Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

DWTNeRF: Boosting Few-shot Neural Radiance Fields via Discrete Wavelet Transform

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Discrete-wavelet loss upweighting low frequencies, paired with attention-based cross-branch interactions, materially improves few-shot neural radiance fields on hash-encoding backbones.

desk verdict Solid incremental few-shot NeRF paper whose headline DW-loss mechanism is confounded with patch sampling and extra rays—worth refereeing, but with a required control. read the letter →

arxiv 2501.12637 v3 pith:PONTFKNN submitted 2025-01-22 cs.CV

Hung Nguyen , Blark Runfa Li , Truong Nguyen This is my paper · ORCID
classification cs.CV
keywords discretewavelettransformfew-shotneuralradiancefieldsInstant-NGPhashencodingmulti-headattentioncross-branchinteractionsfrequencyregularizationnovelviewsynthesis
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

DWTNeRF sets out to make few-shot novel view synthesis work on the fast Instant-NGP-style hash-encoding backbone, where common few-shot tricks fail because the multi-resolution encoding neither maps cleanly to frequencies nor tolerates architectural changes. The key idea is a Discrete Wavelet loss that decomposes rendered and ground-truth views into LL, LH, HL and HH sub-bands and upweights the low-frequency LL term so early training focuses on structure before detail. On the 3-shot LLFF benchmark the paper reports 15.07% higher PSNR, 24.45% higher SSIM and 36.30% lower LPIPS than vanilla INGP, and a modest edge over the CombiNeRF baseline (20.38 vs 20.12 PSNR) that carries over to the 4-shot NeRF-Synthetic setting. It also argues that attention-based cross-branch interactions are the model-based modification that does not degrade INGP.

What carries the argument

The load-bearing object is the Discrete Wavelet (DW) loss, $L_{DW} = \sum_{sb} \lambda_{sb}\| I_{sb} - I^{gt}_{sb}\|_2^2$, computed on $192\times192$ rendered patches with Haar or Daubechies wavelets, with $\lambda_{LL}$ set higher than the other sub-band weights (0.4 vs 0.2 on LLFF). Because the DWT needs a 2D image, the loss forces patch-based ray sampling; it is applied lazily every 10 (LLFF) or 150 (NS) iterations and stopped at iteration 5K to keep the cost down. The second module is multi-head attention inserted at two points: on the concatenation of hash and view encodings before the color MLP, and again on the concatenated density/color outputs. It does this without adding MLP layers, which the paper argues is why it survives INGP's sensitivity to architectural change.

What would settle it

Run the CombiNeRF baseline with patch-based ray sampling and the standard MSE loss (same patch size, same lazy schedule, no DWT) on 3-shot LLFF. If PSNR reaches or exceeds DWTNeRF's 20.25, the wavelet decomposition is not the active ingredient driving the reported gain.

Watch

Extended reading notes

Core claim

The central claim is that frequency-domain supervision can substitute for frequency masking on hash-encoding radiance fields, and that color-density interactions can be learned without touching the MLP architecture. DWTNeRF realizes this by taking the two-dimensional discrete wavelet transform of rendered and ground-truth patches, weighting low-frequency sub-bands more heavily in the training objective, and passing hybrid position/view encodings plus concatenated density-color outputs through multi-head attention. In the paper's experiments this yields results that are competitive with state-of-the-art few-shot methods on 2- to 4-shot LLFF and 4-shot NeRF-Synthetic, with the largest gains in the most extreme few-shot regime and a clear margin over Vanilla INGP.

Load-bearing premise

The paper never compares its wavelet loss against the ordinary pixel-difference loss computed on the same image patches, so the reported improvement could come from patch-based ray sampling itself rather than from the wavelet decomposition or the low-frequency weighting.

