be stated.
  3. [Table 4] The abbreviations PI*, TT*, and IC in Table 4 are not expanded in the caption; the text explains them only later in §4.5.2, which makes the table hard to read in isolation.
  4. [Abstract] The stray '1' after the keywords appears to be a formatting artifact and should be removed.
  5. [§2.1.1] The typo 'V oteNet' should be corrected to 'VoteNet'.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the language-guided scene graph is built from input text and trained 3D-language matching, not from target labels; the sole self-citation is non-load-bearing.

full rationale

LSVG is an empirically evaluated architecture whose central chain—constructing a language-guided scene graph via 3D-language semantic matching, then applying graph attention before a grounding head—does not reduce to its own inputs. The graph edges in Section 3.2.1 are produced by matching point-cloud object features to class-aware prompts and selecting objects whose lemmatized category appears in the description; the ground-truth target label is not used in this construction, and the final prediction comes from a separate trained grounding head over the graph nodes. The contrastive losses of Equation (1) supervise a shared text/3D embedding space, but they do not by themselves determine which object is the target. The only self-citation, SECG [45], appears in a survey-style sentence about prior graph-construction strategies and is not used to justify the method or to set any reported accuracy. The closed-vocabulary dependence noted in Sections 3.1.1 and 3.2.1—instance names come from training-set utterances, so unseen or non-canonical referent nouns may fail to create graph nodes—is a genuine robustness and correctness limitation, and the failure-case discussion in Section 4.4 acknowledges related shortcomings; however, it is an empirical coverage issue rather than a circular derivation. No equation, ablation, or benchmark number is equivalent to a fitted parameter renamed as a prediction, and no uniqueness theorem or load-bearing self-citation chain is invoked. Under the stated rules, the appropriate finding is no significant circularity.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim depends on several hand-chosen hyperparameters (fusion layer, projection dim, graph layers, loss weights) and on domain assumptions about CLIP transfer and MaskClustering proposal quality. No new physical or conceptual entities are postulated.

free parameters (7)
  • loss weights lambda1, lambda2, lambda3 = 0.5, 0.1, 0.5
    Set by hand in Section 4.2 and not ablated in the paper.
  • CLIP fusion layer index = layer 3
    Selected via ablation in Table 5; the paper reports layer 3 as optimal.
  • 2D feature projection dimension = 256
    Chosen to match the output dimension of the point cloud network layer at fusion; Table 5 shows 256 at layer 3 is best.
  • number of graph attention layers = 2
    Selected via ablation in Table 6.
  • CLIP backbone = ViT-B/16
    Selected via ablation in Table 7; no significant gain from larger models.
  • number of 2D views per object = 1 (max projection coverage)
    Design choice in Section 3.1.2; compared against multi-view methods in Table 2.
  • class-aware prompt template = "The object is [instance name]"
    Design choice in Section 3.1.1; no ablation is provided.
assumptions (5)
  • domain assumption Single-view CLIP features from the most projected view adequately represent an object's semantics for grounding.
    Section 3.1.2; the entire 2D enhancement branch depends on this.
  • domain assumption Lemmatized description tokens matched to class-aware prompts identify all relevant referred objects.
    Section 3.2.1; graph edges are built from this match, and unseen categories or attribute-only references break it.
  • domain assumption MaskClustering provides class-agnostic proposals of sufficient quality for ScanRefer grounding.
    Section 3; the two-stage pipeline rests on this segmentation module without quantitative evaluation.
  • domain assumption Contrastive loss with CLIP features transfers 2D-language alignment to 3D point cloud features.
    Section 3.1.1; the training objective assumes that CLIP's embedding geometry is a valid target for 3D features.
  • standard math Standard attention and graph attention formulations are used as given.
    Equations 4 and 5 rely on standard softmax attention and multi-head graph attention.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LSVG: Language-Guided Scene Graphs with 2D-Assisted Multi-Modal Encoding for 3D Visual Grounding." pith.science (2026). https://pith.science/paper/AIVOO2YG

