Pith. sign in

REVIEW 5 major objections 5 minor 34 references

CoDe-NeRF: Neural Rendering via Dynamic Coefficient Decomposition

T0 review · 5 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read CoDe-NeRF decomposes scene appearance into a static neural basis and view-dependent coefficients, claiming sharper specular highlights than prior radiance-field methods on glossy and reflective benchmarks.

desk verdict Plausible architecture, informative ablations, but the Forward-Facing PSNR is a 6 dB outlier with no supporting detail—so the SOTA claim is not established. read the letter →

arxiv 2508.06632 v1 pith:SNZL6AHS submitted 2025-08-08 cs.CV cs.AI

Wenpeng Xing , Jie Chen , Zaifeng Yang , Tiancheng Zhao , Gaolei Li , Changting Lin , Yike Guo , Meng Han This is my paper
classification cs.CVcs.AI
keywords neuralradiancefieldsnovelviewsynthesisspecularreflectiondynamiccoefficientdecompositionFiLMconditioningbasisvolumerenderingview-dependentappearance
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 tries to establish that the hard problem of rendering shiny, reflective scenes from photographs can be handled without choosing between blurry entangled appearance models and unstable physics-based inverse rendering. Its proposal is to split appearance into two learned pieces: a small set of shared neural basis vectors that carry material properties, and dynamic coefficients, generated from view direction and an illumination code, that modulate those bases. A non-linear MLP then combines coefficients and basis into final radiance. If the claims hold, the payoff is sharper specular highlights and reflection details at higher image quality than NeRF, TensoRF, 3D Gaussian Splatting, and Ref-NeRF, while keeping the simple optimisation of a feed-forward radiance field. The reported results include 34.18 dB PSNR on NeRF-Synthetic and 33.33 dB on Forward-Facing.

What carries the argument

The engine is the dynamic coefficient decomposition: appearance features $T_c$ are projected to a compact shared basis $H = W \cdot T_c$; a Coefficient Network $F_\theta$ applies a feature-wise linear modulation (FiLM) $k = W_x \odot d + b_x$ using the view direction and an illumination embedding $z_s$; and a Dynamic Radiance Integrator $G_\phi$, a small MLP, non-linearly blends $k$ and $H$ into outgoing radiance. This replaces the expensive integral over incoming light with a learned, low-rank, non-linear map, and the paper's ablations attribute most of the gain to this three-part division.

What would settle it

Render a glossy scene illuminated by a small, close light that casts a sharp shadow; if CoDe-NeRF reproduces the highlight but the shadow boundary is blurry or absent while an explicit-light method renders it cleanly, the latent-illumination assumption is falsified. Equivalently, on a scene with many distinct materials, check whether PSNR keeps rising when the basis count grows beyond 16; if it plateaus or drops, the fixed basis capacity caps the method.

Watch

Extended reading notes

Core claim

The central discovery is that view-dependent appearance does not need to be baked into a monolithic color field, nor does it require solving a full BRDF inverse-rendering problem. The paper shows that decomposing color into a view-agnostic neural basis $H=\{h_n\}_{n=1}^{N_p}$ and view/light-dependent coefficients $k$, generated by a FiLM-style affine transform $k = W_x \odot d + b_x$ and fused by an MLP integrator $G_\phi(k,H) \approx \int_\Omega f_r(x,\omega_i,d)L_i(x,\omega_i)(\omega_i\cdot n)\,d\omega_i$, produces sharper specular highlights than prior methods on the NeRF-Synthetic, Forward-Facing, and Shiny benchmarks. The decomposition acts as a learned low-rank approximation of the ren

Load-bearing premise

The method assumes that a fixed set of 16 shared neural basis vectors, modulated by a learned affine transform of the view direction and a global illumination code, is expressive enough to reproduce every view- and illumination-dependent specular effect in a scene.

