Tackling View-Dependent Semantics in 3D Language Gaussian Splatting

Changsong Wen; Jiazhong Cen; Jiemin Fang; Lingxi Xie; Qi Tian; Wei Shen; Xiaopeng Zhang; XuDong Zhou

REVIEW 2 major objections 5 minor 1 cited by

Tackling View-Dependent Semantics in 3D Language Gaussian Splatting

T0 review · 2 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read LaGa lifts 3D language segmentation by 18.7% mIoU

desk verdict The view-aggregated descriptor idea is solid and the ablations are convincing, but the headline gain over OpenGaussian rests on an opaque revised evaluation script that the authors do not describe or release. read the letter →

arxiv 2505.24746 v1 pith:IUFCKEJI submitted 2025-05-30 cs.CV

Jiazhong Cen , Xudong Zhou , Jiemin Fang , Changsong Wen , Lingxi Xie , Xiaopeng Zhang , Wei Shen , Qi Tian This is my paper

classification cs.CV

keywords 3DGaussianSplattingopen-vocabularysegmentationview-dependentsemanticslanguage-groundedsceneunderstandingdecompositionviewaggregationCLIPfeaturesSAMmasks

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that the main obstacle to direct 3D language understanding in 3D Gaussian Splatting is not how features are stored but the fact that one physical object carries different visible semantics from different viewpoints. Simply lifting per-view 2D features onto Gaussians leaves each Gaussian with only the semantics visible from the views that trained it, producing both false positives and false negatives at query time. LaGa handles this by first decomposing the 3D scene into objects, then representing each object by an adaptively chosen set of semantic descriptors obtained by clustering its multi-view features, with weights that reward descriptors aligned with the object's global semantics and internally compact. At query time the object's relevance to a text is the maximum weighted response over its descriptors, so a query that matches any reliable view retrieves the whole object. On the LERF-OVS benchmark LaGa reaches 64.0% mIoU, 18.7 points above the previous 3D state of the art and 8.8 points above the best 2D-rendering methods, and it also improves on prior methods for ScanNet point-cloud segmentation.

What carries the argument

The load-bearing object is the view-aggregated semantic representation, built per 3D object in two stages. First, a contrastively trained per-Gaussian affinity feature (32-dimensional, with masked-average-pooled mask prototypes clustered by HDBSCAN, a density-based clustering algorithm) decomposes the scene into objects and assigns every Gaussian to one object, so each object carries all its multi-view 2D semantic features. Second, adaptive K-means with silhouette-score selection partitions those features into one to twenty semantic descriptors, and each descriptor $d$ receives a weight $\omega_d = \langle d, \bar{v}_{S_i}\rangle \cdot \|d\|_2$, where $\bar{v}_{S_i}$ is the mean feature of the object: the first factor rewards directional consistency with the object's global semantics, the second rewards internal compactness of the descriptor's cluster. The object-level relevance to a text query $q$ is then $\mathrm{REL}(G_{S_i}, q) = \max_{d\in D_{G_{S_i}}} \omega_d \cdot \mathrm{Rel}(d,q)$, with Rel following LERF's canonical-phrase normalization. This machinery is what preserves the semantic sides of an object—for example, a book's spine and its title—instead of averaging them away or locking the object into one rigid feature.

What would settle it

The decisive experiment is to run LaGa with ground-truth or manually verified object groupings instead of the learned HDBSCAN decomposition on the LERF-OVS scenes. If the reported 18.7-point mIoU gain over the previous 3D state of the art persists under oracle groupings, the view-aggregated descriptors are the driver; if the gain vanishes, the improvement comes from the decomposition stage rather than from handling view-dependent semantics.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that view-dependent semantics—the same 3D object producing different CLIP features from different viewpoints—is a measurable and fixable cause of degraded Gaussian-level language understanding. The paper quantifies the problem with two analyses: intra-object multi-view features are often less similar to each other than they are to features of other objects, and about half of 2D semantic features cannot fully retrieve their own 3D object. LaGa's answer is to connect views through object-level decomposition and then represent each object with a small set of view-aggregated descriptors rather than a single averaged feature: adaptive K-means selects how many descriptors an object needs, and a weighting scheme suppresses descriptors that point away from the object's dominant semantics or that are internally inconsistent. Inference takes the maximum weighted relevance over these descriptors for each object, producing direct Gaussian-level segmentation without rendering feature maps. The paper's headline evidence is a 18.7-point mIoU improvement over the previous 3D state of the art on LERF-OVS, with the largest gains on the 360-degree scenes where view-dependency is most severe.

Load-bearing premise

The load-bearing premise is that the grouping of multi-view segmentation masks produced by the Segment Anything model into 3D objects is accurate and viewpoint-invariant; if that grouping merges two different objects or splits one object, every semantic descriptor built on those groups inherits the error and the pipeline collapses.

