Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

DGTR: Distributed Gaussian Turbo-Reconstruction for Sparse-View Vast Scenes

T0 review · 5 major / 6 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 DGTR claims that sparse-view vast aerial scenes can be reconstructed in minutes on distributed drones by replacing global COLMAP initialization with feed-forward Gaussian prediction and distillation-based model aggregation.

desk verdict Plausible and genuinely new distributed sparse-view pipeline, but the central alignment equation is mis-specified as written and the evaluation has gaps; still worth serious peer review. read the letter →

arxiv 2411.12309 v2 pith:XP2PITXE submitted 2024-11-19 cs.CV

Hao Li , Yuanyuan Gao , Haosong Peng , Chenming Wu , Weicai Ye , Yufeng Zhan , Chen Zhao , Dingwen Zhang
show 2 more authors
Jingdong Wang Junwei Han
This is my paper · ORCID
classification cs.CV
keywords distributed3DGaussiansplattingsparse-viewreconstructionvastscenenovelviewsynthesisfeed-forwardinitializationmodelaggregationdepthregularizationaerialscenes
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

DGTR sets out to make novel-view synthesis practical for vast aerial scenes when only sparse drone images are available and computation is distributed. The paper's central proposal is to abandon the standard global structure-from-motion (COLMAP) initialization and instead let each drone predict Gaussian primitives from its own image pairs using a pretrained feed-forward model, align those primitives to the drone-supplied poses, train a local 3D Gaussian model, and upload the model rather than the raw images. A central server then stitches the local models into a global one with a distillation-based aggregation step. The intended payoff is that reconstruction finishes in minutes rather than hours and matches or beats centralized and distributed baselines on the tested aerial scenes. The paper introduces a keyframe-sampled sparse-view benchmark built from Mill19 and UrbanScene3D to evaluate this claim.

What carries the argument

The mechanism that carries the argument is the initialization-to-aggregation loop. Concretely: (1) Splatt3R, a pretrained feed-forward Gaussian model, turns co-visible image pairs into pixel-aligned Gaussian primitives; (2) the global alignment objective of Eq. (3), built on DUSt3R-style confidence-weighted pointmaps, rescales and repositions those primitives so that they agree with the drone-provided poses; (3) a local relative-scale correction resizes covariances according to average adjacent-point distance, which the ablations show is worth the largest PSNR jump; (4) parallel local training uses a depth correlation loss with a frozen covariance to avoid overfitting; and (5) a distillation-based model aggregation renders each local model's training views as pseudo ground-truth images and fine-tunes the merged model on them. Eq. (3) is the load-bearing identity because it is the step that converts uncalibrated, scale-ambiguous predictions into a geometrically consistent global initialization.

What would settle it

Run DGTR on a vast aerial scene where consecutive images share very little visual content, so that the sliding-window pairs barely overlap; if Eq. (3) yields a large alignment residual and the rendered model contains holes or doubled structures, the claim that sparse-view vast scenes can be reconstructed in minutes without global COLMAP is falsified.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that the bottleneck in sparse-view vast-scene reconstruction is not training but initialization: replacing COLMAP with a feed-forward Gaussian predictor plus a global alignment step turns a pipeline that takes thousands of seconds into one that takes minutes without sacrificing quality. Each device independently predicts pixel-aligned Gaussian primitives from neighboring image pairs, then solves the optimization in Eq. (3) to fit the pose-free, scale-ambiguous pointmaps to the given drone poses, with an ICP global scale and a relative local scale correction applied to the covariance. Local training is regularized by a differentiable depth rasterizer and a relaxed Pearson correlation loss, and the final global model is produced by rendering pseudo ground-truth views from each local model and distilling them into the merged student model. In the reported experiments on four aerial scenes, this pipeline outperforms the compared centralized and distributed methods on PSNR/SSIM/LPIPS in the sparse-view setting and cuts total latency by up to about 2x.