Editorial extensions

If this is right

  • Glossy and mirror-like surfaces can be rendered at higher fidelity than prior radiance-field methods without explicit BRDF decomposition, since the coefficient-basis split captures specular effects with a single forward pass.
  • The non-linear blending step is load-bearing: swapping the integrator for a dot product drops LPIPS sharply, so linear basis expansion alone is not enough.
  • Dense multi-view training improves the method's quality consistently (from 32.61 dB at 100 views to 36.27 dB at 300 views on the materials scene), while a competing method degrades.
  • The neural basis acts as a reusable dictionary of reflection patterns, so appearance is stored more compactly than feeding raw appearance features directly.
  • The design supports multiple lighting conditions by appending a learnable illumination embedding per condition, at the cost of local effects like cast shadows.

Reading between the lines

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

  • The latent illumination embedding is a global code per lighting condition, so an untested extension is to interpolate or extrapolate these codes to approximate novel lighting; the paper does not run that experiment.
  • The plateau at $N_p=16$ suggests capacity is tied to scene complexity; scenes with many distinct materials may need per-scene or adaptive basis counts, and the paper's fixed choice may under- or over-allocate.
  • Because $G_\phi$ approximates the BRDF integral, the learned bases could be read as a dictionary of effective BRDF lobes; that interpretation suggests a testable route to material editing even though the current framework does not expose physical parameters.
  • The method's gains come without explicit surface normal estimation, so it may complement or compete with normal-aware specular methods; comparing on scenes where normals are known would isolate where the coefficient shortcut fails.
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

5 major / 5 minor

Summary. The paper proposes CoDe-NeRF, a neural radiance field variant that decomposes appearance into a view-invariant neural basis H (projected from TensoRF-style appearance features) and view/illumination-dependent coefficients k produced by a FiLM-modulated MLP (the "Coefficient Network"). A second MLP (the "Dynamic Radiance Integrator") nonlinearly blends k and H to produce radiance. The method is trained with photometric and TV losses, with optional per-condition illumination embeddings. The authors claim state-of-the-art rendering of specular and reflective scenes, reporting PSNR 34.18 on NeRF-Synthetic and 33.33 on Forward-Facing (Table 1), plus ablations and dense-view experiments.

Significance. If the reported results are reproducible, the decomposition idea is a simple and potentially useful alternative to physically-based inverse-rendering approaches, and the ablations provide a reasonable sanity check of the design. The paper is also honest about limitations: no relighting, no editable materials, and latent illumination cannot model cast shadows. However, the central empirical claim is not currently supported: the Forward-Facing PSNR is an unexplained 6 dB outlier relative to all baselines, no per-scene numbers or error bars are provided, the Shiny benchmark—the most direct test for glossy/reflective scenes—is only shown qualitatively, and there are internal inconsistencies between tables. The method's "physics derivation" is motivational rather than a constraint, which weakens the framing but does not by itself invalidate the architecture.