Editorial extensions

If this is right

Gaussian-level open-vocabulary queries become practical for editing, deletion, and transplanting objects across scenes, since segmentation is produced directly on the 3D Gaussians rather than on rendered feature maps.
Per-object descriptor sets replace per-Gaussian high-dimensional language features, which the paper argues reduces storage and accelerates inference while sharing one semantic pool across all Gaussians of an object.
The two diagnostic analyses (intra/inter similarity overlap and the roughly 50% low-recall retrieval rate) give the field a concrete way to measure how strongly view-dependency affects any new 3D language dataset or method.
On forward-facing or nearly saturated benchmarks such as 3D-OVS, gains are small, which follows from the claim: when viewpoint diversity is low, view-dependency is not the bottleneck.

Reading between the lines

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

Beyond the paper: the view-aggregation recipe is not tied to 3D-GS; it could be applied to any per-point feature field lifted from multi-view 2D encoders, including NeRF-based language fields or dynamic 4D reconstructions where visibility changes over time.
Editorial inference: the silhouette-selected cluster count is effectively the number of semantic sides an object has from the training viewpoints; geometry (surface normals or visibility maps) could predict this count more cheaply than clustering features.
Beyond the paper: the retrieval-integrity analysis could be reused as a view-dependency index for scene datasets, letting a practitioner decide in advance whether view aggregation is worth the added pipeline complexity.
Editorial inference: because LaGa inherits CLIP's bag-of-words and context-loss failures, the value of view aggregation should increase as the 2D semantic backbone improves; the method treats the backbone as an exchangeable module.

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 core idea here is worth taking seriously: instead of assigning one rigid semantic vector to a 3D object, LaGa clusters multi-view CLIP features into several descriptors and weights them by directional consistency and internal compactness. That is a sensible response to the view-dependency problem they document, and the ablation study actually supports it—adaptive clustering beats fixed and average pooling, and the weighting terms both matter. The method is simple and the qualitative results look clean. I think the paper does a fair job of situating itself against OpenGaussian, LangSplat, and others, and the failure-case section is honest.

The soft spot is the comparison protocol. The +18.7% mIoU figure on LERF-OVS is computed against the authors' own reimplementation of OpenGaussian (45.3 mIoU, marked ‡), not the published number (38.4). Footnote 2 says they revised OpenGaussian's official evaluation script and gained 6.9 mIoU, but they never say what they revised, and the script is not released. That matters for two reasons: the baseline becomes non-reproducible, and the change could conceivably favor LaGa's binary mask outputs over OpenGaussian's per-Gaussian relevance outputs. The direction of the change is not obvious without seeing the script. So the exact margin is uncertain, even though the method's internal logic is sound.

A second, smaller issue is that inference relies on a chain of hand-tuned thresholds and post-processing steps—0.23 cosine threshold, min-max normalization, 3D bilateral filtering, 0.6 foreground threshold. None of these are ablated, so their contribution to the reported numbers is unknown. That is fixable but should be addressed if the paper is revised.

I would not let the evaluation script issue sink the paper; the method itself is new and the experiments are more extensive than most in this subfield. But if I were citing the +18.7%, I'd add a caveat. A referee should ask for the revised OpenGaussian script and a sensitivity analysis of the thresholds.

For us: this is worth discussing at reading group, and I'd cite it for the descriptor-weighting idea. I'd send it out for peer review if it were under consideration, with the evaluation script question front and center.

Referee Report

2 major / 5 minor

Summary. The paper addresses the degradation of direct 3D language understanding in 3D Gaussian Splatting caused by view-dependent semantics, where an object's semantic appearance varies across viewpoints. The proposed method, LaGa, first decomposes the 3D scene into objects using contrastively trained affinity features and HDBSCAN clustering of multi-view SAM masks, then constructs per-object view-aggregated semantic representations by adaptive K-means clustering of CLIP features and weighting the resulting descriptors via a global-alignment and internal-compactness criterion. At inference, LaGa assigns Gaussians to a query by the maximum weighted descriptor relevance, producing binary Gaussian-level segmentation maps. Experiments on LERF-OVS, 3D-OVS, and ScanNet report improvements over prior 3D methods, including a headline +18.7 mIoU gain over OpenGaussian on LERF-OVS.

Significance. If the results hold, LaGa makes a meaningful contribution by identifying and explicitly modeling view-dependent semantics, a factor largely ignored by previous feature-lifting methods. The paper's strengths include experiments on three benchmarks, ablations of the descriptor extraction and weighting components, qualitative analyses of failure cases, and a released codebase. The view-aggregated representation with adaptive clustering and descriptor weighting is a plausible and well-motivated design. However, the headline quantitative claim currently rests on an undisclosed and unreleased revision of the OpenGaussian evaluation script, which changes the comparison baseline by +6.9 mIoU, and on several hand-tuned inference thresholds that are not ablated. These issues must be resolved before the reported margin can be considered reliable, but they are local to the evaluation protocol and do not invalidate the core methodological idea.

major comments (2)

