Pith. sign in

REVIEW 2 major objections 5 minor 54 references

Deformable Triangle Splatting: Flexible Primitives for Real-Time Radiance Field Rendering

T0 review · 2 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper claims that adding K learnable edge displacements to each triangle in a splatting renderer produces a single primitive that can represent non-convex shapes and improves novel-view quality over rigid triangle splatting.

desk verdict DETRIS is a credible incremental extension of Triangle Splatting with plausible gains, but the 'exactly view-independent by construction' claim does not hold under perspective projection. read the letter →

arxiv 2607.22446 v1 pith:MCXMD3ZN submitted 2026-07-24 cs.CV cs.GR

Oriol Jiménez-Ayguadé , Antonio Agudo This is my paper · ORCID
classification cs.CVcs.GR
keywords deformabletrianglesnon-convexprimitivesbarycentriccoordinatesradiancefieldssplattingdifferentiablerenderingwindingnumbernovelviewsynthesis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to give radiance-field rendering a primitive that is not forced to be convex. Its proposal: start from the standard triangle used in Triangle Splatting, but let each edge be deformed by K control points, each carrying one learnable scalar displacement along the edge normal, evaluated in the triangle's own barycentric coordinate space. A single such deformable triangle can take the shape of a circle, a heart, an arrow, or a crescent, which previously required many rigid triangles or soft Gaussian blobs. The authors report that on real scenes this lowers perceptual error (LPIPS 0.183 vs. 0.191 on Mip-NeRF 360) and improves mesh geometry (DTU Chamfer distance 1.01 vs. 1.06) while keeping rendering real-time. The reason this matters: if true, curved and concave geometry no longer demands excessive primitive proliferation, and a unified primitive can adapt to whatever shape the data requires.

What carries the argument

The load-bearing machinery is the barycentric coordinate map b = T^{-1}(p - p1) with T = [p2 - p1, p3 - p1], which sends every pixel inside the projected triangle to the unit triangle in (u,v) space. Control points are placed at b_{i,k} = b_start + (k/(K+1)) e_i + d_{i,k} nhat_i, where d_{i,k} is the single learnable scalar displacement along the edge normal nhat_i; this is the entire shape-adaptation mechanism, and it preserves the three base vertices that define the 3D plane. Inside/outside decisions use a non-zero winding-number rule via ray casting; the opacity falloff uses a polynomial smooth minimum over point-to-segment distances with smoothing radius g = delta * r_bary, followed by a

What would settle it

Render a single deformed primitive (e.g., a heart or arrow) with fixed displacements and corner smoothness from several camera angles spanning 0°–60° slant, extract the screen-space silhouette each time, and unproject it to 3D using the rendered depth. If the reconstructed 3D boundary differs between views beyond numerical precision, the claimed view-independence fails; equivalently, compute the 3D barycentric coordinates of the ray–plane intersection and compare them to the 2D pixel barycentric coordinates used by the rasterizer, and show the discrepancy grows with slant.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that convexity is not a necessary property of a splatting primitive. By placing K control points per edge and letting each move by a single scalar displacement along the edge normal in barycentric coordinates, the boundary of a triangle becomes a piecewise-linear polygon that can bend inward and outward while the three base vertices still define a 3D plane. The rasterizer classifies pixels with a winding-number test, computes a C1-continuous distance field via a polynomial smooth minimum with a learnable corner-rounding radius, and converts it to opacity through a power-law window normalized by the deformed polygon's inradius. Because all deformatio

Load-bearing premise

The load-bearing premise is that the 2D barycentric coordinates of a pixel, computed from the projected triangle, are an intrinsic and view-independent parameterization of the primitive's surface; because perspective projection is projective, not affine, a fixed deformed polygon in this coordinate space does not correspond to a fixed 3D region when the triangle is viewed obliquely.

