Pith. sign in

REVIEW 3 major objections 5 minor 44 references

Implicit representations are dead. Long live explicit primitives!

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Explicit Gaussian primitives match or beat implicit neural representations for medical images, with far less training time and memory.

desk verdict Useful position/benchmark for Gaussian primitives in medical imaging, but the headline comparison is gated by an early-stopping rule that flatters fast methods; worth refereeing with a convergence-fair retest. read the letter →

arxiv 2608.10001 v1 pith:JWDVO6U5 submitted 2026-08-07 eess.IV cs.CV

Nil Stolt-Ansó , Maik Dannecker , Wenqi Huang , Andras Jakab , Daniel Rueckert This is my paper · ORCID
classification eess.IVcs.CV
keywords explicitrepresentationsGaussianprimitivesimplicitneuralmedicalimagereconstructioncompressionbasisexpansionspartitionofunity
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that explicit Gaussian primitives should replace implicit neural representations (coordinate MLPs) as the default way to represent medical images as continuous functions. It unifies dictionaries, implicit networks, feature grids, and Gaussian primitives under a single basis-expansion formula, then catalogues mathematical properties unique to explicit primitives. The benchmark on 2D histology and 3D lung CT shows Gaussian representations match or exceed the reconstruction quality of implicit methods at all tested compression factors while optimizing 5–100x faster and using 5–200x less memory. If this holds, medical imaging pipelines could become faster and cheaper without sacrificing reconstruction fidelity.

What carries the argument

The central object is the Gaussian primitive basis function $w_k(x)=\exp\left(-\tfrac{1}{2}(x-\mu_k)^\top \Sigma_k^{-1}(x-\mu_k)\right)$, whose center $\mu_k$ and anisotropic covariance $\Sigma_k$ (parameterized through Cholesky factors) are optimized alongside coefficients $c_k$. These primitives are combined through a unified basis expansion $v(x)=\sum_{k\in\mathcal{N}(x)} w_k(x)c_k$ over an active set of neighbors. This expansion is the machinery that compares all paradigms, while the Gaussian's analytic properties (closure under convolution, Fourier duality, explicit Lipschitz constant, closed-form derivatives and volumes) provide the theoretical advantages that the benchmark builds upon.

What would settle it

Retrain every method on BACH and NLST with a fixed large step budget and no early stopping, then compare final PSNR and wall-clock time at compression factors 5, 10, and 20; if any implicit or feature-grid method overtakes the Gaussian baselines, the claim of consistent superiority fails. Alternatively, use a stricter convergence threshold (e.g., 0.05 dB over 1000 steps) and check whether the ordering changes.

Watch

Extended reading notes

Core claim

The central discovery is that the divide between implicit and explicit representations is not a matter of expressive power but of how basis functions are parameterized and supported. The paper establishes that all off-grid methods fit the same basis expansion $v(x)=\sum_{k\in\mathcal{N}(x)} w_k(x)c_k$, and that explicit Gaussian primitives—localized, freely optimizable anisotropic Gaussians with centers $\mu_k$ and covariances $\Sigma_k$—sit at the end of this spectrum where the bases themselves are the free parameters. Empirically, on breast histology and lung CT volumes, Gaussian representations match or surpass implicit neural representations (Fourier-feature MLPs, SIREN, WIRE) and feature grids (Neural Volumes, K-Planes) in PSNR across compression factors 5, 10, and 20, while requiring substantially lower optimization time and per-coordinate backpropagation memory. The paper also claims that explicit primitives offer analytic closure under convolution, Gaussian Fourier duality, closed-form Lipschitz bounds, derivatives, and volumes, properties that implicit networks can only approximate with additional machinery.

Load-bearing premise

The load-bearing premise is that the early-stopping rule—stopping 300 training steps after the last 0.2 dB PSNR improvement—measures each method fairly; if implicit networks improve slowly after long plateaus, the benchmark truncates them and overstates the explicit advantage.

