Pith. sign in

REVIEW 3 major objections 52 references

Reviewed by Pith at T0; open to challenge.

T0 means a machine referee read the full paper against a public rubric. The mark states how deep the mechanical check went, never who wrote it. the ladder, T0–T4 →

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

T0 review · grok-4.3

MLP-Splatting decomposes scenes into a handful of compact neural primitives that align with objects using only RGB supervision.

2026-06-28 11:09 UTC pith:3TL7FGSK

load-bearing objection MLP-Splatting swaps Gaussians for small independent MLPs with local support and claims RGB-only training produces object-aligned primitives, but that central claim rests on an unshown optimization preference. the 3 major comments →

arxiv 2606.03877 v1 pith:3TL7FGSK submitted 2026-06-02 cs.CV

MLP Splatting: Object-Centric Neural Fields

Shinjeong Kim , Yuzhou Cheng , Xin Kong , Paul H. J. Kelly , Andrew J. Davison This is my paper
classification cs.CV
keywords MLP SplattingNeural FieldsScene DecompositionObject-Centric RepresentationNovel View SynthesisVolumetric Rendering3D Scene EditingLight-Field Primitives
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper proposes MLP-Splatting to represent 3D scenes with a small set of expressive light-field primitives instead of many low-level elements or one global field. Each primitive is an independent compact MLP that predicts radiance and opacity inside a localized spatial region. These primitives are trained end-to-end on RGB images alone and, through sparse volumetric compositing, produce decompositions where individual primitives often match objects or object parts. This decomposition supports direct object-level editing by selecting a few primitives and yields lower memory and faster rendering than comparable semantic Gaussian methods. Optional semantic feature distillation further enables open-vocabulary interaction and instant segmentation.

Core claim

MLP-Splatting models each primitive as an independent compact MLP with localized spatial support that predicts radiance and opacity. Rendering is performed through efficient sparse volumetric compositing over ray-primitive interactions. Supervised using RGB supervision alone, the primitives represent local scene regions often corresponding to objects or object parts, enabling interactive object-level editing without segmentation masks by selecting a handful of primitives.

What carries the argument

Independent compact MLP with localized spatial support, rendered via sparse volumetric compositing over ray-primitive interactions.

Load-bearing premise

Independent compact MLPs with localized spatial support trained only on RGB images will naturally produce primitives that correspond to objects or object parts without extra regularization or post-processing.

What would settle it

Training the method on a scene containing clearly separable objects and then observing that most primitives each span multiple objects or that each object is split across dozens of primitives would falsify the emergence claim.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • A scene can be decomposed into a small number of primitives that align with objects using RGB images alone.
  • Object-level editing becomes possible by selecting and manipulating a few primitives without segmentation masks.
  • Optional semantic feature distillation adds open-vocabulary querying and open-set instant segmentation.
  • Memory usage drops to roughly 1/15 of semantic 3D Gaussian Splatting while rendering speed increases by a factor of about 3.
  • Photorealistic novel-view synthesis is retained alongside the decomposition capability.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The localized support of each MLP could reduce blending artifacts across object boundaries compared with global radiance fields.
  • If the primitives remain stable across time, the same representation might support object tracking or dynamic scene editing with minimal changes.
  • Combining the method with sparse depth or instance cues might produce even tighter object boundaries when RGB alone is insufficient.
  • The low primitive count suggests the approach could scale to large environments by adding new primitives on demand rather than retraining a single model.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 0 minor

Summary. The paper introduces MLP-Splatting, a 3D scene representation that decomposes a scene into a small number of independent compact MLPs, each with localized spatial support, that predict radiance and opacity. These primitives are trained end-to-end using only RGB supervision and rendered via sparse volumetric compositing over ray-primitive intersections. The method claims that this architecture produces primitives that naturally align with objects or object parts, enabling interactive object-level editing without masks or additional losses; an optional semantic feature distillation variant further supports open-vocabulary interaction. Experiments are said to show 1/15× lower memory and 3× faster rendering than semantic 3D Gaussian Splatting baselines.