Editorial extensions

If this is right

  • A single deformable triangle can reproduce both convex and concave 2D targets (circle, square, Gaussian, clover, arrow, Pac-Man, heart) that require three rigid triangles or eight Gaussians.
  • On average over Mip-NeRF 360, the method improves LPIPS to 0.183 from Triangle Splatting's 0.191, and on Tanks & Temples to 0.133 from 0.143, with best-in-class perceptual quality among non-volumetric primitives.
  • Zeroing the learned displacements after training increases mean absolute error by +0.052 on Bonsai and +0.029 on Bicycle, showing the deformations are what sharpen fine boundary structures.
  • The per-primitive overhead is small (3K+1 = 10 extra floats, about +17% per primitive), and rendering stays real-time though slower than rigid triangles (47/33 FPS vs. 63/49 on Bonsai/Garden), with the cost tunable through K.
  • Rendered normal maps are geometrically coherent as a by-product, enabling TSDF-based mesh extraction on DTU with Chamfer distance 1.01, better than the rigid-triangle baseline's 1.06.

Reading between the lines

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

  • Inference: The paper leaves implicit that its view-independence claim holds for the affine barycentric map, but perspective projection is projective, not affine; the 2D barycentric coordinates of a pixel do not equal the 3D barycentric coordinates of the point where the viewing ray meets the triangle plane, so a direct slant test would quantify the approximation.
  • Inference: Because each primitive carries a well-defined face normal and a piecewise-linear boundary, the representation could be reused for surface editing, relighting, or level-of-detail by decimating control points; these are natural next steps the paper does not explore.
  • Inference: The curvature regularizer and the tolerance for self-intersecting boundaries suggest the primitive can act as a deformable template in optimization problems beyond radiance fields, such as image segmentation or shape fitting; the paper demonstrates only 2D shape fitting as a proof of concept.
  • Inference: The reported gains over Triangle Splatting are mostly perceptual (LPIPS) rather than PSNR, which the paper attributes to sharper boundaries; an evaluation protocol that weights high-frequency structure would be a more sensitive test of the advantage.
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.

Referee Report

2 major / 5 minor

Summary. The paper proposes DETRIS, a differentiable radiance-field renderer whose primitives are triangles with deformable edges. Each edge carries K learnable control points, each with a scalar displacement along the edge normal, defined in the barycentric coordinate system of the projected triangle. Rendering uses a winding-number inside test, a polynomial smooth-minimum distance field, and a power-law opacity falloff. The authors report improved LPIPS/PSNR/SSIM over Triangle Splatting on Mip-NeRF 360 and Tanks & Temples, improved Chamfer distance on DTU, and ablations showing that the deformation and regularization components contribute.

Significance. Non-convex, deformable triangle primitives are a useful step beyond fixed triangles and convex primitives. The paper is strong on empirical validation: per-scene tables, deformation contribution maps, control-point statistics, and a clear ablation suite. It also describes an efficient CUDA implementation and discusses stability under self-intersection. However, a central theoretical claim—that the barycentric-space formulation is exactly view-independent—is not correct under perspective, so the paper's main conceptual justification needs substantial rework.

major comments (2)
  1. [Sec. 3.1, Eq. (1), ‘Why Barycentric Space?’] The claim that the formulation is ‘exactly view-independent by construction’ is invalid. The map b=T^{-1}(p-p_1) is built from the 2D projected vertices p_i. Under perspective, a fixed 3D point X on the base triangle has 2D barycentric coordinates that change with viewpoint (b_i are projective, not affine, in the 3D barycentrics λ). A fixed deformed polygon in (u,v) therefore corresponds to different 3D regions on the primitive for different cameras. This is exactly the view-dependence the paper attributes to image-space deformations. The Abstract’s ‘view-consistent rendering’, §3.1’s ‘eliminating multi-view gradient conflict’, and the Conclusion’s ‘strictly view-independent’ are therefore unsupported. Figure 3 does not test 3D consistency: it compares two image-space parameterizations. The authors should either formulate the deformation in true 3D barycentric coordinates of the base tri
  2. [Eq. (5)] R=2A/P is not, as stated, the inradius of a general convex polygon; it equals the inradius only for tangential polygons. For a 2×1 rectangle, R=2A/P=4/6≈0.667 while the inradius is 0.5. For non-convex polygons the claim that R is an upper bound on interior distances is not established, and φ≤1 is trivial because of the min(1,·) clamp. Since R sets the spatial scale of the opacity falloff in Eq. (6), the paper should either compute the true polygon inradius or explicitly present R as an area-perimeter heuristic and discuss the consequences.
minor comments (5)
  1. [Sec. 3.3 and Table S1] The loss in Eq. (7) includes L_d with weight λ3, but Table S1 sets λ_dist=0. If the distortion term is disabled, this should be stated in the main text.
  2. [Table 2 and Table S2] FPS and training-time footnotes are ambiguous: TS is reported with two FPS/Train values, and the paper’s average Mip-NeRF 360 FPS (55) is hard to reconcile with the per-scene Bonsai/Garden values (47/33). Please clarify hardware and averaging protocol.
  3. [Fig. 3] The units and color scale for |Δ| are not defined, and Fig. 3-right should state quantitatively what ‘consistent shape’ means; visual comparison alone is not a test of 3D consistency.
  4. [Fig. 5 and Fig. S2] Zeroing δ together with d means the figure does not isolate the displacement contribution; Table 3 is the cleaner ablation. The caption should acknowledge this.
  5. [Sec. 3.1] The statement that barycentric displacements are ‘scale-invariant with respect to the triangle’s image-space size’ is true only for uniform scaling of the projected triangle, not under perspective; the sentence should be qualified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DETRIS's claims rest on fitted parameters and external benchmarks; the view-independence defect is a correctness issue, not a circular derivation.

