Pith. sign in

REVIEW 3 major objections 6 minor 47 references

InstaInpaint: Instant 3D-Scene Inpainting with Masked Large Reconstruction Model

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read A single feed-forward network can inpaint 3D scenes in 0.4 seconds, matching the quality of optimization methods that take hours.

desk verdict Fast feed-forward 3D inpainting with real speed gains and honest ablations; the reference-distribution shift is a real gap but not fatal. read the letter →

arxiv 2506.10980 v1 pith:XSFZGNQR submitted 2025-06-12 cs.CV

classification cs.CV
keywords 3DsceneinpaintinglargereconstructionmodelGaussiansplattingfeed-forwardeditingmaskedfine-tuningreference-basednovelviewsynthesisobjectinsertion
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that 3D scene inpainting—repairing or completing a reconstructed scene after removing or inserting objects—can be executed as a single feed-forward pass rather than a slow per-scene optimization. The proposed model, InstaInpaint, takes four posed images, multi-view masks, and one 2D-inpainted reference view, and outputs a 3D Gaussian splatting scene in about 0.4 seconds. On two standard benchmarks this speed is achieved while matching or beating optimization-based methods that take 40 minutes to 15 hours. If the claim holds, real-time interactive editing of reconstructed scenes becomes practical, and 3D inpainting stops being a bottleneck for VR/AR editing workflows.

What carries the argument

The masked large reconstruction model (masked LRM): a ViT-based transformer with Plücker-ray conditioning that consumes four posed images and predicts per-pixel 3D Gaussian parameters. During masked fine-tuning, the reference view is encoded intact, while the other three input views have masked regions replaced with gray pixels and concatenated with binary masks, so the model learns to copy appearance from the reference while inferring geometry from context. The output tokens decode into pixel-aligned Gaussians that are merged into one scene and supervised by photometric losses on held-out novel views. The three mask-generation strategies—object masks from a video segmentation model, geometric masks obtained by depth-warping random ovals from the reference view, and random image masks—are the mechanism that prevents object bias and teaches the model to handle arbitrary edited regions.

What would settle it

Take a trained InstaInpaint model and feed it a reference image whose inpainted content strongly contradicts the surrounding geometry—for example, a small object drawn floating in mid-air with no depth cue. If the model propagates that appearance into novel views without correcting the geometry, the novel-view renders will show the object unattached to the scene, which would reveal that the model copies reference texture rather than inferring scene-consistent geometry. A quantitative version: measure the disparity or depth discontinuity between the inpainted region and its surroundings in the predicted Gaussians and compare it with the true scene geometry.

Watch

Extended reading notes

Core claim

The central discovery is that an LRM pretrained for reconstruction can be repurposed into an instant 3D inpainter by masked fine-tuning on large-scale real-world video data, without needing before-and-after 3D training pairs. The model treats the single untouched reference view as the source of appearance for the gray-masked regions in the other input views, and learns to infer geometry for those regions from surrounding context. Three generated mask types—video-segmentation object masks, depth-warped geometric masks, and random image masks—force the model to generalize beyond simple object removal. The authors argue that the design choice of masking only the non-reference views, not the reference view, is essential for keeping the reference tokens in the same data distribution as reconstructed tokens.

Load-bearing premise

The model is trained with the reference view showing the true original appearance of the scene, but at test time the reference is a 2D-inpainted image that can contain geometry the model has never seen; the claim rests on the transfer from real appearance to hallucinated appearance being good enough.

Editorial extensions

If this is right

  • Interactive editing of reconstructed scenes becomes feasible: object removal, object insertion, and multi-region inpainting all run in one 0.4-second pass instead of minutes-to-hours of per-scene optimization.
  • The same architecture can switch between removal and insertion by swapping the 2D inpainting model that produces the reference image, including text-driven diffusion inpainting.
  • Since the model is a feed-forward LRM, it can be trained on large-scale real-world video datasets without expensive annotated before/after 3D pairs, so further scaling is relatively cheap.
  • Scene editing and novel-view synthesis are unified into a single network, so the edited scene can be rendered immediately without a separate reconstruction step.