Editorial extensions

If this is right

  • Medical image representation can switch to explicit Gaussian primitives without sacrificing reconstruction fidelity, enabling faster cohort-level processing.
  • Backpropagation memory is cut by 5–15x relative to feature grids and 10–200x relative to fully implicit methods, permitting larger mini-batches or higher-resolution volumes.
  • Optimization times drop by 2–5x versus feature grids and 5–100x versus fully implicit methods, allowing more architectural exploration and hyperparameter tuning.
  • Analytic Gaussian properties make operations like point-spread-function deconvolution and raw Fourier-domain fitting straightforward, without numerical integration or gridding.
  • The unified formulation turns method choice into a design decision about basis parameterization and support, rather than a split between 'implicit' and 'explicit' camps.

Reading between the lines

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

  • If the explicit advantage persists with fixed training budgets and no early stopping, downstream tasks like registration and segmentation could operate directly on the optimized primitive set rather than on reconstructed volumes.
  • The paper leaves generative modeling as an open frontier; one testable extension is whether autoregressive or permutation-equivariant models can generate Gaussian primitive sets for medical images, where ground-truth primitive data is scarce.
  • The reported memory and time figures used a specific early-stopping rule; extending the benchmark to stricter convergence criteria would reveal whether the gap is a true property of the methods or partly an artifact of the stopping protocol.
  • The analytic Lipschitz and volume properties could be exploited for constrained optimization (e.g., volume-preserving registration) more directly than the parameter-space penalties currently applied to MLP fields.
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 argues that explicit Gaussian primitives are preferable to implicit neural representations for medical imaging. It introduces a unified basis-expansion formulation, catalogues mathematical properties of explicit primitives (convolution closure, Fourier duality, Lipschitz bounds, analytic derivatives, controllability, and geometric structure), and reports benchmarks on BACH histology images and NLST lung CT volumes at compression factors 5, 10, and 20. The central claim is that Gaussian methods match or surpass implicit methods in PSNR while being substantially faster and more memory-efficient.

Significance. If the central claim is established, the paper would give medical imaging practitioners a well-supported reason to switch from coordinate MLPs to Gaussian primitives, and the unified formulation in Eq. (1) is a useful conceptual contribution. The benchmark uses independently published baselines and reports concrete memory and time numbers, which is a strength. However, the quantitative comparison is gated by an early-stopping rule that plausibly truncates slow-converging implicit models before their best quality, and the evidence base (n=10 per dataset, no error bars or significance tests) does not yet support the strong wording "consistently match or surpass ... across all compression factors." The theoretical catalog in Section 3 is thought-provoking but several entries are supported by unpublished preprints rather than independently verified results.