full rationale

The paper's derivation chain does not reduce any prediction to its inputs. Eq. (1) defines control-point positions from learnable scalar displacements d_i,k; Eqs. (3), (5), and (6) define opacity from learnable sharpness delta, polygon inradius, opacity o, and falloff sigma. These are optimized against photometric losses on external benchmarks (Mip-NeRF 360, Tanks & Temples, DTU), with ablations (w/o Deformation, w/o Lcurv, w/o Delay) isolating component contributions. The view-independence claim in Sec. 3.1 is asserted rather than derived from fitted values, and the paper's own critique of image-space windows (perspective is projective, not affine) applies to its barycentric frame, which is itself built from projected vertices via T = [p2-p1, p3-p1]. That is a correctness/validity risk, not a circularity: no equation used to support the claim is definitionally equivalent to the target result. Citations to prior work [14,15,48] supply baselines and inspiration but are not used to force the central result. Self-citations appear only in related work. Therefore no load-bearing circular step is present.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

Most model parameters (vertices, SH colors, displacements, sigma, delta, o) are learned end-to-end from the photometric loss; I list the ones that define the new primitive and the hand-set constants that govern it. These are model parameters rather than hidden physics constants, but they are fitted to data and the method's expressiveness scales with them. No new physical entities are introduced.

free parameters (6)
  • Boundary displacement d_{i,k} = learned scalars; mean |d|=0.083 r_bary across scenes
    Central deformation parameters; one scalar per control point per edge (9 floats per primitive for K=3).
  • Falloff exponent sigma = per-primitive learned; init 1.16
    Controls opacity concentration in Eq. (6).
  • Corner smoothness delta = per-primitive learned; mean 0.075
    Sets smoothing radius g=delta*r_bary in Eq. (3).
  • Base opacity o = per-primitive learned; init 0.28
    Multiplicative opacity in Eq. (6).
  • K control points per edge = K=3 (ablation 0-5)
    Hand-chosen hyperparameter balancing quality and cost in Table 1.
  • Initialization and loss hyperparameters = 2.23 r scale, lr rates, lambda_1-6 in Table S1
    Empirically set; affect convergence and final primitive counts.
assumptions (5)
  • standard math Non-zero winding rule (ray casting) classifies inside/outside for non-convex and self-intersecting polygons.
    Used in Sec. 3.1 for deformed polygon containment; standard computational geometry [19].
  • standard math Polynomial smooth minimum (Eq. 3) is C1 and produces localized corner rounding.
    Follows Quilez [41]; the paper uses it to build the SDF fold.
  • ad hoc to paper R=2A/P is a valid opacity normalization: it is the true inradius for convex polygons and upper-bounds interior distances for non-convex polygons.
    Eq. (5). This is only true for tangential convex polygons; for a general rectangle 2A/P exceeds the inradius, and for non-convex shapes the bound is not established.
  • ad hoc to paper The affine image-space map b=T^{-1}(p-p1) is an intrinsic, view-independent coordinate system for the primitive.
    Implicit in Eq. (1) and the 'Why Barycentric Space?' argument; under perspective, 2D barycentric coordinates of a pixel are not equal to 3D barycentric coordinates of the point on the triangle plane.
  • domain assumption SfM point clouds provide reliable triangle initializations and the scenes satisfy photometric consistency.
    Standard in 3DGS-style methods; used in Sec. 3.3 for initialization.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deformable Triangle Splatting: Flexible Primitives for Real-Time Radiance Field Rendering." pith.science (2026). https://pith.science/paper/MCXMD3ZN