Reading between the lines

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

  • Editorially, the same masked-finetuning recipe could turn other reconstruction-first models into editors for relighting or material changes, since the reference view need only supply a new appearance for the masked region.
  • Editorially, the measured FID/KID numbers inherit the ceiling of the 2D diffusion inpainter that produces the reference; swapping in a stronger or multi-view-consistent 2D generator should lift 3D output quality without retraining.
  • Editorially, the declared limitation on dynamic scenes and sparse four-view coverage implies a stress test: fast-moving objects or wide baseline captures would likely break the feed-forward geometry assumption, so practical deployment would need a fallback optimization stage.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. InstaInpaint proposes a feed-forward large reconstruction model for 3D scene inpainting. Given posed input views, multi-view masks, and one 2D-inpainted reference view, the model predicts per-pixel 3D Gaussians and renders the completed scene in about 0.4 s, compared with optimization-based baselines that take tens of minutes to hours. The model is first trained as a GS-LRM reconstruction model and then masked-finetuned on DL3DV video clips; the authors generate three kinds of training masks (video object masks, depth-warped geometric masks, and random masks) and ablate mask encoding and sampling. Experiments on SPIn-NeRF and LLFF compare against NeRF/3DGS inpainting methods and LRM-based baselines using LPIPS, FID, KID and border variants; additional qualitative results show object insertion and multi-region inpainting. The paper's central claims are a 1000x speed-up with state-of-the-art quality and learned geometric consistency for inpainted regions.

Significance. If the claims hold, the paper makes a strong practical contribution: it converts 3D scene inpainting from a per-scene optimization problem into a sub-second feed-forward inference problem, which is directly relevant for interactive VR/AR editing. The self-supervised masked-finetuning scheme is a sensible way to avoid collecting paired before/after 3D data, and the mask-generation taxonomy (object/geometric/random) is a useful analysis. The empirical study is broad: two standard benchmarks, optimization-based and LRM-based baselines, and ablations on mask types, mask encoding, and sampling probabilities. However, the significance is conditional on two things the paper does not yet establish: that the model transfers from real reference views at training to hallucinated diffusion reference views at test time, and that the reported metric differences are statistically reliable. The paper does not claim code release, and the method has no machine-checked derivations; its value rests on the experimental evidence.

