Pith. sign in

REVIEW 4 major objections 4 minor 7 cited by

The visual failures of 3D Gaussian Splatting at novel viewpoints — floaters, broken geometry, black holes — can be repaired by a per-scene fine-tuned diffusion model whose output is distilled back into the 3D scene.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A per-scene fine-tuned latent diffusion model with dual mesh-3DGS conditioning and random mask augmentation improves novel-view repair in Gaussian Splatting, outperforming DIFIX baselines on ScanNet++ and Replica.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Practically useful and mostly honest novel-view repair paper: the per-scene diffusion fine-tuning works and the right control is run, but the mask augmentation is only validated on hand-picked Replica views and the main table omits the closest baseline. the 4 major comments →

arxiv 2508.14717 v1 pith:MHQNXVB5 submitted 2025-08-20 cs.CV

GSFix3D: Diffusion-Guided Repair of Novel Views in Gaussian Splatting

Jiaxin Wei , Stefan Leutenegger , Simon Schaefer This is my paper
classification cs.CV
keywords 3D Gaussian Splattingnovel view repairdiffusion modelslatent diffusion fine-tuningimage inpaintingartifact removalRGB-D SLAMmesh-Gaussian dual conditioning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

GSFix3D is built on the claim that the artifact-laden renderings 3D Gaussian Splatting produces at unseen viewpoints can be repaired by a pretrained image diffusion model that has been lightly fine-tuned on the very scene being reconstructed, without large curated noisy-clean datasets. The core module, GSFixer, conditions a latent diffusion model on two renderings of the target view, one from the Gaussians and one from a mesh, and, through a random-mask augmentation, learns both to erase reconstruction artifacts and to inpaint genuinely missing regions. The repaired frames are then treated as pseudo-images and optimized back into the Gaussian representation, so the fix persists in the scene rather than only in the frame. On ScanNet++, the paper reports GSFixer beating DIFIX and DIFIX-ref on every metric for all three reconstruction backbones tested, including a gain of more than 5 dB PSNR on RTG-SLAM maps, with the full pipeline adding further gains; on Replica's deliberately extreme novel views it leads in PSNR and SSIM, with the baselines sometimes sharper on LPIPS. The practical upshot claimed is that a few hours of fine-tuning on a single consumer GPU adapts one base model to new scenes, new artifact types, and even pose errors.

Core claim

At the paper's core is a fine-tuning protocol that turns a pretrained latent diffusion model into a scene-adapted repair module. From each captured frame the protocol renders the same view twice — once by alpha-blending 3D Gaussians, once by ray-casting the TSDF mesh obtained jointly with the Gaussians — and trains the denoiser to recover the captured image from these two degraded renderings. Randomly overlaid semantic masks from a real-image dataset force the model to inpaint, not just denoise. The central result: GSFixer, applied to novel views of SplaTAM, RTG-SLAM, and GSFusion reconstructions, beats DIFIX and DIFIX-ref on every metric on ScanNet++ (over 5 dB PSNR gain on RTG-SLAM), and r

What carries the argument

GSFixer is the carrying mechanism: the U-Net of Stable Diffusion v2, repurposed as a conditional denoiser by concatenating the latents of the two rendered views with the noisy target latent and duplicating the first-layer weights to absorb the extra input channels. Training minimizes the standard DDPM denoising objective on (3DGS render, mesh render, captured ground truth) triplets, with random semantic masks overlaid on both conditions to teach inpainting; a slight Gaussian blur on the 3DGS mask mimics soft rendering boundaries. Inference uses four-step DDIM sampling. Dual conditioning is what makes the module work: the mesh supplies coherent geometry where the Gaussians are missing, and th

Load-bearing premise

The load-bearing bet is that holes cut with semantic masks from natural photographs look enough like the smears, floaters, and unobserved gaps that really appear in novel 3D renderings — and the Replica evidence only samples hand-picked views, so it is not yet shown that differently shaped holes will be filled plausibly.