@misc{pith2026260722446,
  author       = {Pith},
  title        = {Pith review of: Deformable Triangle Splatting: Flexible Primitives for Real-Time Radiance Field Rendering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MCXMD3ZN}},
  note         = {Machine review of arXiv:2607.22446}
}
abstract

Recent radiance field methods represent scenes with 2D primitives that offer surface alignment and efficient rasterization, from Gaussian disks to triangles, yet all rely on convex boundaries: curved and concave structures demand excessive primitives. We introduce Deformable Triangle Splatting, which augments each triangle with $K$ control points per edge, each parameterized by a single learnable scalar displacement that shifts the boundary inward or outward, enabling non-convex shape representation while preserving the three base vertices that define the 3D plane. To render these non-convex primitives differentiably, we design a rasterization pipeline in the triangle's barycentric coordinate space, ensuring view-consistent rendering. A winding number test determines whether each pixel lies inside the deformed primitive, and a window function controlled by two learnable parameters, sharpness and corner smoothness, together with a per-primitive scalar opacity, produces the smooth opacity transition from interior to boundary. Validation is done in a variety of real-world scenes, outperforming recent works based on non-volumetric primitives in terms of visual quality and versatility while still achieving competitive rendering efficiency.

Figures

Figures reproduced from arXiv: 2607.22446 by the authors.

Figure 1
Figure 1. Method overview. From left to right: A sparse SfM point cloud initial￾izes the primitives; Each triangle is first projected from 3D to image space (pi), then mapped via T −1 to a canonical barycentric coordinate frame, ensuring view￾independent deformations; K learnable Control Points (CP) per edge (green dots) displace the boundary by di,k along edge normals (orange arrows), producing a de￾formable piecewise-linear… view at source ↗
Figure 2
Figure 2. σ and δ interpretation. A deformable heart primitive with K=3 is rendered with increasing falloff exponent σ and corner smoothness δ. Higher σ concentrates opacity near the medial axis; higher δ rounds the sharp corners into smooth curves. per pixel on GPU than the exp/log calls required by LogSumExp; and 3) it op￾erates directly on the minimum of unsigned distances rather than the maximum of signed half-plane dista… view at source ↗
Figure 3
Figure 3. Image-space vs. barycentric formulation under viewpoint change. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison on real scenes. Each row shows a test view from a different scene. Each column shows the ground truth, TS [14], 3DGS [22], and ours. GT Full (K=3) Rigid (d=0) |Full−Rigid| Counter Truck [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Deformation contribution maps. For each scene we render the trained model with learned control-point displacements active (Full) and with all displacements and corner smoothness δ zeroed (Rigid, d=0), keeping all other parameters unchanged. The rightmost column shows t…
Figure 6
Figure 6. Figure 6: Single-primitive shape fitting. In both sides, the same information is dis￾played. Each row represents a target shape and columns compare our method (K=3 and K=7 control points per edge) with TS [14] and 3DGS [22]. N denotes the number of primitives, being N=1 in our c…
Figure 7
Figure 7. Figure 7: Learned parameter analysis (Indoor [2] vs. Outdoor [2] + T&T [25]). Left: Radar profile of six CP application metrics, each normalized to [0, 1]. Middle: Signed CP displacement histogram (d<0: inward, d>0: outward); dashed lines mark ±rbary. In￾door scenes show a clear…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 1 linked inside Pith

  1. [1]

    In: ECCV (2024)

    Bae, J., Kim, S., Yun, Y., Lee, H., Bang, G., Uh, Y.: Per-gaussian embedding- based deformation for deformable 3D gaussian splatting. In: ECCV (2024)

