Pith. sign in

REVIEW 4 major objections 3 minor 3 cited by

Jointly optimizing dense 360-degree views yields globally coherent 3D object inpainting.

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 ↗

ObjFiller3D jointly optimizes a dense 360-degree ring of views to inpaint 3D objects with cross-view-consistent textures, reporting higher PSNR and LPIPS than per-view baselines at much lower runtime.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection A plausible dense-view 3D inpainting system with big claims; the abstract is readable but the full text is garbled, and the headline PSNR gap needs a matched-protocol check. the 4 major comments →

arxiv 2508.18271 v2 pith:LSWSTNB5 submitted 2025-08-25 cs.CV

ObjFiller3D: Scaling 3D Object Inpainting to Dense Multi-View Consistency

Haitang Feng , Xinkai Chen , Jie Liu , Jie Tang , Gangshan Wu , Beiqi Chen , Jianhuang Lai , Guangcong Wang This is my paper
classification cs.CV
keywords 3D object inpaintingmulti-view consistency360-degree trajectorycycle consistencygenerative encoderreference-guided editing3D reconstructionobject completion
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

3D object inpainting usually works by filling each camera view separately, which leaves seams, blur, and geometry breaks when the views are stitched into a 3D model. This paper argues that the right formulation is to optimize all views together along a full 360-degree path, so every completed view has to agree with its neighbors and with the loop as a whole. To make this work it introduces three cooperating encoders—one for temporal dependencies across views, one for object-level semantic completion, and one that enforces cycle consistency around the trajectory—and it also lets a reference image steer the appearance. The paper reports that such joint optimization beats the previous per-view and sparse-view approaches on reconstruction fidelity and perceptual quality and reduces runtime below ten minutes. If true, the result makes coherent 3D object editing practical for real workflows.

Core claim

ObjFiller-3D's central claim is that 3D object inpainting should be solved as one joint optimization over a dense chain of views encircling the object, not as independent per-view 2D completions or sparse-view editing. To make that tractable, the method couples three encoders: a temporal-driven generative encoder that models how appearance changes along the trajectory, a semantic-aware completion encoder that fills the object region at the object level, and a cycle-consistent 3D encoder that enforces agreement when a completion is re-rendered around the full loop. A reference image can be injected to control the inpainted appearance. The paper reports that this design yields globally consist

What carries the argument

The load-bearing mechanism is a closed-loop optimization along a densely sampled 360-degree trajectory. The temporal encoder reads a sequence of neighboring views so completions stay consistent in motion; the semantic encoder produces object-level inpainted texture from a global context; the cycle-consistent encoder re-renders and compares the completed sequence against itself, penalizing disagreement. Together they turn independent 2D inpainting into a multi-view constrained problem, which is what pushes global coherence.

Load-bearing premise

The paper's quantitative advantage over prior work assumes the comparisons used identical masks, object sets, camera trajectories, and scored regions; if those differ, the headline improvements cannot be attributed to the method.

What would settle it

Run the method and the strongest baselines on a common benchmark where masks, objects, camera trajectories, and scored regions are identical, then measure PSNR and LPIPS over exactly the same masked areas. If the PSNR advantage falls to a few dB or the runtime exceeds ten minutes, the central claim fails.

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

If this is right

  • Editing an object once yields a coherent result from every viewpoint, not just from the view that was edited.
  • Reference-guided appearance control becomes available without retraining: a user-supplied image guides the global completion.
  • Reconstructions complete in under 10 minutes, making 3D object inpainting practical compared to prior 40-minute pipelines.
  • Evaluated on the paper's datasets, fidelity and perceptual quality both improve over per-view and sparse-view baselines.
  • The joint dense-view formulation gives a direct route to completing objects with large missing regions, since all sides contribute evidence.