Load-bearing premise

The load-bearing premise is that the pretrained feed-forward Gaussian model Splatt3R, trained on small-scale scenes, produces Gaussian primitives that are complete and scale-consistent enough for sparse aerial image pairs, and that the drone-provided poses used in Eq. (3) are accurate; if either condition fails, the initialization quality collapses and the reported speed and accuracy gains cannot be recovered.

Editorial extensions

If this is right

  • Sparse-view vast-scene reconstruction no longer needs a global SfM/COLMAP step; a drone can start from its own images and poses and still produce a geometrically consistent Gaussian model.
  • Training time on the tested scenes drops to minutes, with overall latency up to about 2x lower than the compared centralized and distributed methods.
  • Because devices upload trained Gaussian models rather than raw images, communication cost scales with model size rather than captured imagery, which is the property that makes the framework suitable for bandwidth-limited drones.
  • The depth-regularized local models converge faster than distributed 3DGS, with the largest early improvement in LPIPS (roughly 20% at the initial stage).
  • The released sparse-view vast-scene benchmark gives later work a common evaluation target for methods that cannot assume dense aerial capture.

Reading between the lines

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

  • The same three-step recipe—feed-forward initialization, local optimization, distillation aggregation—should transfer to other pose-bearing image streams, such as ground robots or handheld captures, if the feed-forward model produces reasonable primitives for their wider baselines.
  • The ablation numbers suggest the scale alignment, not the depth loss, is doing most of the work (local and global scale optimization raise PSNR by roughly 12 dB on the Rubble scene), so improving scale estimation may matter more than adding new regularization.
  • A natural extension the paper does not explore is overlapping region boundaries: the current partition is non-overlapping and the aggregation step handles seams, but overlapping captures would let the global alignment verify consistency across devices rather than taking each region's pose as fixed.
  • The dependence on accurate drone poses is implicit throughout; if pose drift is significant, the Eq. (3) objective will absorb the error into the pointmaps, and the distillation step would have to compensate silently, which is a testable risk in real flights.
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

5 major / 6 minor

Summary. DGTR proposes a distributed 3D Gaussian Splatting pipeline for sparse-view large-scale aerial reconstruction. The scene is divided into non-overlapping regions; each device uses a pretrained feed-forward Gaussian predictor (Splatt3R) on image pairs, followed by a global alignment step intended to fit the predicted pointmaps to drone-provided poses and to rescale the Gaussian covariance. Local models are trained with photometric and depth-correlation losses, and a distillation-based aggregation step merges the local models into a global model using rendered pseudo-views. Experiments are reported on four scenes from Mill-19 and UrbanScene3D, including novel-view-synthesis metrics, latency comparisons, and ablations of the initialization and depth regularization. The paper also introduces a sparse-view benchmark obtained by keyframe sampling from existing datasets.

Significance. If the technical issues in the alignment step and the over-claims in the comparisons can be resolved, DGTR addresses a real gap: it removes the global-COLMAP bottleneck and enables per-device sparse-view reconstruction in minutes. The strengths are concrete: a complete pipeline with off-the-shelf feed-forward Gaussian prediction, a keyframe-sampled benchmark, a latency breakdown, and ablations showing large gains from the proposed initialization and depth regularization. I found no problematic circularity in using frozen DPT depth maps or in the distillation aggregation; the latter is a legitimate self-distillation strategy, although it needs an ablation. However, because the most critical alignment objective (Eq. (3)) is under-specified and the reported table contradicts the headline claims, the contribution is not yet established.