    2024

  2. [2]

    In: CVPR (2022)

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

    2022

  3. [3]

    In: ICCV (2023)

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

    2023

  4. [4]

    In: CVPR (2022)

    Chan, E., Lin, C., Chan, M., Nagano, K., Pan, B., de Mello, S., Gallo, O., Guibas, L., Tremblay, J., Khamis, S., Karras, T., Wetzstein, G.: Efficient geometry-aware 3D generative adversarial networks. In: CVPR (2022)

    2022

  5. [5]

    In: ECCV (2022)

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

    2022

  6. [6]

    arXiv preprint arXiv:2411.12440 (2024)

    Chen, H., Chen, R., Qu, Q., Wang, Z., Liu, T., Chen, X., Chung, Y.: Beyond gaussians: Fast and high-fidelity 3D splatting with linear kernels. arXiv preprint arXiv:2411.12440 (2024)

    arXiv 2024

  7. [7]

    In: CVPR (2023)

    Chen, Z., Funkhouser, T., Hedman, P., Tagliasacchi, A.: MobileNeRF: Exploiting the polygon rasterization pipeline for efficient neural field rendering on mobile architectures. In: CVPR (2023)

    2023

  8. [8]

    In: CVPR (2023)

    Du, Y., Smith, C., Tewari, A., Sitzmann, V.: Learning to render novel views from wide-baseline stereo pairs. In: CVPR (2023)

    2023

Show all 54 references
  1. [9]

    In: CVPR (2025)

    Fan, C.D., Chang, C.W., Liu, Y.R., Lee, J.Y., Huang, J.L., Tseng, Y.C., Liu, Y.L.: SpectroMotion: Dynamic 3D reconstruction of specular scenes. In: CVPR (2025)

    2025

  2. [10]

    In: CVPR (2023)

    Fridovich-Keil, S., Meanti, G., Warburg, F., Recht, B., Kanazawa, A.: K-planes: Explicit radiance fields in space, time, and appearance. In: CVPR (2023)

    2023

  3. [11]

    In: CVPR (2022)

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

    2022

  4. [12]

    In: ECCV (2024)

    Gue´don, A., Lepetit, V.: Gaussianfrosting: Editable complex radiance fields with real-time rendering. In: ECCV (2024)

    2024

  5. [13]

    In: CVPR (2024)

    Hamdi, A., Melas-Kyriazi, L., Mai, J., Qian, G., Liu, R., Vondrick, C., Ghanem, B., Vedaldi, A.: GES: Generalized exponential splatting for efficient radiance field rendering. In: CVPR (2024)

    2024

  6. [14]

    In: 3DV (2026)

    Held, J., Vandeghen, R., Deliege, A., Hamdi, A., Giancola, S., Cioppa, A., Vedaldi, A., Ghanem, B., Tagliasacchi, A., Droogenbroeck, M.V.: Triangle splatting for real- time radiance field rendering. In: 3DV (2026)

    2026

  7. [15]

    In: CVPR (2025)

    Held, J., Vandeghen, R., Hamdi, A., Deliege, A., Cioppa, A., Giancola, S., Vedaldi, A., Ghanem, B., Droogenbroeck, M.V.: 3D convex splatting: Radiance field ren- dering with 3D smooth convexes. In: CVPR (2025)

    2025

  8. [16]

    In: CVPR (2024)

    Hu, S., Hu, T., Liu, Z.: Articulated gaussian splatting from monocular human videos. In: CVPR (2024)

    2024

  9. [17]

    SIGGRAPH (2024)

    Huang, B., Yu, Z., Chen, A., Geiger, A., Gao, S.: 2D gaussian splatting for geo- metrically accurate radiance fields. SIGGRAPH (2024)

    2024

  10. [18]

    In: CVPR (2025)

    Huang, Y.H., Lin, M.X., Sun, Y.T., Yang, Z., Lyu, X., Cao, Y.P., Qi, X.: De- formable radial kernel splatting. In: CVPR (2025)

    2025

  11. [19]

    SIGGRAPH (2013)

