LESV: Language Embedded Sparse Voxel Fusion for Open-Vocabulary 3D Scene Understanding

Dzmitry Tsishkou; Fabien Moutarde; Fusang Wang; Luis Rold\~ao; Moussab Bennehar; Nathan Piasco

REVIEW 2 major objections 5 minor 31 references

LESV: Language Embedded Sparse Voxel Fusion for Open-Vocabulary 3D Scene Understanding

T0 review · 2 major / 5 minor · reviewed 2026-07-13 · grok-4.5

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

Pith's one-line read Sparse voxels plus dense foundation features turn open-vocabulary 3D queries into a deterministic volume fusion problem, cutting bleeding and hierarchy cost while lifting accuracy on LERF and ScanNet.

desk verdict Solid engineering paper: SVRaster + dense RADIO features + confidence fusion cleanly beats 3DGS registration on LERF/ScanNet and cuts preprocessing time; monocular-prior sensitivity is a real but secondary soft spot, not a collapse of the claim. read the letter →

arxiv 2604.01388 v2 pith:EQZ3NK7R submitted 2026-04-01 cs.CV

Fusang Wang , Nathan Piasco , Moussab Bennehar , Luis Roldão , Dzmitry Tsishkou , Fabien Moutarde This is my paper · ORCID

classification cs.CV

keywords open-vocabulary3DunderstandingsparsevoxelrasterizationlanguagefeatureregistrationAM-RADIOsemanticbleedingpoint-cloudLERFScanNet

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

Open-vocabulary 3D scene understanding has been built mostly on 3D Gaussian Splatting, which registers language features into overlapping, unstructured primitives. That geometry forces probabilistic assignment and produces semantic bleeding; separate multi-level mask pipelines are then needed to handle part-to-whole queries. This paper replaces the Gaussian backbone with Sparse Voxel Rasterization, regularizes it with monocular depth and normal priors, and fuses dense language-aligned tokens from an agglomerative foundation model with a confidence gate derived from multi-level TSDF mesh depth. The result is a single, deterministic 3D language field that answers fine-grained and global queries without hierarchical mask training. On LERF 3D object retrieval and ScanNet open-vocabulary point-cloud understanding the method sets new state-of-the-art numbers, while feature preprocessing drops by roughly an order of magnitude.

What carries the argument

Confidence-aware sparse-voxel fusion: each voxel aggregates multi-view dense foundation features weighted by a Gaussian kernel on depth discrepancy and by a continuous geometric confidence map obtained from multi-level TSDF mesh rendering; the explicit, disjoint voxel grid makes the mapping deterministic and memory-partitionable.

What would settle it

On a LERF or ScanNet scene where monocular depth is known to be systematically wrong near thin structures or grazing angles, disable the confidence gate and measure whether 3D mIoU collapses relative to the gated version; if the gated model still wins by a large margin the surface assumption is not load-bearing.

Watch

Extended reading notes

Core claim

A monocular-prior-regularized Sparse Voxel Rasterization volume, combined with confidence-aware fusion of dense AM-RADIO language tokens, yields a deterministic open-vocabulary 3D feature field that eliminates the spatial ambiguity of overlapping Gaussians and the multi-level mask overhead of hierarchical methods, producing state-of-the-art retrieval and point-cloud scores.

Load-bearing premise

The monocular depth and normal priors plus multi-level TSDF mesh must give a surface accurate enough that the depth-discrepancy kernel and confidence gate correctly suppress bleeding; systematic bias in those priors would be inherited by every registered feature.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The useful takeaway is simple: they replace probabilistic 3DGS registration with an explicit sparse-voxel volume, gate the fusion with multi-level TSDF confidence, and drop dense AM-RADIO tokens instead of hierarchical SAM+CLIP. That combination produces the numbers they advertise—LERF 3D mIoU 56.11, ScanNet 19-class 53.22 (+21 over Dr. Splat)—and the qualitative figures really do show less bleeding and cleaner part-level hits.

