Pith. sign in

REVIEW 3 major objections 7 minor 1 cited by

Learning Radiance Fields from a Single Snapshot Compressive Image

T0 review · 3 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Two algorithms, SCINeRF and SCISplat, recover an underlying 3D scene and camera poses from a single snapshot compressive image, with SCISplat achieving 35.94 dB average PSNR and 205 FPS rendering.

desk verdict A competent 3DGS extension of SCINeRF with an honest ablation, but the paper's success depends on a fragile initialization that it doesn't quantify. read the letter →

arxiv 2412.19483 v1 pith:JY3BFUPG submitted 2024-12-27 cs.CV

Yunhao Li , Xiang Liu , Xiaodong Wang , Xin Yuan , Peidong Liu This is my paper · ORCID
classification cs.CV
keywords snapshotcompressiveimagingneuralradiancefields3DGaussiansplattingnovelviewsynthesiscameraposeestimationcompressedsensingcomputational
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

Snapshot compressive imaging (SCI) records a whole video of masked frames as a single 2D photo by summing them in one exposure. The paper argues that this single measurement, a sequence of random masks plus one summed image, carries enough information to reconstruct the 3D scene a moving camera was looking at, not just to decode the individual frames. To do this it proposes SCINeRF, which folds the SCI formation model into NeRF training while jointly solving for camera poses, and SCISplat, a 3D Gaussian Splatting version that starts from a learned structure-from-motion initialization and refines both scene and poses. If correct, a low-cost camera equipped with a DMD could capture, transmit, and store 3D scene content compactly, and SCISplat's 205 FPS rendering makes real-time multi-view viewing possible.

What carries the argument

The load-bearing object is the forward model of video SCI used as a training objective: given masks $M_i$ and rendered frames $\hat{X}_i$, the synthesized measurement $\hat{Y} = \sum_i \hat{X}_i \odot M_i$ must match the captured $Y$. The second device is the pose parameterization: SCINeRF restricts the virtual cameras to a linear trajectory through $T_i = T_1 \exp(\frac{i}{N} \log(T_1^{-1} T_N))$ with near-identity starting poses, while SCISplat obtains initial point clouds and poses by interpolating degraded frames from the measurement and passing them through VGGSfM, then jointly optimizes all Gaussians and poses. A third device, MCMC densification, replaces the standard 3DGS adaptive density control so that opacity does not spike, which would otherwise push the ill-posed fitting toward bright noisy pixels and make poses drift.

What would settle it

Render a novel view far outside the encoded camera trajectory from a SCISplat reconstruction and compare against a real photograph of that viewpoint; disagreement would show the recovered object is not the true 3D scene. A sharper test is to solve the forward model backwards to build two distinct scenes and trajectories that produce exactly the same $Y$; if the optimization converges to the wrong one, single-measurement 3D recovery is not identifiable.

Watch

Extended reading notes

Core claim

The central claim is that the ill-posed SCI decoding problem can be lifted from 2D frame recovery to 3D scene estimation. The paper renders frames from a radiance field, modulates them by the same binary masks used during capture, and sums them to synthesize a measurement $\hat{Y}$ that is compared directly with the real snapshot $Y$. Optimizing this photometric loss while also refining camera poses yields multi-view consistent images that existing frame-level decoders cannot produce. On six synthetic scenes, SCISplat reports 35.94 dB average PSNR for reconstruction, 2.3 dB above SCINeRF and more than 4.3 dB above the best prior method, along with 205 FPS rendering; on real data captured by the authors' SCI system, both methods recover scene detail that prior decoders blur or lose.

Load-bearing premise

The method assumes the jointly optimized camera poses and scene can converge to the true configuration from the chosen initialization, near-identity poses for SCINeRF and VGGSfM's estimates for SCISplat, since the paper shows random and COLMAP starts fail outright; without a proof of identifiability, the success of the entire approach rests on that starting basin.

Editorial extensions

If this is right

  • A single coded snapshot can substitute for a multi-view image capture for 3D reconstruction, provided the camera moves during the exposure.
  • SCISplat's 205 FPS rendering means the recovered 3D scene can be deployed for real-time interactive viewing and downstream vision tasks.
  • The test-time optimization approach sidesteps the generalization gap that limits pre-trained deep SCI decoders when applied to real hardware data.
  • The same masks that enable compression also act as a privacy-preserving encoding, since the raw measurement is unintelligible without them.
  • Joint pose optimization removes the need for conventional structure-from-motion on the compressed image, which the paper shows fails outright.