    Jacobson, A., Kavan, L., Sorkine-Hornung, O.: Robust inside-outside segmentation using generalized winding numbers. SIGGRAPH (2013)

    2013

  12. [20]

    In: CVPR (2014)

    Jensen, R., Dahl, A., Vogiatzis, G., Tola, E., Aanæs, H.: Large scale multi-view stereopsis evaluation. In: CVPR (2014)

    2014

  13. [21]

    In: CVPR (2018) 16 O

    Kato, H., Ushiku, Y., Harada, T.: Neural 3D mesh renderer. In: CVPR (2018) 16 O. Jiménez-Ayguadé and A. Agudo

    2018

  14. [22]

    TOG (2023)

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

    2023

  15. [23]

    In: NeurIPS (2024)

    Kheradmand, S., Rebain, D., Sharma, G., Sun, W., Tseng, J., Isack, H., Kar, A., Tagliasacchi, A., Yi, K.M.: 3D gaussian splatting as markov chain monte carlo. In: NeurIPS (2024)

    2024

  16. [24]

    In: ICLR (2015)

    Kingma, D., Ba, J.: Adam: A method for stochastic optimization. In: ICLR (2015)

    2015

  17. [25]

    TOG (2017)

    Knapitsch, A., Park, J., Zhou, Q.Y., Koltun, V.: Tanks and temples: Benchmarking large-scale scene reconstruction. TOG (2017)

    2017

  18. [26]

    In: ICCV (2023)

    Kulhanek, J., Sattler, T.: Tetra-NeRF: Representing neural radiance fields using tetrahedra. In: ICCV (2023)

    2023

  19. [27]

    In: CVPR (2024)

    Lee,J.C.,Rho,D.,Sun,X.,Ko,J.H.,Park,E.:Compact3Dgaussianrepresentation for radiance field. In: CVPR (2024)

    2024

  20. [28]

    ICLR (2026)

    Liu, R., Gao, Z., Planche, B., Chen, M., Nguyen, V.N., Zheng, M., Choudhuri, A., Chen, T., Wang, Y., Feng, A., Wu, Z.: Universal beta splatting. ICLR (2026)

    2026

  21. [29]

    In: SIGGRAPH (2025)

    Liu, R., Sun, D., Chen, M., Wang, Y., Feng, A.: Deformable beta splatting. In: SIGGRAPH (2025)

    2025

  22. [30]

    In: ICCV (2019)

    Liu, S., Chen, W., Li, T., Li, H.: Soft rasterizer: A differentiable renderer for image- based 3D reasoning. In: ICCV (2019)

    2019

  23. [31]

    In: CVPR (2024)

    Lu, T., Yu, M., Xu, L., Xiangli, Y., Wang, L., Lin, D., Dai, B.: Scaffold-GS: Struc- tured 3D gaussians for view-adaptive rendering. In: CVPR (2024)

    2024

  24. [32]

    In: NeurIPS (2025)

    von Lützow, N., Nießner, M.: Linprim: Linear primitives for differentiable volu- metric rendering. In: NeurIPS (2025)

    2025

  25. [33]

    A K Peters/CRC Press (2021)

    Marschner, S., Shirley, P.: Fundamentals of Computer Graphics. A K Peters/CRC Press (2021)

    2021

  26. [34]

    In: CVPR (2021)

    Martin-Brualla, R., Radwan, N., Sajjadi, M., Barron, J., Dosovitskiy, A., Duck- worth, D.: NeRF in the wild: Neural radiance fields for unconstrained photo col- lections. In: CVPR (2021)

    2021

  27. [35]

    In: ECCV (2020)

    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: ECCV (2020)

    2020

  28. [36]

    SIGGRAPH (2022)

    Müller, T., Evans, A., Schied, C., Keller, A.: Instant neural graphics primitives with a multiresolution hash encoding. SIGGRAPH (2022)

    2022

  29. [37]

    In: CVPR (2022)

    Niemeyer, M., Barron, J., Mildenhall, B., Sajjadi, M., Geiger, A., Radwan, N.: RegNeRF:Regularizingneuralradiancefieldsforviewsynthesisfromsparseinputs. In: CVPR (2022)

    2022

  30. [38]

    Cambridge university press (1998)