What is new is the pipeline, not any single ingredient. SVRaster, monocular depth/normal losses, RADIO dense tokens, and sliding-window upsampling all exist. The contribution is wiring them so that feature weights become local geometric proximity plus a confidence map, which also lets them batch the fusion and drop the 256 GB RAM tax of Dr. Splat. Ablations (Table 4) and the efficiency table are clean; the multi-level TSDF mesh extraction is a practical detail that works. Citations are appropriate and the free parameters are ordinary kernel bandwidths, not hidden fits.

The stress-test note is half-right. Section 3.2 admits vanilla SVRaster geometry is hollow, so the priors and confidence gate are load-bearing. Table 4 shows the confidence term itself adds almost nothing (+0.05 mIoU), which means the real lift is the structured voxels plus RADIO, not the fancy gate. They never ablate monocular estimator choice or outdoor/reflective failure modes, so that assumption stays unprobed. Still, the central claim does not collapse: the geometry is regularized enough for the reported indoor benchmarks, and the bleeding reduction is visible. Missing code and a few underspecified thresholds are the usual reproducibility tax, not circularity.

This is for people building open-vocab 3D systems who care about both accuracy and wall-clock preprocessing. It is not a theory paper. I would send it to peer review; the gains and ablations are real enough that referees can demand the prior-sensitivity experiment and code. Worth reading if you work in the area; I would cite the ScanNet numbers and the fusion formulation.

Referee Report

2 major / 5 minor

Summary. LESV proposes open-vocabulary 3D scene understanding by registering dense language-aligned features from AM-RADIO onto a Sparse Voxel Rasterization (SVRaster) geometry rather than unstructured 3D Gaussians. The authors argue that overlapping Gaussians force probabilistic registration and produce semantic bleeding, while SAM-hierarchy mask pooling dilutes multi-level semantics. They regularize SVRaster with patch-wise monocular depth (Eq. 3) and analytic normal (Eq. 4) losses, extract a multi-level TSDF mesh for a depth-confidence gate (Eqs. 5–6), and fuse features deterministically by local geometric proximity (Eq. 2). Multi-level ambiguity is addressed by projecting AM-RADIO spatial tokens through a language head, with sliding-window upsampling plus SCRA/SCGA denoising. Reported results claim SOTA on LERF 3D object retrieval (avg. mIoU 56.11 %) and ScanNet open-vocabulary point-cloud understanding (19-class mIoU 53.22 %, +21.5 over Dr. Splat), competitive 2D retrieval, and large reductions in feature preprocessing and fusion time.

Significance. If the results hold under fair comparison, the work is a clear practical advance for open-vocabulary 3D understanding. Replacing probabilistic 3DGS registration with an explicit, disjoint voxel volume plus confidence-aware fusion is a well-motivated architectural shift; the large ScanNet gains and the qualitative elimination of spillover are useful for robotics and AR. Using AM-RADIO dense tokens to avoid hierarchical SAM pipelines is an orthogonal efficiency contribution that is well supported by the timing table. The paper supplies standard external benchmarks, component ablations, and qualitative evidence of multi-level localization, which strengthens the claim relative to pure distillation baselines.

major comments (2)

Section 3.2 and Eqs. 2–6: the central claim that registration is deterministic and bleeding-free rests on monocular depth/normal priors plus multi-level TSDF mesh extraction producing a sufficiently accurate, view-consistent surface. The manuscript itself states that vanilla SVRaster yields fragmented/hollow geometry that disrupts fusion; yet no sensitivity experiment is reported (different monocular estimators, outdoor/reflective scenes, or deliberate prior noise). Table 4 shows the confidence term alone adds only +0.05 mIoU, so the geometric advantage over Dr. Splat is thinner than the narrative and remains unprobed. A short prior-sensitivity or failure-mode analysis is needed to support the load-bearing geometric claim.
Section 4.2 and Table 1: because AM-RADIO features have a different score distribution from SAM+CLIP baselines, the authors introduce normalized cosine similarity for 3D thresholding on all methods. This is reasonable for fairness, but the paper does not report the un-normalized Dr. Splat numbers under the original protocol, nor a threshold-sweep. Without that, it is hard to isolate how much of the +3.4 mIoU average gain is attributable to the representation versus the evaluation alignment. A brief protocol appendix or dual-threshold column would make the SOTA claim fully transparent.