Reading between the lines

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

  • Not established in the paper is identifiability: nothing proves that a single measurement has a unique 3D scene behind it, and a two-scene, one-measurement counterexample would delimit the claim and is testable with the same pipeline.
  • A testable stress test: perturb the initialization away from the VGGSfM estimate or the near-identity start and chart when recovery flips from the true scene to an alternative that still matches $Y$; the paper's own ablation shows random and COLMAP starts fail, so the basin of attraction is the real load-bearing feature.
  • The same formation model should extend to hyperspectral SCI, where the coded frames are spectral bands rather than time steps, giving a route to recovering a 4D scene from a single snapshot.
  • The mask design itself becomes a tunable parameter of 3D recovery: the overlap-rate ablation suggests that the masks, not just the decoder, control how much 3D information a single snapshot can carry.
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 / 7 minor

Summary. The paper addresses the problem of recovering a 3D scene representation from a single snapshot compressive imaging (SCI) measurement. It proposes SCINeRF, which formulates the SCI formation model within NeRF training and jointly optimizes a neural radiance field with camera poses (interpolated along a linear trajectory), and SCISplat, a 3D Gaussian Splatting variant that uses an initialization protocol based on pixel interpolation followed by VGGSfM for initial point clouds and poses, together with MCMC-based densification. The methods are evaluated on six synthetic scenes and on real data captured by the authors' SCI system, with comparisons to prior SCI restoration methods and to two-stage NeRF/3DGS baselines. The paper reports that SCISplat achieves an average PSNR of 35.94 dB on synthetic SCI reconstruction and a rendering speed of 205 FPS.

Significance. If the reported results are reproducible, the paper demonstrates a practically valuable capability: recovering a 3D scene from a single coded snapshot, with real-time rendering enabled by 3DGS. The evaluation gives ground-truth poses to the novel-view baseline methods, which is a favorable treatment of the baselines, and the inclusion of real hardware experiments strengthens the empirical contribution. The main significance is therefore the new task formulation and the strong quantitative results. The significance is tempered, however, by the paper's heavy reliance on the initialization choice and by the absence of robustness or identifiability analysis, which leaves the scope of the central claim unclear.

major comments (3)
  1. [Section 4.3.1, Table 6] The success of SCISplat is critically dependent on the initialization from VGGSfM. Table 6 shows that random initialization and COLMAP fail completely (N/A), and that VGGSfM outperforms DUSt3R by 3.8 dB PSNR. This is direct evidence that the optimization landscape of the loss in Eq. (15) is not benign and that the favorable initialization is load-bearing for the reported results. The paper provides no identifiability analysis of the single-measurement inverse problem, nor any characterization of the basin of attraction of the joint (scene, pose) optimization. To support the abstract's claim that the 'underlying 3D scene information' is recovered from a single snapshot compressive image, the authors should either provide conditions under which the joint optimization has the intended global optimum, or empirically quantify the robustness to initialization (e.g., by perturbing the VGGSfM poses and reporting PSNR as a function of perturbation magnitude). Without such an analysis, it is unclear whether the method generalizes beyond the specific initialization regime tested.
  2. [Tables 1-4] All quantitative results are reported as single runs without error bars or multiple seeds. Given the ill-posed nature of SCI reconstruction and the demonstrated sensitivity to initialization (Section 6.3), single-run numbers do not establish that the method reliably recovers the scene. The authors should report mean and standard deviation over multiple trials (e.g., different mask realizations, random seeds, and VGGSfM initializations) for at least the synthetic benchmark.
  3. [Sections 6.1 and 6.2] The mask overlapping rate and the interpolation threshold are selected based on the same synthetic and real scenes that are later used for the final evaluation. This constitutes hyperparameter tuning on the test data and can inflate the reported performance. The paper should separate a validation set from the test scenes, or show that the results are stable over a range of these hyperparameters, before claiming that the reported PSNR and SSIM numbers reflect the method's intrinsic quality.
minor comments (7)
  1. [Section 6.5] The text says 'The results presented in Table 7 demonstrate the significant improvements of the new MCMC densification strategy,' but the MCMC results appear in Table 8; Table 7 reports the initial-point-count ablation.
  2. [Table 1] In the Factory column, the GAP-TV LPIPS value is listed as '0.0.515', which appears to be a typo and should likely read '0.515'.
  3. [Section 6.1] The sentence 'we empirically select an overlapping rate of 0.25 as the upper limit of all the experiments' is confusing; since 0.25 is the best rate in Table 5, it should be described as the selected operating point rather than an upper limit.
  4. [Section 1, Contribution 1] The claim of being 'the first method to restore 3D aware multi-view images from a single snapshot compressive image' needs qualification, because SCINeRF was already published at CVPR 2024; the novel contribution of the present manuscript is primarily SCISplat.
  5. [Section 5.3] The real-data evaluation is entirely qualitative. A quantitative proxy (e.g., no-reference image quality metrics, or a controlled real scene with known ground truth) would help support the claim of practical viability.
  6. [Equation (15)] The loss weights λ_o and λ_s are not ablated; a sensitivity analysis would clarify their influence on the final results and on the stability of the optimization.
  7. [Abstract] The URL 'https://github.com/WU- CVGL/SCISplat' contains an unintended space and should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the reported novel-view and reconstruction results are not restatements of the single SCI measurement loss, and no load-bearing claim reduces to a self-citation.