What would settle it

Take a 3DGS scene reconstructed from an uneven trajectory and evaluate on novel views chosen automatically (a held-out sequence slice) rather than hand-selected ones, including frames where one continuous unobserved region covers a third or more of the image — a shape no natural-image semantic mask resembles. If GSFixer's inpainted content in those regions loses its advantage over DIFIX, the mask-augmentation premise is the bottleneck; if the advantage holds, the model generalizes beyond its training mask shapes.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • The same captured frames that build a 3DGS scene also provide the fine-tuning signal, so novel-view repair no longer depends on curated noisy-clean pairs: pretraining on two synthetic datasets plus 400–800 scene-specific iterations suffices.
  • Because GSFixer operates on rendered images, it works as a plug-in for any 3DGS pipeline — demonstrated on SplaTAM, RTG-SLAM, and GSFusion — and the extra mesh condition is optional, not required.
  • Lifting repaired frames back into the 3D representation (GSFix3D) improves PSNR and SSIM over 2D-only repair, so multi-view consistency turns per-view fixes into a better scene model that benefits nearby viewpoints as well.
  • Four-step DDIM sampling and a few hours of fine-tuning on one consumer GPU make the repair stage fast and cheap enough for ordinary reconstruction workflows.
  • Demonstrated robustness to pose-induced floaters on uncontrolled real-world data implies the adaptation is not locked to one pipeline's artifact signature.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The paper's clearest generalization bet is the random-mask augmentation: it assumes natural-image semantic masks approximate the shape distribution of real novel-view holes. A testable extension would augment with masks derived from rendered depth discontinuities or occupancy gaps and compare extreme-view inpainting quality.
  • The dual-conditioning recipe should transfer to other representation pairs with complementary failure modes — a NeRF or depth-sensor mesh paired with photometric Gaussians — provided both are geometrically aligned, since the paper's own ablation shows the gain comes from the combination rather than either input alone.
  • The fine-tuning appetite is small enough that GSFixer could plausibly move inside the mapping loop, repairing and distilling during SLAM rather than after; the paper does not explore online adaptation.
  • Because the repaired views are generated, not measured, consistency across them is enforced only by the post-hoc multi-view optimization; very sparse repaired viewpoints could still harbor conflicting hallucinated content where no captured frame constrains them.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes GSFix3D, a pipeline for improving novel-view renderings from 3D Gaussian Splatting reconstructions. A latent diffusion model (GSFixer) is fine-tuned per scene to remove artifacts and inpaint missing regions, conditioned on renderings from both 3DGS and a mesh representation. The repaired images are then lifted back into 3DGS via photometric loss. The authors introduce a random-mask augmentation to teach inpainting, pretrain on synthetic data, and evaluate on ScanNet++ and Replica against DIFIX and DIFIX-ref, with additional real-world qualitative tests. The central claim is that GSFixer/GsFix3D achieves state-of-the-art novel-view repair with only minimal per-scene fine-tuning.

Significance. If the stated results hold, the paper would make a useful contribution: a plug-and-play diffusion-based repair module that adapts to scene-specific artifact statistics without large curated training pairs. The DIFIX-finetune control in the supplementary is a valuable step toward isolating the benefit of per-scene adaptation, and the real-world tests on pose-error-prone data are informative. The authors also commit to releasing code, data, and the curated Replica novel views, which would help reproducibility. However, the current evidence base is narrower than the claims: the main results rest on four ScanNet++ scenes and hand-selected Replica views, no error bars are reported, and the only quantitative validation of the mask augmentation is on those hand-selected Replica views. The 'state-of-the-art' language is also stronger than the comparison set supports.