major comments (3)
  1. [Sec. 3.2 (Masked Finetuning), Eq. (2); Tables 1-3] The central transfer assumption is unmeasured. During masked finetuning, the reference view is always a real, geometrically consistent frame of the same video clip (Eq. 2 keeps I_ref intact), and all photometric losses are computed against real held-out views. At test time the reference is the output of a 2D diffusion inpainter (Flux), which has no true 3D geometry and can violate scale, orientation, and occlusion. The reported benchmarks (Tables 1, 2, 4, 5) contain only image-space metrics; no depth, point-cloud, or multi-view geometric consistency metric is reported for the inpainted volume, and Table 3 uses a ground-truth reference, which cannot probe this distribution shift. The limitations paragraph in Section 5 mentions dynamic scenes and sparse coverage but not this shift. I would need either (a) a direct test of geometry transfer (e.g., rendering depth/point clouds of the inpainted region from novel views and comparing them with the reference's implied geometry, for real vs. diffusion references), or (b) training/evaluation on references with intentionally inconsistent geometry, before accepting the claim that the model learned to identify geometry from the surrounding context for hallucinated content.
  2. [Tables 1, 2, 4, 5] No error bars, confidence intervals, or multiple-seed results are reported. Some of the headline differences are small: in Table 2, LPIPS is 0.4122 for MVInpainter+LRM versus 0.4147 for InstaInpaint, and several ablations in Table 4 differ by less than 1 FID point (e.g., 84.535 vs. 85.645). Without variance estimates it is not possible to know whether the state-of-the-art and ablation conclusions reflect systematic improvements or noise. The GScream column in Table 1 is also missing on the LLFF side ('- -') without a footnote, which makes the benchmark comparison incomplete. Please add multiple seeds or confidence intervals, and either fill in or explain the missing GScream LLFF numbers.
  3. [Sec. 4.2 / Table 4 / Fig. 11] The claim that object masks improve geometric consistency for inserted instances is supported only by qualitative images (the tilted traffic cone and twisted minion in Fig. 11) and by roughly equal quantitative scores. Since geometric consistency is one of the paper's explicit design goals, this should be measured (e.g., depth-map consistency across novel views, or a chamfer/point-cloud metric in the masked region). Without such a metric, the design choice is not quantitatively validated.
minor comments (6)
  1. [Table 1] The GScream LLFF entries are missing without explanation; please add the numbers or a footnote stating why they are unavailable.
  2. [Section 3.3] The text says 'Insection-over-Union' and should read 'intersection-over-union'; Eq. (3) would also benefit from an explicit statement of the homogeneous-coordinate conversion when applying the inverse extrinsic.
  3. [Throughout] There are several typos: 'freeforwardly' in Section 1, 'MVInapinter' in the Figure 7 caption, 'Boarder Impact' and 'Mechenisms' in Appendix D, and 'PNSR' in Table 3; reference [15] also has a garbled author string ('Sanja annd Williams').
  4. [Section 3.2] The phrase 'quartile frames' is not the right description for the frame selection example (1st, 5th, 10th, 15th of 15 frames); please rephrase to something like 'four evenly spaced frames'.
  5. [Section 4 / Table 5] The mask-sampling probabilities in Table 5 appear to be selected on the same evaluation datasets used for the final numbers; please state whether a validation split was used to avoid selection bias.
  6. [Section 5] The Limitations paragraph should explicitly acknowledge the reference-distribution shift between training (real reference views) and inference (diffusion-inpainted references) and note that no geometric metric is currently reported.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the masked-finetuning objective is supervised by held-out views, and the reported speed/quality results are measured against external benchmarks and baselines.

full rationale

The paper's central claim is a feed-forward 3D inpainter trained by masked fine-tuning of an LRM. I checked the derivation chain for places where an output is defined as an input, a fitted parameter is renamed a prediction, or a load-bearing premise rests on a self-citation. None of these appear. In Sec. 3.2, training masks regions of non-reference views with gray pixels while the reference view stays intact; supervision is computed on held-out views from the same video clips (Fig. 4), so the rendered novel views are not part of the input. This is a standard self-supervised masked-reconstruction objective and is not equivalent to copying the reference by construction. The geometric masks in Sec. 3.3 are created by warping random ovals through depth from a first-stage GS-LRM, but they only define the inpainted region; the photometric supervision still comes from real held-out views, so the final quality is not forced by the mask-generation model. The 1000x speed-up is a wall-clock comparison against per-scene optimization baselines (Table 1), and the SOTA claim is supported by external benchmarks with a shared 2D diffusion reference for all reference-based methods (Sec. 4), not by a fitted constant. The only self-citations (e.g., MALD-NeRF [28], which shares two authors with this paper) appear as a comparison baseline and are not load-bearing for the method. The limitations paragraph (Sec. 5) identifies dynamic-scene and sparse-view weaknesses; a separate concern is the train/test gap between real references and diffusion-inpainted references, but that is a generalization/validity risk, not a circular derivation, since the paper does not define the output in terms of the training references.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The model introduces no new physical entities. The load-bearing assumptions are the quality of the training dataset, the accuracy of the first-stage depth predictions, and the sufficiency of the photometric supervision. The tunable hyperparameters (mask mix, view counts) are ablations, not constants fitted to the evaluation outcome.