major comments (5)
  1. [Table 1] The Forward-Facing PSNR of 33.33 dB is 6.07 dB above the best baseline (Nex, 27.26) and 6.83 dB above NeRF (26.50). This is far outside the range reported by existing work on this benchmark and implies a ~60% reduction in MSE. No per-scene breakdown, error bars, or evaluation-protocol details are given, and Figure 2 shows only selected crops. Since the paper's central 'state-of-the-art' claim rests largely on this number, the claim is not established. Please provide per-scene results, the exact train/test split, and ideally code or a reproducibility statement.
  2. [Figure 3 / Shiny dataset] The Shiny dataset is the most direct benchmark for the paper's stated goal of rendering glossy and reflective scenes, yet it is only shown qualitatively. Quantitative PSNR/SSIM/LPIPS comparisons on Shiny against NeX, Ref-NeRF, and other baselines are needed to support the abstract's claim of 'several challenging benchmarks' and the SOTA claim.
  3. [Table 3 vs Table 1] The ablation study in Table 3 reports 'Ours (Full Model)' with PSNR 32.54 and LPIPS 0.041, while Table 1 reports the same method on NeRF-Synthetic with PSNR 34.18 and LPIPS 0.037. The text does not specify which scene or subset the ablations use, making it impossible to judge the magnitude of each component's contribution. This needs to be stated clearly; otherwise the ablations are uninterpretable.
  4. [Table 2 / Experimental protocol] Table 2 reports results on the 'materials' scene of NeRF-Synthetic trained with 100, 200, and 300 views. The standard NeRF-Synthetic dataset provides 100 training views per scene. Please explain how 200- and 300-view training sets were constructed (e.g., rendered from the NeRF model, subsampled differently, or obtained from another source) and confirm no test-view contamination. This is necessary to validate the claim that the method 'scales effectively with view count.'
  5. [Eqs. (10)-(11)] Equation (10) states that the learnable illumination embedding z_s approximates the lighting integral, but z_s is a per-condition vector with no dependence on surface point x or incoming direction ω_i, so it cannot represent the spatially varying integral in Eq. (5). If this is intended only as motivation, the text should say so. The phrase 'constrained by BRDF-inspired priors' in the Coefficient Network section is not backed by any explicit constraint in the loss or architecture.
minor comments (5)
  1. [Figure 2 caption] Typo: 'F ortress' and 'Horn' should be 'Fortress' and 'Horns'.
  2. [Page 1/Table 1] Grammar: 'Tables 1 present' should be 'Table 1 presents'.
  3. [Eq. (16)] Please define Δ²V and Δ²M. The denominator P as 'total number of parameters' is ambiguous given the sum is over parameters; clarify indexing.
  4. [Eq. (7)] H is derived from per-point appearance features T_c(x), so calling H 'static' is misleading. It is view-invariant, not spatially constant; the terminology should be clarified.
  5. [Implementation Details] The text says training runs for 100k iterations, then states 'training a NeRF-Synthetic scene for 10k iterations takes approximately 2 hours.' Please reconcile these numbers or clarify that one is a typo.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method is an empirical neural rendering fit; the physics equations are motivational, not derived predictions.

full rationale

The paper's central claim is an empirical architecture for novel-view synthesis, not a first-principles derivation. All components—neural basis H, dynamic coefficients k, the radiance integrator Gφ, and illumination embeddings zs—are learned jointly from the photometric loss on the same scenes. The physics discussion in Eqs. (5)–(11) explicitly replaces the BRDF integral with an MLP approximation; no physical constraint is enforced, so there is no quantity that is 'predicted' from fitted inputs and hence no reduction of a prediction to a fit by construction. The ablation study is internal evidence about the architecture, not a circular validation. The only self-citation (Xing et al. 2022, in the Related Work) is a background statement about light-field limitations and is not load-bearing for the paper's contributions. The unusually large Forward-Facing PSNR gain is a legitimate empirical-validity and reproducibility concern, but it is not a circularity issue. The paper's own Limitations section honestly states that the latent illumination model does not support relighting or material editing, further indicating that no physical prediction is being claimed. Overall, no circular step is identifiable from the text.

Assumptions & free parameters 4 free parameters · 5 assumptions · 2 invented entities

The central contribution is entirely learned: the basis, the coefficients, and the illumination codes are all optimized against a photometric loss on the target scenes. No parameter-free derivation, external physical constraint, or independent prediction is provided. The 'neural approximation' of the BRDF integral is an assertion that an MLP can approximate it, not a derivation of the decomposition.