minor comments (5)

Many equations and figure captions in the provided text are corrupted by encoding artifacts (e.g., Eq. 1–9, Fig. 1–5 labels). The camera-ready version must restore clean math and legible figures.
Table 2: LESV is second-best on average 2D mIoU and Loc; the abstract’s phrasing of “highly competitive” is accurate, but the main text should avoid overstating 2D SOTA.
Free parameters (σ, τ, sliding-window size, SCRA/SCGA threshold) are listed only implicitly; a short hyperparameter table or default values would aid reproducibility.
Related-work discussion of ProFuse and OpenGaussian is brief; a clearer contrast on how multi-level TSDF confidence differs from their mask-proposal aggregation would help.
Supplementary Algorithm 1 and multi-level TSDF fusion are important for the confidence gate; a short pointer or one-sentence summary in the main text would improve self-containment.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: geometry-driven fusion and external LERF/ScanNet metrics are independent of the language queries being evaluated.

full rationale

The paper's derivation chain is self-contained and non-circular. SVRaster geometry is optimized with monocular depth/normal priors (Eqs. 3-4) and multi-level TSDF confidence (Eq. 5), then used to deterministically fuse pre-extracted AM-RADIO features via local proximity weights (Eqs. 2, 6). These steps produce a 3D feature field that is subsequently queried with open-vocabulary text embeddings; the fusion weights depend only on rendered vs. mesh depth discrepancy and never on the language labels or evaluation masks. All reported numbers (LERF 3D/2D mIoU, ScanNet 19/15/10-class mIoU) are obtained under fixed external protocols against held-out ground truth, with no parameters fitted to the test queries. Self-citations (e.g., [26] for the confidence formulation) supply a reusable engineering motif, not a uniqueness theorem or a quantity that is later re-predicted. Ablations (Table 4) and efficiency timings are likewise measured, not definitional. Consequently no step reduces a claimed prediction to its own inputs by construction.

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

The central empirical claim rests on standard CV assumptions (foundation-model features transfer, monocular geometric cues are useful) plus a handful of ordinary kernel and temperature parameters; no new physical entities are postulated.

free parameters (4)

Gaussian kernel bandwidth σ (depth discrepancy)
Controls how sharply fusion weights fall off with depth error (Eq. 2); chosen by hand / validation.
confidence decay temperature τ
Scales the absolute depth discrepancy into a continuous confidence map (Eq. 5); free hyper-parameter.
sliding-window crop size / Gaussian attenuation width
Determines spatial resolution and boundary blending of the dense RADIO feature maps; selected empirically.
SCRA/SCGA cosine-similarity threshold
Hard mask for self-correcting attention; value not derived, set to suppress noise.

assumptions (3)

domain assumption AM-RADIO patch tokens remain language-aligned after projection through the SigLIP head without dense supervision
Invoked in §3.3; taken from emergent-property observations in the RADIO literature.
domain assumption Monocular depth and normal estimators supply locally accurate relative geometry for SVRaster regularization
Used in the patch-wise depth and analytic normal losses (Eqs. 3–4).
domain assumption Sparse voxels are strictly disjoint and admit deterministic surface-to-voxel mapping
Core geometric premise of SVRaster that enables the claimed non-probabilistic registration.

how reviews work

0 comments

Cite this review

Pith. "Pith review of LESV: Language Embedded Sparse Voxel Fusion for Open-Vocabulary 3D Scene Understanding." pith.science (2026). https://pith.science/paper/EQZ3NK7R