major comments (5)
  1. [IV-A2, Eq. (3)] Equation (3) writes the alignment objective as min_{χ,σ} Σ C ||χ^v_i − σ_e P_e X^{v,e}_i|| with P_e ∈ R^{3×4}. Because σ_e multiplies the full pose matrix, the translation part of P_e is scaled as well, yielding σ_e R_e X + σ_e t_e instead of σ_e R_e X + t_e. This introduces a bias proportional to the camera center, so the optimized pointmaps are not in the drone coordinate frame unless t_e = 0. If the implementation in fact scales the local pointmap before applying the pose, the objective should be stated correctly and validated on synthetic scale/pose data; as written, the central initialization claim is under-specified.
  2. [V-B.1, Table I] The abstract and Section V-B.1 claim that DGTR outperforms both centralized and distributed methods and achieves state-of-the-art on all scenes. Table I does not support this: on Building, Mega-NeRF (19.25 PSNR) and Fed-3DGS (18.66 PSNR, 0.602 SSIM, 0.362 LPIPS) exceed DGTR (18.47 PSNR, 0.532 SSIM, 0.392 LPIPS); on Campus, Mega-NeRF (22.58 PSNR) and Fed-3DGS (20.00 PSNR, 0.665 SSIM) exceed DGTR (19.95 PSNR, 0.517 SSIM); on Residence, 3DGS has better SSIM (0.715 vs 0.677) and LPIPS (0.335 vs 0.339) than DGTR. Please revise the claims to match the table, or extend the evaluation so the claims are supported.
  3. [Related work and Section V-B, Table I] DoGaussian [10] and the Multiplane-prior method [8] are directly relevant to the claimed contributions, but neither appears in Table I. DoGaussian is a distributed Gaussian-splatting baseline in the same setting, and Multiplane prior is a sparse-view aerial rendering method; their absence makes the claim of state-of-the-art in 'distributed' and 'few-shot vast scene' settings incomplete. Please add these baselines or justify their exclusion.
  4. [V-A, Fig. 4] The experiments simulate multi-drone collaboration by partitioning four aerial datasets into M non-overlapping regions and running all training on a single A100 GPU. There is no actual multi-device deployment, no measurement of communication or synchronization cost, and no test with noisy or inaccurate drone poses. Because the paper's core motivation is 'truly distributed systems' (Section I), the reported latency and quality results do not yet validate the distributed claim. A small synthetic pose-noise experiment and a two-device deployment, or a precise communication model, would materially strengthen the paper.
  5. [IV-C, Table I] Model aggregation by distillation is described as a main contribution, but the experiments never remove or replace this module. The reader cannot tell whether the aggregation step helps, hurts, or is neutral relative to simply concatenating the filtered local models. Please add an ablation with and without distillation aggregation, and report boundary-region quality separately.
minor comments (6)
  1. [IV-A2, Eq. (4)] Equation (4) should clarify that s_g^m and s_l^m are applied only to the covariance scales, not to the already-aligned positions; as written, the formula suggests rescaling the whole Gaussian set after alignment.
  2. [IV-A2] The quantity s_r^m is introduced as R^{2×H×W} but never defined, and Eq. (4) uses a different symbol s_l^m; please make the notation consistent.
  3. [V-A3] The sliding-window construction is not fully specified; with step 2 and N_m images, state exactly which pairs are formed and how many pointmaps are concatenated.
  4. [Tables I-III] No error bars or repeated-run statistics are reported, and the color-based best/second/third markers in Table I are not visible in grayscale; please report mean±std over at least three runs or state the single-run policy explicitly.
  5. [VI and Abstract] There are several typos ('Code will released', 'on on', 'multi-drones multi-drones', 'Campuse'); please proofread the manuscript.
  6. [IV-B, Eq. (5)] The exponent on the Pearson-correlation term is unclear, and the text should define D and D_est precisely and explain why freezing Σ in Table II is beneficial.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DGTR's initialization, depth supervision, and aggregation are anchored to external data or pretrained models, and the central speed/quality claims are evaluated on held-out views.