major comments (4)
  1. [Sec. 4.2, Table 1, Supp. Tables 5-6] The statement that 'GSFixer consistently outperforms DIFIX and DIFIX-ref across all metrics on ScanNet++' is contradicted by Table 1 itself: for GSFusion(gs)+GSFixer, LPIPS is 0.196, worse than DIFIX (0.193) and DIFIX-ref (0.184). Against the fine-tuned control in Supp. Table 5, GSFixer wins PSNR in most rows but loses LPIPS in several, e.g., ScanNet++ GSFusion(gs): 0.196 vs 0.142 for DIFIX-finetune; Replica SplaTAM: 0.215 vs 0.149. Please qualify the claim (e.g., 'best PSNR/SSIM among compared methods'), report per-scene results and error bars, and discuss the LPIPS trade-off explicitly.
  2. [Sec. 3.1.2, Table 4, Supp. Sec. 6.1] The random-mask augmentation is the stated mechanism for inpainting, but its only quantitative validation is on Replica novel views that were manually selected to contain large unobserved regions (Supp. Sec. 6.1). This is a distribution-matched test for the augmentation, not evidence of transfer to non-hand-picked ScanNet++ evaluation views. No mask on/off ablation is reported for ScanNet++, where the main 5 dB PSNR claim is made. Add a ScanNet++ mask ablation or restrict the claim to hand-selected extreme views.
  3. [Sec. 4.1, Tables 1 and 4] Only four ScanNet++ scenes are used, no standard deviations or per-scene numbers are given, and the Replica evaluation set is hand-selected. The word 'consistently' is therefore not statistically supported. Please provide per-scene breakdowns, variance across scenes or runs, and ideally additional scenes or randomly sampled novel views, so the reader can judge the stability of the reported gains.
  4. [Abstract, Sec. 1, Sec. 4.2] The abstract and introduction claim 'state-of-the-art performance', but the experiments compare only against the DIFIX family (DIFIX and DIFIX-ref, plus the fine-tuned DIFIX in the supplementary). Baselines mentioned in Sec. 2.2, such as 3DGS-Enhancer, GenFusion, RI3D, and SGD, are not evaluated. If those methods cannot be applied to this setting, that should be stated and justified; otherwise the state-of-the-art claim is unsupported and should be restricted to 'compared baselines'.
minor comments (4)
  1. [Sec. 4.1] The text says fine-tuning takes 'typically 4 hours for ScanNet++ and 2 hours for Replica' while calling this 'minimal fine-tuning'. Clarify that 'minimal' refers to data curation and not compute, since 2-4 hours per scene is non-trivial in an online/robotics setting.
  2. [Supp. Sec. 7.1] DIFIX-finetune is trained on an NVIDIA A40 (48 GB) while GSFixer uses an RTX 4500 Ada (24 GB). This is a reasonable practical advantage, but it should be stated in the main paper if comparisons with DIFIX-finetune are used to support efficiency claims.
  3. [Supp. Sec. 6.2] For the ship dataset, the text says a nearby captured training view is used as a reference 'ground truth'. This is not ground truth; please use 'reference view' to avoid overstatement.
  4. [Sec. 4.4, Table 4] The mask ablation is reported only on Replica. Please state in the main text that the augmentation's effect on ScanNet++ is not directly measured, or add the corresponding ablation.

Circularity Check

0 steps flagged

No significant circularity: predictions are evaluated on held-out viewpoints with a fine-tuning-controlled baseline, and the GSFusion self-citation is not load-bearing.

full rationale

Walking the paper's derivation chain from fine-tuning through inference to 3D lifting, no step reduces a reported prediction to a fitted input or to a self-citation. GSFixer is fine-tuned on source-view captured data and conditional renderings (Sec. 3.1); the ScanNet++ evaluation uses an independent held-out camera trajectory (Sec. 4.1), and the Replica novel views are manually rendered ground-truth viewpoints (Supp. 6.1), none of which are in the fine-tuning set. The per-scene fine-tuning advantage is controlled by the DIFIX-finetune comparisons in Supp. Tables 5-6, where DIFIX receives the same scene-specific fine-tuning and still trails GSFixer. The dual-input conditioning uses GSFusion [33] to supply mesh and 3DGS maps, and [33] is first-authored by the paper's first author; however, this is a tool/system citation, not a theorem or uniqueness argument, and the repair mechanism is also demonstrated on SplaTAM and RTG-SLAM reconstructions that do not involve GSFusion, so the central claim does not depend on GSFusion's correctness. The random-mask augmentation (Sec. 3.1.2) is a training strategy to simulate holes; the Replica stress test deliberately selects views with large unobserved regions (Supp. 6.1). This raises a legitimate generalization/selection-bias concern about whether natural-image semantic masks match real novel-view hole statistics, but it is not circular in the formal sense: the augmentation masks are not the evaluation renderings, and no evaluation metric is optimized or fitted during training. Equations (1)-(5) are standard diffusion and photometric losses connecting the same quantities used in training and inference; no equation is defined in terms of the result it is used to support. Overall: no self-definitional, fitted-input, uniqueness-import, or ansatz-smuggling circularity.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The paper introduces no new physical entities; GSFixer is a trained neural network module, not a postulated mechanism. The free parameters are hyperparameters of the training and optimization pipeline; the axioms are the domain assumptions underpinning the fine-tuning and dual-input design.