major comments (3)
  1. [Section 4, "Optimization duration"] The early-stopping criterion (300 training steps without a PSNR gain of 0.2 dB) defines both the reported final PSNR and the reported optimization time, so the central comparison is conditional on this rule. Since the fully implicit baselines are trained with learning rates of 1e-5 (SIREN, WIRE) and 1e-4 (FF-ReLU), while Gaussian methods use 5e-2, a 300-step plateau is plausibly a temporary plateau rather than convergence for the implicit models. If an implicit run would improve by more than 0.2 dB after additional training, its final PSNR is understated and the headline claim that Gaussian methods match or surpass implicit methods is not supported by the measured numbers. The authors should report convergence curves or checkpoints after early stopping, or rerun the comparison with a fixed step budget and with a stricter stopping criterion, to show the conclusion is not an artifact of the stopping rule.
  2. [Section 4, Figures 2 and 3] The results are reported without error bars, per-sample variance, or significance tests, despite the claim of consistency "across all compression factors" on n=10 samples per dataset. The scatter plots may show individual runs, but the text does not report means and standard deviations; the observed PSNR gaps (often around 1 dB or less) cannot be distinguished from run-to-run variability on this evidence. The authors should report the full distribution of PSNR, time, and memory over the 10 BACH images and 10 NLST volumes, and ideally include a pairwise significance test between Gaussian and implicit methods.
  3. [Section 4, "Baselines" and figure captions] The compression-factor matching procedure is not specified. The figure captions list parameter counts (e.g., 78.6K Gaussians versus SIREN 2x789 units), but the text does not state how a compression factor of 5, 10, or 20 is converted into Gaussian counts, MLP widths, feature-grid channels, or whether all methods have exactly equal total trainable parameters at each factor. This is load-bearing for the phrase "across all compression factors," because unequal parameter budgets would confound the quality and efficiency comparison. Please state the matching rule explicitly and confirm that all methods at a given factor have the same number of trainable parameters.
minor comments (5)
  1. [Section 4, "Baselines"] The hyperparameter search is described only as "grid-search," with no grid ranges, number of trials, or selection criterion; please provide these details so the tuning effort is reproducible.
  2. [Figures 2 and 3, captions] The caption states that point grayness indicates the number of steps before early stopping (black=earlier, white=later), but no color bar or numeric scale is provided; please add a legend or color bar.
  3. [Section 3.1, "Analytical Lipschitz continuity"] The claim that "the Lipschitz constant of an explicit Gaussian representation is analytically defined" should be qualified: for a sum of Gaussians one can compute an upper bound from covariances and amplitudes, but the exact Lipschitz constant is not generally available in closed form. The current wording overstates the mathematical guarantee.
  4. [Section 3 and references] Several application claims in Section 3 are supported by author-group arXiv preprints (e.g., [6], [10], [14], [42]) that are not independently verified; for a journal publication, please mark these as unpublished or preliminary and clarify what has undergone peer review.
  5. [Title and abstract] The rhetorical title "Implicit representations are dead" is stronger than the evidence presented; even if the benchmark is corrected, the results show that explicit methods are competitive and faster, not that implicit methods should be discarded. A more measured title or a qualifying sentence in the abstract would better match the content.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central benchmark is self-contained against external baselines, with only minor non-load-bearing self-citations in the applications overview.

full rationale

The central empirical claim — Gaussian representations match or surpass implicit methods in PSNR while using less time and memory — is evaluated against independently published baselines: SIREN [33], WIRE [28], Fourier-features [36], Neural Volumes [16], K-Planes [8], GaussianImage [43], and Image-GS [44]. None of these baselines are the authors' own, and the Gaussian variants tested are implementations of those external methods, not new models derived from the paper's equations. The early-stopping rule (300 steps without +0.2 dB PSNR gain) is a uniform, fixed criterion applied to all methods; it may bias the comparison against slow-converging implicit networks, but it is not defined in terms of the Gaussian methods' behavior and therefore does not make the time advantage true by construction. The unified formulation in Eq. (1) is an observation about existing paradigms rather than a derivation whose output is fed back as input. The Section 3 catalog of 'explicit advantages' cites several author-group preprints ([6], [10], [20], [30], [34]), but these support the applications overview and do not supply the benchmark numbers or the main conclusion. No fitted parameter is renamed as a prediction, and no uniqueness theorem is imported from the authors' prior work. Accordingly, the paper's central result does not reduce to its own definitions or citations; the self-citations present are minor and not load-bearing, consistent with a score of 2.

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

The paper introduces no new entities. The main external inputs are the unified formulation, the Gaussian mathematical properties, and a set of experimental design choices, including hyperparameters and the early-stopping rule, that the central comparison depends on.