full rationale

The derivation chain is not circular. The feed-forward initialization (Eq. 2) uses the externally pretrained Splatt3R model; no parameter of DGTR is fitted to the reported benchmark and then renamed a prediction. The global alignment (Eq. 3) fits predicted pointmaps to drone-provided poses P_e, which are external inputs, and the per-edge scale σ_e is a fitted nuisance parameter, not the target reconstruction. The resulting Gaussians are then trained with photometric losses against real images (Eq. 7) and a depth regularizer (Eq. 5) whose depth maps come from the frozen pretrained DPT network. The model aggregation step uses synthetic views rendered from the local teacher models as pseudo ground truth for a student global model; this is a self-distillation training strategy rather than a claim whose verification reduces to its own output, and the final numbers in Table I are computed on held-out evaluation views from Mill19 and UrbanScene3D. The only self-citation, [24] in the related-work list of feed-forward Gaussian methods, is not load-bearing for any technical claim. A reviewer-level concern that Eq. (3) as written scales the translation part of P_e, making it not a proper similarity alignment, is a correctness/typo question rather than circularity: the objective still depends on externally given poses instead of on the paper's own conclusions. No step in the paper is equivalent to its input by construction, so the appropriate verdict is no significant circularity.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim rests less on new math than on transfer of pretrained components and on hyperparameter choices; there are no new physical entities. The load-bearing assumptions are the transferability of Splatt3R and DPT to vast aerial scenes and the accuracy of input poses.

free parameters (7)
  • Per-pair scale sigma_e in Eq. (3) = optimized per edge during global alignment
    Introduced as a variable in the projection-error objective (Eq. 3); it is fitted per image pair to resolve scale ambiguity.
  • Per-region global scale s_g^m = estimated per region via ICP
    Used in Eq. (4) to rescale covariance after pose alignment; fitted to each region's data.
  • Per-region local scale s_l^m = estimated from average adjacent-point distances
    Used in Eq. (4) to adjust near and far region scales; the paper gives no closed-form formula, only a qualitative description.
  • Supervision weights lambda_1, lambda_2, lambda_3 = 0.8, 0.2, 0.05
    Chosen constants in Eq. (7) and the distillation loss; no sensitivity study is reported.
  • Sliding window step = 2
    Hyperparameter for forming Splatt3R image pairs; affects coverage and redundancy.
  • Device training steps and densification interval = 10,000 steps, every 300
    Training schedule for each local model; chosen without reported sensitivity analysis.
  • Aggregation epochs = 5
    Number of distillation epochs for the merged global model; chosen without reported sensitivity analysis.
assumptions (5)
  • domain assumption Splatt3R transfers from its training distribution to sparse aerial vast-scene image pairs
    Section IV-A.1 uses the pretrained feed-forward model as the sole initializer; no aerial-domain validation for Splatt3R is provided.
  • domain assumption Drone-provided camera poses P_e in Eq. (3) are accurate enough to align pose-free Splatt3R pointmaps
    Section IV-A.2 builds the connectivity graph and projection error objective on these poses; pose error propagates directly into Gaussian positions.
  • domain assumption The pretrained DPT depth prior gives reliable relative depth on aerial scenes
    Section IV-B uses DPT in the correlation loss of Eq. (5); no aerial-specific validation of DPT is reported.
  • standard math The weighted least-squares objective in Eq. (3) has a unique, well-behaved global solution
    The global alignment step relies on this optimization without discussing initialization, local minima, or degeneracies.
  • standard math Standard 3DGS rasterization and photometric losses optimize Gaussian attributes correctly
    Sections III and IV-B assume the standard differentiable rasterization, L1, and SSIM losses behave as in 3D Gaussian Splatting.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DGTR: Distributed Gaussian Turbo-Reconstruction for Sparse-View Vast Scenes." pith.science (2026). https://pith.science/paper/XP2PITXE

@misc{pith2026241112309,
  author       = {Pith},
  title        = {Pith review of: DGTR: Distributed Gaussian Turbo-Reconstruction for Sparse-View Vast Scenes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XP2PITXE}},
  note         = {Machine review of arXiv:2411.12309}
}
read the original abstract