Significance. If the central claims hold after verification, the work would offer a meaningful step toward object-centric neural fields that combine photorealistic rendering with built-in decomposability, reducing reliance on post-hoc segmentation or grouping. The reported memory and speed advantages, if reproducible under controlled conditions, would also be practically relevant for interactive applications.

major comments (3)
  1. Abstract: performance claims of 1/15× memory reduction and 3× rendering speedup are stated without equations, training details, ablation studies, or verification that the gains persist after controlling for implementation differences; central claims rest on unshown experiments.
  2. Abstract and method description: the assertion that RGB-only supervision on independent compact MLPs with localized support yields primitives corresponding to objects or parts (without additional regularization, losses, or post-processing) lacks any derivation, initialization analysis, loss-term examination, or ablation demonstrating why the optimization landscape favors semantic partitions over arbitrary spatial ones.
  3. Abstract: the claim of object correspondence enabling editing by selecting a handful of primitives is presented as an emergent property, yet no quantitative metrics (e.g., correspondence accuracy, stability across seeds, or comparison to random decompositions) are referenced to substantiate it.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below with clarifications drawn from the full paper and indicate planned revisions where appropriate.

read point-by-point responses

  1. Referee: Abstract: performance claims of 1/15× memory reduction and 3× rendering speedup are stated without equations, training details, ablation studies, or verification that the gains persist after controlling for implementation differences; central claims rest on unshown experiments.

    Authors: The performance numbers are substantiated by the controlled comparisons in Section 5 (Tables 2 and 3), which report memory footprints and rendering times against semantic 3DGS baselines using identical hardware and the same training protocol. We agree the abstract would be clearer with an explicit pointer to these results. In revision we will add a parenthetical reference to Section 5 and include a short implementation-control ablation in the supplementary material. revision: partial

  2. Referee: Abstract and method description: the assertion that RGB-only supervision on independent compact MLPs with localized support yields primitives corresponding to objects or parts (without additional regularization, losses, or post-processing) lacks any derivation, initialization analysis, loss-term examination, or ablation demonstrating why the optimization landscape favors semantic partitions over arbitrary spatial ones.

    Authors: The emergence of object-aligned primitives is presented as an empirical outcome supported by the visualizations and editing examples in Sections 4 and 5. The method section and supplementary material already contain the initialization procedure and loss-term breakdown; we will expand the main-text method description with a concise paragraph discussing the role of localized support and independent optimization in favoring coherent partitions. revision: yes

  3. Referee: Abstract: the claim of object correspondence enabling editing by selecting a handful of primitives is presented as an emergent property, yet no quantitative metrics (e.g., correspondence accuracy, stability across seeds, or comparison to random decompositions) are referenced to substantiate it.

    Authors: Editing results are demonstrated qualitatively in Section 5. We concur that quantitative measures would strengthen the claim and will add correspondence accuracy, seed-stability statistics, and a random-decomposition baseline in the revised experiments section. revision: yes

Circularity Check

0 steps flagged

No significant circularity; central claims are empirical modeling outcomes, not reductions by construction.

full rationale

The paper introduces MLP-Splatting as a new architecture of independent compact MLPs with localized support, rendered via sparse volumetric compositing, and asserts that RGB-only photometric supervision produces object-corresponding primitives. This is presented as an observed behavior of the method rather than a derived prediction from equations or fitted parameters. No self-citations are invoked as load-bearing uniqueness theorems, no ansatz is smuggled, and no quantity is renamed or fitted then called a prediction. The assumption that localization and independence will favor semantic partitions is a modeling hypothesis, not a tautological reduction of the result to its inputs. The derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Abstract provides no explicit free parameters, axioms, or external evidence for the new primitives; the central claim rests on the unstated premise that RGB supervision plus localized MLPs suffice for object-level decomposition.

axioms (1)
  • domain assumption RGB supervision alone is sufficient to produce object-corresponding primitives
    Stated directly in the abstract as the supervision regime that yields the editing capability.
invented entities (1)
  • MLP primitive no independent evidence
    purpose: Independent compact MLP with localized spatial support that predicts radiance and opacity for local scene regions
    Core new representation introduced to replace Gaussians or global fields; no independent evidence supplied in abstract.

pith-pipeline@v0.9.1-grok · 5791 in / 1230 out tokens · 37299 ms · 2026-06-28T11:09:27.388796+00:00 · methodology