@misc{pith2026260401388,
  author       = {Pith},
  title        = {Pith review of: LESV: Language Embedded Sparse Voxel Fusion for Open-Vocabulary 3D Scene Understanding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EQZ3NK7R}},
  note         = {Machine review of arXiv:2604.01388}
}

read the original abstract

Recent advancements in open-vocabulary 3D scene understanding heavily rely on 3D Gaussian Splatting (3DGS) to register vision-language features into 3D space. However, we identify two critical limitations in these approaches: the spatial ambiguity arising from unstructured, overlapping Gaussians which necessitates probabilistic feature registration, and the multi-level semantic ambiguity caused by pooling features over object-level masks, which dilutes fine-grained details. To address these challenges, we present a novel framework that leverages Sparse Voxel Rasterization (SVRaster) as a structured, disjoint geometry representation. By regularizing SVRaster with monocular depth and normal priors, we establish a stable geometric foundation. This enables a deterministic, confidence-aware feature registration process and suppresses the semantic bleeding artifact common in 3DGS. Furthermore, we resolve multi-level ambiguity by exploiting the emerging dense alignment properties of the AM-RADIO foundation model, avoiding the computational overhead of hierarchical training methods. Our approach achieves state-of-the-art performance on Open Vocabulary Point Cloud Understanding, and highly competitive results on 3D and 2D Object Retrieval benchmarks.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 2 linked inside Pith

[1]

In: International Conf

Alama, O., Bhattacharya, A., He, H., Kim, S., Qiu, Y., Wang, W., Ho, C., Keetha, N.V., Scherer, S.A.: RayFronts: Open-set semantic ray frontiers for online scene understanding and exploration. In: International Conf. on Intelligent Robots and Systems (IROS) (2025)

2025
[2]

arXiv:2511.19704 (2025)

Alama, O., Jariwala, D., Bhattacharya, A., Kim, S., Wang, W., Scherer, S.: RAD- Seg: Unleashing parameter and compute efficient zero-shot open-vocabulary seg- mentation using agglomerative models. arXiv:2511.19704 (2025)

arXiv 2025
[3]

arXiv:2601.04754 (2026)

Chiou, Y.J., Cheng, W.T., Yang, Y.F.: ProFuse: Efficient cross-view context fusion for open-vocabulary 3D gaussian splatting. arXiv:2601.04754 (2026)

arXiv 2026
[4]

In: IEEE Conf

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

2017
[5]

In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

Ding, R., Yang, J., Xue, C., Zhang, W., Bai, S., Qi, X.: Pla: Language-driven open- vocabulary 3d scene understanding. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 7010–7019 (2023)

2023
[6]

IEEE Robotics and Automation Letters (2025)

Halacheva, A.M., Zaech, J.N., Wang, X., Paudel, D.P., Van Gool, L.: Gaussianvlm: Scene-centric 3d vision-language models using language-aligned gaussian splats for embodied reasoning and beyond. IEEE Robotics and Automation Letters (2025)

2025
[7]

Splat: Directlyreferring3Dgaussiansplattingviadirectlanguageembeddingregistration

Jun-Seong, K., Kim, G., Kim, Y.J., Wang, Y.C.F., Choe, J., Oh, T.H.: Dr. Splat: Directlyreferring3Dgaussiansplattingviadirectlanguageembeddingregistration. In: IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) (2025)

2025
[8]

ACM Trans

Kerbl, B., Kopanas, G., Leimkuehler, T., Drettakis, G.: 3D gaussian splatting for real-time radiance field rendering. ACM Trans. on Graphics (TOG) (2023)

2023

Show all 31 references

[9]

In: Proc

Kerr, J., Kim, C.M., Goldberg, K., Kanazawa, A., Tancik, M.: LERF: Language embedded radiance fields. In: Proc. of the IEEE International Conf. on Computer Vision (ICCV) (2023)

2023
[10]

In: Proceedings of the IEEE/CVF international conference on computer vision

Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A.C., Lo, W.Y., et al.: Segment anything. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 4015–4026 (2023)

2023
[11]

In: IEEE Conf

Li, J., Zhang, J., Bai, X., Zheng, J., Ning, X., Zhou, J., Gu, L.: Dngaussian: Op- timizing sparse-view 3d gaussian radiance fields with global-local depth normal- ization. In: IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) (2024)

2024
[12]

In: Advances in Neural Information Processing Systems (NeurIPS) (2025)