Novel-view synthesis (NVS) approaches play a critical role in vast scene reconstruction. However, these methods rely heavily on dense image inputs and prolonged training times, making them unsuitable where computational resources are limited. Additionally, few-shot methods often struggle with poor reconstruction quality in vast environments. This paper presents DGTR, a novel distributed framework for efficient Gaussian reconstruction for sparse-view vast scenes. Our approach divides the scene into regions, processed independently by drones with sparse image inputs. Using a feed-forward Gaussian model, we predict high-quality Gaussian primitives, followed by a global alignment algorithm to ensure geometric consistency. Synthetic views and depth priors are incorporated to further enhance training, while a distillation-based model aggregation mechanism enables efficient reconstruction. Our method achieves high-quality large-scale scene reconstruction and novel-view synthesis in significantly reduced training times, outperforming existing approaches in both speed and scalability. We demonstrate the effectiveness of our framework on vast aerial scenes, achieving high-quality results within minutes. Code will released on our [https://3d-aigc.github.io/DGTR].

Figures

Figures reproduced from arXiv: 2411.12309 by the authors.

Figure 1
Figure 1. Our proposed DGTR can rapidly reconstruct sparse-view vast scenes in a distributed manner. Compared with the standard central 3DGS training method, we achieve better visual appearance and geometry accuracy at a faster speed. Abstract— Novel-view synthesis (NVS) approaches play a critical role in vast scene reconstruction. However, these meth￾ods rely heavily on dense image inputs and prolonged training times, making… view at source ↗
Figure 2
Figure 2. DGTR (Ours) Overview: Given M individual devices (drones), we aim to perform sparse-view vast scene reconstruction with fast speed in a multi-device collaboration manner. The whole pipeline can be divided into three steps: 1) each device explores a non-overlap region and conducts Gaussian initialization using the off-the-shelf feed-forward Gaussian method and global alignment strategy; 2) each device performs sparse… view at source ↗
Figure 3
Figure 3. Overview of our Model Aggregation Algorithm. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Partitions of different scenes. ‘□’ denote partition areas for each drone. ‘·’ denote camera positions. The back￾ground is the sparse points produced by global COLMAP. the rendered depth Dras and the estimated depth Dest. Specifically, we utilize the alpha-blending ren…
Figure 5
Figure 5. Figure 5: Training curves for both ours and distributed-3DGS [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results on on Mill19 [32] dataset and UrbanScene3D [41] dataset. 3DGS-VastGS-DGTR Rubble 3DGS-VastGS-DGTR Building 3DGS-VastGS-DGTR Campus 3DGS-VastGS-DGTR Residence 0 2000 4000 6000 Overall Latency (s) 2003.5 2438.4 4908.1 5588.3 1487.6 1932.4 4432.1 5075.…
Figure 7
Figure 7. Figure 7: The overall latency comparison of our method, [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 9
Figure 9. Figure 9: Visual comparisons in scale S optimization of Σ. improvements by 9.87 dB and 2.47 dB. 2) Depth Regularisation: We conduct ablations of our depth regularisation method on the Rubble Scene. As shown in Tab. II, our depth supervision achieves better results in novel-view …

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. Proxy-GS: Unified Occlusion Priors for Training and Inference in Structured 3D Gaussian Splatting

    cs.CV 2025-09 conditional novelty 5.0 of 10

    A proxy mesh rendered through hardware rasterization provides a cheap occlusion depth prior that culls hidden anchors at inference and guides densification at training, giving Octree-GS-like MLP splatting a 3 to 4x sp...