[Section 6.2, Table 1, footnote 2] The claim of '+18.7% mIoU over the previous SOTA on LERF-OVS' is computed against the authors' reimplementation of OpenGaussian (45.3, marked ‡) rather than the published OpenGaussian result (38.4, marked †). Footnote 2 states that the evaluation script was revised, but the revision is not described and the revised script is not released. Because the same revised protocol is applied to LaGa, the headline margin could change if the revision is incorrect, inconsistent with the original benchmark, or systematically favors LaGa's binary Gaussian-mask outputs. Please specify the exact revisions, release the script, and report LaGa's performance against the published numbers for direct comparison.
[Appendix B (Implementation Details), inference thresholds] Inference uses several hand-set components: the auxiliary cosine similarity threshold (0.23), min-max normalization, 3D bilateral filtering, and the foreground relevance threshold (0.6). None of these is ablated. If these thresholds were tuned on the evaluation sets, the reported gains could reflect threshold tuning rather than the view-aggregated semantic representation. Please ablate each component or demonstrate that the thresholds are fixed a priori and applied identically to all baselines and all scenes.

minor comments (5)

[Abstract and Section 6.2] The phrase '+18.7% mIoU' is ambiguous: it should be stated as '+18.7 percentage points of mIoU' or as a relative improvement, since the reported numbers are absolute mIoU values.
[Section 6.1] The description of how binary Gaussian segmentation maps are rendered into 2D views for evaluation would benefit from a precise statement of whether this rendering protocol matches the original LERF-OVS benchmark protocol used by the baselines, especially given the revision to OpenGaussian's evaluation script.
[Section 5.3, 'Why Is 3D Scene Decomposition Unaffected by View-Dependency?'] This paragraph argues that SAM masks remain stable across viewpoints, but it does not provide quantitative evidence. Given that the whole pipeline depends on the decomposition grouping masks into objects correctly, a measurable analysis of multi-view mask grouping stability would strengthen the claim.
[Section 6.4, Table 4] The use of a gray background to mark superior results within each 'DW' group is difficult to discern in print; consider using bold or a footnote instead.
[Various] Typos and duplicated text: 'negative affect' in the introduction, 'assemantic' in Section 5.4, and a duplicated sentence in Appendix C.5 ('A more intriguing example involves transplanting a “cup” ...').

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the mIoU gain is an external measurement; the unreleased revised evaluation script is a validity risk, not a circular reduction.

full rationale

No equation-level circularity is present. The descriptors and weights (Eqs. 7-12) are computed from training-view CLIP features of SAM masks, and the headline LERF-OVS / 3D-OVS / ScanNet numbers are evaluated against external ground-truth annotations. The object-level relevance score (Eq. 9) uses a text query that is not an input to the descriptor extraction, so the result is not defined in terms of the prediction. The self-citation to the authors' SAGA for feature-norm regularization (Appendix B.2) is auxiliary to affinity-feature training and does not justify the central view-aggregation claim, so it is not load-bearing. The main risk is the footnote-2 'revised evaluation script' that raises the OpenGaussian baseline from 38.4 to 45.3 and is neither described nor released; this is a benchmark-validity and reproducibility concern, not a logical dependence of the method's output on its own inputs. Inference thresholds (0.23, 0.6, bilateral filtering) could overfit the benchmark, but they are post-processing choices rather than fitted parameters renamed as predictions.

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

The central claim rests on borrowed components (SAM masks, CLIP features, HDBSCAN, LERF relevance) and on the paper's own decomposition and weighting choices. No new physical entities are introduced. Several inference thresholds are tuned rather than derived, and the decomposition quality is assumed to be viewpoint-invariant.

free parameters (7)

Kmax = 20
Maximum number of K-means clusters in adaptive descriptor selection (Algorithm 1); fixed at 20 for all experiments; ablated in Table 6 (63.4 to 64.1 mIoU), so stable but still user-set.
HDBSCAN epsilon = 0.1 (subpart), 0.2 (part), 0.3 (whole); 0.1 for single-level
Cluster selection threshold for grouping masks into 3D objects; chosen per level, ablated in Table 7 (60.6 to 64.0), and 0.4 causes unintended merging.
auxiliary cosine threshold = 0.23
Cosine similarity threshold applied at inference to remove unwanted regions (Appendix B); chosen empirically and stated as helpful rather than derived.
foreground relevance threshold = 0.6
Threshold on min-max normalized relevance scores to classify Gaussians as foreground (Appendix B); follows the LangSplat strategy and is not validated on a held-out split.
semantic retrieval threshold = 0.75
Cosine similarity threshold in the semantic retrieval integrity analysis (Section 4 and Appendix B.3); chosen from a precision-recall trade-off, not part of LaGa inference.
affinity feature dimension C' = 32
Dimension of per-Gaussian affinity features in the contrastive decomposition (Section 5.3); fixed hyperparameter with no sensitivity study.
bilateral filter bandwidths = not reported
3D bilateral filtering applied to relevance maps (Appendix B); spatial and range sigmas are not specified, which is a reproducibility gap.