0 comments
read the original abstract

3D representations are fundamental to scene rendering, understanding, and interaction. Recent approaches, such as 3D Gaussian Splatting and Neural Radiance Fields, achieve impressive photorealistic novel-view synthesis, but lack the ability to easily decompose scene elements into a few primitives, requiring additional segmentation or grouping for object-level manipulation. We present MLP-Splatting, a method that enables scene decomposition via a few expressive light-field primitives while providing photorealistic novel-view synthesis. MLP-Splatting models each primitive as an independent compact MLP with localized spatial support that predicts radiance and opacity. In contrast to low-level Gaussian primitives or a single global radiance field, our neural primitives provide greater expressive capacity while remaining spatially localized. Rendering is performed through efficient sparse volumetric compositing over ray-primitive interactions. Our primitives are supervised using RGB supervision alone, which yields primitives that represent local scene regions often corresponding to objects or object parts, enabling interactive object-level editing without segmentation masks by selecting a handful of primitives. Our method, augmented with optional semantic feature distillation, enables open-vocabulary scene interaction and open-set instant segmentation. Compared to state-of-the-art methods, we achieve substantially lower memory usage (1/15$\times$) and faster rendering (3$\times$), as we show in our experiments compared to semantic 3DGS methods. Project Page: https://shinjeongkim.com/mlp-splatting

Figures

Figures reproduced from arXiv: 2606.03877 by Andrew J. Davison, Paul H. J. Kelly, Shinjeong Kim, Xin Kong, Yuzhou Cheng.

Figure 1
Figure 1. Figure 1: MLP-Splatting Overview. (a) Conceptual comparison of 3D scene represen￾tations along neurality, atomicity, and semantic awareness. Our method combines the expressiveness of neural fields with the atomicity of primitives, enabling representations that better align with semantic structure. (b) Example scene and learned primitives corresponding to objects or object parts. From another perspective, scene repre… view at source ↗
Figure 2
Figure 2. Figure 2: Procedure of MLP-Splatting pipeline. (Left) For each tile, sorting for MLPs is performed per tile, based on the depth calculated with respect to a virtual ray passing through the center of the tile; only MLPs for which the support region is splatted to the image plane that overlap with the tile are considered. (Center) Based on the sorted order, we perform the MLP weight loading and forward pass cooperativ… view at source ↗
Figure 3
Figure 3. Figure 3: Interactive scene editing demonstrations. In row (a), we show object scaling demonstrated on the bonsai tree. In row (b), we show the candle’s object deletion. In row (c), color editing is shown on the bonsai tree. In row (d), we show a rigid body transformation applied to the rolling pin primitives and how it renders correctly from the new location. Baselines. We compare against (i) NeRF-DFF [17], which d… view at source ↗
Figure 4
Figure 4. Figure 4: 3D semantic segmentation with LSeg-guided embeddings rendered on novel views. The top two rows are from the Replica dataset [37], and the bottom two rows are from the ScanNet [6] dataset. As shown in Tab. 3, evaluated under the protocol proposed in [17, 46], our method outperforms [17] across all semantic segmentation metrics, despite each primitive modeling a light-field-like radiance field, and performs … view at source ↗
Figure 5
Figure 5. Figure 5: Novel view semantic segmentation with SAM embeddings [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

52 extracted references · 2 canonical work pages · 1 internal anchor

  1. [1]

    Barron, J.T., Mildenhall, B., Tancik, M., Hedman, P., Martin-Brualla, R., Srini- vasan, P.P.: Mip-nerf: A multiscale representation for anti-aliasing neural radiance fields. In: Int. Conf. Comput. Vis. pp. 5855–5864 (October 2021)

    2021

  2. [2]

    In: IEEE Conf

    Barron, J.T., Mildenhall, B., Verbin, D., Srinivasan, P.P., Hedman, P.: Mip-nerf 360: Unbounded anti-aliased neural radiance fields. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 5470–5479 (2022)

    2022

  3. [3]

    Barron, J.T., Mildenhall, B., Verbin, D., Srinivasan, P.P., Hedman, P.: Zip-nerf: Anti-aliased grid-based neural radiance fields. In: Int. Conf. Comput. Vis. pp. 19697–19705 (2023)