free parameters (6)
  • SIREN frequency omega = 250 (BACH), 30 (NLST)
    Chosen by grid search per dataset to maximize reconstruction quality; affects the spectral behavior of the baseline.
  • WIRE frequency omega = 250 (BACH), 10 (NLST)
    Chosen by grid search; combined with sigma controls the wavelet scale of the baseline.
  • WIRE scale sigma = 0.1 (BACH), 1 (NLST)
    Chosen by grid search; affects the width of the wavelet bases.
  • Fourier feature count = 256 (BACH), 512 (NLST)
    Set per dataset to balance expressivity and parameter budget.
  • Per-method learning rates = 1e-4 (FF), 1e-5 (SIREN/WIRE), 1e-3 (feature grids), 5e-2 (Gaussians)
    Tuned via grid search; different rates across methods make efficiency comparisons dependent on tuning quality.
  • Early-stopping patience and PSNR threshold = 300 steps, +0.2 dB
    Chosen by the authors; this rule determines when training stops and therefore affects final PSNR and optimization time.
assumptions (5)
  • domain assumption All off-grid methods can be written as v(x) = sum_k w_k(x) c_k with active set N(x) (Eq. 1)
    Stated without proof and used to argue the paradigms differ only in basis parameterization, support, and optimized quantities.
  • domain assumption Gaussian primitives are representative of explicit primitive methods and their closed-form properties transfer to practical medical tasks
    Section 3 assumes mathematical properties such as convolution closure and analytic Lipschitz continuity yield real-world advantages; several are supported only by citation to preprints.
  • ad hoc to paper The chosen early-stopping criterion estimates the plateau of optimization time fairly
    Section 4 sets 300 steps without +0.2 dB PSNR as the stopping rule; if implicit methods improve after long plateaus, the comparison is biased.
  • domain assumption Matched compression factors provide a fair parameter budget across methods
    The paper runs all methods at compression factors 5, 10, and 20 but does not specify byte-level accounting or show all methods use the same number of trainable parameters.
  • ad hoc to paper PSNR on n=10 samples from BACH and NLST is sufficient to support general conclusions about medical imaging
    The datasets cover only two modalities, and no error bars or significance tests are reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Implicit representations are dead. Long live explicit primitives!." pith.science (2026). https://pith.science/paper/JWDVO6U5

@misc{pith2026260810001,
  author       = {Pith},
  title        = {Pith review of: Implicit representations are dead. Long live explicit primitives!},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JWDVO6U5}},
  note         = {Machine review of arXiv:2608.10001}
}
read the original abstract

Continuous parameterization of medical data has emerged as a powerful paradigm for resolution-independent image representation. While Implicit Neural Representations offer high fidelity and compact storage, their reliance on global Multi-Layer Perceptrons incurs sizeable computational costs, large memory requirements, and extensive optimization times. As medical imaging trends towards ever-more detailed, high-resolution volumes, these costs impose significant bottlenecks in the applicability of implicit approaches. Recently, explicit Gaussian-based primitives have revolutionized the representation learning paradigm by trading deep network evaluations for localized, rasterization-friendly primitives. In this paper, we present a comprehensive, cross-dimensional evaluation of Gaussian representations against implicit approaches for medical imaging applications. First, we outline a theoretical overview on the mathematical properties offered by explicit primitives beyond what is capable under the implicit neural paradigm. Subsequently, we benchmark the computational performance on two demanding image datasets: 2D microscopy histology and 3D lung Computed Tomography (CT). Our experiments demonstrate that Gaussian representations consistently match or surpass reconstruction metrics compared to implicit methods across all compression factors, while displaying significantly lower optimization times, and memory requirements. Together with the compelling mathematical properties offered by explicit primitives, these findings motivate the wider adoption of Gaussian representations and position them as an attractive direction for future research in medical imaging.

Figures

Figures reproduced from arXiv: 2608.10001 by the authors.

Figure 1
Figure 1. Representation paradigms positioned by domain (discrete→continuous) and basis type (implicit→explicit). Red arrows indicate the direction of research in recent years: (1-dotted) movement from explicit dictionaries towards neural representations, (2-dashed) shift towards feature-grids, (3-solid) switch to explicit primitives. and temporal resolution in that a scan’s intrinsic detail is directly coupled to its memory … view at source ↗
Figure 2
Figure 2. Performance and efficiency metrics for various implicit and explicit approaches. Scatter point grayness indicates the number of steps before a run failed to increase PSNR by 0.2 and was early-stopped (black=earlier, white=later). Implicit representations. We choose three popular fully-implicit baselines: Fourier-features [36], SIREN [33], and WIRE [28]. We used a learning rate of 10−4 for Fourier-features, while SIR… view at source ↗
Figure 3
Figure 3. Performance and efficiency metrics for various implicit and explicit approaches. Scatter point grayness indicates the number of steps before a run failed to increase PSNR by 0.2 and was early-stopped (black=earlier, white=later). 5 Conclusion In this work we have consolidated implicit and explicit representations into the same mathematical framework of dictionary methods. Furthermore, we have outlined the various ma… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 35 canonical work pages