Where Pith is reading between the lines

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

  • The cycle-consistency encoder is likely the component that does the heavy lifting for global agreement; an ablation removing just that encoder would reveal whether the other two already prevent cross-view drift.
  • The same joint-trajectory optimization could transfer to video object removal or 3D scene editing with a moving camera, provided a semantic object prior exists.
  • The reported fidelity gap may shrink under strictly matched evaluation; a public benchmark with fixed masks and identical camera trajectories would make the comparison decisive.
  • Varying the density of sampled views would map the trade-off between runtime and consistency, potentially showing that far fewer views suffice once cycle consistency is enforced.
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 / 3 minor

Summary. The abstract proposes ObjFiller-3D, a method for 3D object inpainting that jointly optimizes a sequence of densely sampled views along a 360-degree trajectory using three complementary encoders: a Temporal-Driven Generative Encoder, a Semantic-Aware Completion Encoder, and a Cycle-Consistent 3D Encoder. The paper claims higher reconstruction fidelity and perceptual quality than NeRFiller and Instant3dit, plus a runtime reduction from over 40 minutes to under 10 minutes, and reports concrete numbers (PSNR 26.6 vs. 15.9, LPIPS 0.19 vs. 0.25). The supplied full text, however, is heavily corrupted and unreadable, so the architecture, equations, datasets, ablations, and evaluation protocol cannot be audited from the submitted materials.

Significance. If the reported quantitative improvements hold under matched evaluation conditions, the contribution would be significant: dense-view consistency is a known weakness of per-view 2D inpainting, and a method that jointly optimizes 360-degree views with cycle-consistent encoders could be practically useful for 3D editing. The provision of a project page and code links is a positive reproducibility step. However, in the submitted copy, the evidence needed to evaluate these claims is absent; the experimental protocol and even the technical description are unreadable. The significance therefore remains conditional on a clean, complete manuscript.

major comments (4)
  1. [Full text (provided copy)] The supplied full text is unreadable due to encoding corruption and includes a stray header for arXiv:2508.18278. No legible equations, tables, or algorithm blocks are present. This is load-bearing: the central claim of state-of-the-art quantitative results rests on an experimental section that cannot be inspected. The authors must provide a clean, complete PDF before the paper can be substantively reviewed.
  2. [Abstract] The abstract reports PSNR 26.6 vs. 15.9 for NeRFiller and LPIPS 0.19 vs. 0.25 for Instant3dit, but no evaluation protocol is disclosed. A 10.7 dB PSNR advantage is unusually large across independent implementations; differences in evaluation region (masked pixels vs. full image), object/mask sets, input view density, and baseline adaptation could account for much of the gap. The manuscript must specify datasets, masks, camera trajectories, number of views, evaluation pixels, and ideally error bars; otherwise the reported superiority is unverified.
  3. [Full text (architecture)] The claimed contribution is the combination of a Temporal-Driven Generative Encoder, a Semantic-Aware Completion Encoder, and a Cycle-Consistent 3D Encoder. The provided copy contains no legible description of these components, no equations, and no visible ablation study. Since the method's value is exactly this combination, an ablation isolating each component and demonstrating its marginal contribution is load-bearing.
  4. [Abstract (runtime)] The runtime claim (from over 40 minutes to under 10 minutes) lacks hardware details, resolution, number of iterations, convergence criteria, and whether the baseline runtimes are from the original papers or reproduced on the same machine. Runtime comparisons are sensitive to early stopping and optimization schedules; the manuscript should report timings under controlled conditions with the same hardware and view set.
minor comments (3)
  1. [Title/abstract] The title uses 'ObjFiller3D' while the abstract uses 'ObjFiller-3D'; please standardize the name.
  2. [Abstract] The sentence 'PSNR 26.6 vs. 15.9 of NeRFiller' is ambiguous; specify whether these are masked-region or full-image numbers and state units for all metrics.
  3. [Reproducibility] The project page and code links are appreciated, but no version or commit hash is provided, so the code cannot be used to verify the claims in the submitted copy.