free parameters (3)
  • mask sampling probabilities = object 25%, geometric 25%, random 50%
    Tuned via ablation in Table 5; the chosen distribution gives best FID/KID on both benchmarks.
  • number of input views N = 4
    Architectural choice following GS-LRM; the model is trained with exactly 4 input views and this is not ablated.
  • supervision views M = 8
    Training uses 8 supervision views for photometric loss; a fixed design choice.
assumptions (4)
  • standard math Pinhole camera model with known intrinsics/extrinsics and Plücker ray coordinates is a valid parametrization of viewing geometry.
    Used throughout for encoding input views and for depth-based mask warping (Eq. 3-4).
  • domain assumption DL3DV-10K provides a sufficiently diverse and geometrically accurate set of real-world scenes for training a generalizable 3D inpainting model.
    No independent validation of pose or depth quality; the method relies on COLMAP poses and GS-LRM depth.
  • domain assumption The first-stage GS-LRM predicts metric depth accurate enough for cross-view mask warping.
    Geometric mask generation (Section 3.3) projects points from the reference depth map; inaccurate depth would misalign masks across views.
  • domain assumption Photometric loss on unmasked novel views is a sufficient training signal for learning reference-appearance propagation into masked regions.
    The model is never directly supervised with 3D geometry of the inpainted region; correctness relies on the rendering loss.

how reviews work

0 comments
Cite this review

Pith. "Pith review of InstaInpaint: Instant 3D-Scene Inpainting with Masked Large Reconstruction Model." pith.science (2026). https://pith.science/paper/XSFZGNQR

@misc{pith2026250610980,
  author       = {Pith},
  title        = {Pith review of: InstaInpaint: Instant 3D-Scene Inpainting with Masked Large Reconstruction Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XSFZGNQR}},
  note         = {Machine review of arXiv:2506.10980}
}
read the original abstract

Recent advances in 3D scene reconstruction enable real-time viewing in virtual and augmented reality. To support interactive operations for better immersiveness, such as moving or editing objects, 3D scene inpainting methods are proposed to repair or complete the altered geometry. However, current approaches rely on lengthy and computationally intensive optimization, making them impractical for real-time or online applications. We propose InstaInpaint, a reference-based feed-forward framework that produces 3D-scene inpainting from a 2D inpainting proposal within 0.4 seconds. We develop a self-supervised masked-finetuning strategy to enable training of our custom large reconstruction model (LRM) on the large-scale dataset. Through extensive experiments, we analyze and identify several key designs that improve generalization, textural consistency, and geometric correctness. InstaInpaint achieves a 1000x speed-up from prior methods while maintaining a state-of-the-art performance across two standard benchmarks. Moreover, we show that InstaInpaint generalizes well to flexible downstream applications such as object insertion and multi-region inpainting. More video results are available at our project page: https://dhmbb2.github.io/InstaInpaint_page/.

Figures

Figures reproduced from arXiv: 2506.10980 by the authors.