Reference graph

Works this paper leans on

43 extracted references · 16 canonical work pages · cited by 1 Pith paper

  1. [10]

    Dogaussian: Distributed-oriented gaussian splatting for large-scale 3d reconstruction via gaussian consensus,

    Y . Chen and G. H. Lee, “Dogaussian: Distributed-oriented gaussian splatting for large-scale 3d reconstruction via gaussian consensus,” arXiv preprint arXiv:2405.13943 , 2024

    arXiv 2024

  2. [8]

    Multiplane prior guided few-shot aerial scene rendering,

    Z. Gao, L. Jiao, L. Li, X. Liu, F. Liu, P. Chen, and Y . Guo, “Multiplane prior guided few-shot aerial scene rendering,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 5009–5019

    work page 2024

  3. [1]

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

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoor- thi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,” Communications of the ACM , 2021

    2021

  4. [2]

    3d gaussian splatting for real-time radiance field rendering,

    B. Kerbl, G. Kopanas, T. Leimk ¨uhler, and G. Drettakis, “3d gaussian splatting for real-time radiance field rendering,” ACM Transactions on Graphics, vol. 42, no. 4, pp. 1–14, 2023

    2023

  5. [3]

    Waymo open dataset: An autonomous driving dataset,

    “Waymo open dataset: An autonomous driving dataset,” 2019

    work page 2019

  6. [4]

    Nerf2real: Sim2real transfer of vision-guided bipedal motion skills using neural radiance fields,

    A. Byravan, J. Humplik, L. Hasenclever, A. Brussee, F. Nori, T. Haarnoja, B. Moran, S. Bohez, F. Sadeghi, B. Vujatovic, et al. , “Nerf2real: Sim2real transfer of vision-guided bipedal motion skills using neural radiance fields,” in 2023 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2023, pp. 9362–9369

    2023

  7. [5]

    Renderable neural radiance map for visual navigation,

    O. Kwon, J. Park, and S. Oh, “Renderable neural radiance map for visual navigation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 9099–9108

    work page 2023

  8. [6]

    Dngaussian: Optimizing sparse-view 3d gaussian radiance fields with global-local depth normalization,

    J. Li, J. Zhang, X. Bai, J. Zheng, X. Ning, J. Zhou, and L. Gu, “Dngaussian: Optimizing sparse-view 3d gaussian radiance fields with global-local depth normalization,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 20 775–20 785

    2024

Show all 43 references
  1. [7]

    Fsgs: Real-time few-shot view synthesis using gaussian splatting,

    Z. Zhu, Z. Fan, Y . Jiang, and Z. Wang, “Fsgs: Real-time few-shot view synthesis using gaussian splatting,” arXiv preprint arXiv:2312.00451 , 2023

    2023 arXiv

  2. [9]

    Vastgaussian: Vast 3d gaussians for large scene reconstruction,

    J. Lin, Z. Li, X. Tang, J. Liu, S. Liu, J. Liu, Y . Lu, X. Wu, S. Xu, Y . Yan, et al. , “Vastgaussian: Vast 3d gaussians for large scene reconstruction,” arXiv preprint arXiv:2402.17427 , 2024

    2024 arXiv

  3. [11]

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

    G. Wang, Z. Chen, C. C. Loy, and Z. Liu, “Sparsenerf: Distilling depth ranking for few-shot novel view synthesis,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 9065–9076

    2023

  4. [12]

    Freenerf: Improving few-shot neural rendering with free frequency regularization,

    J. Yang, M. Pavone, and Y . Wang, “Freenerf: Improving few-shot neural rendering with free frequency regularization,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2023, pp. 8254–8263

    2023

  5. [13]

    Learning to render novel views from wide-baseline stereo pairs,

    Y . Du, C. Smith, A. Tewari, and V . Sitzmann, “Learning to render novel views from wide-baseline stereo pairs,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 4970–4980

    2023

  6. [14]

    Regnerf: Regularizing neural radiance fields for view synthesis from sparse inputs,

    M. Niemeyer, J. T. Barron, B. Mildenhall, M. S. Sajjadi, A. Geiger, and N. Radwan, “Regnerf: Regularizing neural radiance fields for view synthesis from sparse inputs,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 5480–5490

    2022

  7. [15]

    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 Transactions on Graphics (ToG) , vol. 38, no. 4, pp. 1–14, 2019