  1. [1]

    IEEE Trans

    Aharon,M.,Elad,M.,Bruckstein,A.:K-SVD:Analgorithmfordesigningovercom- plete dictionaries for sparse representation. IEEE Trans. Signal Process.54(11), 4311–4322 (2006)

    work page 2006

  2. [2]

    Aresta, G., et al.: BACH: Grand challenge on breast cancer histology images. Med. Image Anal.56, 122–139 (2019)

    work page 2019

  3. [3]

    In: ECCV (2020)

    Chabra, R., Lenssen, J.E., Ilg, E., Schmidt, T., Straub, J., Lovegrove, S., New- combe, R.: Deep local shapes: Learning local SDF priors for detailed 3d recon- struction. In: ECCV (2020)

    work page 2020

  4. [4]

    In: ECCV (2022)

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

    work page 2022

  5. [5]

    Coiffier, G., Béthune, L.: 1-Lipschitz neural distance fields. Comput. Graph. Forum 43(5), e15128 (2024)

    work page 2024

  6. [6]

    arXiv preprint arXiv:2512.11624 (2025)

    Dannecker, M., et al.: Fast and explicit: Slice-to-volume reconstruction via 3d Gaussian primitives with analytic point spread function modeling. arXiv preprint arXiv:2512.11624 (2025)

    arXiv 2025

  7. [7]

    In: ICLR (2023)

    De Luigi, L., Cardace, A., Spezialetti, R., Zama Ramirez, P., Salti, S., Di Stefano, L.: Deep learning on implicit neural representations of shapes. In: ICLR (2023)

    work page 2023

  8. [8]

    In: CVPR (2023)

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

    work page 2023

Show all 44 references
  1. [9]

    Inverse Probl.20(5), 1621–1638 (2004)

    Haber, E., Modersitzki, J.: Numerical methods for volume preserving image regis- tration. Inverse Probl.20(5), 1621–1638 (2004)

    2004

  2. [10]

    arXiv preprint arXiv:2603.05681 (2026)

    Huang, W., et al.: Gabor primitives for accelerated cardiac cine MRI reconstruc- tion. arXiv preprint arXiv:2603.05681 (2026)

    2026

  3. [11]

    In: NeurIPS (2024)

    Jain, U., Mirzaei, A., Gilitschenski, I.: GaussianCut: Interactive segmentation via graph cut for 3d Gaussian splatting. In: NeurIPS (2024)

    2024

  4. [12]

    ACM Trans

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3d Gaussian splatting for real-time radiance field rendering. ACM Trans. Graph.42(4), 1–14 (2023)

    2023

  5. [13]

    In: ICLR (2025)

    Lan, Y., Zhou, S., Lyu, Z., Hong, F., Yang, S., Dai, B., Pan, X., Loy, C.C.: Gaus- sianAnything: Interactive point cloud flow matching for 3d generation. In: ICLR (2025)

    2025

  6. [14]

    arXiv preprint arXiv:2406.03394 (2024)

    Li, X., et al.: Gaussian primitives for deformable image registration. arXiv preprint arXiv:2406.03394 (2024)

    2024 arXiv

  7. [15]

    In: ACM SIGGRAPH (2022)

    Liu, H.T.D., Williams, F., Jacobson, A., Fidler, S., Litany, O.: Learning smooth neural functions via Lipschitz regularization. In: ACM SIGGRAPH (2022)

    2022

  8. [16]

    arXiv preprint arXiv:1906.07751 (2019)