Li, W., Zhao, Y., Qin, M., Liu, Y., Cai, Y., Gan, C., Pfister, H.: LangSplatV2: High-dimensional 3D language gaussian splatting with 450+ fps. In: Advances in Neural Information Processing Systems (NeurIPS) (2025)

2025
[13]

In: IEEE Conf

Li, Y., Ma, Q., Yang, R., Li, H., Ma, M., Ren, B., Popovic, N., Sebe, N., Konukoglu, E., Gevers, T., et al.: Scenesplat: Gaussian splatting-based scene understanding with vision-language pretraining. In: IEEE Conf. on Computer Vision and Pattern Recognition (CVPR (2025)

2025
[14]

IEEE Transactions on Pattern Analysis and Machine Intelligence45, 3292–3310 (2021),��

Liao, Y., Xie, J., Geiger, A.: Kitti-360: A novel dataset and benchmarks for urban scene understanding in 2d and 3d. IEEE Transactions on Pattern Analysis and Machine Intelligence45, 3292–3310 (2021),��

2021
[15]

Advances in Neural Information Processing Systems36, 53433–53456 (2023) Abbreviated paper title 9

Liu, K., Zhan, F., Zhang, J., Xu, M., Yu, Y., El Saddik, A., Theobalt, C., Xing, E., Lu, S.: Weakly supervised 3d open-vocabulary segmentation. Advances in Neural Information Processing Systems36, 53433–53456 (2023) Abbreviated paper title 9

2023
[16]

In: Proc

Mildenhall, B., Srinivasan, P.P., Tancik, M., Barron, J.T., Ramamoorthi, R., Ng, R.: NeRF: Representing scenes as neural radiance fields for view synthesis. In: Proc. of the European Conf. on Computer Vision (ECCV) (2020)

2020
[17]

arXiv preprint arXiv:2304.07193 (2023)

Oquab, M., Darcet, T., Moutakanni, T., Vo, H., Szafraniec, M., Khalidov, V., Fernandez, P., Haziza, D., Massa, F., El-Nouby, A., et al.: Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193 (2023)

2023 arXiv
[18]

In: Proc

Peng, Y., Wang, H., Liu, Y., Wen, C., Dong, Z., Yang, B.: GAGS: Granularity- aware feature distillation for language gaussian splatting. In: Proc. of the Conf. on Artificial Intelligence (AAAI) (2026)

2026
[19]

In: IEEE Conf

Qin, M., Li, W., Zhou, J., Wang, H., Pfister, H.: LangSplat: 3D language gaussian splatting. In: IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) (2024)

2024
[20]

In: Proc

Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., Sutskever, I.: Learning transferable visual models from natural language supervision. In: Proc. of the International Conf. on Machine Learning (I...

2021
[21]

In: International conference on machine learning

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

2021
[22]

In: IEEE Conf

Ranzinger,M.,Heinrich,G.,Kautz,J.,Molchanov,P.:AM-RADIO:Agglomerative vision foundation model reduce all domains into one. In: IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) (2024)

2024
[23]

ArXivabs/2511.16301(2025),��

Seo, M., Hamilton, M., Kim, C.: Upsample anything: A simple and hard to beat baseline for feature upsampling. ArXivabs/2511.16301(2025),��

2025
[24]

In: IEEE Conf

Shi, J.C., Wang, M., Duan, H.B., Guan, S.H.: Language embedded 3D gaussians for open-vocabulary scene understanding. In: IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) (2024)

2024
[25]

In: IEEE Conf

Sun, C., Choe, J., Loop, C., Ma, W.C., Wang, Y.C.F.: Sparse voxels rasteriza- tion: Real-time high-fidelity radiance field rendering. In: IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) (2025)

2025
[26]

arXiv preprint arXiv:2501.03932 (2025)

Wang, F., Djeghim, H., Piasco, N., Bennehar, M., Roldão, L., Wu, Y., Moutarde, F., Sidibé, D., Tsishkou, D.: J-neus: Joint field optimization for neural sur- face reconstruction in urban scenes with limited image overlap. arXiv preprint arXiv:2501.03932 (2025)