free parameters (5)
  • Photometric loss weight lambda = not reported
    Weights L1 and SSIM in the 3DGS optimization loss (Sec. 3.3); value chosen by hand and not specified in the paper.
  • Fine-tuning iterations = 800 (ScanNet++), 400 (Replica)
    Chosen by hand; longer fine-tuning could overfit to training viewpoints, shorter might underfit artifacts.
  • Pretraining iterations = 6000
    Chosen by hand for the two synthetic datasets (Hypersim, Virtual KITTI).
  • DDIM inference steps = 4
    Accelerated sampling schedule for inference; affects output quality.
  • Gradient accumulation steps = 16
    Effective batch size adjustment on a 24GB GPU.
axioms (4)
  • domain assumption Per-scene fine-tuning on rendered training views generalizes to novel viewpoints.
    Central to the method (Sec. 3.1); if artifact distribution shifts between training and novel views, repair fails.
  • domain assumption Stable Diffusion v2's pretrained visual prior transfers to rendered image repair.
    The method relies on the pretrained U-Net initialization preserving generative quality (Sec. 3.1.1).
  • domain assumption Mesh and 3DGS renderings from GSFusion are geometrically aligned and complementary.
    Dual conditioning (Sec. 3.1) assumes both representations describe the same scene without contradictory cues.
  • standard math DDPM/DDIM sampling mathematics.
    Used in Eqs. 1-4 for forward noising and reverse denoising.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of GSFix3D: Diffusion-Guided Repair of Novel Views in Gaussian Splatting." pith.science (2026). https://pith.science/paper/MHQNXVB5

@misc{pith2026250814717,
  author       = {Pith},
  title        = {Pith review of: GSFix3D: Diffusion-Guided Repair of Novel Views in Gaussian Splatting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MHQNXVB5}},
  note         = {Machine review of arXiv:2508.14717}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Recent developments in 3D Gaussian Splatting have significantly enhanced novel view synthesis, yet generating high-quality renderings from extreme novel viewpoints or partially observed regions remains challenging. Meanwhile, diffusion models exhibit strong generative capabilities, but their reliance on text prompts and lack of awareness of specific scene information hinder accurate 3D reconstruction tasks. To address these limitations, we introduce GSFix3D, a novel framework that improves the visual fidelity in under-constrained regions by distilling prior knowledge from diffusion models into 3D representations, while preserving consistency with observed scene details. At its core is GSFixer, a latent diffusion model obtained via our customized fine-tuning protocol that can leverage both mesh and 3D Gaussians to adapt pretrained generative models to a variety of environments and artifact types from different reconstruction methods, enabling robust novel view repair for unseen camera poses. Moreover, we propose a random mask augmentation strategy that empowers GSFixer to plausibly inpaint missing regions. Experiments on challenging benchmarks demonstrate that our GSFix3D and GSFixer achieve state-of-the-art performance, requiring only minimal scene-specific fine-tuning on captured data. Real-world test further confirms its resilience to potential pose errors. Our code and data will be made publicly available. Project page: https://gsfix3d.github.io.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 7 Pith papers

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

  1. CORGI: Consistency-Aware 3D Dog Reconstruction from a Single Image in the Wild

    cs.CV 2026-07 unverdicted novelty 7.0

    A new pipeline using canonical LoRAs for view synthesis, deformable 3D Gaussian splatting anchored on D-SMAL, and generative repair to produce animatable 3D dogs from single wild images without 3D supervision.