Editorial extensions

If this is right

  • If the central claim holds, few-shot INGP can be improved by an image-space frequency-prioritizing loss that replaces the positional-encoding masking used by FreeNeRF.
  • The gains are concentrated at 2–4 training views and largely fade at 6–9 views, so the method targets extreme few-shot capture rather than general view synthesis.
  • Removing frequency masking from CombiNeRF already improves PSNR from 20.12 to 20.19, so part of the benefit comes from undoing an INGP-incompatible regularizer rather than from the wavelet loss alone.
  • Because the attention modules sit outside the MLP, the framework could be transplanted onto other hash-encoding renderers, and the paper suggests 3DGS and SDF representations as future targets.

Reading between the lines

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

  • An untested but decisive check: run CombiNeRF with patch-based ray sampling and the ordinary MSE loss under the same lazy schedule; if PSNR reaches 20.25, the reported DW-loss gain is a sampling effect, not a frequency-prioritization effect.
  • The fact that Haar, db2 and db3 give nearly identical results suggests the specific frequency decomposition is not the active ingredient; a spatially structured loss over patches may be what helps.
  • The attention modules only handle input and output interactions; lightweight alternatives such as cross-attention or small graph layers might achieve the same INGP-compatible gains.
  • If low-frequency emphasis reduces hallucinated geometry, a similar wavelet-based supervision on depth or density maps could extend the benefit beyond color.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes DWTNeRF, a few-shot NeRF method built on Instant-NGP and CombiNeRF. Its two contributions are (1) a Discrete Wavelet (DW) loss that decomposes rendered and ground-truth patches into sub-bands and weights low frequencies more heavily, applied with patch-based ray sampling, and (2) multi-head attention modules at the input and output levels of the color/density branches to model cross-branch interactions without modifying the MLP architecture. The method is evaluated on LLFF (2/3/6/9 views) and NeRF-Synthetic (4 views), reporting consistent but small improvements over CombiNeRF, and ablations are provided for the DW loss, attention modules, and wavelet choice.

Significance. If the improvements are attributable to the proposed components, the paper would make a useful contribution to few-shot INGP by showing that frequency-domain supervision can be adapted to hash-based encodings and that attention-based cross-branch interactions avoid the fragility of architectural changes. The paper also provides a solid negative result that FreeNeRF-style frequency masking and mi-MLP-style modifications degrade or fail to improve INGP. However, the central claim that the DW loss drives the improvement is not cleanly established because the +DW ablation changes patch sampling and ray budget simultaneously, and the reported gains are small with no repeated-run statistics. The significance is therefore conditional on additional controlled experiments.