    2019

  8. [16]

    Objaverse: A universe of annotated 3d objects,

    M. Deitke, D. Schwenk, J. Salvador, L. Weihs, O. Michel, E. Vander- Bilt, L. Schmidt, K. Ehsani, A. Kembhavi, and A. Farhadi, “Objaverse: A universe of annotated 3d objects,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 13 142–13 153

    2023

  9. [17]

    Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images,

    Y . Chen, H. Xu, C. Zheng, B. Zhuang, M. Pollefeys, A. Geiger, T.- J. Cham, and J. Cai, “Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images,” arXiv preprint arXiv:2403.14627 , 2024

    2024 arXiv

  10. [18]

    Reconx: Reconstruct any scene from sparse views with video diffusion model,

    F. Liu, W. Sun, H. Wang, Y . Wang, H. Sun, J. Ye, J. Zhang, and Y . Duan, “Reconx: Reconstruct any scene from sparse views with video diffusion model,” arXiv preprint arXiv:2408.16767 , 2024

    2024 arXiv

  11. [19]

    pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction,

    D. Charatan, S. L. Li, A. Tagliasacchi, and V . Sitzmann, “pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 19 457–19 467

    2024

  12. [20]

    Flash3d: Feed-forward generalisable 3d scene reconstruction from a single image,

    S. Szymanowicz, E. Insafutdinov, C. Zheng, D. Campbell, J. F. Henriques, C. Rupprecht, and A. Vedaldi, “Flash3d: Feed-forward generalisable 3d scene reconstruction from a single image,” arXiv preprint arXiv:2406.04343, 2024

    2024 arXiv

  13. [21]

    Splatter image: Ultra-fast single-view 3d reconstruction,

    S. Szymanowicz, C. Rupprecht, and A. Vedaldi, “Splatter image: Ultra-fast single-view 3d reconstruction,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 10 208–10 217

    2024

  14. [22]

    Gps-gaussian: Generalizable pixel-wise 3d gaussian splatting for real- time human novel view synthesis,

    S. Zheng, B. Zhou, R. Shao, B. Liu, S. Zhang, L. Nie, and Y . Liu, “Gps-gaussian: Generalizable pixel-wise 3d gaussian splatting for real- time human novel view synthesis,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 19 680–19 690

    2024

  15. [23]

    Splatt3r: Zero- shot gaussian splatting from uncalibarated image pairs,

    B. Smart, C. Zheng, I. Laina, and V . A. Prisacariu, “Splatt3r: Zero- shot gaussian splatting from uncalibarated image pairs,” arXiv preprint arXiv:2408.13912, 2024

    2024 arXiv

  16. [24]

    Ggrt: Towards generalizable 3d gaussians without pose priors in real-time,

    H. Li, Y . Gao, D. Zhang, C. Wu, Y . Dai, C. Zhao, H. Feng, E. Ding, J. Wang, and J. Han, “Ggrt: Towards generalizable 3d gaussians without pose priors in real-time,” in ECCV, 2024

    2024

  17. [25]

    Attention is all you need,

    A. Vaswani, “Attention is all you need,” Advances in Neural Informa- tion Processing Systems , 2017

    2017

  18. [26]

    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

  19. [27]

    Building rome in a day,

    S. Agarwal, Y . Furukawa, N. Snavely, I. Simon, B. Curless, S. M. Seitz, and R. Szeliski, “Building rome in a day,” Communications of the ACM, vol. 54, no. 10, pp. 105–112, 2011

    2011

  20. [28]

    An automated method for large-scale, ground- based city model acquisition,

    C. Fr ¨uh and A. Zakhor, “An automated method for large-scale, ground- based city model acquisition,” International Journal of Computer Vision, vol. 60, pp. 5–24, 2004

    2004

  21. [29]

    Structure-from-motion revisited,

    J. L. Schonberger and J.-M. Frahm, “Structure-from-motion revisited,” in CVPR, 2016

    2016

  22. [30]