Figure 1
Figure 1. Overview. InstaInpaint generates an inpainted 3D scene from a set of posed images, a set of multi-view foreground instance masks, and a 2D reference image. The 3D reconstruction and inpainting process takes only 0.4 seconds. InstaInpaint is a generalizable framework supporting background inpainting, object insertion and multi-region inpainting simultaneously. Abstract Recent advances in 3D scene reconstruction enabl… view at source ↗
Figure 2
Figure 2. 3D Inpainting quality and speed. Left: Our proposed method reconstructs the inpainted scene at a much faster speed with more competitive quality compared with existing approaches. Right: Our proposed method takes only 0.4s for reconstruction and 3D inpainting. 1 Introduction Recent advances in neural reconstruction [1, 2, 3, 4, 5] achieve photorealistic and real-time rendering in virtual and augmented reality, enabl… view at source ↗
Figure 4
Figure 4. Overall pipeline of Masked Finetuning. Given a video clip, a reference view and three inpaint views are selected. The reference view remains intact - its RGB values and corresponding Plücker ray coordinates are directly tokenized. For the inpainting views, we first apply multi-view masks to the images, then concatenate these masked images with their Plücker coordinates and the binary masks before tokenization. Token… view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: Overview of mask generation methods. a) Object Mask Generation. We propagate scene videos through a video segmentation model to obtain object masks. b) Random Image Mask. We randomly mask the same area for all input views. c) Geometric Mask Generation. We use a GS-LRM …
Figure 6
Figure 6. Figure 6: Qualitative comparisons with state-of-the-art methods.. InstaInpaint obtains more plausible inpainted texture, better consistency and smoother transition at the inpainting border. The top 4 rows are from SPIn-NeRF and the bottom 4 rows are from LLFF [PITH_FULL_IMAGE:f…
Figure 7
Figure 7. Figure 7: Qualitative comparisons with LRM-based methods. InstaInpaint gives cleaner and sharper results with better cross-view consistency. Camera Pose InFusion InstaInpaint [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 9
Figure 9. Figure 9: Qualitative comparisons of object insertion ability. InstaInpaint predicts plausible geometry for inserted objects and seamlessly integrates inpaint regions with reconstruction regions. 1000× speedup. InstaInpaint also provides a smoother transition at the inpainting b…
Figure 10
Figure 10. Figure 10: Multiple region inpainting. InstaInpaint supports multiple disjoint inpainting regions. NV1 NV2 NV1 NV2 w/o Object Mask w/ Object Mask [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 33 canonical work pages

  1. [1]

    Nerf: Representing scenes as neural radiance fields for view synthesis

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

  2. [2]

    Tensorf: Tensorial radiance fields

    Anpei Chen, Zexiang Xu, Andreas Geiger, Jingyi Yu, and Hao Su. Tensorf: Tensorial radiance fields. In ECCV, 2022. 2

  3. [3]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph., 42(4):139–1, 2023. 2

  4. [4]

    Scaffold-gs: Structured 3d gaussians for view-adaptive rendering

    Tao Lu, Mulin Yu, Linning Xu, Yuanbo Xiangli, Limin Wang, Dahua Lin, and Bo Dai. Scaffold-gs: Structured 3d gaussians for view-adaptive rendering. In CVPR, pages 20654–20664, 2024. 2

  5. [5]

    2d gaussian splatting for geometrically accurate radiance fields

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

  6. [6]

    LRM: Large reconstruction model for single image to 3d

    Yicong Hong, Kai Zhang, Jiuxiang Gu, Sai Bi, Yang Zhou, Difan Liu, Feng Liu, Kalyan Sunkavalli, Trung Bui, and Hao Tan. LRM: Large reconstruction model for single image to 3d. In ICLR, 2024. 2, 3, 4

  7. [7]

    Gs-lrm: Large reconstruction model for 3d gaussian splatting

    Kai Zhang, Sai Bi, Hao Tan, Yuanbo Xiangli, Nanxuan Zhao, Kalyan Sunkavalli, and Zexiang Xu. Gs-lrm: Large reconstruction model for 3d gaussian splatting. In European Conference on Computer Vision, pages 1–19, 2024. 2, 3, 4, 6

  8. [8]

    Stereo magnification: Learning view synthesis using multiplane images

    Tinghui Zhou, Richard Tucker, John Flynn, Graham Fyffe, and Noah Snavely. Stereo magnification: Learning view synthesis using multiplane images. ACM Trans. Graph. (Proc. SIGGRAPH), 37, 2018. 2, 4

Show all 47 references
  1. [9]

    Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision

    Lu Ling, Yichen Sheng, Zhi Tu, Wentian Zhao, Cheng Xin, Kun Wan, Lantao Yu, Qianyu Guo, Zixun Yu, Yawen Lu, et al. Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision. In CVPR, 2024. 2, 4, 6, 12

  2. [10]

    Mvinpainter: Learning multi-view consistent inpainting to bridge 2d and 3d editing