full rationale

The paper's derivation chain is a standard forward-model fitting procedure: it assumes the SCI formation model Y = sum_i X_i ⊙ M_i (Eq. 1), renders virtual frames \widehat{X}_i from a NeRF or 3DGS scene representation, synthesizes \widehat{Y} = sum_i \widehat{X}_i ⊙ M_i (Eq. 13), and optimizes the scene and camera poses by minimizing \|\widehat{Y} - Y\|^2 (Eqs. 14-15). The individual masked frames and the novel views are not given as inputs; they are latent variables recovered through the forward model, so reporting PSNR/SSIM/LPIPS on rendered novel views (Table 2) is not a restatement of the training objective. No uniqueness theorem or central premise is imported from the authors' own prior work: the CVPR 2024 SCINeRF paper [19] is used as a baseline and extension point, while the initialization and densification components (VGGSfM, DUSt3R, COLMAP, MCMC) are external. The ablation in Sec. 6.3 shows sensitivity to initialization, with random and COLMAP initializations failing while VGGSfM succeeds; this is a robustness and identifiability limitation, not circularity, because the successful configuration still predicts out-of-sample views rather than merely echoing the fitted measurement. Mask overlapping rate and interpolation threshold are selected via ablations, which is a hyperparameter-selection concern rather than a circular derivation; these choices do not make the reported results equivalent to the inputs by construction.

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

The central method is an optimization that uses the measurement itself as supervision, so the model parameters are not free parameters in the fitting sense. The true free parameters are the experimental and initialization choices that are tuned to the evaluation datasets (OR, tau, n) plus hidden weights, and the load-bearing assumptions are the forward model fidelity, static scene, and the reliability of the pretrained SfM initiator.

free parameters (5)
  • Mask overlapping rate = 0.25
    Chosen via ablation (Table 5) as the value giving best PSNR on the synthetic benchmark and used for all reported experiments; it sets the information content of the measurement, a design choice favorable to the method.
  • Interpolation threshold tau = 0.8 (real), 1.0 (synthetic)
    Selected by visual inspection of the degraded frames (Fig. 8) as a balance between removing noise and retaining pixels; affects the quality of the VGGSfM initialization in SCISplat.
  • Initial point cloud size n = Varies per scene (e.g., 10000 for Airplants; no downsampling for Factory)
    Downsampling threshold is tuned per dataset in Sec. 6.4 to trade off noise against scene coverage; final numbers use the best configuration.
  • Loss weights lambda_o, lambda_s = Not reported
    Eq. (15) defines opacity and scale regularizers but the paper never gives the weight values used in experiments; this is an unstated hyperparameter affecting the optimization.
  • Training iterations / convergence criterion = 100K-200K for SCINeRF; not stated for SCISplat
    Implementation details (Sec. 5.1.4) specify a range for SCINeRF but do not report the number of iterations used for SCISplat, making reproduction lossy.
assumptions (6)
  • domain assumption The video SCI formation model in Eq. (1), with precisely known binary masks, is the correct forward model for both synthetic and real systems.
    Invoked throughout (Sec. 3, Sec. 4.4). Real masks deviate from binary and noise is present (Sec. 5.3), so this is an idealization the method relies on.
  • domain assumption The captured scene is static and view-consistent, so NeRF/3DGS reconstruction is well-posed.
    NeRF and 3DGS both assume a static scene with consistent appearance across views; no dynamic or specular variations are modeled beyond view-dependent color.
  • ad hoc to paper The camera trajectory is smooth and, for SCINeRF, can be represented as linear interpolation in SE(3) between two endpoint poses (Eq. 9).
    Sec. 4.2. The paper acknowledges this breaks on snake/arc trajectories (Table 4); SCISplat optimizes per-frame poses but still starts from a learning-based SfM guess.
  • ad hoc to paper VGGSfM, a pretrained model, returns sufficiently accurate poses and point clouds from the degraded interpolated frames.
    Sec. 4.3.1. The ablation (Table 6) shows random initialization and COLMAP fail completely; the whole SCISplat pipeline depends on VGGSfM's generalization to SCI-degraded images.
  • ad hoc to paper The photometric loss on the single compressed measurement has a basin of attraction that, from the chosen init, contains the true scene and poses.
    This is the core feasibility assumption of joint pose-scene optimization from one image; the paper provides no proof, only empirical success on selected scenes (Sec. 4.2, Sec. 6.3).
  • ad hoc to paper MCMC densification is more stable than adaptive density control for this ill-posed inverse problem.
    Sec. 4.3.2; supported only by a 2.2 dB ablation on the same datasets, no theoretical justification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Radiance Fields from a Single Snapshot Compressive Image." pith.science (2026). https://pith.science/paper/JY3BFUPG

@misc{pith2026241219483,
  author       = {Pith},
  title        = {Pith review of: Learning Radiance Fields from a Single Snapshot Compressive Image},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JY3BFUPG}},
  note         = {Machine review of arXiv:2412.19483}
}
read the original abstract