assumptions (5)

domain assumption SAM produces reliable and viewpoint-stable 2D masks that capture object boundaries
Used to build cross-view connections via 3D scene decomposition (Section 5.3, 'Why Is 3D Scene Decomposition Unaffected by View-Dependency?'). If SAM over- or under-segments, the decomposition and thus view aggregation are corrupted.
domain assumption CLIP image features of masked crops provide meaningful per-view semantics for object recognition
The method relies on CLIP features extracted from SAM crops (Appendix B.1) as the semantic signal; the paper itself lists CLIP's bag-of-words and context-loss limitations in Section 6.5.
domain assumption Contrastive training on rendered affinity features yields Gaussian-level features whose cosine similarity reflects object membership
Section 5.3 and Eq (4) and (13) to (16) assume that the learned affinity features cluster by object so that HDBSCAN on mask prototypes yields correct 3D objects.
domain assumption The relevance formulation of Kerr et al. (2023) with canonical phrases ('object', 'thing', 'texture', 'stuff') is a valid text-to-feature scoring for open-vocabulary queries
Eq (10) is imported from LERF and used unchanged; the paper relies on this scoring for all results.
ad hoc to paper The weighting rule in Eq (11), cosine to global mean times L2 norm, is a valid reliability measure for descriptors
The internal compactness heuristic (norm of centroid) is introduced without theoretical justification; the empirical ablations in Table 4 provide the only support.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Tackling View-Dependent Semantics in 3D Language Gaussian Splatting." pith.science (2026). https://pith.science/paper/IUFCKEJI

@misc{pith2026250524746,
  author       = {Pith},
  title        = {Pith review of: Tackling View-Dependent Semantics in 3D Language Gaussian Splatting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IUFCKEJI}},
  note         = {Machine review of arXiv:2505.24746}
}

read the original abstract

Recent advancements in 3D Gaussian Splatting (3D-GS) enable high-quality 3D scene reconstruction from RGB images. Many studies extend this paradigm for language-driven open-vocabulary scene understanding. However, most of them simply project 2D semantic features onto 3D Gaussians and overlook a fundamental gap between 2D and 3D understanding: a 3D object may exhibit various semantics from different viewpoints--a phenomenon we term view-dependent semantics. To address this challenge, we propose LaGa (Language Gaussians), which establishes cross-view semantic connections by decomposing the 3D scene into objects. Then, it constructs view-aggregated semantic representations by clustering semantic descriptors and reweighting them based on multi-view semantics. Extensive experiments demonstrate that LaGa effectively captures key information from view-dependent semantics, enabling a more comprehensive understanding of 3D scenes. Notably, under the same settings, LaGa achieves a significant improvement of +18.7% mIoU over the previous SOTA on the LERF-OVS dataset. Our code is available at: https://github.com/SJTU-DeepVisionLab/LaGa.

Figures

Figures reproduced from arXiv: 2505.24746 by the authors.

**Figure 1.** Pipeline comparison of existing 2D methods (a) and the direct 3D scene understanding paradigm (b). While 2D methods excel in pixel-wise understanding via rendered feature maps, they fail unexpectedly when 3D Gaussians are directly retrieved by matching learned 3D features with CLIP text embeddings. vocabulary scene understanding. To this end, recent studies (Qin et al., 2024; Shi et al., 2024; Bhalgat et al., 2024;… view at source ↗

**Figure 2.** An illustration of view-dependent 3D semantics. The passport exhibits different semantics from different viewpoints. as 3D object editing, AI-driven interaction, and precise 3D localization. We identify a fundamental issue behind this limitation: the view-dependency of 3D semantics. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Distribution of intra- and inter-object cosine similarities. that the multi-view semantics of a single 3D object often exhibit lower cosine similarities than those of semantics from different 3D objects. This statistical finding provides strong evidence of the existence of view-dependent semantics [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (12 more)

**Figure 4.** Figure 4: Distribution of recall rates for 2D semantic features. Semantic Retrieval Integrity. We project multi-view 2D semantics onto 3D Gaussians and use them to retrieve 3D Gaussians from the scene (implementation details in Appendix B.3). Since each 2D mask corresponds to a…

**Figure 5.** Figure 5: Overall pipeline of LaGa. LaGa first establishes cross-view semantic connections through contrastive 3D scene decomposition and then constructs view-aggregated semantic representations by adaptively clustering semantic descriptors and reweighting them. features serve a…

**Figure 6.** Figure 6: Qualitative comparison on the LERF-OVS dataset. LaGa delivers more precise 3D segmentations within 3D-GS. Although corresponding views are provided with colored bounding boxes for clarity, neither viewpoint information nor visual prompts are used. both LEGaussians and …

**Figure 7.** Figure 7 [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

**Figure 8.** Figure 8: Failure cases of LaGa, with prompts shown below and predicted regions highlighted. • Bag-of-Words Effect in CLIP. Prompts such as “pirate hat on the rubber duck” or “cookies in the plate” require models to resolve compositional semantics. However, CLIP tends to activa…

**Figure 9.** Figure 9: presents the results, which lead to conclusions similar to those in Section 6.3. One phenomenon worth clarifying is the needle-like borders around segmented objects. This effect arises from an inherent flaw of the 3D-GS representation, where 3D Gaussians are trained to…

**Figure 10.** Figure 10: Multi-view segmentation results on the LERF-OVS dataset. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]

**Figure 11.** Figure 11: Multi-view semantic segmentation results on the 3D-OVS dataset. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]

**Figure 12.** Figure 12: Multi-granularity semantic segmentation results on the LERF-OVS dataset [PITH_FULL_IMAGE:figures/full_fig_p020_12.png]

**Figure 13.** Figure 13: Scene editing examples with the help of LaGa. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_13.png]

**Figure 14.** Figure 14: Visualization of mask groups obtained from the 3D scene decomposition phase. Highlighted regions indicate the corresponding masks. Groups of masks are visually separated by dotted lines. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_14.png]

**Figure 15.** Figure 15: View-dependent semantic variations within the same mask group. Each row corresponds to a different text prompt, shown in the top-left corner. The relevance scores for individual 2D semantics are displayed above each figure. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_15.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Vision-Language Memory for Spatial Reasoning
cs.CV 2025-11 conditional novelty 5.0 of 10

A video-based vision-language model with 3D-aligned visual features and bounded dual memory achieves state-of-the-art scores on four spatial reasoning benchmarks.