Circularity Check

0 steps flagged

No significant circularity: the paper's central claims are empirical comparisons against external benchmarks and an architectural system, not quantities defined from the method's own fitted inputs.

full rationale

The paper is an empirical systems paper. Its core claims are quantitative results — PSNR 26.6 vs 15.9 (NeRFiller), LPIPS 0.19 vs 0.25 (Instant3dit), and runtime under 10 minutes — measured against ground-truth views and clock time. These are external metrics, not quantities constructed from the method's own fitted parameters. The proposed components (Temporal-Driven Generative Encoder, Semantic-Aware Completion Encoder, Cycle-Consistent 3D Encoder) are architectural and training mechanisms; nothing in the readable abstract or visible text defines a predicted quantity as equal to a fitted variable, nor does any equation show a parameter fit to a subset of data being reported as a prediction of that same data. The named baselines are external prior works, not self-citations, and no uniqueness theorem or ansatz is imported from the authors' own prior publications. The cycle-consistency and joint-optimization language describes an optimization objective, not a derivation that assumes its own conclusion. The supplied full text is corrupted and contains a stray arXiv header, so the internal equations cannot be fully audited; however, that is a source-quality limitation, not evidence of circularity. Evaluation-protocol concerns (matched masks, scoring regions, input view density) are validity concerns, not definitional circularity. Under the hard rules, a paper whose central claims are external comparisons and whose derivation is not definitionally tied to its inputs should receive a score of 0. Therefore no circular steps are listed.

Axiom & Free-Parameter Ledger

3 free parameters · 3 axioms · 0 invented entities

All content in this ledger is taken from the abstract, since the full text is unreadable. The free parameters listed are standard for deep multi-objective systems but are not disclosed in the abstract. The axioms are the load-bearing domain assumptions on which the claimed gains rest. No invented physical or ontological entities are introduced; the three encoders are architectural components of a trained system. The evaluation-protocol assumption is the most consequential, as it conditions the headline benchmark numbers.

free parameters (3)
  • Composite loss weights for the three encoder objectives
    Deep multi-objective frameworks require hand-tuned balancing weights for temporal, semantic, and cycle-consistency terms; values are not reported in the abstract.
  • Number and spacing of dense sampled views
    The density of the 360-degree trajectory is a hand-chosen hyperparameter that directly trades consistency against compute; not disclosed in the abstract.
  • Trained model weights
    All learned parameters are fitted to training data. Standard for this class of method, listed for completeness because the central performance claim depends on them.
axioms (3)
  • domain assumption A sequence of densely sampled views around a 360-degree trajectory, jointly optimized, is sufficient to enforce global 3D coherence.
    The central modeling premise of the abstract. If dense-view joint optimization cannot resolve cross-view inconsistencies, for example in textureless or occluded regions, the core claim fails.
  • domain assumption Pre-trained 2D inpainting knowledge transfers to 3D-consistent multi-view inpainting under the proposed training objectives.
    The framework builds on a temporal-driven generative encoder with 2D priors; the validity of transfer to a globally consistent 3D object is asserted, not proven in the abstract.
  • domain assumption Benchmark protocols for the NeRFiller and Instant3dit comparisons are reproduced fairly, with identical masks, viewpoints, and datasets.
    Load-bearing for the quantitative claims. The abstract does not disclose whether the 15.9 and 0.25 baseline numbers were re-run under the authors' protocol or quoted from prior papers.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of ObjFiller3D: Scaling 3D Object Inpainting to Dense Multi-View Consistency." pith.science (2026). https://pith.science/paper/LSWSTNB5

@misc{pith2026250818271,
  author       = {Pith},
  title        = {Pith review of: ObjFiller3D: Scaling 3D Object Inpainting to Dense Multi-View Consistency},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LSWSTNB5}},
  note         = {Machine review of arXiv:2508.18271}
}
Share X Bluesky LinkedIn Reddit HN
abstract