    Switch-nerf: Learning scene decomposition with mixture of experts for large-scale neural radiance fields,

    M. Zhenxing and D. Xu, “Switch-nerf: Learning scene decomposition with mixture of experts for large-scale neural radiance fields,” in The Eleventh International Conference on Learning Representations , 2022

    2022

  23. [31]

    Efficient large-scale scene represen- tation with a hybrid of high-resolution grid and plane features,

    Y . Zhang, G. Chen, and S. Cui, “Efficient large-scale scene represen- tation with a hybrid of high-resolution grid and plane features,” arXiv preprint arXiv:2303.03003, 2023

    2023 arXiv

  24. [32]

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

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

    2022

  25. [33]

    Block-nerf: Scalable large scene neural view synthesis,

    M. Tancik, V . Casser, X. Yan, S. Pradhan, B. Mildenhall, P. P. Srinivasan, J. T. Barron, and H. Kretzschmar, “Block-nerf: Scalable large scene neural view synthesis,” in CVPR, 2022, pp. 8248–8258

    2022

  26. [34]

    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

  27. [35]

    Grid-guided neural radiance fields for large urban scenes,

    L. Xu, Y . Xiangli, S. Peng, X. Pan, N. Zhao, C. Theobalt, B. Dai, and D. Lin, “Grid-guided neural radiance fields for large urban scenes,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 8296–8306

    2023

  28. [36]

    Dgnr: Density-guided neural point rendering of large driving scenes,

    Z. Li, C. Wu, L. Zhang, and J. Zhu, “Dgnr: Density-guided neural point rendering of large driving scenes,” IEEE Transactions on Au- tomation Science and Engineering , 2024

    2024

  29. [37]

    Citygaussian: Real-time high-quality large-scale scene rendering with gaussians,

    Y . Liu, H. Guan, C. Luo, L. Fan, J. Peng, and Z. Zhang, “Citygaussian: Real-time high-quality large-scale scene rendering with gaussians,” in Proceedings of the European conference on computer vision (ECCV) , 2024

    2024

  30. [38]

    HO-Gaussian: Hybrid optimization of 3d gaussian splatting for urban scenes,

    Z. Li, Y . Zhang, C. Wu, J. Zhu, and L. Zhang, “HO-Gaussian: Hybrid optimization of 3d gaussian splatting for urban scenes,” in Proceedings of the European conference on computer vision (ECCV) , 2024

    2024

  31. [39]

    Fed3dgs: Scalable 3d gaussian splatting with federated learning,

    T. Suzuki, “Fed3dgs: Scalable 3d gaussian splatting with federated learning,” arXiv preprint arXiv:2403.11460 , 2024

    2024 arXiv

  32. [40]

    Vision transformers for dense prediction,

    R. Ranftl, A. Bochkovskiy, and V . Koltun, “Vision transformers for dense prediction,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 12 179–12 188

    2021

  33. [41]

    Capturing, reconstructing, and simulating: the urbanscene3d dataset,

    L. Lin, Y . Liu, Y . Hu, X. Yan, K. Xie, and H. Huang, “Capturing, reconstructing, and simulating: the urbanscene3d dataset,” in European Conference on Computer Vision . Springer, 2022, pp. 93–109

    2022

  34. [42]

    Drone-nerf: Efficient nerf based 3d scene reconstruction for large-scale drone survey,

    Z. Jia, B. Wang, and C. Chen, “Drone-nerf: Efficient nerf based 3d scene reconstruction for large-scale drone survey,” Image and Vision Computing, vol. 143, p. 104920, 2024

    2024

  35. [43]

    Neuralrecon: Real-time coherent 3d reconstruction from monocular video,

    J. Sun, Y . Xie, L. Chen, X. Zhou, and H. Bao, “Neuralrecon: Real-time coherent 3d reconstruction from monocular video,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 15 598–15 607

    2021

Pith tools

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