Reference graph

Works this paper leans on

65 extracted references · 56 canonical work pages · cited by 1 Pith paper

[1]

T., Mildenhall, B., Tancik, M., Hedman, P., Martin-Brualla, R., and Srinivasan, P

Barron, J. T., Mildenhall, B., Tancik, M., Hedman, P., Martin-Brualla, R., and Srinivasan, P. P. Mip-nerf: A multiscale representation for anti-aliasing neural radiance fields. In ICCV, 2021

work page 2021
[2]

T., Mildenhall, B., Verbin, D., Srinivasan, P

Barron, J. T., Mildenhall, B., Verbin, D., Srinivasan, P. P., and Hedman, P. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. In CVPR, 2022

work page 2022
[3]

T., Mildenhall, B., Verbin, D., Srinivasan, P

Barron, J. T., Mildenhall, B., Verbin, D., Srinivasan, P. P., and Hedman, P. Zip-nerf: Anti-aliased grid-based neural radiance fields. In ICCV, 2023

work page 2023
[4]

Bhalgat, Y., Laina, I., Henriques, J. a. F., Vedaldi, A., and Zisserman, A. Contrastive lift: 3d object instance segmentation by slow-fast contrastive fusion. In NeurIPS, 2023

work page 2023
[5]

F., Zisserman, A., and Vedaldi, A

Bhalgat, Y., Laina, I., Henriques, J. F., Zisserman, A., and Vedaldi, A. N2f2: Hierarchical scene understanding with nested neural feature fields. In ECCV, 2024

work page 2024
[6]

Dm-nerf: 3d scene geometry decomposition and manipulation from 2d images

Bing, W., Chen, L., and Yang, B. Dm-nerf: 3d scene geometry decomposition and manipulation from 2d images. In ICLR, 2023

work page 2023
[7]

Emerging properties in self-supervised vision transformers

Caron, M., Touvron, H., Misra, I., J \'e gou, H., Mairal, J., Bojanowski, P., and Joulin, A. Emerging properties in self-supervised vision transformers. In ICCV, 2021

2021
[8]

Segment anything in 3d with nerfs

Cen, J., Zhou, Z., Fang, J., Yang, C., Shen, W., Xie, L., Jiang, D., Zhang, X., and Tian, Q. Segment anything in 3d with nerfs. In NeurIPS, 2023

work page 2023

Show all 65 references

[9]

Segment any 3d gaussians

Cen, J., Fang, J., Yang, C., Xie, L., Zhang, X., Shen, W., and Tian, Q. Segment any 3d gaussians. In AAAI, 2025 a

2025
[10]

Segment anything in 3d with radiance fields

Cen, J., Zhou, Z., Fang, J., Yang, C., Shen, W., Xie, L., Jiang, D., Zhang, X., and Tian, Q. Segment anything in 3d with radiance fields. IJCV, 2025 b

2025
[11]

Tensorf: Tensorial radiance fields

Chen, A., Xu, Z., Geiger, A., Yu, J., and Su, H. Tensorf: Tensorial radiance fields. In ECCV, 2022

2022
[12]

Gaussianeditor: Swift and controllable 3d editing with gaussian splatting

Chen, Y., Chen, Z., Zhang, C., Wang, F., Yang, X., Wang, Y., Cai, Z., Yang, L., Liu, H., and Lin, G. Gaussianeditor: Swift and controllable 3d editing with gaussian splatting. In CVPR, 2024

2024
[13]

V., and Paudel, D

Cheng, J., Zaech, J.-N., Gool, L. V., and Paudel, D. P. Occam's lgs: A simple approach for language gaussian splatting. arXiv preprint arXiv:2412.01807, 2024

2024 arXiv
[14]

X., Savva, M., Halber, M., Funkhouser, T., and Nie ner, M

Dai, A., Chang, A. X., Savva, M., Halber, M., Funkhouser, T., and Nie ner, M. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In CVPR, 2017

2017
[15]

4d-rotor gaussian splatting: Towards efficient novel view synthesis for dynamic scenes

Duan, Y., Wei, F., Dai, Q., He, Y., Chen, W., and Chen, B. 4d-rotor gaussian splatting: Towards efficient novel view synthesis for dynamic scenes. In SIGGRAPH, 2024

2024
[16]

Nerf-sos: Any-view self-supervised object segmentation on complex scenes

Fan, Z., Wang, P., Jiang, Y., Gong, X., Xu, D., and Wang, Z. Nerf-sos: Any-view self-supervised object segmentation on complex scenes. In ICLR, 2023