  2. ConFixGS: Learning to Fix Feedforward 3D Gaussian Splatting with Confidence-Aware Diffusion Priors in Driving Scenes

    cs.CV 2026-05 unverdicted novelty 7.0

    ConFixGS repairs feedforward 3D Gaussian Splatting with confidence-aware diffusion priors, delivering up to 3.68 dB PSNR gains and halved FID scores on Waymo, nuScenes, and KITTI novel view synthesis tasks.

  3. Latent Riemannian Flow Matching for Geometry-Grounded 3D Foundation Models

    cs.CV 2026-07 conditional novelty 6.0

    Using a product-of-hyperspheres Riemannian flow matching model on VGGT's latent codes, the authors generate plausible depth, point maps, and RGB for target views from one to four unposed context images.

  4. CORGI: Consistency-Aware 3D Dog Reconstruction from a Single Image in the Wild

    cs.CV 2026-07 conditional novelty 6.0

    CORGI reconstructs high-fidelity, animatable 3D dogs from a single in-the-wild image via canonical orbital generation, deformable 3DGS anchored to D-SMAL, and self-supervised generative repair, without 3D supervision.

  5. RealityBridge: Bridging Editable 3D Gaussian Splatting Driving Simulations and Real-World Videos

    cs.CV 2026-06 conditional novelty 6.0

    A multimodal gated video model with targeted 3DGS-to-real data and reward post-training reduces artifacts, lighting mismatch, and flicker in edited driving simulations better than prior restorers.

  6. Nix and Fix: Targeting 1000x Compression of 3D Gaussian Splatting with Diffusion Models

    cs.CV 2026-02 unverdicted novelty 6.0

    NiFi applies artifact-aware, diffusion-based one-step distillation to compress 3D Gaussian Splatting to 0.1 MB while claiming state-of-the-art perceptual quality and up to 1000x rate reduction.

  7. StereoSplat+: Feed-Forward Stereo Gaussian Splatting with Diffusion-Assisted Progressive Inference

    cs.CV 2026-07 conditional novelty 5.5

    A dual-branch feed-forward 3DGS estimator plus one-shot diffusion-refined pseudo-view reinjection improves single-stereo novel-view and depth quality on KITTI-360 over prior feed-forward baselines.

Reference graph

Works this paper leans on

45 extracted references · 30 canonical work pages · cited by 6 Pith papers

  1. [1]

    Vir- tual kitti 2

    Yohann Cabon, Naila Murray, and Martin Humenberger. Vir- tual kitti 2. arXiv preprint arXiv:2001.10773, 2020. 1, 5

    Pith/arXiv arXiv 2001

  2. [2]

    A volumetric method for building complex models from range images

    Brian Curless and Marc Levoy. A volumetric method for building complex models from range images. InProceedings of the 23rd annual conference on Computer graphics and interactive techniques, pages 303–312, 1996. 2

    1996

  3. [3]

    Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering

    Antoine Gu ´edon and Vincent Lepetit. Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5354–5363, 2024. 3

    work page 2024

  4. [4]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 3

    2020

  5. [5]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models. ICLR, 1(2):3, 2022. 1

    work page 2022

  6. [6]

    2d gaussian splatting for geometrically ac- curate radiance fields

    Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically ac- curate radiance fields. In ACM SIGGRAPH 2024 conference papers, pages 1–11, 2024. 3

    work page 2024

  7. [7]

    Marigold: Affordable adaptation of diffusion- based image generators for image analysis

    Bingxin Ke, Kevin Qu, Tianfu Wang, Nando Metzger, Shengyu Huang, Bo Li, Anton Obukhov, and Konrad Schindler. Marigold: Affordable adaptation of diffusion- based image generators for image analysis. arXiv preprint arXiv:2505.09358, 2025. 3

    Pith/arXiv arXiv 2025

  8. [8]

    Splatam: Splat track & map 3d gaussians for dense rgb-d slam

    Nikhil Keetha, Jay Karhade, Krishna Murthy Jatavallabhula, Gengshan Yang, Sebastian Scherer, Deva Ramanan, and Jonathon Luiten. Splatam: Splat track & map 3d gaussians for dense rgb-d slam. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 21357–21366, 2024. 2, 5

    work page 2024

  9. [9]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph., 42(4):139–1,

  10. [10]