In this paper, we explore the potential of Snapshot Compressive Imaging (SCI) technique for recovering the underlying 3D scene structure from a single temporal compressed image. SCI is a cost-effective method that enables the recording of high-dimensional data, such as hyperspectral or temporal information, into a single image using low-cost 2D imaging sensors. To achieve this, a series of specially designed 2D masks are usually employed, reducing storage and transmission requirements and offering potential privacy protection. Inspired by this, we take one step further to recover the encoded 3D scene information leveraging powerful 3D scene representation capabilities of neural radiance fields (NeRF). Specifically, we propose SCINeRF, in which we formulate the physical imaging process of SCI as part of the training of NeRF, allowing us to exploit its impressive performance in capturing complex scene structures. In addition, we further integrate the popular 3D Gaussian Splatting (3DGS) framework and propose SCISplat to improve 3D scene reconstruction quality and training/rendering speed by explicitly optimizing point clouds into 3D Gaussian representations. To assess the effectiveness of our method, we conduct extensive evaluations using both synthetic data and real data captured by our SCI system. Experimental results demonstrate that our proposed approach surpasses the state-of-the-art methods in terms of image reconstruction and novel view synthesis. Moreover, our method also exhibits the ability to render high frame-rate multi-view consistent images in real time by leveraging SCI and the rendering capabilities of 3DGS. Codes will be available at: https://github.com/WU- CVGL/SCISplat.

Figures

Figures reproduced from arXiv: 2412.19483 by the authors.

Figure 1
Figure 1. Given a single snapshot compressive image, our method can recover its underlying 3D scene representation. Leveraging the strong 3D [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of Proposed Methods. Both methods take the real SCI measurement Y and modulation masks M as input to recover the compressed images and the underlying 3D scene structure. For SCINeRF, camera poses Ti are constrained by a spline. The scene information, including scene volumetric density σ and RGB color c, is encoded in a lightweight MLP, which is then used to render the compressed multi-view images Xˆ through… view at source ↗
Figure 3
Figure 3. Experimental setup for real dataset collection. This SCI imag￾ing system contains a CCD camera to record snapshot measurement, primary and rely lens, and a DMD to modulate input frames. with existing techniques. We further prove that incorporat￾ing 3DGS with SCI delivers higher performance in image restoration quality and achieves significantly faster training and rendering speed. 5.1 Experimental Setup 5.1.1 Synthe… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Qualitative evaluations of our methods against SOTA SCI image restoration methods on the synthetic dataset. Top to bottom shows the results for different scenes, including Airplants, Hotdog, Cozy2room, Factory, Tanabata, and Vender. The experimental results demonstrate…
Figure 5
Figure 5. Figure 5: Qualitative evaluations of our methods against naive two-stage baselines. We compared the quality of synthesized novel-view images from our methods against that of vanilla 3DGS from SOTA methods. Top to bottom shows different scenes. The qualitative comparisons demonst…
Figure 6
Figure 6. Figure 6: Qualitative evaluations of our methods against SOTA SCI image restoration methods on real grayscale dataset captured by our system. Top to bottom shows the results for different scenes. Since the compressed pixel-wise aligned ground truth images are unavailable, we cap…
Figure 7
Figure 7. Figure 7: Qualitative evaluations of our methods against SOTA SCI image restoration methods with real color dataset. Top to bottom shows different scenes. Due to the generalization gap, the prior SOTA methods experience significant performance degradation on real dataset, leadin…
Figure 8
Figure 8. Figure 8: Ablation studies on different thresholds τ for interpolating real data. At a low threshold τ = 0.1, the interpolated image is overwhelmed by measurement noise. At a high threshold τ = 1, the image retains only a vague outline of the captured object. Therefore, we selec…
Figure 9
Figure 9. Figure 9: Qualitative results with different densification strategies. The original densification strategy in 3DGS introduces high opacity noise around the edges and causes noticeable deficiencies. On the other hand, our pipeline applies MCMC densification strategy in 3DGS train…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. GS$^{2}$CI: Robust Gaussian Splatting For Snapshot Compressive Imaging via Large Vision Model Priors

    cs.CV 2026-08 conditional novelty 5.0 of 10

    GS2CI recovers a static 3D Gaussian scene and its camera trajectory from one SCI snapshot by combining vision-foundation-model initialization with an opacity-guided densification strategy.