2023
[17]

Plenoxels: Radiance fields without neural networks

Fridovich - Keil, S., Yu, A., Tancik, M., Chen, Q., Recht, B., and Kanazawa, A. Plenoxels: Radiance fields without neural networks. In CVPR, 2022

2022
[18]

Panoptic nerf: 3d-to-2d label transfer for panoptic urban scene segmentation

Fu, X., Zhang, S., Chen, T., Lu, Y., Zhu, L., Zhou, X., Geiger, A., and Liao, Y. Panoptic nerf: 3d-to-2d label transfer for panoptic urban scene segmentation. In 3DV, 2022

2022
[19]

Interactive segmentation of radiance fields

Goel, R., Sirikonda, D., Saini, S., and Narayanan, P. Interactive segmentation of radiance fields. In CVPR, 2023

2023
[20]

P., Mildenhall, B., Reiser, C., Barron, J

Hedman, P., Srinivasan, P. P., Mildenhall, B., Reiser, C., Barron, J. T., and Debevec, P. Baking neural radiance fields for real-time view synthesis. IEEE TPAMI, 2024

2024
[21]

Nerf-rpn: A general framework for object detection in nerfs

Hu, B., Huang, J., Liu, Y., Tai, Y.-W., and Tang, C.-K. Nerf-rpn: A general framework for object detection in nerfs. In CVPR, 2023

2023
[22]

Segment anything in 3d gaussians

Hu, X., Wang, Y., Fan, L., Fan, J., Peng, J., Lei, Z., Li, Q., and Zhang, Z. Segment anything in 3d gaussians. arXiv preprint arXiv:2401.17857, 2024

2024 arXiv
[23]

3d gaussian splatting for real-time radiance field rendering

Kerbl, B., Kopanas, G., Leimk \"u hler, T., and Drettakis, G. 3d gaussian splatting for real-time radiance field rendering. ACM TOG, 2023

2023
[24]

M., Goldberg, K., Kanazawa, A., and Tancik, M

Kerr, J., Kim, C. M., Goldberg, K., Kanazawa, A., and Tancik, M. Lerf: Language embedded radiance fields. In ICCV, 2023

2023
[25]

M., Wu, M., Kerr, J., Tancik, M., Goldberg, K., and Kanazawa, A

Kim, C. M., Wu, M., Kerr, J., Tancik, M., Goldberg, K., and Kanazawa, A. Garfield: Group anything with radiance fields. In CVPR, 2024

2024
[26]

C., Lo, W.-Y., et al

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 ICCV, 2023

2023
[27]

Decomposing nerf for editing via feature field distillation

Kobayashi, S., Matsumoto, E., and Sitzmann, V. Decomposing nerf for editing via feature field distillation. In NeurIPS, 2022

2022
[28]

Rethinking open-vocabulary segmentation of radiance fields in 3d space

Lee, H., Yun, Y., Bae, J., Kim, S., and Uh, Y. Rethinking open-vocabulary segmentation of radiance fields in 3d space. In AAAI, 2025

2025
[29]

Supergseg: Open-vocabulary 3d segmentation with structured super-gaussians

Liang, S., Wang, S., Li, K., Niemeyer, M., Gasperini, S., Navab, N., and Tombari, F. Supergseg: Open-vocabulary 3d segmentation with structured super-gaussians. arXiv preprint arXiv:2412.10231, 2024

2024
[30]

B., Martel, J

Lindell, D. B., Martel, J. N. P., and Wetzstein, G. Autoint: Automatic integration for fast neural volume rendering. In CVPR, 2021

2021
[31]

E., Theobalt, C., Xing, E., and Lu, S

Liu, K., Zhan, F., Zhang, J., XU, M., Yu, Y., Saddik, A. E., Theobalt, C., Xing, E., and Lu, S. Weakly supervised 3d open-vocabulary segmentation. In NeurIPS, 2023 a

2023
[32]

Instance neural radiance field

Liu, Y., Hu, B., Huang, J., Tai, Y.-W., and Tang, C.-K. Instance neural radiance field. In ICCV, 2023 b

2023
[33]

Gaga: Group any gaussians via 3d-aware memory bank

Lyu, W., Li, X., Kundu, A., Tsai, Y.-H., and Yang, M.-H. Gaga: Group any gaussians via 3d-aware memory bank. arXiv preprint arXiv:2404.07977, 2024

2024 arXiv
[34]

S., Barron, J

Martin-Brualla, R., Radwan, N., Sajjadi, M. S., Barron, J. T., Dosovitskiy, A., and Duckworth, D. Nerf in the wild: Neural radiance fields for unconstrained photo collections. In CVPR, 2021

2021
[35]

hdbscan: Hierarchical density based clustering

McInnes, L., Healy, J., and Astels, S. hdbscan: Hierarchical density based clustering. Journal of Open Source Software, 2017

2017
[36]

Splatfields: Neural gaussian splats for sparse 3d and 4d reconstruction

Mihajlovic, M., Prokudin, S., Tang, S., Maier, R., Bogo, F., Tung, T., and Boyer, E. Splatfields: Neural gaussian splats for sparse 3d and 4d reconstruction. In ECCV, 2025