    Auto-encoding vari- ational bayes, 2013

    Diederik P Kingma, Max Welling, et al. Auto-encoding vari- ational bayes, 2013. 3

    2013

  11. [11]

    Gaussian- lic: Real-time photo-realistic slam with gaussian splat- ting and lidar-inertial-camera fusion

    Xiaolei Lang, Laijian Li, Chenming Wu, Chen Zhao, Lina Liu, Yong Liu, Jiajun Lv, and Xingxing Zuo. Gaussian- lic: Real-time photo-realistic slam with gaussian splat- ting and lidar-inertial-camera fusion. arXiv preprint arXiv:2404.06926, 2024. 7, 3, 5

    Pith/arXiv arXiv 2024

  12. [12]

    Okvis2: Realtime scalable visual-inertial slam with loop closure

    Stefan Leutenegger. Okvis2: Realtime scalable visual-inertial slam with loop closure. arXiv preprint arXiv:2202.09199, 2022. 7, 1

    Pith/arXiv arXiv 2022

  13. [13]

    Deceptive-nerf/3dgs: Diffusion- generated pseudo-observations for high-quality sparse-view reconstruction

    Xinhang Liu, Jiaben Chen, Shiu-Hong Kao, Yu-Wing Tai, and Chi-Keung Tang. Deceptive-nerf/3dgs: Diffusion- generated pseudo-observations for high-quality sparse-view reconstruction. In European Conference on Computer Vi- sion, pages 337–355. Springer, 2024. 2

    work page 2024

  14. [14]

    3dgs-enhancer: Enhancing unbounded 3d gaussian splatting with view- consistent 2d diffusion priors

    Xi Liu, Chaoyi Zhou, and Siyu Huang. 3dgs-enhancer: Enhancing unbounded 3d gaussian splatting with view- consistent 2d diffusion priors. Advances in Neural Informa- tion Processing Systems, 37:133305–133327, 2024. 2

    work page 2024

  15. [15]

    Gaussian splatting slam

    Hidenobu Matsuki, Riku Murai, Paul HJ Kelly, and An- drew J Davison. Gaussian splatting slam. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 18039–18048, 2024. 2

    work page 2024

  16. [16]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. Communications of the ACM, 65(1):99–106, 2021. 1, 2

    2021

  17. [17]

    T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models

    Chong Mou, Xintao Wang, Liangbin Xie, Yanze Wu, Jian Zhang, Zhongang Qi, and Ying Shan. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In Proceedings of the AAAI conference on artificial intelligence, pages 4296–4304, 2024. 1

    work page 2024

  18. [18]

    Kinectfusion: Real-time dense surface mapping and track- ing

    Richard A Newcombe, Shahram Izadi, Otmar Hilliges, David Molyneaux, David Kim, Andrew J Davison, Pushmeet Kohi, Jamie Shotton, Steve Hodges, and Andrew Fitzgibbon. Kinectfusion: Real-time dense surface mapping and track- ing. In 2011 10th IEEE international symposium on mixed and augmented reality, pages 127–136. Ieee, 2011. 2

    work page 2011

  19. [19]

    Glide: Towards photorealistic image generation and editing with text-guided diffusion models.arXiv preprint arXiv:2112.10741, 2021

    Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew, Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with text-guided diffusion models.arXiv preprint arXiv:2112.10741, 2021. 3

    Pith/arXiv arXiv 2021

  20. [20]

    Real-time 3d reconstruction at scale us- ing voxel hashing

    Matthias Nießner, Michael Zollh ¨ofer, Shahram Izadi, and Marc Stamminger. Real-time 3d reconstruction at scale us- ing voxel hashing. ACM Transactions on Graphics (ToG), 32 (6):1–11, 2013. 2

    work page 2013

  21. [21]

    V oxblox: Incremental 3d eu- clidean signed distance fields for on-board mav planning

    Helen Oleynikova, Zachary Taylor, Marius Fehr, Roland Siegwart, and Juan Nieto. V oxblox: Incremental 3d eu- clidean signed distance fields for on-board mav planning. In 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 1366–1373. IEEE, 2017. 2

    work page 2017

  22. [22]