    O’Rourke, J.: Computational geometry in C. Cambridge university press (1998)

    1998

  31. [39]

    In: ACCV (2024)

    de Paco, S., Agudo, A.: 4DPV: 4D pet from videos by coarse-to-fine non-rigid radiance fields. In: ACCV (2024)

    2024

  32. [40]

    Park, K., Sinha, U., Barron, J.T., Bouaziz, S., Goldman, D.B., Seitz, S.M., Martin- Brualla, R.: Nerfies: Deformable neural radiance fields (2021)

    2021

  33. [41]

    Quilez, I.: Smooth minimum.https://iquilezles.org/articles/smin/(2013), accessed: 2025-01-15

    2013

  34. [42]

    TOG (2023)

    Reiser, C., Szeliski, R., Verbin, D., Srinivasan, P., Mildenhall, B., Geiger, A., Bar- ron, J., Hedman, P.: MERF: Memory-efficient radiance fields for real-time view synthesis in unbounded scenes. TOG (2023)

    2023

  35. [43]

    In: WACV (2025)

    Rong, V., Chen, J., Bahmani, S., Kutulakos, K., Lindell, D.B.: Gstex: Per-primitive texturing of 2D gaussian splatting for decoupled appearance and geometry model- ing. In: WACV (2025)

    2025

  36. [44]

    In: ISBI (2026) DETRIS 17

    Salort-Benejam, L., Agudo, A.: Nerfscopy: Neural radiance fields for in-vivo time- varying tissues from endoscopy. In: ISBI (2026) DETRIS 17

    2026

  37. [45]

    In: CVPR (2016)

    Schönberger, J.L., Frahm, J.M.: Structure-from-motion revisited. In: CVPR (2016)

    2016

  38. [46]

    In: ECCV (2022)

    Shen, J., Agudo, A., Moreno-Noguer, F., Ruiz, A.: Conditional-flow NeRF: Accu- rate 3D modelling with reliable uncertainty quantification. In: ECCV (2022)

    2022

  39. [47]

    In: ICCV (2025)

    Svitov, D., Morerio, P., Agapito, L., Del Bue, A.: Billboard splatting (BBSplat): Learnable textured primitives for novel view synthesis. In: ICCV (2025)

    2025

  40. [48]

    In: SIGGRAPH (1987)

    Terzopoulos, D., Platt, J., Barr, A., Fleischer, K.: Elastically deformable models. In: SIGGRAPH (1987)

    1987

  41. [49]

    In: NeurIPS (2023)

    Uy, M.A., Nakayama, G.K., Yang, G., Thomas, R.K., Guibas, L., Li, K.: NeRF revisited: Fixing quadrature instability in volume rendering. In: NeurIPS (2023)

    2023

  42. [50]

    In: CVPR (2022)

    Verbin, D., Hedman, P., Mildenhall, B., Zickler, T., Barron, J., Srinivasan, P.: Ref-NeRF: Structured view-dependent appearance for neural radiance fields. In: CVPR (2022)

    2022

  43. [51]

    In: AAAI (2026)

    Wu, R., Zhang, Z., Chen, M., Yan, Z., Zuo, W.: Deblur4dgs: 4D gaussian splatting from blurry monocular video. In: AAAI (2026)

    2026

  44. [52]

    In: CVPR (2022)

    Xu,Q.,Xu,Z.,Philip,J.,Bi,S.,Shu,Z.,Sunkavalli,K.,Neumann,U.:Point-NeRF: Point-based neural radiance fields. In: CVPR (2022)

    2022

  45. [53]

    In: CVPR (2022)

    Zhang, X., Bi, S., Sunkavalli, K., Su, H., Xu, Z.: NeRFusion: Fusing radiance fields for large-scale scene reconstruction. In: CVPR (2022)

    2022

  46. [54]

    In: ICCV (2025) 18 O

    Zhang, Z., Huang, B., Jiang, H., Zhou, L., Xiang, X., Shen, S.: Quadratic gaussian splatting: High quality surface reconstruction with second-order geometric primi- tives. In: ICCV (2025) 18 O. Jiménez-Ayguadé and A. Agudo Supplementary Material Deformable Triangle Splatting: ...

    2025

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.