major comments (3)
  1. [Table 3 and Section 4.2 (DW loss, patch-based rendering)] The ablation isolating the DW loss is confounded. Comparing 'CombiNeRF (-freq mask.)' at 20.19 PSNR with 'DWTNeRF (+DW.)' at 20.25 changes three things at once: the +DW configuration renders an extra 192x192 patch (36,864 rays) every 10 iterations until iteration 5K, those rays are adjacent rather than random, and the loss applied to the patch is a weighted wavelet MSE. Because the Haar/Daubechies DWT is orthogonal, the DW loss is mathematically equivalent to ordinary MSE with frequency-dependent subband weights; hence the only component unique to the DW loss is the subband weighting. Without a control that applies plain MSE to the same patches with the same extra-ray schedule, the 0.06 dB gain in Table 3, and the larger gaps over CombiNeRF in Table 1, can be explained by patch-based sampling or simply by the optimizer receiving roughly 45% more gradients. This control must be added before the paper's title-level claim about the wavelet loss can be accepted.
  2. [Tables 1-5 and Section 5.3-5.4] No repeated runs, seeds, or error bars are reported, yet the headline improvements are very small: e.g., 3-shot LLFF PSNR improves from 20.12 to 20.38, and 4-shot NeRF-Synthetic PSNR from 19.15 to 19.25. Given that the ablation rows in Table 3 differ by 0.01-0.06 dB PSNR and 0.001-0.007 SSIM, single-run results cannot establish that these differences are not optimization noise. The authors should report mean and standard deviation over at least three seeds for the main tables and the key ablations, or otherwise justify why single runs are sufficient.
  3. [Tables 1, 3, 5 and Section 5.1 (hyperparameters)] The DW loss depends on several per-dataset hyperparameters (lambda_LL=0.4 vs 0.04, other subband weights 0.2 vs 0.02, K_DW=10 vs 150, T_DW=5K, patch size 192, and 2 vs 1 attention heads), and no sensitivity analysis is provided. Since the final result is only 0.1-0.3 dB above the baseline, it is unclear whether the reported configuration is a robust operating point or a carefully selected one. A sensitivity study over the subband weights and the loss schedule would substantially strengthen the empirical claim.
minor comments (5)
  1. [Section 5.4] The text contains a typo: 'cross-banch interactions' should be 'cross-branch interactions'.
  2. [Figure 7 caption] The caption says 'much noiser' and 'noiser' should be 'noisier'.
  3. [Table 1 venue column] The venue column formats some entries inconsistently (e.g., 'ARXIV‘22', 'ICML ‘23', 'CVPR ‘23'); these should be normalized.
  4. [Section 4.2] The sentence 'This is a concern in applying the DW loss, which we will explain later' is vague; the concern about sub-band size could be stated explicitly at that point.
  5. [Section 5.4 and Table 5] The wavelet comparison reports differences of 0.01-0.02 dB PSNR without any variance information; the conclusion that the Haar wavelet is sufficient would be more convincing with repeated runs.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the DW loss and attention modules are defined directly on rendered and ground-truth images and evaluated against external baselines; the patch-sampling confound is an experimental-design concern, not a circular derivation.

full rationale

The DWTNeRF paper's central claim—that a wavelet-domain loss with low-frequency subband weighting plus attention-based cross-branch interactions improves few-shot INGP—is derived empirically and is not circular. The DW loss (Section 4.2) is defined directly as a weighted sum of squared differences between rendered and ground-truth wavelet subbands; no parameter is fitted to the target metric, and no prediction is constructed from the same data it claims to predict. The base method CombiNeRF and the auxiliary regularization losses (distortion, full geometry, depth smoothness, KL divergence) are external prior work, not self-citations. The ablation in Table 3 compares DWTNeRF variants against the CombiNeRF baseline and shows incremental gains; although the +DW row changes patch sampling and ray budget simultaneously with the wavelet weighting, that is an experimental confound affecting attribution of the gain, not a case where the result is equivalent to its input by construction. No uniqueness theorem or ansatz is imported from the authors' own prior work, and the method is self-contained against external benchmarks (LLFF, NeRF-Synthetic) and external baselines. The paper's own limitation statement concedes that the standard photometric MSE remains in the objective, making the DW loss an auxiliary reweighting rather than a fully validated mechanism; however, this is a question of experimental isolation and strength of evidence, not circularity. Therefore no circular step is exhibited and the score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several unproven modeling choices: the usefulness of a weighted DWT L2 loss as a training signal, the incompatibility of frequency masking with hash encodings, the safety of external attention for fast-converging INGP, and the continued validity of CombiNeRF's regularization losses. Each is plausible and partially supported by ablations, but none is derived from first principles. The DW loss hyperparameters (sub-band weights, schedule, patch size, attention heads) are free parameters selected per benchmark without a held-out validation split.