2025
[37]

P., Tancik, M., Barron, J

Mildenhall, B., Srinivasan, P. P., Tancik, M., Barron, J. T., Ramamoorthi, R., and Ng, R. Nerf: Representing scenes as neural radiance fields for view synthesis. In ECCV, 2020

2020
[38]

Instant neural graphics primitives with a multiresolution hash encoding

M \" u ller, T., Evans, A., Schied, C., and Keller, A. Instant neural graphics primitives with a multiresolution hash encoding. ACM TOG , 2022

2022
[39]

H., Chaitanya, C

Neff, T., Stadlbauer, P., Parger, M., Kurz, A., Mueller, J. H., Chaitanya, C. R. A., Kaplanyan, A., and Steinberger, M. Donerf: Towards real-time rendering of compact neural radiance fields using depth oracle networks. In Computer Graphics Forum, 2021

2021
[40]

3d vision-language gaussian splatting

Peng, Q., Planche, B., Gao, Z., Zheng, M., Choudhuri, A., Chen, T., Chen, C., and Wu, Z. 3d vision-language gaussian splatting. ICLR, 2025

2025
[41]

Langsplat: 3d language gaussian splatting

Qin, M., Li, W., Zhou, J., Wang, H., and Pfister, H. Langsplat: 3d language gaussian splatting. In CVPR, 2024

2024
[42]

Goi: Find 3d gaussians of interest with an optimizable open-vocabulary semantic-space hyperplane

Qu, Y., Dai, S., Li, X., Lin, J., Cao, L., Zhang, S., and Ji, R. Goi: Find 3d gaussians of interest with an optimizable open-vocabulary semantic-space hyperplane. In ACM MM, 2024

2024
[43]

W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., and Sutskever, I

Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., and Sutskever, I. Learning transferable visual models from natural language supervision. In ICML, 2021

2021
[44]

Kilonerf: Speeding up neural radiance fields with thousands of tiny mlps

Reiser, C., Peng, S., Liao, Y., and Geiger, A. Kilonerf: Speeding up neural radiance fields with thousands of tiny mlps. In ICCV, 2021

2021
[45]

C., Schwing, A

Ren, Z., Agarwala, A., Russell, B. C., Schwing, A. G., and Wang, O. Neural volumetric object selection. In CVPR, 2022

2022
[46]

Language embedded 3d gaussians for open-vocabulary scene understanding

Shi, J.-C., Wang, M., Duan, H.-B., and Guan, S.-H. Language embedded 3d gaussians for open-vocabulary scene understanding. In CVPR, 2024

2024
[47]

R., M \"u ller, N., Nie ner, M., Dai, A., and Kontschieder, P

Siddiqui, Y., Porzi, L., Bul \'o , S. R., M \"u ller, N., Nie ner, M., Dai, A., and Kontschieder, P. Panoptic lifting for 3d scene understanding with neural fields. In CVPR, 2023

2023
[48]

Stelzner, K., Kersting, K., and Kosiorek, A. R. Decomposing 3d scenes into objects via unsupervised volume segmentation. arXiv preprint arXiv:2104.01148, 2021

2021 arXiv
[49]

Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction

Sun, C., Sun, M., and Chen, H. Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction. In CVPR, 2022

2022
[50]

Lgm: Large multi-view gaussian model for high-resolution 3d content creation

Tang, J., Chen, Z., Chen, X., Wang, T., Zeng, G., and Liu, Z. Lgm: Large multi-view gaussian model for high-resolution 3d content creation. In European Conference on Computer Vision, pp.\ 1--18. Springer, 2025

2025
[51]

Neural feature fusion fields: 3d distillation of self-supervised 2d image representations

Tschernezki, V., Laina, I., Larlus, D., and Vedaldi, A. Neural feature fusion fields: 3d distillation of self-supervised 2d image representations. In 3DV, 2022

2022
[52]

S., Pot, E., Tagliasacchi, A., and Duckworth, D

Vora, S., Radwan, N., Greff, K., Meyer, H., Genova, K., Sajjadi, M. S., Pot, E., Tagliasacchi, A., and Duckworth, D. Nesf: Neural semantic fields for generalizable semantic segmentation of 3d scenes. TMLR, 2022

2022
[53]

Gaussianeditor: Editing 3d gaussians delicately with text instructions

Wang, J., Fang, J., Zhang, X., Xie, L., and Tian, Q. Gaussianeditor: Editing 3d gaussians delicately with text instructions. In CVPR, 2024

2024
[54]

Nex: Real-time view synthesis with neural basis expansion

Wizadwongsa, S., Phongthawee, P., Yenphraphai, J., and Suwajanakorn, S. Nex: Real-time view synthesis with neural basis expansion. In CVPR, 2021

2021
[55]

4d gaussian splatting for real-time dynamic scene rendering

Wu, G., Yi, T., Fang, J., Xie, L., Zhang, X., Wei, W., Liu, W., Tian, Q., and Wang, X. 4d gaussian splatting for real-time dynamic scene rendering. In CVPR, 2024 a

2024
[56]