    2023

  4. [4]

    Cen, J., Zhou, X., Fang, J., Wen, C., Xie, L., Zhang, X., Shen, W., Tian, Q.: Tackling view-dependentsemantics in3d language gaussian splatting. In: Int. Conf. Mach. Learn. (2025)

    2025

  5. [5]

    In: 2025 IEEE International Conference on Robotics and Automation (ICRA)

    Cheng, Y., Jiao, J., Wang, Y., Kanoulas, D.: Logs: Visual localization via gaussian splatting with fewer training images. In: 2025 IEEE International Conference on Robotics and Automation (ICRA). pp. 15029–15036. IEEE (2025)

    2025

  6. [6]

    Dai, A., Chang, A.X., Savva, M., Halber, M., Funkhouser, T., Nießner, M.: Scan- Net:Richly-annotated3dreconstructionsofindoorscenes.In:IEEEConf.Comput. Vis. Pattern Recog. (July 2017)

    2017

  7. [7]

    Transactions of the American Mathematical So- ciety93(3), 418–491 (1959)

    Federer, H.: Curvature measures. Transactions of the American Mathematical So- ciety93(3), 418–491 (1959)

    1959

  8. [8]

    In: IEEE Conf

    Fridovich-Keil, S., Yu, A., Tancik, M., Chen, Q., Recht, B., Kanazawa, A.: Plenox- els: Radiance fields without neural networks. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 5501–5510 (June 2022)

    2022

  9. [9]

    Govindarajan, S., Rebain, D., Yi, K.M., Tagliasacchi, A.: Radiant foam: Real-time differentiable ray tracing. In: Int. Conf. Comput. Vis. pp. 4135–4145 (2025)

    2025

  10. [10]

    Birkhäuser, 2 edn

    Gray, A.: Tubes. Birkhäuser, 2 edn. (2004)

    2004

  11. [11]

    Object-Centric Neural Scene Rendering

    Guo, M., Fathi, A., Wu, J., Funkhouser, T.: Object-centric neural scene rendering. arXiv preprint arXiv:2012.08503 (2020)

    work page arXiv 2012

  12. [12]

    In: ACM SIGGRAPH

    Huang, B., Yu, Z., Chen, A., Geiger, A., Gao, S.: 2d gaussian splatting for geo- metrically accurate radiance fields. In: ACM SIGGRAPH. pp. 1–11 (2024)

    2024

  13. [13]

    ACM Trans

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G., et al.: 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph.42(4), 139–1 (2023)

    2023

  14. [14]

    Kerr, J., Kim, C.M., Goldberg, K., Kanazawa, A., Tancik, M.: Lerf: Language embedded radiance fields. In: Int. Conf. Comput. Vis. pp. 19729–19739 (2023)

    2023

  15. [15]

    In: IEEE Conf

    Kim, C.M., Wu, M., Kerr, J., Goldberg, K., Tancik, M., Kanazawa, A.: Garfield: Group anything with radiance fields. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 21530–21539 (2024)

    2024

  16. [16]

    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: Int. Conf. Comput. Vis. pp. 4015–4026 (2023)

    2023

  17. [17]

    Kobayashi, S., Matsumoto, E., Sitzmann, V.: Decomposing nerf for editing via featurefielddistillation.Adv.NeuralInform.Process.Syst.35,23311–23330(2022)

    2022

  18. [18]

    In: IEEE Conf

    Kong, X., Liu, S., Taher, M., Davison, A.J.: vmap: Vectorised object mapping for neural field slam. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 952–961 (2023)

    2023

  19. [19]

    Birkhäuser, 2 edn

    Krantz, S.G., Parks, H.R.: A Primer of Real Analytic Functions. Birkhäuser, 2 edn. (2002) MLP-Splatting 17

    2002

  20. [20]

    In: IEEE Conf

    Kundu, A., Genova, K., Yin, X., Fathi, A., Pantofaru, C., Guibas, L.J., Tagliasac- chi, A., Dellaert, F., Funkhouser, T.: Panoptic neural fields: A semantic object- aware neural scene representation. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 12871–12881 (2022)

    2022

  21. [21]