    Lombardi, S., Simon, T., Saragih, J., Schwartz, G., Lehrmann, A., Sheikh, Y.: Neural volumes: Learning dynamic renderable volumes from images. arXiv preprint arXiv:1906.07751 (2019)

    2019 arXiv

  9. [17]

    arXiv preprint arXiv:2603.26661 (2026)

    von Lützow, N., Rößle, B., Schmid, K., Nießner, M.: GaussianGPT: Towards au- toregressive 3d Gaussian scene generation. arXiv preprint arXiv:2603.26661 (2026)

    2026 arXiv

  10. [18]

    IEEE Trans

    Maas, K.W., Ruijters, D., Vilanova, A., Pezzotti, N.: NeRF-CA: Dynamic recon- struction of x-ray coronary angiography with extremely sparse views. IEEE Trans. Vis. Comput. Graph. (2025)

    2025

  11. [19]

    In: ICML (2009)

    Mairal, J., Bach, F., Ponce, J., Sapiro, G.: Online dictionary learning for sparse coding. In: ICML (2009)

    2009

  12. [20]

    In: ICLR (2026) Implicit representations are dead

    McGinnis, J., Shit, S., Hölzl, F.A., Friedrich, P., Büschl, P., Sideri-Lampretsa, V., Mühlau, M., Cattin, P.C., Menze, B., Rueckert, D., et al.: Beyond uniformity: Regularizing implicit neural representations through a Lipschitz lens. In: ICLR (2026) Implicit representations a...

    2026

  13. [21]

    In: ICCV (2021)

    Mehta, I., Gharbi, M., Barnes, C., Shechtman, E., Ramamoorthi, R., Chandraker, M.: Modulated periodic activations for generalizable local functional representa- tions. In: ICCV (2021)

    2021

  14. [22]

    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

  15. [23]

    Radiology258(1), 243–253 (2011)

    National Lung Screening Trial Research Team: The national lung screening trial: Overview and study design. Radiology258(1), 243–253 (2011)

    2011

  16. [24]

    In: CVPR (2019)

    Park, J.J., Florence, P., Straub, J., Newcombe, R., Lovegrove, S.: DeepSDF: Learn- ing continuous signed distance functions for shape representation. In: CVPR (2019)

    2019

  17. [25]

    In: ICCV (2021)

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

    2021

  18. [26]

    In: CVPR (2017)

    Qi, C.R., Su, H., Mo, K., Guibas, L.J.: PointNet: Deep learning on point sets for 3d classification and segmentation. In: CVPR (2017)

    2017

  19. [27]

    In: IROS (2025)

    Qi, Z., Ma, J., Xu, J., Zhou, Z., Cheng, L., Xiong, G.: GSPR: Multimodal place recognition using 3d Gaussian splatting for autonomous driving. In: IROS (2025)

    2025

  20. [28]

    In: CVPR (2023)

    Saragadam, V., LeJeune, D., Tan, J., Balakrishnan, G., Veeraraghavan, A., Bara- niuk, R.G.: WIRE: Wavelet implicit neural representations. In: CVPR (2023)

    2023

  21. [29]

    arXiv preprint arXiv:2604.04874 (2026)

    Shabanov, A., Hedman, P., Weber, E., Li, Z., Rozumny, D., Lan, G.L., Dhingra, N., Luo, L., Vedaldi, A., Richardt, C., et al.: Free-range gaussians: Non-grid-aligned generative 3d gaussian reconstruction. arXiv preprint arXiv:2604.04874 (2026)

    2026 arXiv

  22. [30]

    In: MIDL (2024)

    Sideri-Lampretsa, V., McGinnis, J., Qiu, H., Paschali, M., Simson, W., Rueckert, D.: SINR: Spline-enhanced implicit neural representation for multi-modal registra- tion. In: MIDL (2024)

    2024

  23. [31]

    In: ISMRM Workshop on Data Sampling and Image Reconstruction (2026)