@misc{pith2026250504058,
  author       = {Pith},
  title        = {Pith review of: LSVG: Language-Guided Scene Graphs with 2D-Assisted Multi-Modal Encoding for 3D Visual Grounding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AIVOO2YG}},
  note         = {Machine review of arXiv:2505.04058}
}
read the original abstract

3D visual grounding aims to localize the unique target described by natural languages in 3D scenes. The significant gap between 3D and language modalities makes it a notable challenge to distinguish multiple similar objects through the described spatial relationships. Current methods attempt to achieve cross-modal understanding in complex scenes via a target-centered learning mechanism, ignoring the modeling of referred objects. We propose a novel 3D visual grounding framework that constructs language-guided scene graphs with referred object discrimination to improve relational perception. The framework incorporates a dual-branch visual encoder that leverages pre-trained 2D semantics to enhance and supervise the multi-modal 3D encoding. Furthermore, we employ graph attention to promote relationship-oriented information fusion in cross-modal interaction. The learned object representations and scene graph structure enable effective alignment between 3D visual content and textual descriptions. Experimental results on popular benchmarks demonstrate our superior performance compared to state-of-the-art methods, especially in handling the challenges of multiple similar distractors.

Figures

Figures reproduced from arXiv: 2505.04058 by the authors.

Figure 1
Figure 1. An example of text-guided localization in a scene. In the text, the target is marked in yellow and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The overall architecture of our proposed LSVG. The input scene point clouds are segmented into object point clouds, which are then processed by a dual-branch object encoder. We leverage 2D multi-modal features from the frozen CLIP model to guide additional feature encoding. Object-level feature alignment is calculated exclusively during training. Instance identifiers are separated from the original textual descripti… view at source ↗
Figure 3
Figure 3. [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Visualization of our method on Nr3D and ScanRefer. The number under the text entry in the “Description” column indicates the number of target-class objects in the scene. Yellow bounding boxes surround the correctly predicted targets, and the error predicted results are…
Figure 5
Figure 5. Figure 5: Visualization of the results of ablation experiments. The “baseline” denotes the complete LSVG method we propose. The last three columns present the localization results of the model without the three important components. The incorrectly predicted 3D bounding boxes ar…
Figure 6
Figure 6. Figure 6: Visualization of graph attention results. The third column shows the heat maps of the attention weights of the last graph attention layer, where only the top 10 responsive node pairs are retained to represent object scores. In the second column, Orange ellipses demarca…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 45 canonical work pages

  1. [1]

    D. Liu, Y . Liu, W. Huang, W. Hu, A survey on text-guided 3-d visual grounding: Elements, recent advances, and future directions, IEEE Transactions on Neural Networks and Learning Systems (2025)

    work page 2025

  2. [2]

    Achlioptas, A

    P. Achlioptas, A. Abdelreheem, F. Xia, M. Elhoseiny, L. Guibas, Referit3d: Neu- ral listeners for fine-grained 3d object identification in real-world scenes, in: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, Au- gust 23–28, 2020, Proceedings, Part I 16, Springer, 2020, pp. 422–440

    work page 2020

  3. [3]

    D. He, Y . Zhao, J. Luo, T. Hui, S. Huang, A. Zhang, S. Liu, Transrefer3d: Entity- and-relation aware transformer for fine-grained 3d visual grounding, in: Proceed- ings of the 29th ACM International Conference on Multimedia, 2021, pp. 2344– 2352

    work page 2021

  4. [4]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural infor- mation processing systems 30 (2017)

    2017

  5. [5]

    Z. Yang, S. Zhang, L. Wang, J. Luo, Sat: 2d semantics assisted training for 3d visual grounding, in: Proceedings of the IEEE /CVF International Conference on Computer Vision, 2021, pp. 1856–1866

    work page 2021

  6. [6]