free parameters (6)
  • DW sub-band weights for LLFF (lambda_LL, lambda_LH, lambda_HL, lambda_HH) = 0.4, 0.2, 0.2, 0.2
    Chosen per benchmark to prioritize low frequencies; no held-out validation.
  • DW sub-band weights for NeRF-Synthetic = 0.04, 0.02, 0.02, 0.02
    Scaled differently for the synthetic dataset; selected without a separate validation split.
  • DW loss interval K_DW = 10 (LLFF), 150 (NS)
    Controls how often the patch-based DWT loss is computed; tuned to balance overhead and effect.
  • DW loss stopping iteration T_DW = 5000 (both)
    Stops the low-frequency emphasis at half of training; a temporal schedule hyperparameter.
  • Number of attention heads = 2 (LLFF), 1 (NS)
    Chosen per dataset; no analysis of sensitivity.
  • Patch size for DW loss = 192x192 (36864 rays)
    Required to make wavelet sub-bands visually meaningful; adds a substantial second ray batch.
assumptions (4)
  • domain assumption Minimizing a weighted L2 loss between DWT sub-bands of rendered and ground-truth views improves final render quality in the few-shot regime.
    This is the core premise of the DW loss; the paper motivates it via overfitting of high frequencies but provides no theoretical justification or held-out validation.
  • domain assumption INGP's multi-resolution hash encodings do not map monotonically to spatial frequencies, so frequency masking (FreeNeRF) is inappropriate for INGP.
    Supported by the ablation in Table 3 (removing frequency mask improves PSNR from 20.12 to 20.19), but demonstrated only on 3-shot LLFF.
  • domain assumption Multi-head attention applied to the input and output tensors captures density-color interactions without triggering INGP's sensitivity to MLP architectural changes.
    Supported by ablations (Table 3), but the mechanism is not analyzed and the SSIM slightly degrades with attention alone (0.676 to 0.670).
  • domain assumption The regularization losses from CombiNeRF (distortion, full geometry, depth smoothness, KL divergence) remain beneficial when adding the DW loss and attention.
    Assumed, not ablated; all CombiNeRF losses are kept fixed in every DWTNeRF configuration.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DWTNeRF: Boosting Few-shot Neural Radiance Fields via Discrete Wavelet Transform." pith.science (2026). https://pith.science/paper/PONTFKNN

@misc{pith2026250112637,
  author       = {Pith},
  title        = {Pith review of: DWTNeRF: Boosting Few-shot Neural Radiance Fields via Discrete Wavelet Transform},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PONTFKNN}},
  note         = {Machine review of arXiv:2501.12637}
}
read the original abstract

Neural Radiance Fields (NeRF) has achieved superior performance in novel view synthesis and 3D scene representation, but its practical applications are hindered by slow convergence and reliance on dense training views. To this end, we present DWTNeRF, a unified framework based on Instant-NGP's fast-training hash encoding. It is coupled with regularization terms designed for few-shot NeRF, which operates on sparse training views. Our DWTNeRF additionally includes a novel Discrete Wavelet loss that allows explicit prioritization of low frequencies directly in the training objective, reducing few-shot NeRF's overfitting on high frequencies in earlier training stages. We also introduce a model-based approach, based on multi-head attention, that is compatible with INGP, which are sensitive to architectural changes. On the 3-shot LLFF benchmark, DWTNeRF outperforms Vanilla INGP by 15.07% in PSNR, 24.45% in SSIM and 36.30% in LPIPS. Our approach encourages a re-thinking of current few-shot approaches for fast-converging implicit representations like INGP or 3DGS.

Figures

Figures reproduced from arXiv: 2501.12637 by the authors.

Figure 1
Figure 1. Comparison of frequency masking as applied to sine [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Architecture diagram of DWTNeRF. Firstly, we encode 3D positions with INGP’s [ [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Random (left) and patch-based ray sampling (right). [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Qualitative results for Vanilla INGP (a), CombiNeRF [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results for CombiNeRF (a), DWTNeRF with [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: mi-MLP’s [46] model-based methods: “Residual Con￾nections” (a) and “Element-wise Cross-branch Interactions” (b) 7 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Decomposition using the Haar (a & b) and db2 (c). We [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Frequency-based scene decompositions with different wavelets. From top to bottom: the LL, LH, HL and HH sub-bands. The first [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Qualitative results for Vanilla INGP (a), CombiNeRF (b), DWTNeRF with only cross-branch interactions (c), DWTNeRF with full [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: Qualitative results for Vanilla INGP (a), CombiNeRF (b), DWTNeRF with only cross-branch interactions (c), DWTNeRF [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Speech2Grasp: Data-Efficient Transfer of Text-Conditioned Grasp Detection to Speech in Humanoid Robots

    cs.RO 2026-07 conditional novelty 4.0 of 10

    A lightweight MLP projector transfers ALBEF-based text-conditioned grasp detection to speech, matching text-only performance and beating an ASR cascade on a humanoid robot.