Reference graph

Works this paper leans on

70 extracted references · 54 canonical work pages · cited by 1 Pith paper

  1. [1]

    Snapshot compres- sive imaging: Theory, algorithms, and applications,

    X. Yuan, D. J. Brady, and A. K. Katsaggelos, “Snapshot compres- sive imaging: Theory, algorithms, and applications,” IEEE Signal Processing Magazine, vol. 38, no. 2, pp. 65–88, 2021. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 17

    work page 2021

  2. [2]

    Coded aperture compressive temporal imaging,

    P . Llull, X. Liao, X. Yuan, J. Yang, D. Kittle, L. Carin, G. Sapiro, and D. J. Brady, “Coded aperture compressive temporal imaging,” Opt. Express, vol. 21, no. 9, pp. 10 526–10 545, May 2013

    work page 2013

  3. [3]

    Robust uncertainty principles: Exact signal reconstruction from highly incomplete frequency information,

    E. J. Cand `es, J. Romberg, and T. Tao, “Robust uncertainty principles: Exact signal reconstruction from highly incomplete frequency information,” IEEE T ransactions on information theory , vol. 52, no. 2, pp. 489–509, 2006

    work page 2006

  4. [4]

    Compressed sensing,

    D. L. Donoho, “Compressed sensing,” IEEE T ransactions on infor- mation theory, vol. 52, no. 4, pp. 1289–1306, 2006

    work page 2006

  5. [5]

    Generalized alternating projection based total variation minimization for compressive sensing,

    X. Yuan, “Generalized alternating projection based total variation minimization for compressive sensing,” in 2016 IEEE International conference on image processing (ICIP) . IEEE, 2016, pp. 2539–2543

    work page 2016

  6. [6]

    Generalized alternating projection for weighted-2,1 minimization with applications to model-based compressive sensing,

    X. Liao, H. Li, and L. Carin, “Generalized alternating projection for weighted-2,1 minimization with applications to model-based compressive sensing,” SIAM Journal on Imaging Sciences , vol. 7, no. 2, pp. 797–823, 2014

    work page 2014

  7. [7]

    Rank minimization for snapshot compressive imaging,

    Y. Liu, X. Yuan, J. Suo, D. J. Brady, and Q. Dai, “Rank minimization for snapshot compressive imaging,” IEEE transactions on pattern analysis and machine intelligence, vol. 41, no. 12, pp. 2990–3006, 2018

    work page 2018

  8. [8]

    Deep learning for video compressive sensing,

    M. Qiao, Z. Meng, J. Ma, and X. Yuan, “Deep learning for video compressive sensing,” Apl Photonics, vol. 5, no. 3, 2020

    work page 2020

Show all 70 references
  1. [9]

    Birnat: Bidirectional recurrent neural networks with adversarial training for video snapshot compressive imaging,

    Z. Cheng, R. Lu, Z. Wang, H. Zhang, B. Chen, Z. Meng, and X. Yuan, “Birnat: Bidirectional recurrent neural networks with adversarial training for video snapshot compressive imaging,” in European Conference on Computer Vision . Springer, 2020, pp. 258– 275

    2020

  2. [10]

    Memory-efficient network for large-scale video com- pressive sensing,

    Z. Cheng, B. Chen, G. Liu, H. Zhang, R. Lu, Z. Wang, and X. Yuan, “Memory-efficient network for large-scale video com- pressive sensing,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 16 246–16 255

    2021

  3. [11]

    Deep tensor admm-net for snapshot compressive imaging,

    J. Ma, X.-Y. Liu, Z. Shou, and X. Yuan, “Deep tensor admm-net for snapshot compressive imaging,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 10 223–10 232

    2019

  4. [12]

    Spatial-temporal transformer for video snapshot compressive imaging,

    L. Wang, M. Cao, Y. Zhong, and X. Yuan, “Spatial-temporal transformer for video snapshot compressive imaging,”IEEE T rans- actions on Pattern Analysis and Machine Intelligence , 2022

    2022

  5. [13]

    Efficientsci: Densely connected network with space-time factorization for large-scale video snap- shot compressive imaging,

    L. Wang, M. Cao, and X. Yuan, “Efficientsci: Densely connected network with space-time factorization for large-scale video snap- shot compressive imaging,” in Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , 2023, pp. 18 477– 18 486

    2023

  6. [14]

    Hybrid cnn-transformer architecture for efficient large-scale video snapshot compressive imaging,

    M. Cao, L. Wang, M. Zhu, and X. Yuan, “Hybrid cnn-transformer architecture for efficient large-scale video snapshot compressive imaging,” International Journal of Computer Vision , pp. 1–20, 2024

    2024

  7. [15]

    Nerf: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P . P . Srinivasan, M. Tancik, J. T. Barron, R. Ra- mamoorthi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,” Communications of the ACM , vol. 65, no. 1, pp. 99–106, 2021

    2021

  8. [16]

    Structure-from-motion revis- ited,

    J. L. Schonberger and J.-M. Frahm, “Structure-from-motion revis- ited,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 4104–4113

    2016

  9. [17]

    3d gaus- sian splatting for real-time radiance field rendering

    B. Kerbl, G. Kopanas, T. Leimk ¨uhler, and G. Drettakis, “3d gaus- sian splatting for real-time radiance field rendering.” ACM T rans. Graph., vol. 42, no. 4, pp. 139–1, 2023

    2023

  10. [18]

    3d gaussian splatting as markov chain monte carlo,

    S. Kheradmand, D. Rebain, G. Sharma, W. Sun, J. Tseng, H. Isack, A. Kar, A. Tagliasacchi, and K. M. Yi, “3d gaussian splatting as markov chain monte carlo,” arXiv preprint arXiv:2404.09591 , 2024