    J. Roh, K. Desingh, A. Farhadi, D. Fox, Languagerefer: Spatial-language model for 3d visual grounding, in: Conference on Robot Learning, PMLR, 2022, pp. 1046–1056

    work page 2022

  7. [7]

    Huang, Y

    S. Huang, Y . Chen, J. Jia, L. Wang, Multi-view transformer for 3d visual ground- ing, in: Proceedings of the IEEE /CVF Conference on Computer Vision and Pat- tern Recognition, 2022, pp. 15524–15533

    work page 2022

  8. [8]

    Z. Guo, Y . Tang, R. Zhang, D. Wang, Z. Wang, B. Zhao, X. Li, Viewrefer: Grasp the multi-view knowledge for 3d visual grounding, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 15372– 15383. 25

    work page 2023

Show all 50 references
  1. [9]

    Chen, P.-L

    S. Chen, P.-L. Guhur, M. Tapaswi, C. Schmid, I. Laptev, Language conditioned spatial relation reasoning for 3d object grounding, Advances in neural informa- tion processing systems 35 (2022) 20522–20535

    2022

  2. [10]

    C. Xu, Y . Han, R. Xu, L. Hui, J. Xie, J. Yang, Multi-attribute interactions matter for 3d visual grounding, in: Proceedings of the IEEE /CVF Conference on Com- puter Vision and Pattern Recognition, 2024, pp. 17253–17262

    2024

  3. [11]

    Chang, S

    C.-P. Chang, S. Wang, A. Pagani, D. Stricker, Mikasa: Multi-key-anchor & scene- aware transformer for 3d visual grounding, in: Proceedings of the IEEE /CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 14131– 14140

    2024

  4. [12]

    Chang, P

    X. Chang, P. Ren, P. Xu, Z. Li, X. Chen, A. Hauptmann, A comprehensive sur- vey of scene graphs: Generation and application, IEEE Transactions on Pattern Analysis and Machine Intelligence 45 (1) (2021) 1–26

    2021

  5. [13]

    Zhang, Z

    H. Zhang, Z. Li, J. Liu, Scenellm: Implicit language reasoning in llm for dynamic scene graph generation, Pattern Recognition 170 (2026) 111992

    2026

  6. [14]

    M. Feng, Z. Li, Q. Li, L. Zhang, X. Zhang, G. Zhu, H. Zhang, Y . Wang, A. Mian, Free-form description guided 3d visual graph network for object grounding in point cloud, in: Proceedings of the IEEE /CVF international conference on com- puter vision, 2021, pp. 3722–3731

    2021

  7. [15]

    Y . Li, Z. Wang, W. Liang, R2g: Reasoning to ground in 3d scenes, Pattern Recog- nition (2025) 111728

    2025

  8. [16]

    Huang, H.-H

    P.-H. Huang, H.-H. Lee, H.-T. Chen, T.-L. Liu, Text-guided graph neural net- works for referring 3d instance segmentation, in: Proceedings of the AAAI Con- ference on Artificial Intelligence, V ol. 35, 2021, pp. 1610–1618

    2021

  9. [17]

    Z. Yuan, X. Yan, Y . Liao, R. Zhang, S. Wang, Z. Li, S. Cui, Instancerefer: Cooper- ative holistic understanding for visual grounding on point clouds through instance multi-level contextual referring, in: Proceedings of the IEEE /CVF International Conference on Computer Vision...