  2. From Coarse to Fine: Learnable Discrete Wavelet Transforms for Efficient 3D Gaussian Splatting

    cs.CV 2025-06 conditional novelty 4.0 of 10

    AutoOpti3DGS uses learnable discrete wavelet transforms on input images to train 3DGS from coarse to fine, reducing peak Gaussian counts by about 18 to 23 percent with modest quality trade-offs.

Reference graph

Works this paper leans on

51 extracted references · 43 canonical work pages · cited by 2 Pith papers

  1. [1]

    Barron, Ben Mildenhall, Dor Verbin, Pratul P

    Jonathan T. Barron, Ben Mildenhall, Dor Verbin, Pratul P. Srinivasan, and Peter Hedman. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. In 2022 IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR), pages 5460–5469, 2022. 2, 4

    work page 2022

  2. [2]

    CombiNeRF: A Combination of Regularization Techniques for Few-Shot Neural Radiance Field View Synthesis

    Matteo Bonotto, Luigi Sarrocco, Daniele Evangelista, Marco Imperoli, and Alberto Pretto. CombiNeRF: A Combination of Regularization Techniques for Few-Shot Neural Radiance Field View Synthesis. In International Conference on 3D Vision (3DV), 2024. 2, 3, 4, 6, 1

    work page 2024

  3. [3]

    Chan and Wei Zhu

    T. Chan and Wei Zhu. Level set based shape prior segmen- tation. In 2005 IEEE Computer Society Conference on Com- puter Vision and Pattern Recognition (CVPR’05), pages 1164– 1170 vol. 2, 2005. 8

    work page 2005

  4. [4]

    Cunerf: Cube-based neural radiance field for zero- shot medical image arbitrary-scale super resolution

    Zixuan Chen, Lingxiao Yang, Jian-Huang Lai, and Xiao- hua Xie. Cunerf: Cube-based neural radiance field for zero- shot medical image arbitrary-scale super resolution. In 2023 IEEE/CVF International Conference on Computer Vision (ICCV), pages 21128–21138, 2023. 1

    work page 2023

  5. [5]

    Abril Corona-Figueroa, Jonathan Frawley, Sam Bond Taylor, Sarath Bethapudi, Hubert P. H. Shum, and Chris G. Willcocks. Mednerf: Medical neural radiance fields for reconstructing 3d-aware ct-projections from a single x-ray. In 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), pages 3843–3848,

    work page 2022

  6. [6]

    Ten Lectures on Wavelets

    Ingrid Daubechies. Ten Lectures on Wavelets. Society for Industrial and Applied Mathematics, 1992. 8, 1

    work page 1992

  7. [7]

    Nerf: Neural radiance field in 3d vision, a comprehensive review, 2023

    Kyle Gao, Yina Gao, Hongjie He, Dening Lu, Linlin Xu, and Jonathan Li. Nerf: Neural radiance field in 3d vision, a comprehensive review, 2023. 2

    work page 2023

  8. [8]

    Sparsenerf: Distilling depth ranking for few-shot novel view synthesis

    Guangcong, Zhaoxi Chen, Chen Change Loy, and Ziwei Liu. Sparsenerf: Distilling depth ranking for few-shot novel view synthesis. IEEE/CVF International Conference on Computer Vision (ICCV), 2023. 2 8

    work page 2023

Show all 51 references
  1. [9]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016. 2

    2016

  2. [10]

    Weinberger

    Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kil- ian Q. Weinberger. Densely connected convolutional net- works. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 2261–2269, 2017. 2