    2024 arXiv

  11. [19]

    SCINeRF: Neural Radiance Fields from a Snapshot Compressive Image,

    Y. Li, X. Wang, P . Wang, X. Yuan, and P . Liu, “SCINeRF: Neural Radiance Fields from a Snapshot Compressive Image,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 10 542–10 552

    2024

  12. [20]

    Shearlet enhanced snapshot compressive imaging,

    P . Yang, L. Kong, X.-Y. Liu, X. Yuan, and G. Chen, “Shearlet enhanced snapshot compressive imaging,” IEEE T ransactions on Image Processing, vol. 29, pp. 6466–6481, 2020

    2020

  13. [21]

    Dis- tributed optimization and statistical learning via the alternating direction method of multipliers,

    S. Boyd, N. Parikh, E. Chu, B. Peleato, J. Eckstein et al. , “Dis- tributed optimization and statistical learning via the alternating direction method of multipliers,” Foundations and T rends® in Ma- chine learning, vol. 3, no. 1, pp. 1–122, 2011

    2011

  14. [22]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P . Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Pro- ceedings, Part III 18...

    2015

  15. [23]

    Generative adversarial nets,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative adversarial nets,” Advances in neural information processing systems , vol. 27, 2014

    2014

  16. [24]

    Perceptual losses for real-time style transfer and super-resolution,

    J. Johnson, A. Alahi, and L. Fei-Fei, “Perceptual losses for real-time style transfer and super-resolution,” in Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part II 14 . Springer, 2016, pp. 694–711

    2016

  17. [25]

    l-net: Reconstruct hyper- spectral images from a snapshot measurement,

    X. Miao, X. Yuan, Y. Pu, and V . Athitsos, “l-net: Reconstruct hyper- spectral images from a snapshot measurement,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 4059–4069

    2019

  18. [26]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

    2016

  19. [27]

    Plug-and-play algorithms for large-scale snapshot compressive imaging,

    X. Yuan, Y. Liu, J. Suo, and Q. Dai, “Plug-and-play algorithms for large-scale snapshot compressive imaging,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 1447–1457

    2020

  20. [28]

    Plug-and-play algorithms for video snapshot compressive imaging,

    X. Yuan, Y. Liu, J. Suo, F. Durand, and Q. Dai, “Plug-and-play algorithms for video snapshot compressive imaging,” IEEE T rans- actions on Pattern Analysis and Machine Intelligence , vol. 44, no. 10, pp. 7093–7111, 2021

    2021

  21. [29]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

    2017

  22. [30]

    Stylenerf: A style-based 3d-aware generator for high-resolution image synthesis,

    J. Gu, L. Liu, P . Wang, and C. Theobalt, “Stylenerf: A style-based 3d-aware generator for high-resolution image synthesis,” 2021

    2021

  23. [31]

    Bungeenerf: Progressive neural radiance field for extreme multi-scale scene rendering,

    Y. Xiangli, L. Xu, X. Pan, N. Zhao, A. Rao, C. Theobalt, B. Dai, and D. Lin, “Bungeenerf: Progressive neural radiance field for extreme multi-scale scene rendering,” in European conference on computer vision. Springer, 2022, pp. 106–122

    2022

  24. [32]

    Mega-nerf: Scal- able construction of large-scale nerfs for virtual fly-throughs,

    H. Turki, D. Ramanan, and M. Satyanarayanan, “Mega-nerf: Scal- able construction of large-scale nerfs for virtual fly-throughs,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 12 922–12 931

    2022

  25. [33]

    Fourier features let networks learn high frequency functions in low dimensional domains,

    M. Tancik, P . Srinivasan, B. Mildenhall, S. Fridovich-Keil, N. Raghavan, U. Singhal, R. Ramamoorthi, J. Barron, and R. Ng, “Fourier features let networks learn high frequency functions in low dimensional domains,” Advances in Neural Information Process- ing Systems, vol. 33, ...

    2020

  26. [34]

    Rign- erf: Fully controllable neural 3d portraits,

    S. Athar, Z. Xu, K. Sunkavalli, E. Shechtman, and Z. Shu, “Rign- erf: Fully controllable neural 3d portraits,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition , 2022, pp. 20 364–20 373

    2022

  27. [35]

    Dynamic neural radiance fields for monocular 4d facial avatar reconstruction,

    G. Gafni, J. Thies, M. Zollhofer, and M. Nießner, “Dynamic neural radiance fields for monocular 4d facial avatar reconstruction,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 8649–8658

    2021

  28. [36]

    Animatable neural radiance fields for modeling dynamic human bodies,

    S. Peng, J. Dong, Q. Wang, S. Zhang, Q. Shuai, X. Zhou, and H. Bao, “Animatable neural radiance fields for modeling dynamic human bodies,” in Proceedings of the IEEE/CVF International Confer- ence on Computer Vision , 2021, pp. 14 314–14 323

    2021

  29. [37]

    Neural body: Implicit neural representations with structured latent codes for novel view synthesis of dynamic humans,

    S. Peng, Y. Zhang, Y. Xu, Q. Wang, Q. Shuai, H. Bao, and X. Zhou, “Neural body: Implicit neural representations with structured latent codes for novel view synthesis of dynamic humans,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021,...