3D object inpainting is commonly achieved via multi-view 2D image completion, yet independently inpainted views often suffer from cross-view inconsistencies, leading to blurred textures, geometric discontinuities, and visual artifacts in the reconstructed 3D objects. To overcome these limitations, we propose ObjFiller-3D, a novel method designed for the completion and editing of high-quality and consistent 3D objects. Instead of relying on sparse-view editing or per-view 2D inpainting, our method jointly optimizes a sequence of densely sampled views along a $360^\circ$ trajectory, enabling global coherence across viewpoints. We design a new framework with three complementary components: a Temporal-Driven Generative Encoder for modeling dense-view dependencies, a Semantic-Aware Completion Encoder for object-level inpainting, and a Cycle-Consistent 3D Encoder that enforces global coherence through a closed-loop formulation. Our framework also supports reference-guided 3D inpainting, allowing fine-grained control over appearance. Extensive experiments on diverse datasets demonstrate that ObjFiller-3D significantly outperforms prior methods, achieving higher reconstruction fidelity (PSNR 26.6 vs.\ 15.9 of NeRFiller) and perceptual quality (LPIPS 0.19 vs.\ 0.25 of Instant3dit), while reducing reconstruction time from over 40 minutes to under 10 minutes. These results highlight the effectiveness and practical potential of our approach for real-world 3D editing applications. Project page: https://objfiller3d.github.io/ Code: https://github.com/objfiller3d/ObjFiller-3D .

discussion (0)

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

Forward citations

Cited by 3 Pith papers

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

  1. Arbor: Explicit Geometric Conditioning for Controllable 3D Asset Generation

    cs.CV 2026-06 unverdicted novelty 7.0

    Arbor attaches constraint mesh tokens to a frozen text-to-3D denoiser to enable controllable generation obeying hull, avoidance, and touch constraints.

  2. Axolotl3D: a Unified Framework for Faithful 3D Shape Completion

    cs.CV 2026-07 conditional novelty 5.0

    A single diffusion model, fine-tuned from Hunyuan3D, completes 3D shapes from any mix of images, masks, cameras, and point clouds, reaching state-of-the-art or competitive accuracy on Toys4K and OmniObject3D.

  3. GENA3D: Generative Amodal 3D Modeling by Bridging 2D Priors and 3D Coherence

    cs.CV 2025-11 conditional novelty 5.0

    A generative model reconstructs complete, occlusion-free 3D objects from sparse unposed views by combining 2D amodal inpainting with stereo-point-cloud-conditioned cross-attention.

Reference graph

Works this paper leans on

60 extracted references · 26 canonical work pages · cited by 3 Pith papers

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Gpt-4 technical report

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023

    Pith/arXiv arXiv 2023

  3. [3]

    Instant3dit: Multiview inpainting for fast editing of 3d objects

    Amir Barda, Matheus Gadelha, Vladimir G Kim, Noam Aigerman, Amit H Bermano, and Thibault Groueix. Instant3dit: Multiview inpainting for fast editing of 3d objects. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 16273--16282, 2025

    work page 2025

  4. [4]

    Skyreels-v2: Infinite-length film generative model

    Guibin Chen, Dixuan Lin, Jiangping Yang, Chunze Lin, Junchen Zhu, Mingyuan Fan, Hao Zhang, Sheng Chen, Zheng Chen, Chengcheng Ma, et al. Skyreels-v2: Infinite-length film generative model. arXiv preprint arXiv:2504.13074, 2025

    Pith/arXiv arXiv 2025

  5. [5]

    Proedit: Simple progression is all you need for high-quality 3d scene editing

    Jun-Kun Chen and Yu-Xiong Wang. Proedit: Simple progression is all you need for high-quality 3d scene editing. Advances in Neural Information Processing Systems, 37: 0 4934--4955, 2024

    work page 2024

  6. [6]

    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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 21476--21485, 2024

    work page 2024

  7. [7]

    Objaverse: A universe of annotated 3d objects

    Matt Deitke, Dustin Schwenk, Jordi Salvador, Luca Weihs, Oscar Michel, Eli VanderBilt, Ludwig Schmidt, Kiana Ehsani, Aniruddha Kembhavi, and Ali Farhadi. Objaverse: A universe of annotated 3d objects. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 13142--13153, 2023

    work page 2023

  8. [8]