free parameters (4)
  • Per-condition illumination embedding z_s (D=32) = R^32, one per lighting condition
    Learned latent code introduced in Eq. (10) to absorb lighting; fitted jointly with scene parameters via photometric loss.
  • Neural basis projection matrix W (Eq. 7) = Np=16 basis vectors
    Maps appearance features Tc to a compact basis H; learned from data, controls expressiveness as shown in Table 4.
  • FiLM weight matrix W_x and bias b_x (Eq. 12) = Nw=32
    Affine transform producing coefficients k from view direction; learned.
  • Hyperparameters (Np=16, Nw=32, freq=2, lambda1, lambda2) = 16, 32, 2, unspecified
    Chosen by hand or small ablation; affecting model capacity and regularization strength.
assumptions (5)
  • standard math The differentiable volume rendering equation (Eq. 3-4) correctly maps radiance samples to pixel colors.
    Standard NeRF assumption, not proven here.
  • domain assumption The vector-matrix tensor factorization (Eq. 1) can represent the scene's density and appearance features.
    Adopted from TensoRF as the backbone; no analysis of its capacity for reflective scenes.
  • ad hoc to paper A linear projection of appearance features captures sufficient material information (H = W*Tc).
    Core to the decomposition; no independent evidence that a small basis encodes material properties.
  • ad hoc to paper The learned illumination embedding z_s is a sufficient proxy for the lighting integral (Eq. 10).
    Assumes a global latent code models lighting; limitations admit it fails on cast shadows.
  • ad hoc to paper The FiLM affine transform k = W_x (dot) d + b_x can encode view-dependent specular variation.
    A design choice; the paper argues concatenation dilutes the direction signal but provides no proof.
invented entities (2)
  • Shared neural basis H
    purpose: Encodes reusable reflection patterns
    Learned feature vectors; no physical interpretation or external falsifiable handle.
  • Illumination embedding z_s
    purpose: Latent code representing lighting conditions
    Learned per-condition code; the paper's own limitations say it cannot model local lighting like cast shadows.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CoDe-NeRF: Neural Rendering via Dynamic Coefficient Decomposition." pith.science (2026). https://pith.science/paper/SNZL6AHS

@misc{pith2026250806632,
  author       = {Pith},
  title        = {Pith review of: CoDe-NeRF: Neural Rendering via Dynamic Coefficient Decomposition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SNZL6AHS}},
  note         = {Machine review of arXiv:2508.06632}
}
read the original abstract

Neural Radiance Fields (NeRF) have shown impressive performance in novel view synthesis, but challenges remain in rendering scenes with complex specular reflections and highlights. Existing approaches may produce blurry reflections due to entanglement between lighting and material properties, or encounter optimization instability when relying on physically-based inverse rendering. In this work, we present a neural rendering framework based on dynamic coefficient decomposition, aiming to improve the modeling of view-dependent appearance. Our approach decomposes complex appearance into a shared, static neural basis that encodes intrinsic material properties, and a set of dynamic coefficients generated by a Coefficient Network conditioned on view and illumination. A Dynamic Radiance Integrator then combines these components to synthesize the final radiance. Experimental results on several challenging benchmarks suggest that our method can produce sharper and more realistic specular highlights compared to existing techniques. We hope that this decomposition paradigm can provide a flexible and effective direction for modeling complex appearance in neural scene representations.

Figures

Figures reproduced from arXiv: 2508.06632 by the authors.

Figure 1
Figure 1. Overall pipeline of our method, centered on dynamic coefficient decomposition. This approach decomposes scene [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Qualitative comparisons on test views from the Forward-Facing dataset. For each scene (Horns, Fern, Fortress), the [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison on the CD scene from the Shiny dataset (Wizadwongsa et al. 2021). Our method bet￾ter preserves specular highlights and structural details com￾pared to prior methods. Datasets We evaluate our method on three established benchmarks: NeRF-Synthetic (Mildenhall et al. 2020), Forward-Facing (Mildenhall et al. 2019), and Shiny (Wiz￾adwongsa et al. 2021). Competing Methods We compare our method again… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 29 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    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 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Attal, B.; Huang, J.-B.; Zollh \"o fer, M.; Kopf, J.; and Kim, C. 2022. Learning Neural Light Fields With Ray-Space Embedding. In CVPR, 19819--19829

    work page 2022

  4. [4]

    T.; Liu, C.; and Lensch, H

    Boss, M.; Braun, R.; Jampani, V.; Barron, J. T.; Liu, C.; and Lensch, H. 2021 a . Nerd: Neural reflectance decomposition from image collections. In ICCV, 12684--12694

    work page 2021

  5. [5]

    Boss, M.; Jampani, V.; Braun, R.; Liu, C.; Barron, J.; and Lensch, H. 2021 b . Neural-pil: Neural pre-integrated lighting for reflectance decomposition. Advances in Neural Information Processing Systems, 34: 10691--10704

    work page 2021

  6. [6]