    2021

  30. [38]

    Hdr- nerf: High dynamic range neural radiance fields,

    X. Huang, Q. Zhang, Y. Feng, H. Li, X. Wang, and Q. Wang, “Hdr- nerf: High dynamic range neural radiance fields,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 18 398–18 408

    2022

  31. [39]

    Learning object-compositional neural radiance field for editable scene rendering,

    B. Yang, Y. Zhang, Y. Xu, Y. Li, H. Zhou, H. Bao, G. Zhang, and Z. Cui, “Learning object-compositional neural radiance field for editable scene rendering,” in Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision , 2021, pp. 13 779–13 788

    2021

  32. [40]

    Nerf- editing: geometry editing of neural radiance fields,

    Y.-J. Yuan, Y.-T. Sun, Y.-K. Lai, Y. Ma, R. Jia, and L. Gao, “Nerf- editing: geometry editing of neural radiance fields,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, 2022, pp. 18 353–18 364

    2022

  33. [41]

    Removing objects from neural radiance fields,

    S. Weder, G. Garcia-Hernando, A. Monszpart, M. Pollefeys, G. J. Brostow, M. Firman, and S. Vicente, “Removing objects from neural radiance fields,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 16 528–16 538

    2023

  34. [42]

    Clutter detection and removal in 3d scenes with view-consistent inpainting,

    F. Wei, T. Funkhouser, and S. Rusinkiewicz, “Clutter detection and removal in 3d scenes with view-consistent inpainting,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 18 131–18 141

    2023

  35. [43]

    Neu- JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 18 ral reflectance fields for appearance acquisition,

    S. Bi, Z. Xu, P . Srinivasan, B. Mildenhall, K. Sunkavalli, M. Ha ˇsan, Y. Hold-Geoffroy, D. Kriegman, and R. Ramamoorthi, “Neu- JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 18 ral reflectance fields for appearance acquisition,” arXiv preprint arXiv:2008.03824, 2020

    2021 arXiv

  36. [44]

    Nerd: Neural reflectance decomposition from image collections,

    M. Boss, R. Braun, V . Jampani, J. T. Barron, C. Liu, and H. Lensch, “Nerd: Neural reflectance decomposition from image collections,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 12 684–12 694

    2021

  37. [45]

    Nerf–: Neural radiance fields without known camera parameters,

    Z. Wang, S. Wu, W. Xie, M. Chen, and V . A. Prisacariu, “Nerf–: Neural radiance fields without known camera parameters,” arXiv preprint arXiv:2102.07064, 2021

    2021 arXiv

  38. [46]

    Self-calibrating neural radiance fields,

    Y. Jeong, S. Ahn, C. Choy, A. Anandkumar, M. Cho, and J. Park, “Self-calibrating neural radiance fields,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 5846–5854

    2021

  39. [47]

    imap: Implicit map- ping and positioning in real-time,

    E. Sucar, S. Liu, J. Ortiz, and A. J. Davison, “imap: Implicit map- ping and positioning in real-time,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 6229–6238

    2021

  40. [48]

    Gnerf: Gan-based neural radiance field without posed camera,

    Q. Meng, A. Chen, H. Luo, M. Wu, H. Su, L. Xu, X. He, and J. Yu, “Gnerf: Gan-based neural radiance field without posed camera,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 6351–6361

    2021

  41. [49]

    Barf: Bundle- adjusting neural radiance fields,

    C.-H. Lin, W.-C. Ma, A. Torralba, and S. Lucey, “Barf: Bundle- adjusting neural radiance fields,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 5741–5751

    2021

  42. [50]

    Bad-nerf: Bundle adjusted deblur neural radiance fields,

    P . Wang, L. Zhao, R. Ma, and P . Liu, “Bad-nerf: Bundle adjusted deblur neural radiance fields,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2023, pp. 4170–4179

    2023

  43. [51]

    TensoRF: Tensorial Radiance Fields,

    A. Chen, Z. Xu, A. Geiger, J. Yu, and H. Su, “TensoRF: Tensorial Radiance Fields,” in ECCV. Springer, 2022, pp. 333–350

    2022

  44. [52]

    Plenoxels: Radiance Fields without Neural Net- works,

    S. Fridovich-Keil, A. Yu, M. Tancik, Q. Chen, B. Recht, and A. Kanazawa, “Plenoxels: Radiance Fields without Neural Net- works,” in CVPR, 2022, pp. 5501–5510

    2022

  45. [53]

    Hexplane: A fast representation for dynamic scenes,

    A. Cao and J. Johnson, “Hexplane: A fast representation for dynamic scenes,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 130–141