Opengaussian: Towards point-level 3d gaussian-based open vocabulary understanding

Wu, Y., Meng, J., Li, H., Wu, C., Shi, Y., Cheng, X., Zhao, C., Feng, H., Ding, E., Wang, J., and Zhang, J. Opengaussian: Towards point-level 3d gaussian-based open vocabulary understanding. In NeurIPS, 2024 b

2024
[57]

Gaussianobject: High-quality 3d object reconstruction from four views with gaussian splatting

Yang, C., Li, S., Fang, J., Liang, R., Xie, L., Zhang, X., Shen, W., and Tian, Q. Gaussianobject: High-quality 3d object reconstruction from four views with gaussian splatting. ACM Transactions on Graphics (TOG), 2024 a

2024
[58]

Deformable 3d gaussians for high-fidelity monocular dynamic scene reconstruction

Yang, Z., Gao, X., Zhou, W., Jiao, S., Zhang, Y., and Jin, X. Deformable 3d gaussians for high-fidelity monocular dynamic scene reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 20331--20341, 2024 b

2024
[59]

Gaussiandreamer: Fast generation from text to 3d gaussians by bridging 2d and 3d diffusion models

Yi, T., Fang, J., Wang, J., Wu, G., Xie, L., Zhang, X., Liu, W., Tian, Q., and Wang, X. Gaussiandreamer: Fast generation from text to 3d gaussians by bridging 2d and 3d diffusion models. In CVPR, 2024

2024
[60]

Omniseg3d: Omniversal 3d segmentation via hierarchical contrastive learning

Ying, H., Yin, Y., Zhang, J., Wang, F., Yu, T., Huang, R., and Fang, L. Omniseg3d: Omniversal 3d segmentation via hierarchical contrastive learning. In CVPR, 2024

2024
[61]

Nerf++: Analyzing and improving neural radiance fields

Zhang, K., Riegler, G., Snavely, N., and Koltun, V. Nerf++: Analyzing and improving neural radiance fields. arXiv preprint arXiv:2010.07492, 2020

2010 arXiv
[62]

Zhi, S., Laidlow, T., Leutenegger, S., and Davison, A. J. In-place scene labelling and understanding with implicit scene representation. In ICCV, 2021

2021
[63]

Feature 3dgs: Supercharging 3d gaussian splatting to enable distilled feature fields

Zhou, S., Chang, H., Jiang, S., Fan, Z., Zhu, Z., Xu, D., Chari, P., You, S., Wang, Z., and Kadambi, A. Feature 3dgs: Supercharging 3d gaussian splatting to enable distilled feature fields. In CVPR, 2024

2024
[64]

Pcf-lift: Panoptic lifting by probabilistic contrastive fusion

Zhu, R., Qiu, S., Wu, Q., Hui, K., Heng, P., and Fu, C. Pcf-lift: Panoptic lifting by probabilistic contrastive fusion. In ECCV, 2024

2024
[65]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

T., Mildenhall, B., Tancik, M., Hedman, P., Martin-Brualla, R., and Srinivasan, P

Barron, J. T., Mildenhall, B., Tancik, M., Hedman, P., Martin-Brualla, R., and Srinivasan, P. P. Mip-nerf: A multiscale representation for anti-aliasing neural radiance fields. In ICCV, 2021

work page 2021

[2] [2]

T., Mildenhall, B., Verbin, D., Srinivasan, P

Barron, J. T., Mildenhall, B., Verbin, D., Srinivasan, P. P., and Hedman, P. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. In CVPR, 2022

work page 2022

[3] [3]

T., Mildenhall, B., Verbin, D., Srinivasan, P

Barron, J. T., Mildenhall, B., Verbin, D., Srinivasan, P. P., and Hedman, P. Zip-nerf: Anti-aliased grid-based neural radiance fields. In ICCV, 2023

work page 2023

[4] [4]

Bhalgat, Y., Laina, I., Henriques, J. a. F., Vedaldi, A., and Zisserman, A. Contrastive lift: 3d object instance segmentation by slow-fast contrastive fusion. In NeurIPS, 2023

work page 2023

[5] [5]

F., Zisserman, A., and Vedaldi, A

Bhalgat, Y., Laina, I., Henriques, J. F., Zisserman, A., and Vedaldi, A. N2f2: Hierarchical scene understanding with nested neural feature fields. In ECCV, 2024

work page 2024

[6] [6]

Dm-nerf: 3d scene geometry decomposition and manipulation from 2d images

Bing, W., Chen, L., and Yang, B. Dm-nerf: 3d scene geometry decomposition and manipulation from 2d images. In ICLR, 2023

work page 2023

[7] [7]

Emerging properties in self-supervised vision transformers

Caron, M., Touvron, H., Misra, I., J \'e gou, H., Mairal, J., Bojanowski, P., and Joulin, A. Emerging properties in self-supervised vision transformers. In ICCV, 2021

2021

[8] [8]

Segment anything in 3d with nerfs

Cen, J., Zhou, Z., Fang, J., Yang, C., Shen, W., Xie, L., Jiang, D., Zhang, X., and Tian, Q. Segment anything in 3d with nerfs. In NeurIPS, 2023

work page 2023