    R.; Monteiro, M.; Kellnhofer, P.; Wu, J.; and Wetzstein, G

    Chan, E. R.; Monteiro, M.; Kellnhofer, P.; Wu, J.; and Wetzstein, G. 2021. pi-gan: Periodic implicit generative adversarial networks for 3d-aware image synthesis. In CVPR, 5799--5809

    work page 2021

  7. [7]

    Chen, A.; Xu, Z.; Geiger, A.; Yu, J.; and Su, H. 2022. TensoRF: Tensorial Radiance Fields. In ECCV, 333--350

    work page 2022

  8. [8]

    Chen, G.; Han, K.; and Wong, K.-Y. K. 2018. PS-FCN: A flexible learning framework for photometric stereo. In ECCV, 3--18

    work page 2018

Show all 34 references
  1. [9]

    d.; Courville, A.; and Bengio, Y

    Dumoulin, V.; Perez, E.; Schucher, N.; Strub, F.; Vries, H. d.; Courville, A.; and Bengio, Y. 2018. Feature-wise transformations. Distill, 3(7): e11

    2018

  2. [10]

    R.; Recht, B.; and Kanazawa, A

    Fridovich-Keil, S.; Meanti, G.; Warburg, F. R.; Recht, B.; and Kanazawa, A. 2023. K-planes: Explicit radiance fields in space, time, and appearance. In CVPR, 12479--12488

    2023

  3. [11]

    Fridovich-Keil, S.; Yu, A.; Tancik, M.; Chen, Q.; Recht, B.; and Kanazawa, A. 2022. Plenoxels: Radiance Fields Without Neural Networks. In CVPR, 5501--5510

    2022

  4. [12]

    Gardner, M.-A.; Hold-Geoffroy, Y.; Sunkavalli, K.; Gagn \'e , C.; and Lalonde, J.-F. 2019. Deep parametric indoor lighting estimation. In ICCV, 7175--7183

    2019

  5. [13]

    Jin, H.; Liu, I.; Xu, P.; Zhang, X.; Han, S.; Bi, S.; Zhou, X.; Xu, Z.; and Su, H. 2023. TensoIR: Tensorial Inverse Rendering. In CVPR

    2023

  6. [14]

    Kerbl, B.; Kopanas, G.; Leimk \"u hler, T.; and Drettakis, G. 2023. 3D Gaussian Splatting for Real-Time Radiance Field Rendering. ACM Transactions on Graphics, 42(4)

    2023

  7. [15]

    Kulhanek, J.; and Sattler, T. 2023. Tetra-nerf: Representing neural radiance fields using tetrahedra. In ICCV, 18458--18469

    2023

  8. [16]

    Kwon, W.; Sung, J.; Jeon, M.; Eom, C.; and Oh, J. 2025. R3eVision: A Survey on Robust Rendering, Restoration, and Enhancement for 3D Low-Level Vision. arXiv preprint arXiv:2506.16262

    2025 arXiv

  9. [17]

    Li, Z.; Shafiei, M.; Ramamoorthi, R.; Sunkavalli, K.; and Chandraker, M. 2020. Inverse rendering for complex indoor scenes: Shape, spatially-varying lighting and svbrdf from a single image. In CVPR, 2475--2484

    2020

  10. [18]

    Liu, Y.; Wang, P.; Lin, C.; Long, X.; Wang, J.; Liu, L.; Komura, T.; and Wang, W. 2023. Nero: Neural geometry and brdf reconstruction of reflective objects from multiview images. ACM Transactions on Graphics (ToG), 42(4): 1--22

    2023

  11. [19]

    S.; Barron, J

    Martin-Brualla, R.; Radwan, N.; Sajjadi, M. S.; Barron, J. T.; Dosovitskiy, A.; and Duckworth, D. 2021. Nerf in the wild: Neural radiance fields for unconstrained photo collections. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 7210--7219

    2021

  12. [20]