    2023

  46. [54]

    Instant Neural Graphics Primitives with a Multiresolution Hash Encoding,

    T. M ¨uller, A. Evans, C. Schied, and A. Keller, “Instant Neural Graphics Primitives with a Multiresolution Hash Encoding,” vol. 41, no. 4, pp. 102:1–102:15, Jul. 2022. [Online]. Available: https://doi.org/10.1145/3528223.3530127

    2022

  47. [55]

    Instantsplat: Un- bounded sparse-view pose-free gaussian splatting in 40 seconds,

    Z. Fan, W. Cong, K. Wen, K. Wang, J. Zhang, X. Ding, D. Xu, B. Ivanovic, M. Pavone, G. Pavlakos et al. , “Instantsplat: Un- bounded sparse-view pose-free gaussian splatting in 40 seconds,” arXiv preprint arXiv:2403.20309 , 2024

    2024 arXiv

  48. [56]

    Dust3r: Geometric 3d vision made easy,

    S. Wang, V . Leroy, Y. Cabon, B. Chidlovskii, and J. Revaud, “Dust3r: Geometric 3d vision made easy,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 20 697–20 709

    2024

  49. [57]

    Vggsfm: Visual geometry grounded deep structure from motion,

    J. Wang, N. Karaev, C. Rupprecht, and D. Novotny, “Vggsfm: Visual geometry grounded deep structure from motion,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 21 686–21 697

    2024

  50. [58]

    Radiance fields from vggsfm and mast3r, and their comparison,

    H. Heo, “Radiance fields from vggsfm and mast3r, and their comparison,” 2024. [Online]. Available: https://github.com/ hwanhuh/Radiance-Fields-from-VGGSfM-Mast3r

    2024

  51. [59]

    Revising densification in gaussian splatting,

    S. R. Bul `o, L. Porzi, and P . Kontschieder, “Revising densification in gaussian splatting,” arXiv preprint arXiv:2404.06109 , 2024

    2024 arXiv

  52. [60]

    Pixel-gs: Density control with pixel-aware gradient for 3d gaussian splatting,

    Z. Zhang, W. Hu, Y. Lao, T. He, and H. Zhao, “Pixel-gs: Density control with pixel-aware gradient for 3d gaussian splatting,” arXiv preprint arXiv:2403.15530, 2024

    2024 arXiv

  53. [61]

    Absgs: Recovering fine details in 3d gaussian splatting,

    Z. Ye, W. Li, S. Liu, P . Qiao, and Y. Dou, “Absgs: Recovering fine details in 3d gaussian splatting,” in ACM Multimedia 2024 , 2024

    2024

  54. [62]

    Image quality assessment: From error visibility to structural similarity,

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P . Simoncelli, “Image quality assessment: From error visibility to structural similarity,” IEEE T ransactions on Image Processing , vol. 13, no. 4, pp. 600–612, 2004

    2004

  55. [63]

    Loss functions for image restoration with neural networks,

    H. Zhao, O. Gallo, I. Frosio, and J. Kautz, “Loss functions for image restoration with neural networks,” IEEE T ransactions on Computational Imaging, vol. 3, no. 1, pp. 47–57, 2017

    2017

  56. [64]

    Local light field fusion: Practical view synthesis with prescriptive sampling guidelines,

    B. Mildenhall, P . P . Srinivasan, R. Ortiz-Cayon, N. K. Kalantari, R. Ramamoorthi, R. Ng, and A. Kar, “Local light field fusion: Practical view synthesis with prescriptive sampling guidelines,” ACM T ransactions on Graphics (TOG), vol. 38, no. 4, pp. 1–14, 2019

    2019

  57. [65]

    Deblur-nerf: Neural radiance fields from blurry images,

    L. Ma, X. Li, J. Liao, Q. Zhang, X. Wang, J. Wang, and P . V . Sander, “Deblur-nerf: Neural radiance fields from blurry images,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 12 861–12 870

    2022

  58. [66]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P . Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR) , 2018, pp. 586–595

    2018

  59. [67]

    Nerf-pytorch,

    L. Yen-Chen, “Nerf-pytorch,” https://github.com/yenchenlin/ nerf-pytorch/, 2020

    2020

  60. [68]

    Adam: A method for stochastic optimization,

    D. P . Kingma, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  61. [69]

    Exploring learning rate scaling rules for distributed ml training on transient resources,

    J. Andr ´e, F. Strati, and A. Klimovic, “Exploring learning rate scaling rules for distributed ml training on transient resources,” in Proceedings of the 3rd International Workshop on Distributed Machine Learning. ACM, 2022, pp. 1–8. Yunhao Li received his B.Eng. degree in Inf...

    2022

  62. [2015]

    His research interests include signal processing, computational imaging, and machine learning. He has been an Associate Editor for Pattern Recognition since 2019, International Journal of Pattern Recog- nition and Artificial Intelligence since 2020, a Topic Editor for Chinese ...

    2019

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.