    2021

  10. [18]

    Y . Lei, Z. Wang, F. Chen, G. Wang, P. Wang, Y . Yang, Recent advances in multi- modal 3d scene understanding: A comprehensive survey and evaluation, arXiv preprint arXiv:2310.15676 (2023)

    2023 arXiv

  11. [19]

    D. Z. Chen, A. X. Chang, M. Nießner, Scanrefer: 3d object localization in rgb- d scans using natural language, in: European conference on computer vision, Springer, 2020, pp. 202–221

    2020

  12. [20]

    L. Zhao, D. Cai, L. Sheng, D. Xu, 3dvg-transformer: Relation modeling for vi- sual grounding on point clouds, in: Proceedings of the IEEE /CVF International Conference on Computer Vision, 2021, pp. 2928–2937

    2021

  13. [21]

    J. Luo, J. Fu, X. Kong, C. Gao, H. Ren, H. Shen, H. Xia, S. Liu, 3d-sps: Single- stage 3d visual grounding via referred point progressive selection, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 16454–16463

    2022

  14. [22]

    D. Cai, L. Zhao, J. Zhang, L. Sheng, D. Xu, 3djcg: A unified framework for joint dense captioning and visual grounding on 3d point clouds, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 16464–16473

    2022

  15. [23]

    D. Z. Chen, Q. Wu, M. Nießner, A. X. Chang, D 3 net: A unified speaker-listener architecture for 3d dense captioning and visual grounding, in: European Confer- ence on Computer Vision, Springer, 2022, pp. 487–505

    2022

  16. [24]

    Zhang, Z

    Y . Zhang, Z. Gong, A. X. Chang, Multi3drefer: Grounding text description to multiple 3d objects, in: Proceedings of the IEEE /CVF International Conference on Computer Vision, 2023, pp. 15225–15236

    2023

  17. [25]

    C. R. Qi, O. Litany, K. He, L. J. Guibas, Deep hough voting for 3d object detection in point clouds, in: proceedings of the IEEE /CVF International Conference on Computer Vision, 2019, pp. 9277–9286. 27

    2019

  18. [26]

    Jiang, H

    L. Jiang, H. Zhao, S. Shi, S. Liu, C.-W. Fu, J. Jia, Pointgroup: Dual-set point grouping for 3d instance segmentation, in: Proceedings of the IEEE/CVF confer- ence on computer vision and Pattern recognition, 2020, pp. 4867–4876

    2020

  19. [27]

    Z. Liu, Z. Zhang, Y . Cao, H. Hu, X. Tong, Group-free 3d object detection via transformers, in: Proceedings of the IEEE/CVF international conference on com- puter vision, 2021, pp. 2949–2958

    2021

  20. [28]

    X. Shi, Z. Wu, S. Lee, Aware visual grounding in 3d scenes, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 14056–14065

    2024

  21. [29]

    Paszke, A

    A. Paszke, A. Chaurasia, S. Kim, E. Culurciello, Enet: A deep neural network ar- chitecture for real-time semantic segmentation, arXiv preprint arXiv:1606.02147 (2016)

    2016 arXiv

  22. [30]

    Salvador, X

    A. Salvador, X. Gir ´o-i Nieto, F. Marqu ´es, S. Satoh, Faster r-cnn features for in- stance search, in: Proceedings of the IEEE conference on computer vision and pattern recognition workshops, 2016, pp. 9–16

    2016

  23. [31]

    E. Bakr, Y . Alsaedy, M. Elhoseiny, Look around and refer: 2d synthetic semantics knowledge distillation for 3d visual grounding, Advances in neural information processing systems 35 (2022) 37146–37158

    2022

  24. [32]

    X. Ke, P. Xu, W. Guo, Language–image consistency augmentation and distillation network for visual grounding, Pattern Recognition 166 (2025) 111663

    2025

  25. [33]

    Chen, D.-Z

    F.-L. Chen, D.-Z. Zhang, M.-L. Han, X.-Y . Chen, J. Shi, S. Xu, B. Xu, Vlp: A survey on vision-language pre-training, Machine Intelligence Research 20 (1) (2023) 38–56

    2023

  26. [34]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al., Learning transferable visual models from natural language supervision, in: International conference on machine learning, PmLR, 2021, pp. 8748–8763. 28

    2021

  27. [35]

    C. Jia, Y . Yang, Y . Xia, Y .-T. Chen, Z. Parekh, H. Pham, Q. Le, Y .-H. Sung, Z. Li, T. Duerig, Scaling up visual and vision-language representation learning with noisy text supervision, in: International conference on machine learning, PMLR, 2021, pp. 4904–4916