    2017

  3. [11]

    Putting nerf on a diet: Semantically consistent few-shot view synthesis

    Ajay Jain, Matthew Tancik, and Pieter Abbeel. Putting nerf on a diet: Semantically consistent few-shot view synthesis. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 5885–5894, 2021. 2, 6

    2021

  4. [12]

    Cg-nerf: Conditional generative neu- ral radiance fields for 3d-aware image synthesis

    Kyungmin Jo, Gyumin Shim, Sanghun Jung, Soyoung Yang, and Jaegul Choo. Cg-nerf: Conditional generative neu- ral radiance fields for 3d-aware image synthesis. In 2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 724–733, 2023. 2

    2023

  5. [13]

    Kajiya and Brian V on Herzen

    James T. Kajiya and Brian V on Herzen. Ray tracing volume densities. Proceedings of the 11th annual conference on Computer graphics and interactive techniques, 1984. 3

    1984

  6. [14]

    Analyzing and improving the image quality of stylegan

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and improving the image quality of stylegan. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 8107–8116, 2020. 4

    2020

  7. [15]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 42(4), 2023. 8

    2023

  8. [16]

    Infonerf: Ray entropy minimization for few-shot neural volume render- ing

    Mijeong Kim, Seonguk Seo, and Bohyung Han. Infonerf: Ray entropy minimization for few-shot neural volume render- ing. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12902–12911, 2022. 2, 4, 6

    2022

  9. [17]

    Geconerf: Few-shot neural radiance fields via geometric consistency

    Minseop Kwak, Jiuhn Song, and Seungryong Kim. Geconerf: Few-shot neural radiance fields via geometric consistency. arXiv preprint arXiv:2301.10941, 2023. 2, 6

    2023 arXiv

  10. [18]

    Splatsdf: Boosting neural implicit sdf via gaussian splatting fusion, 2024

    Runfa Blark Li, Keito Suzuki, Bang Du, Ki Myung Brian Le, Nikolay Atanasov, and Truong Nguyen. Splatsdf: Boosting neural implicit sdf via gaussian splatting fusion, 2024. 8

    2024

  11. [19]

    Are nerfs ready for autonomous driving? towards closing the real-to-simulation gap

    Carl Lindstr¨om, Georg Hess, Adam Lilja, Maryam Fatemi, Lars Hammarstrand, Christoffer Petersson, and Lennart Svensson. Are nerfs ready for autonomous driving? towards closing the real-to-simulation gap. In 2024 IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition ...

    2024

  12. [20]

    Srinivasan, Rodrigo Ortiz-Cayon, Nima Khademi Kalantari, Ravi Ramamoorthi, Ren Ng, and Abhishek Kar

    Ben Mildenhall, Pratul P. Srinivasan, Rodrigo Ortiz-Cayon, Nima Khademi Kalantari, Ravi Ramamoorthi, Ren Ng, and Abhishek Kar. Local light field fusion: Practical view synthe- sis with prescriptive sampling guidelines. ACM Transactions on Graphics (TOG), 2019. 6, 7, 4

    2019

  13. [21]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthe- sis. In ECCV, 2020. 1, 2, 6, 7, 3

    2020

  14. [22]

    Instant neural graphics primitives with a multires- olution hash encoding

    Thomas M¨uller, Alex Evans, Christoph Schied, and Alexan- der Keller. Instant neural graphics primitives with a multires- olution hash encoding. ACM Trans. Graph., 41(4):102:1– 102:15, 2022. 1, 2, 4

    2022

  15. [23]

    Barron, Ben Mildenhall, Mehdi S

    Michael Niemeyer, Jonathan T. Barron, Ben Mildenhall, Mehdi S. M. Sajjadi, Andreas Geiger, and Noha Radwan. Regnerf: Regularizing neural radiance fields for view syn- thesis from sparse inputs. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), page...