    Singh, I., Dupuis, A., Kukran, S., Badve, C., Griswold, M.: Back to basis: Spatially continuous MRI with adaptive Gaussians. In: ISMRM Workshop on Data Sampling and Image Reconstruction (2026)

    2026

  24. [32]

    In: NeurIPS (2020)

    Sitzmann, V., Chan, E.R., Tucker, R., Snavely, N., Wetzstein, G.: MetaSDF: Meta- learning signed distance functions. In: NeurIPS (2020)

    2020

  25. [33]

    In: NeurIPS (2020)

    Sitzmann, V., Martel, J.N., Bergman, A.W., Lindell, D.B., Wetzstein, G.: Implicit neural representations with periodic activation functions. In: NeurIPS (2020)

    2020

  26. [34]

    In: MICCAI (2023)

    Stolt-Ansó, N., McGinnis, J., Pan, J., Hammernik, K., Rueckert, D.: NISF: Neural implicit segmentation functions. In: MICCAI (2023)

    2023

  27. [35]

    In: CVPR (2021)

    Tancik, M., Mildenhall, B., Wang, T., Schmidt, D., Srinivasan, P.P., Barron, J.T., Ng, R.: Learned initializations for optimizing coordinate-based neural representa- tions. In: CVPR (2021)

    2021

  28. [36]

    In: NeurIPS (2020)

    Tancik, M., Srinivasan, P.P., Mildenhall, B., Fridovich-Keil, S., Raghavan, N., Sing- hal, U., Ramamoorthi, R., Barron, J.T., Ng, R.: Fourier features let networks learn high frequency functions in low dimensional domains. In: NeurIPS (2020)

    2020

  29. [37]

    In: ICASSP (2012)

    Van Nguyen, H., Patel, V.M., Nasrabadi, N.M., Chellappa, R.: Kernel dictionary learning. In: ICASSP (2012)

    2012

  30. [38]

    In: MICCAI (2025)

    Vyas, K., Veeraraghavan, A., Balakrishnan, G.: Fit pixels, get labels: Meta-learned implicit networks for image segmentation. In: MICCAI (2025)

    2025

  31. [39]

    IEEE Trans

    Xu, J., Moyer, D., Gagoski, B., Iglesias, J.E., Grant, P.E., Golland, P., Adalsteins- son, E.: NeSVoR: Implicit neural representation for slice-to-volume reconstruction in MRI. IEEE Trans. Med. Imaging42(6), 1707–1719 (2023)

    2023

  32. [40]

    In: CVPR (2023)

    Yang, J., Pavone, M., Wang, Y.: FreeNeRF: Improving few-shot neural rendering with free frequency regularization. In: CVPR (2023)

    2023

  33. [41]

    In: NeurIPS (2017)

    Zaheer, M., Kottur, S., Ravanbakhsh, S., Póczos, B., Salakhutdinov, R., Smola, A.J.: Deep sets. In: NeurIPS (2017)

    2017

  34. [42]

    In: NeurIPS (2024) 12 N

    Zha, R., Lin, T.J., Cai, Y., Cao, J., Zhang, Y., Li, H.: R2-Gaussian: Rectifying radiative Gaussian splatting for tomographic reconstruction. In: NeurIPS (2024) 12 N. Stolt-Ansó & M. Dannecker et al

    2024

  35. [43]

    In: ECCV (2024)

    Zhang, X., Ge, X., Xu, T., He, D., Wang, Y., Qin, H., Lu, G., Geng, J., Zhang, J.: GaussianImage: 1000 FPS image representation and compression by 2d Gaussian splatting. In: ECCV (2024)

    2024

  36. [44]

    In: ACM SIGGRAPH (2025)

    Zhang, Y., Li, B., Kuznetsov, A., Jindal, A., Diolatzis, S., Chen, K., Sochenov, A., Kaplanyan, A., Sun, Q.: Image-GS: Content-adaptive image representation via 2d Gaussians. In: ACM SIGGRAPH (2025)

    2025

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.