    Liu, S., Zeng, Z., Ren, T., Li, F., Zhang, H., Yang, J., Jiang, Q., Li, C., Yang, J., Su, H., et al.: Grounding dino: Marrying dino with grounded pre-training for open-set object detection. In: Eur. Conf. Comput. Vis. pp. 38–55. Springer (2024)

    2024

  22. [22]

    In: Seminal graphics: pioneering efforts that shaped the field, pp

    Lorensen, W.E., Cline, H.E.: Marching cubes: A high resolution 3d surface con- struction algorithm. In: Seminal graphics: pioneering efforts that shaped the field, pp. 347–353. ACM New York, NY, USA (1998)

    1998

  23. [23]

    In: IEEE Conf

    Lüddecke, T., Ecker, A.: Image segmentation using text and image prompts. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 7086–7096 (2022)

    2022

  24. [24]

    von Lützow, N., Nießner, M.: Linprim: Linear primitives for differentiable volu- metric rendering. In: Adv. Neural Inform. Process. Syst. (2025)

    2025

  25. [25]

    Mai, A., Hedman, P., Kopanas, G., Verbin, D., Futschik, D., Xu, Q., Kuester, F., Barron, J.T., Zhang, Y.: Ever: Exact volumetric ellipsoid rendering for real-time view synthesis. In: Int. Conf. Comput. Vis. pp. 4930–4939 (2025)

    2025

  26. [26]

    In: IEEE Conf

    Martin-Brualla, R., Radwan, N., Sajjadi, M.S.M., Barron, J.T., Dosovitskiy, A., Duckworth, D.: Nerf in the wild: Neural radiance fields for unconstrained photo collections. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 7210–7219 (June 2021)

    2021

  27. [27]

    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: Eur. Conf. Comput. Vis. (2020)

    2020

  28. [28]

    ACM Trans

    Müller, T., Evans, A., Schied, C., Keller, A.: Instant neural graphics primitives with a multiresolution hash encoding. ACM Trans. Graph.41(4), 1–15 (2022)

    2022

  29. [29]

    In: IEEE Conf

    Niemeyer, M., Geiger, A.: Giraffe: Representing scenes as compositional generative neural feature fields. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 11453–11464 (June 2021)

    2021

  30. [30]

    In: IEEE Conf

    Peng, S., Genova, K., Jiang, C., Tagliasacchi, A., Pollefeys, M., Funkhouser, T., et al.: Openscene: 3d scene understanding with open vocabularies. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 815–824 (2023)

    2023

  31. [31]

    Neural Networks108, 296–330 (2018)

    Petersen, P., Voigtlaender, F.: Optimal approximation of piecewise smooth func- tions using deep ReLU neural networks. Neural Networks108, 296–330 (2018)

    2018

  32. [32]

    In: IEEE Conf

    Qin, M., Li, W., Zhou, J., Wang, H., Pfister, H.: Langsplat: 3d language gaussian splatting. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 20051–20060 (2024)

    2024

  33. [33]

    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: Int. Conf. Mach. Learn. pp. 8748–8763. PmLR (2021)

    2021

  34. [34]

    Reiser, C., Peng, S., Liao, Y., Geiger, A.: Kilonerf: Speeding up neural radiance fields with thousands of tiny mlps. In: Int. Conf. Comput. Vis. pp. 14335–14345 (October 2021)

    2021

  35. [35]

    In: IEEE Conf

    Schönberger, J.L., Frahm, J.M.: Structure-from-motion revisited. In: IEEE Conf. Comput. Vis. Pattern Recog. (2016)

    2016

  36. [36]

    Foundations of Computational Mathematics 24(2), 481–537 (2024)

    Siegel, J.W., Xu, J.: Sharp bounds on the approximation rates, metric entropy, and n-widths of shallow neural networks. Foundations of Computational Mathematics 24(2), 481–537 (2024)

    2024

  37. [37]

    The Replica Dataset: A Digital Replica of Indoor Spaces

    Straub, J., Whelan, T., Ma, L., Chen, Y., Wijmans, E., Green, S., Engel, J.J., Mur-Artal, R., Ren, C., Verma, S., et al.: The replica dataset: A digital replica of indoor spaces. arXiv preprint arXiv:1906.05797 (2019) 18 S. Kim et al

    work page internal anchor Pith review Pith/arXiv arXiv 1906

  38. [38]