    2022

  16. [24]

    Physically Based Rendering: From Theory to Implementation

    Matt Pharr, Wenzel Jakob, and Greg Humphreys. Physically Based Rendering: From Theory to Implementation. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 3rd edition, 2016. 3

    2016

  17. [25]

    D-NeRF: Neural Radiance Fields for Dynamic Scenes

    Albert Pumarola, Enric Corona, Gerard Pons-Moll, and Francesc Moreno-Noguer. D-NeRF: Neural Radiance Fields for Dynamic Scenes. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, 2020. 2

    2020

  18. [26]

    Hamprecht, Yoshua Bengio, and Aaron C

    Nasim Rahaman, Aristide Baratin, Devansh Arpit, Felix Draxler, Min Lin, Fred A. Hamprecht, Yoshua Bengio, and Aaron C. Courville. On the spectral bias of neural networks. In ICML, pages 5301–5310. PMLR, 2019. 1

    2019

  19. [27]

    Modeling illumination variation with spherical harmonics

    Ravi Ramamoorthi. Modeling illumination variation with spherical harmonics. 2005. 4, 1

    2005

  20. [28]

    Plenoxels: Radiance fields without neural networks

    Sara Fridovich-Keil and Alex Yu, Matthew Tancik, Qinhong Chen, Benjamin Recht, and Angjoo Kanazawa. Plenoxels: Radiance fields without neural networks. In CVPR, 2022. 1

    2022

  21. [29]

    Flipnerf: Flipped reflection rays for few-shot novel view synthesis

    Seunghyeon Seo, Yeonjin Chang, and Nojun Kwak. Flipnerf: Flipped reflection rays for few-shot novel view synthesis. In 2023 IEEE/CVF International Conference on Computer Vision (ICCV), pages 22826–22836, 2023. 1

    2023

  22. [30]

    Wavelets and filter banks

    Gilbert Strang and Truong Nguyen. Wavelets and filter banks. SIAM, 1996. 3, 1

    1996

  23. [31]

    Torch-ngp: a pytorch implementation of instant-ngp, 2022

    Jiaxiang Tang. Torch-ngp: a pytorch implementation of instant-ngp, 2022. https://github.com/ashawkey/torch-ngp. 6

    2022

  24. [32]

    Neurad: Neural rendering for autonomous driving

    Adam Tonderski, Carl Lindstr ¨om, Georg Hess, William Ljungbergh, Lennart Svensson, and Christoffer Petersson. Neurad: Neural rendering for autonomous driving. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14895–14904, 2024. 1

    2024

  25. [33]

    Sparf: Neural radiance fields from sparse and noisy poses

    Prune Truong, Marie-Julie Rakotosaona, Fabian Manhardt, and Federico Tombari. Sparf: Neural radiance fields from sparse and noisy poses. IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR, 2023. 2

    2023

  26. [34]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in neural information processing systems, pages 5998–6008, 2017. 4, 5

    2017

  27. [35]

    Bovik, H.R

    Zhou Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing, 13(4): 600–612, 2004. 6

    2004

  28. [36]

    Srinivasan, Jonathan T

    Chung-Yi Weng, Brian Curless, Pratul P. Srinivasan, Jonathan T. Barron, and Ira Kemelmacher-Shlizerman. Hu- manNeRF: Free-viewpoint rendering of moving people from monocular video. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR), p...

    2022

  29. [37]

    Diffusionerf: Regularizing neural radiance fields with denoising diffusion models

    Jamie Wynn and Daniyar Turmukhambetov. Diffusionerf: Regularizing neural radiance fields with denoising diffusion models. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4180–4189, 2023. 2, 4, 6

    2023

  30. [38]

    VR-NeRF: High-fidelity virtualized walkable spaces

    Linning Xu, Vasu Agrawal, William Laney, Tony Garcia, Aayush Bansal, Changil Kim, Samuel Rota Bul `o, Lorenzo Porzi, Peter Kontschieder, Aljaˇz Boˇziˇc, Dahua Lin, Michael Zollh¨ofer, and Christian Richardt. VR-NeRF: High-fidelity virtualized walkable spaces. In SIGGRAPH Asia ...