    Ri3d: Few- shot gaussian splatting with repair and inpainting diffusion priors

    Avinash Paliwal, Xilong Zhou, Wei Ye, Jinhui Xiong, Rakesh Ranjan, and Nima Khademi Kalantari. Ri3d: Few- shot gaussian splatting with repair and inpainting diffusion priors. 2025. 2

    work page 2025

  23. [23]

    Rtg-slam: Real-time 3d re- construction at scale using gaussian splatting

    Zhexi Peng, Tianjia Shao, Yong Liu, Jingke Zhou, Yin Yang, Jingdong Wang, and Kun Zhou. Rtg-slam: Real-time 3d re- construction at scale using gaussian splatting. In ACM SIG- GRAPH 2024 Conference Papers, pages 1–11, 2024. 2, 5

    work page 2024

  24. [24]

    Hypersim: A photorealistic syn- thetic dataset for holistic indoor scene understanding

    Mike Roberts, Jason Ramapuram, Anurag Ranjan, Atulit Kumar, Miguel Angel Bautista, Nathan Paczan, Russ Webb, and Joshua M Susskind. Hypersim: A photorealistic syn- thetic dataset for holistic indoor scene understanding. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10912–10922, 2021. 1, 5 9

    work page 2021

  25. [25]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 1, 3

    work page 2022

  26. [26]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020. 3, 4

    Pith/arXiv arXiv 2010

  27. [27]

    Generative modeling by esti- mating gradients of the data distribution

    Yang Song and Stefano Ermon. Generative modeling by esti- mating gradients of the data distribution. Advances in neural information processing systems, 32, 2019. 3

    2019

  28. [28]

    Large-scale multi-resolution surface reconstruction from rgb-d sequences

    Frank Steinbrucker, Christian Kerl, and Daniel Cremers. Large-scale multi-resolution surface reconstruction from rgb-d sequences. In Proceedings of the IEEE International Conference on Computer Vision, pages 3264–3271, 2013. 2

    work page 2013

  29. [29]

    The replica dataset: A digital replica of indoor spaces

    Julian Straub, Thomas Whelan, Lingni Ma, Yufan Chen, Erik Wijmans, Simon Green, Jakob J Engel, Raul Mur-Artal, Carl Ren, Shobhit Verma, et al. The replica dataset: A digital replica of indoor spaces. arXiv preprint arXiv:1906.05797,

    Pith/arXiv arXiv 1906

  30. [30]

    imap: Implicit mapping and positioning in real-time

    Edgar Sucar, Shikun Liu, Joseph Ortiz, and Andrew J Davi- son. imap: Implicit mapping and positioning in real-time. In Proceedings of the IEEE/CVF international conference on computer vision, pages 6229–6238, 2021. 2

    work page 2021

  31. [31]

    Efficient octree-based volumetric slam supporting signed-distance and occupancy mapping

    Emanuele Vespa, Nikolay Nikolov, Marius Grimm, Luigi Nardi, Paul HJ Kelly, and Stefan Leutenegger. Efficient octree-based volumetric slam supporting signed-distance and occupancy mapping. IEEE Robotics and Automation Letters, 3(2):1144–1151, 2018. 2

    work page 2018

  32. [32]

    Paint by inpaint: Learning to add image objects by removing them first

    Navve Wasserman, Noam Rotstein, Roy Ganz, and Ron Kimmel. Paint by inpaint: Learning to add image objects by removing them first. In Proceedings of the Computer Vision and Pattern Recognition Conference , pages 18313–18324,

    work page

  33. [33]

    Gsfusion: Online rgb-d mapping where gaussian splatting meets tsdf fusion

    Jiaxin Wei and Stefan Leutenegger. Gsfusion: Online rgb-d mapping where gaussian splatting meets tsdf fusion. IEEE Robotics and Automation Letters, 2024. 2, 3, 5, 1

    work page 2024

  34. [34]

    Foundationstereo: Zero- shot stereo matching

    Bowen Wen, Matthew Trepte, Joseph Aribido, Jan Kautz, Orazio Gallo, and Stan Birchfield. Foundationstereo: Zero- shot stereo matching. In Proceedings of the Computer Vi- sion and Pattern Recognition Conference, pages 5249–5260,

    work page

  35. [35]