2025
[27]

In: International Conf

Wimmer, T., Truong, P., Rakotosaona, M.J., Oechsle, M., Tombari, F., Schiele, B., Lenssen, J.E.: Anyup: Universal feature upsampling. In: International Conf. on Learning Representations (ICLR) Workshop (2026)

2026
[28]

In: Advances in Neural Information Processing Systems (NeurIPS) (2024)

Wu, Y., Meng, J., Li, H., Wu, C., Shi, Y., Cheng, X., Zhao, C., Feng, H., Ding, E., Wang, J., Zhang, J.: OpenGaussian: Towards point-level 3D gaussian-based open vocabulary understanding. In: Advances in Neural Information Processing Systems (NeurIPS) (2024)

2024
[29]

In: Advances in Neural Information Processing Systems (NeurIPS) (2022)

Yu,Z.,Peng,S.,Niemeyer,M.,Sattler,T.,Geiger,A.:Monosdf:Exploringmonocu- lar geometric cues for neural implicit surface reconstruction. In: Advances in Neural Information Processing Systems (NeurIPS) (2022)

2022
[30]

Wang et al

Zhai, X., Mustafa, B., Kolesnikov, A., Beyer, L.: Sigmoid loss for language image pre-training.In:Proc.oftheIEEEInternationalConf.onComputerVision(ICCV) (2023) 10 F. Wang et al

2023
[31]

Zhou, S., Chang, H., Jiang, S., Fan, Z., Zhu, Z., Xu, D., Chari, P., You, S., Wang, Z., Kadambi, A.: Feature 3DGS: Supercharging 3D gaussian splatting to enable distilledfeaturefields.In:IEEEConf.onComputerVisionandPatternRecognition (CVPR) (2024)

2024

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.

[1] [1]

In: International Conf

Alama, O., Bhattacharya, A., He, H., Kim, S., Qiu, Y., Wang, W., Ho, C., Keetha, N.V., Scherer, S.A.: RayFronts: Open-set semantic ray frontiers for online scene understanding and exploration. In: International Conf. on Intelligent Robots and Systems (IROS) (2025)

2025

[2] [2]

arXiv:2511.19704 (2025)

Alama, O., Jariwala, D., Bhattacharya, A., Kim, S., Wang, W., Scherer, S.: RAD- Seg: Unleashing parameter and compute efficient zero-shot open-vocabulary seg- mentation using agglomerative models. arXiv:2511.19704 (2025)

arXiv 2025

[3] [3]

arXiv:2601.04754 (2026)

Chiou, Y.J., Cheng, W.T., Yang, Y.F.: ProFuse: Efficient cross-view context fusion for open-vocabulary 3D gaussian splatting. arXiv:2601.04754 (2026)

arXiv 2026

[4] [4]

In: IEEE Conf

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

2017

[5] [5]

In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

Ding, R., Yang, J., Xue, C., Zhang, W., Bai, S., Qi, X.: Pla: Language-driven open- vocabulary 3d scene understanding. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 7010–7019 (2023)

2023

[6] [6]

IEEE Robotics and Automation Letters (2025)

Halacheva, A.M., Zaech, J.N., Wang, X., Paudel, D.P., Van Gool, L.: Gaussianvlm: Scene-centric 3d vision-language models using language-aligned gaussian splats for embodied reasoning and beyond. IEEE Robotics and Automation Letters (2025)

2025

[7] [7]

Splat: Directlyreferring3Dgaussiansplattingviadirectlanguageembeddingregistration

Jun-Seong, K., Kim, G., Kim, Y.J., Wang, Y.C.F., Choe, J., Oh, T.H.: Dr. Splat: Directlyreferring3Dgaussiansplattingviadirectlanguageembeddingregistration. In: IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) (2025)

2025

[8] [8]

ACM Trans

Kerbl, B., Kopanas, G., Leimkuehler, T., Drettakis, G.: 3D gaussian splatting for real-time radiance field rendering. ACM Trans. on Graphics (TOG) (2023)

2023