    2023

  31. [39]

    Freenerf: Im- proving few-shot neural rendering with free frequency reg- ularization

    Jiawei Yang, Marco Pavone, and Yue Wang. Freenerf: Im- proving few-shot neural rendering with free frequency reg- ularization. In Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2023. 1, 2, 6

    2023

  32. [40]

    Featurenerf: Learning generalizable nerfs by distilling foundation models

    Jianglong Ye, Naiyan Wang, and Xiaolong Wang. Featurenerf: Learning generalizable nerfs by distilling foundation models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 8962–8973, 2023. 2

    2023

  33. [41]

    PlenOctrees for real-time rendering of neural radiance fields

    Alex Yu, Ruilong Li, Matthew Tancik, Hao Li, Ren Ng, and Angjoo Kanazawa. PlenOctrees for real-time rendering of neural radiance fields. In ICCV, 2021. 2

    2021

  34. [42]

    pixelNeRF: Neural radiance fields from one or few images

    Alex Yu, Vickie Ye, Matthew Tancik, and Angjoo Kanazawa. pixelNeRF: Neural radiance fields from one or few images. In CVPR, 2021. 2

    2021

  35. [43]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, 2018. 6

    2018

  36. [44]

    Nerf in the palm of your hand: Correc- tive augmentation for robotics via novel-view synthesis

    Allan Zhou, Moo Jin Kim, Lirui Wang, Pete Florence, and Chelsea Finn. Nerf in the palm of your hand: Correc- tive augmentation for robotics via novel-view synthesis. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 17907–17917, 2023. 1

    2023

  37. [45]

    Multimodal neural radiance field

    Haidong Zhu, Yuyin Sun, Chi Liu, Lu Xia, Jiajia Luo, Nan Qiao, Ram Nevatia, and Cheng–Hao Kuo. Multimodal neural radiance field. In 2023 IEEE International Conference on Robotics and Automation (ICRA), pages 9393–9399, 2023. 1

    2023

  38. [46]

    Hanxin Zhu, Tianyu He, Xin Li, Bingchen Li, and Zhibo Chen. Is vanilla mlp in neural radiance field enough for few-shot view synthesis? In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 20288–20298, 2024. 1, 2, 6, 7 10 DWTNeRF: ...

    2024

  39. [47]

    More Visualizations We provide more qualitative results on the 3-shot LLFF and 4-shot NeRF-Synthetic benchmarks at Figures 10 and

  40. [48]

    hallucinated

    There are 6 scenes in total. In each scene, we would like to direct attention towards the regions enclosed by the red box, which highlights the effects of cross-branch interactions (b → c) and DW loss (c → d). They provide finer details, preserve better structures and reduce r...

  41. [49]

    Alternating Flip

    Wavelet Coefficients Table 6. Low-pass filters of the Daubechies [6] wavelets ORDER 1 2 3 ℓk 1 1 + √ 3 0.3327 1 3 + √ 3 0.8069 3 − √ 3 0.4599 1 − √ 3 -0.1350 -0.0854 0.0352 FACTOR √ 2 4 √ 2 1 Table 6 shows the coefficients ℓk of the 1D low-pass filters ℓ that are used in our e...

  42. [50]

    Residual Connections

    Model-based methods In this section, we compared DWTNeRF’s cross-branch interactions against model-based methods as introduced by mi-MLP [ 46]: “Residual Connections” and “Element- wise Cross-branch Interactions”. The architectures of each method are provided in Figure 6. In g...

  43. [51]

    chair” and “hotdog

    with 16 resolution levels. The color branch has 3 MLP layers, and its inputs are encoded using 64 bases of the spherical harmonics [27]. Residual Connections . This provides a shorter path between the inputs and intermediate MLP layers, helping to reduce overfitting. For our i...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.