    P.; Ortiz-Cayon, R.; Kalantari, N

    Mildenhall, B.; Srinivasan, P. P.; Ortiz-Cayon, R.; Kalantari, N. K.; Ramamoorthi, R.; Ng, R.; and Kar, A. 2019. Local Light Field Fusion: Practical View Synthesis with Prescriptive Sampling Guidelines. ACM Transactions on Graphics, 38(4): 1--14

    2019

  13. [21]

    P.; Tancik, M.; Barron, J

    Mildenhall, B.; Srinivasan, P. P.; Tancik, M.; Barron, J. T.; Ramamoorthi, R.; and Ng, R. 2020. NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis. In ECCV, 405--421

    2020

  14. [22]

    M\"uller, T.; Evans, A.; Schied, C.; and Keller, A. 2022. Instant Neural Graphics Primitives with a Multiresolution Hash Encoding. ACM Transactions on Graphics, 41(4): 102:1--102:15

    2022

  15. [23]

    Munkberg, J.; Hasselgren, J.; Shen, T.; Gao, J.; Chen, W.; Evans, A.; M \"u ller, T.; and Fidler, S. 2022. Extracting Triangular 3D Models, Materials, and Lighting From Images. In CVPR, 8280--8290

    2022

  16. [24]

    Perez, E.; Strub, F.; De Vries, H.; Dumoulin, V.; and Courville, A. 2018. Film: Visual reasoning with a general conditioning layer. In AAAI, volume 32

    2018

  17. [25]

    L.; Zheng, E.; Pollefeys, M.; and Frahm, J.-M

    Sch\" o nberger, J. L.; Zheng, E.; Pollefeys, M.; and Frahm, J.-M. 2016. Pixelwise View Selection for Unstructured Multi-View Stereo. In ECCV, 501--518

    2016

  18. [26]

    Sitzmann, V.; Zollh \"o fer, M.; and Wetzstein, G. 2019. Scene representation networks: Continuous 3 D -structure-aware neural scene representations. In Proc. Advances in Neural Information Processing Systems, 1121--1132

    2019

  19. [27]

    Sun, C.; Sun, M.; and Chen, H.-T. 2022. Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction. In CVPR, 5459--5469

    2022

  20. [28]

    T.; and Srinivasan, P

    Verbin, D.; Hedman, P.; Mildenhall, B.; Zickler, T.; Barron, J. T.; and Srinivasan, P. P. 2022. Ref-nerf: Structured view-dependent appearance for neural radiance fields. In CVPR, 5481--5490

    2022

  21. [29]

    Wizadwongsa, S.; Phongthawee, P.; Yenphraphai, J.; and Suwajanakorn, S. 2021. Nex: Real-time view synthesis with neural basis expansion. In CVPR, 8534--8543

    2021

  22. [30]

    Xing, W.; Chen, J.; Yang, Z.; Wang, Q.; and Guo, Y. 2022. Scale-Consistent Fusion: from Heterogeneous Local Sampling to Global Immersive Rendering. IEEE Transactions on Image Processing

    2022

  23. [31]

    Xu, B.; Wang, N.; Chen, T.; and Li, M. 2015. Empirical evaluation of rectified activations in convolutional network. arXiv preprint arXiv:1505.00853

    2015 arXiv

  24. [32]

    Xu, Q.; Xu, Z.; Philip, J.; Bi, S.; Shu, Z.; Sunkavalli, K.; and Neumann, U. 2022. Point-nerf: Point-based neural radiance fields. In CVPR, 5438--5448

    2022

  25. [33]

    Yu, A.; Li, R.; Tancik, M.; Li, H.; Ng, R.; and Kanazawa, A. 2021. Plenoctrees for real-time rendering of neural radiance fields. In ICCV, 5752--5761

    2021

  26. [34]

    Zhang, K.; Luan, F.; Wang, Q.; Bala, K.; and Snavely, N. 2021. Physg: Inverse rendering with spherical gaussians for physics-based material editing and relighting. In CVPR, 5453--5462

    2021

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.