    Chenjie Cao, Chaohui Yu, Fan Wang, Xiangyang Xue, and Yanwei Fu. Mvinpainter: Learning multi-view consistent inpainting to bridge 2d and 3d editing. In NeurIPS, 2024. 2, 3, 6

  3. [11]

    SAM 2: Segment anything in images and videos

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, H Khedr, R Rädle, C Rolland, L Gustafson, et al. SAM 2: Segment anything in images and videos. In ICLR, 2025. 2, 5

  4. [12]

    Lgm: Large multi-view gaussian model for high-resolution 3d content creation

    Jiaxiang Tang, Zhaoxi Chen, Xiaokang Chen, Tengfei Wang, Gang Zeng, and Ziwei Liu. Lgm: Large multi-view gaussian model for high-resolution 3d content creation. In ECCV, 2024. 3

  5. [13]

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

    Yuedong Chen, Haofei Xu, Chuanxia Zheng, Bohan Zhuang, Marc Pollefeys, Andreas Geiger, Tat-Jen Cham, and Jianfei Cai. Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images. In ECCV,

  6. [14]

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

    David Charatan, Sizhe Lester Li, Andrea Tagliasacchi, and Vincent Sitzmann. pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction. In CVPR, pages 19457–19467, 2024. 3

  7. [15]

    Scube: Instant large-scale scene reconstruction using voxsplats

    Xuanchi Ren, Yifan Lu, Hanxue Liang, Jay Zhangjie Wu, Huan Ling, Mike Chen, Francis Fidler, Sanja annd Williams, and Jiahui Huang. Scube: Instant large-scale scene reconstruction using voxsplats. In NeurIPS, 2024. 3

  8. [16]

    Lvsm: A large view synthesis model with minimal 3d inductive bias

    Haian Jin, Hanwen Jiang, Hao Tan, Kai Zhang, Sai Bi, Tianyuan Zhang, Fujun Luan, Noah Snavely, and Zexiang Xu. Lvsm: A large view synthesis model with minimal 3d inductive bias. In ICLR, 2025. 3

  9. [17]

    3d mesh editing using masked lrms

    William Gao, Dilin Wang, Yuchen Fan, Aljaž Boži ˇc, Tuur Stuyck, Zhengqin Li, Zhao Dong, Rakesh Ranjan, and Nikolaos Sarafianos. 3d mesh editing using masked lrms. arXiv preprint arXiv:2412.08641,

  10. [18]

    Nerf-in: Free-form nerf inpainting with rgb-d priors

    Hao-Kang Liu, I Shen, Bing-Yu Chen, et al. Nerf-in: Free-form nerf inpainting with rgb-d priors. arXiv preprint arXiv:2206.04901, 2022. 3

  11. [19]

    Removing objects from neural radiance fields

    Silvan Weder, Guillermo Garcia-Hernando, Aron Monszpart, Marc Pollefeys, Gabriel J Brostow, Michael Firman, and Sara Vicente. Removing objects from neural radiance fields. In CVPR, pages 16528–16538,

  12. [20]

    Or-nerf: Object removing from 3d scenes guided by multiview segmentation with neural radiance fields

    Youtan Yin, Zhoujie Fu, Fan Yang, and Guosheng Lin. Or-nerf: Object removing from 3d scenes guided by multiview segmentation with neural radiance fields. arXiv preprint arXiv:2305.10503, 2023. 3 10

  13. [21]

    Mvip-nerf: Multi-view 3d inpainting on nerf scenes via diffusion prior

    Honghua Chen, Chen Change Loy, and Xingang Pan. Mvip-nerf: Multi-view 3d inpainting on nerf scenes via diffusion prior. In CVPR, 2024. 3

  14. [22]

    Gaussianeditor: Swift and controllable 3d editing with gaussian splatting

    Yiwen Chen, Zilong Chen, Chi Zhang, Feng Wang, Xiaofeng Yang, Yikai Wang, Zhongang Cai, Lei Yang, Huaping Liu, and Guosheng Lin. Gaussianeditor: Swift and controllable 3d editing with gaussian splatting. In CVPR, 2023. 3