    In: IEEE Int

    Taher, M., Alzugaray, I., Davison, A.J.: Fit-ngp: Fitting object models to neural graphics primitives. In: IEEE Int. Conf. Robot. Autom. pp. 18186–18192. IEEE (2024)

    2024

  39. [39]

    Wu, Q., Wang, K., Li, K., Zheng, J., Cai, J.: Objectsdf++: Improved object- compositional neural implicit surfaces. In: Int. Conf. Comput. Vis. pp. 21764–21774 (October 2023)

    2023

  40. [40]

    Yang, B., Zhang, Y., Xu, Y., Li, Y., Zhou, H., Bao, H., Zhang, G., Cui, Z.: Learning object-compositional neural radiance field for editable scene rendering. In: Int. Conf. Comput. Vis. pp. 13779–13788 (October 2021)

    2021

  41. [41]

    Neural networks94, 103–114 (2017)

    Yarotsky, D.: Error bounds for approximations with deep relu networks. Neural networks94, 103–114 (2017)

    2017

  42. [42]

    Ye, M., Danelljan, M., Yu, F., Ke, L.: Gaussian grouping: Segment and edit any- thing in 3d scenes. In: Eur. Conf. Comput. Vis. pp. 162–179. Springer (2024)

    2024

  43. [43]

    In: 2018 IEEE/ACM 26th international symposium on quality of service (IWQoS)

    Zhang, Z.: Improved adam optimizer for deep neural networks. In: 2018 IEEE/ACM 26th international symposium on quality of service (IWQoS). pp. 1–2. Ieee (2018)

    2018

  44. [44]

    Zhi, S., Laidlow, T., Leutenegger, S., Davison, A.J.: In-place scene labelling and understanding with implicit scene representation. In: Int. Conf. Comput. Vis. pp. 15838–15847 (2021)

    2021

  45. [45]

    IEEE Robot

    Zhi, S., Sucar, E., Mouton, A., Haughton, I., Laidlow, T., Davison, A.J.: ilabel: Revealing objects in neural fields. IEEE Robot. and Auto. Letters (RAL)8(2), 832–839 (2022)

    2022

  46. [46]

    In: IEEE Conf

    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 distilled feature fields. In: IEEE Conf. Comput. Vis. Pattern Recog. pp. 21676–21685 (2024)

    2024

  47. [47]

    Zhou, X., Nguyen, B.H., Magne, L., Golyanik, V., Leimkühler, T., Theobalt, C.: Splat the net: Radiance fields with splattable neural primitives. In: Int. Conf. Learn. Represent. (2026) MLP-Splatting 19 A Ablation Study A.1 Novel-view Synthesis without Semantic Guidance T able 4:Effect of co-training on the Replica dataset. Replica PSNR ↑ SSIM ↑ LPIPS ↓ ML...

    2026

  48. [48]

    eachU i lies entirely in one smooth regionΩℓ of Definition 3

  49. [49]

    the overlap multiplicity is uniformly bounded: sup x∈D #{i:x∈U i} ≤C

  50. [50]

    Proof.Start from a regular cubic grid of mesh size comparable tohon a box containingD

    the number of patches satisfies N=Θ(h −3). Proof.Start from a regular cubic grid of mesh size comparable tohon a box containingD. For cubes that do not intersectΓ, keep them unchanged. For cubes intersectingΓ, use theC 2 regularity ofΓto split the cube along the interface into finitely many subpatches, each lying entirely on one side of the interface. The...

  51. [51]

    Gaussian splatting requires Ω(ε−2) parameters to achieve image errorε

  52. [52]

    In particular, fors≥2, ε−3/s ≤ε −3/2, which is asymptotically smaller than the Gaussian rateε−2

    MLP-Splatting with fixed-size neural primitives requires Θ(ε−3/s) parameters to achieve image errorε. In particular, fors≥2, ε−3/s ≤ε −3/2, which is asymptotically smaller than the Gaussian rateε−2. Proof.The Gaussian statement follows immediately from Theorem 3: if CK −1/2 ≤ε, then necessarily K≥C ′ε−2. The MLP-Splatting statement follows from Theorem 4,...