    Difix3d+: Improving 3d reconstruc- tions with single-step diffusion models

    Jay Zhangjie Wu, Yuxuan Zhang, Haithem Turki, Xuanchi Ren, Jun Gao, Mike Zheng Shou, Sanja Fidler, Zan Goj- cic, and Huan Ling. Difix3d+: Improving 3d reconstruc- tions with single-step diffusion models. In Proceedings of the Computer Vision and Pattern Recognition Conference , pages 26024–26035, 2025. 1, 2, 5

    work page 2025

  36. [36]

    Genfusion: Closing the loop between recon- struction and generation via videos

    Sibo Wu, Congrong Xu, Binbin Huang, Andreas Geiger, and Anpei Chen. Genfusion: Closing the loop between recon- struction and generation via videos. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 6078–6088, 2025. 2

    work page 2025

  37. [37]

    Gs-slam: Dense visual slam with 3d gaussian splatting

    Chi Yan, Delin Qu, Dan Xu, Bin Zhao, Zhigang Wang, Dong Wang, and Xuelong Li. Gs-slam: Dense visual slam with 3d gaussian splatting. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 19595–19604, 2024. 2

    2024

  38. [38]

    Scannet++: A high-fidelity dataset of 3d in- door scenes

    Chandan Yeshwanth, Yueh-Cheng Liu, Matthias Nießner, and Angela Dai. Scannet++: A high-fidelity dataset of 3d in- door scenes. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 12–22, 2023. 5

    work page 2023

  39. [39]

    Sgd: Street view synthesis with gaussian splatting and diffusion prior

    Zhongrui Yu, Haoran Wang, Jinze Yang, Hanzhang Wang, Jiale Cao, Zhong Ji, and Mingming Sun. Sgd: Street view synthesis with gaussian splatting and diffusion prior. In2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 3812–3822. IEEE, 2025. 2

    work page 2025

  40. [40]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3836–3847, 2023. 1

    2023

  41. [41]

    Fast-livo: Fast and tightly- coupled sparse-direct lidar-inertial-visual odometry

    Chunran Zheng, Qingyan Zhu, Wei Xu, Xiyuan Liu, Qizhi Guo, and Fu Zhang. Fast-livo: Fast and tightly- coupled sparse-direct lidar-inertial-visual odometry. In 2022 IEEE/RSJ international conference on intelligent robots and systems (IROS), pages 4003–4009. IEEE, 2022. 7, 3, 5

    work page 2022

  42. [42]

    Taming video diffusion prior with scene- grounding guidance for 3d gaussian splatting from sparse in- puts

    Yingji Zhong, Zhihao Li, Dave Zhenyu Chen, Lanqing Hong, and Dan Xu. Taming video diffusion prior with scene- grounding guidance for 3d gaussian splatting from sparse in- puts. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 6133–6143, 2025. 2

    work page 2025

  43. [43]

    Nice-slam: Neural implicit scalable encoding for slam

    Zihan Zhu, Songyou Peng, Viktor Larsson, Weiwei Xu, Hu- jun Bao, Zhaopeng Cui, Martin R Oswald, and Marc Polle- feys. Nice-slam: Neural implicit scalable encoding for slam. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 12786–12796, 2022. 2, 5, 1 10 GSFix3D: Diffusion-Guided Repair of Novel Views in Gaussian ...

    work page 2022

  44. [44]

    ground truth

    Data Preparation 6.1. Novel View Selection for Replica The Replica dataset [29] contains high-quality reconstruc- tions of diverse indoor scenes, featuring clean dense ge- ometry and high-resolution textures. We leverage the pro- vided 3D models and the official Replica SDK to render novel view images, which serve as ground truth for quan- titative evalua...

    work page

  45. [45]

    Additional Results 7.1. More DIFIX Variants DIFIX and DIFIX-ref [35] are diffusion models pretrained on 80k noisy-clean real image pairs created using their pro- posed dataset curation strategies, whereas our GSFixer is only pretrained on two synthetic datasets with randomly added Gaussian noise and blur, followed by fine-tuning on a small amount of clean...

    work page

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.