  15. [23]

    Zero-1-to-3: Zero-shot one image to 3d object

    Ruoshi Liu, Rundi Wu, Basile Van Hoorick, Pavel Tokmakov, Sergey Zakharov, and Carl V ondrick. Zero-1-to-3: Zero-shot one image to 3d object. In CVPR, pages 9298–9309, 2023. 3

  16. [24]

    Mvdream: Multi-view diffusion for 3d generation

    Yichun Shi, Peng Wang, Jianglong Ye, Mai Long, Kejie Li, and Xiao Yang. Mvdream: Multi-view diffusion for 3d generation. arXiv preprint arXiv:2308.16512, 2023. 3

  17. [25]

    Zero123++: a single image to consistent multi-view diffusion base model

    Ruoxi Shi, Hansheng Chen, Zhuoyang Zhang, Minghua Liu, Chao Xu, Xinyue Wei, Linghao Chen, Chong Zeng, and Hao Su. Zero123++: a single image to consistent multi-view diffusion base model. arXiv preprint arXiv:2310.15110, 2023. 3

  18. [26]

    Spin-nerf: Multiview segmentation and perceptual inpainting with neural radiance fields

    Ashkan Mirzaei, Tristan Aumentado-Armstrong, Konstantinos G Derpanis, Jonathan Kelly, Marcus A Brubaker, Igor Gilitschenski, and Alex Levinshtein. Spin-nerf: Multiview segmentation and perceptual inpainting with neural radiance fields. In CVPR, pages 20669–20679, 2023. 3, 6, 12

  19. [27]

    Inpaintnerf360: Text-guided 3d inpainting on unbounded neural radiance fields

    Dongqing Wang, Tong Zhang, Alaa Abboud, and Sabine Süsstrunk. Inpaintnerf360: Text-guided 3d inpainting on unbounded neural radiance fields. In CVPR, 2023. 3

  20. [28]

    Taming latent diffusion model for neural radiance field inpainting

    Chieh Hubert Lin, Changil Kim, Jia-Bin Huang, Qinbo Li, Chih-Yao Ma, Johannes Kopf, Ming-Hsuan Yang, and Hung-Yu Tseng. Taming latent diffusion model for neural radiance field inpainting. In ECCV, pages 149–165, 2024. 3, 6, 13

  21. [29]

    Reference-guided controllable inpainting of neural radiance fields

    Ashkan Mirzaei, Tristan Aumentado-Armstrong, Marcus A Brubaker, Jonathan Kelly, Alex Levinshtein, Konstantinos G Derpanis, and Igor Gilitschenski. Reference-guided controllable inpainting of neural radiance fields. In ICCV, pages 17815–17825, 2023. 3

  22. [30]

    Reffusion: Reference adapted diffusion models for 3d scene inpainting

    Ashkan Mirzaei, Riccardo De Lutio, Seung Wook Kim, David Acuna, Jonathan Kelly, Sanja Fidler, Igor Gilitschenski, and Zan Gojcic. Reffusion: Reference adapted diffusion models for 3d scene inpainting. arXiv preprint arXiv:2404.10765, 2024. 3

  23. [31]

    Learning 3d geometry and feature consistent gaussian splatting for object removal

    Yuxin Wang, Qianyi Wu, Guofeng Zhang, and Dan Xu. Learning 3d geometry and feature consistent gaussian splatting for object removal. In ECCV, pages 1–17, 2024. 3, 6, 13

  24. [32]

    Gaussian grouping: Segment and edit anything in 3d scenes

    Mingqiao Ye, Martin Danelljan, Fisher Yu, and Lei Ke. Gaussian grouping: Segment and edit anything in 3d scenes. In ECCV, pages 162–179, 2024. 3

  25. [33]