    2021

  28. [36]

    J. Li, R. Selvaraju, A. Gotmare, S. Joty, C. Xiong, S. C. H. Hoi, Align before fuse: Vision and language representation learning with momentum distillation, Advances in neural information processing systems 34 (2021) 9694–9705

    2021

  29. [37]

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

    2023

  30. [38]

    B. Jia, Y . Chen, H. Yu, Y . Wang, X. Niu, T. Liu, Q. Li, S. Huang, Sceneverse: Scaling 3d vision-language learning for grounded scene understanding, in: Euro- pean Conference on Computer Vision, Springer, 2024, pp. 289–310

    2024

  31. [39]

    L. Xue, M. Gao, C. Xing, R. Mart´ın-Mart´ın, J. Wu, C. Xiong, R. Xu, J. C. Niebles, S. Savarese, Ulip: Learning a unified representation of language, images, and point clouds for 3d understanding, in: Proceedings of the IEEE /CVF conference on computer vision and pattern recog...

    2023

  32. [40]

    M. Yan, J. Zhang, Y . Zhu, H. Wang, Maskclustering: View consensus based mask graph clustering for open-vocabulary 3d instance segmentation, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 28274–28284

    2024

  33. [41]

    C. R. Qi, L. Yi, H. Su, L. J. Guibas, Pointnet ++: Deep hierarchical feature learn- ing on point sets in a metric space, Advances in neural information processing systems 30 (2017)

    2017

  34. [42]

    C. Yang, Z. An, L. Huang, J. Bi, X. Yu, H. Yang, B. Diao, Y . Xu, Clip-kd: An empirical study of clip model distillation, in: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, 2024, pp. 15952–15962. 29

    2024

  35. [43]

    J. D. M.-W. C. Kenton, L. K. Toutanova, Bert: Pre-training of deep bidirectional transformers for language understanding, in: Proceedings of naacL-HLT, V ol. 1, 2019, p. 2

    2019

  36. [44]

    Zhang, S

    T. Zhang, S. He, T. Dai, Z. Wang, B. Chen, S.-T. Xia, Vision-language pre- training with object contrastive learning for 3d scene understanding, in: Pro- ceedings of the AAAI Conference on Artificial Intelligence, V ol. 38, 2024, pp. 7296–7304

    2024

  37. [45]

    F. Xiao, H. Xu, Q. Wu, W. Kang, Secg: Semantic-enhanced 3d visual grounding via cross-modal graph attention, arXiv preprint arXiv:2403.08182 (2024)

    2024 arXiv

  38. [46]

    A. Dai, A. X. Chang, M. Savva, M. Halber, T. Funkhouser, M. Nießner, Scan- net: Richly-annotated 3d reconstructions of indoor scenes, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 5828– 5839

    2017

  39. [47]

    Y . Wu, X. Cheng, R. Zhang, Z. Cheng, J. Zhang, Eda: Explicit text-decoupling and dense alignment for 3d visual grounding, in: Proceedings of the IEEE /CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 19231– 19242

    2023

  40. [48]

    A. Jain, N. Gkanatsios, I. Mediratta, K. Fragkiadaki, Bottom up top down detec- tion transformers for language grounding in images and point clouds, in: Euro- pean Conference on Computer Vision, Springer, 2022, pp. 417–433

    2022

  41. [49]

    Z. Qian, Y . Ma, Z. Lin, J. Ji, X. Zheng, X. Sun, R. Ji, Multi-branch collaborative learning network for 3d visual grounding, in: European Conference on Computer Vision, Springer, 2024, pp. 381–398

    2024

  42. [50]

    Z. Jin, M. Hayat, Y . Yang, Y . Guo, Y . Lei, Context-aware alignment and mutual masking for 3d-language pre-training, in: Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2023, pp. 10984–10994. 30

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.