    Imfine: 3d inpainting via geometry-guided multi-view refinement

    Zhihao Shi, Dong Huo, Yuhongze Zhou, Kejia Yin, Yan Min, Juwei Lu, and Xinxin Zuo. Imfine: 3d inpainting via geometry-guided multi-view refinement. In CVPR, 2025. 3

  26. [34]

    Aurafusion360: Augmented unseen region alignment for reference-based 360 unbounded scene inpainting

    Chung-Ho Wu, Yang-Jung Chen, Ying-Huan Chen, Jie-Ying Lee, Bo-Hsu Ke, Chun-Wei Tuan Mu, Yi- Chuan Huang, Chin-Yang Lin, Min-Hung Chen, Yen-Yu Lin, and Yu-Lun Liu. Aurafusion360: Augmented unseen region alignment for reference-based 360 unbounded scene inpainting. In CVPR, 2025. 3

  27. [35]

    Infusion: Inpainting 3d gaussians via learning depth completion from diffusion prior

    Zhiheng Liu, Hao Ouyang, Qiuyu Wang, Ka Leong Cheng, Jie Xiao, Kai Zhu, Nan Xue, Yu Liu, Yujun Shen, and Yang Cao. Infusion: Inpainting 3d gaussians via learning depth completion from diffusion prior. arXiv preprint arXiv:2404.11613, 2024. 3, 6, 13

  28. [36]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv...

  29. [37]

    Julius Plucker. Xvii. on a new geometry of space. Philosophical Transactions of the Royal Society of London, (155):725–791, 1865. 4

  30. [38]

    Photographic image synthesis with cascaded refinement networks

    Qifeng Chen and Vladlen Koltun. Photographic image synthesis with cascaded refinement networks. In ICCV, pages 1511–1520, 2017. 4

  31. [39]

    Very deep convolutional networks for large-scale image recogni- tion

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recogni- tion. arXiv preprint arXiv:1409.1556, 2014. 4

  32. [40]

    Structure-from-motion revisited

    Johannes L Schonberger and Jan-Michael Frahm. Structure-from-motion revisited. In CVPR, pages 4104–4113, 2016. 5, 6 11

  33. [41]

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

    Ben Mildenhall, Pratul P. Srinivasan, Rodrigo Ortiz-Cayon, Nima Khademi Kalantari, Ravi Ramamoorthi, Ren Ng, and Abhishek Kar. Local light field fusion: Practical view synthesis with prescriptive sampling guidelines. ACM Transactions on Graphics (TOG), 2019. 6, 12

  34. [42]

    Resolution-robust large mask inpainting with fourier convolutions

    Roman Suvorov, Elizaveta Logacheva, Anton Mashikhin, Anastasia Remizova, Arsenii Ashukha, Aleksei Silvestrov, Naejin Kong, Harshith Goka, Kiwoong Park, and Victor Lempitsky. Resolution-robust large mask inpainting with fourier convolutions. arXiv preprint arXiv:2109.07161, 2021. 6

  35. [43]

    Repaint: Inpainting using denoising diffusion probabilistic models

    Andreas Lugmayr, Martin Danelljan, Andres Romero, Fisher Yu, Radu Timofte, and Luc Van Gool. Repaint: Inpainting using denoising diffusion probabilistic models. In CVPR, pages 11461–11471, 2022. 6

  36. [44]

    Black Forest Labs. Flux. https://github.com/black-forest-labs/flux , 2024. 6

  37. [45]

    The unreasonable effectiveness of deep features as a perceptual metric

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

  38. [46]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium. NeurIPS, 30, 2017. 6

  39. [47]

    Demystifying mmd gans

    Mikołaj Bi´nkowski, Danica J Sutherland, Michael Arbel, and Arthur Gretton. Demystifying mmd gans. arXiv preprint arXiv:1801.01401, 2018. 6 Supplementary Material A More Training Details Camera Normalization and Selection. For SPIn-NeRF [ 26] and LLFF [ 41], we normalize all c...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.