    Flow-edge guided video completion

    Chen Gao, Ayush Saraf, Jia-Bin Huang, and Johannes Kopf. Flow-edge guided video completion. In European Conference on Computer Vision, pages 713--729. Springer, 2020

    work page 2020

  9. [9]

    Cat3d: Create anything in 3d with multi-view diffusion models

    Ruiqi Gao, Aleksander Holynski, Philipp Henzler, Arthur Brussee, Ricardo Martin-Brualla, Pratul Srinivasan, Jonathan T Barron, and Ben Poole. Cat3d: Create anything in 3d with multi-view diffusion models. arXiv preprint arXiv:2405.10314, 2024

    Pith/arXiv arXiv 2024

  10. [10]

    World models

    David Ha and J \"u rgen Schmidhuber. World models. arXiv preprint arXiv:1803.10122, 2 0 (3), 2018

    Pith/arXiv arXiv 2018

  11. [11]

    Instruct-nerf2nerf: Editing 3d scenes with instructions

    Ayaan Haque, Matthew Tancik, Alexei A Efros, Aleksander Holynski, and Angjoo Kanazawa. Instruct-nerf2nerf: Editing 3d scenes with instructions. In Proceedings of the IEEE/CVF international conference on computer vision, pages 19740--19750, 2023

    work page 2023

  12. [12]

    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. Neural Information Processing Systems,Neural Information Processing Systems, 2017

    work page 2017

  13. [13]

    Classifier-free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598, 2022

    Pith/arXiv arXiv 2022

  14. [14]

    Imagen video: High definition video generation with diffusion models

    Jonathan Ho, William Chan, Chitwan Saharia, Jay Whang, Ruiqi Gao, Alexey Gritsenko, Diederik P Kingma, Ben Poole, Mohammad Norouzi, David J Fleet, et al. Imagen video: High definition video generation with diffusion models. arXiv preprint arXiv:2210.02303, 2022

    Pith/arXiv arXiv 2022

  15. [15]

    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. arXiv preprint arXiv:2311.04400, 2023

    Pith/arXiv arXiv 2023

  16. [16]

    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 0 (2): 0 3, 2022

    2022

  17. [17]

    Dreamtime: An improved optimization strategy for text-to-3d content creation

    Yukun Huang, Jianan Wang, Yukai Shi, Xianbiao Qi, Zheng-Jun Zha, and Lei Zhang. Dreamtime: An improved optimization strategy for text-to-3d content creation. arXiv preprint arXiv:2306.12422, 3 0 (5): 0 14, 2023

    Pith/arXiv arXiv 2023

  18. [18]

    Mv-adapter: Multi-view consistent image generation made easy

    Zehuan Huang, Yuan-Chen Guo, Haoran Wang, Ran Yi, Lizhuang Ma, Yan-Pei Cao, and Lu Sheng. Mv-adapter: Multi-view consistent image generation made easy. arXiv preprint arXiv:2412.03632, 2024

    Pith/arXiv arXiv 2024

  19. [19]

    Vace: All-in-one video creation and editing

    Zeyinzi Jiang, Zhen Han, Chaojie Mao, Jingfeng Zhang, Yulin Pan, and Yu Liu. Vace: All-in-one video creation and editing. arXiv preprint arXiv:2503.07598, 2025

    Pith/arXiv arXiv 2025

  20. [20]

    Ray tracing volume densities

    James T Kajiya and Brian P Von Herzen. Ray tracing volume densities. ACM SIGGRAPH computer graphics, 18 0 (3): 0 165--174, 1984

    work page 1984

  21. [21]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk \"u hler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph., 42 0 (4): 0 139--1, 2023

    2023

  22. [22]

    Hunyuanvideo: A systematic framework for large video generative models

    Weijie Kong, Qi Tian, Zijian Zhang, Rox Min, Zuozhuo Dai, Jin Zhou, Jiangfeng Xiong, Xin Li, Bo Wu, Jianwei Zhang, et al. Hunyuanvideo: A systematic framework for large video generative models. arXiv preprint arXiv:2412.03603, 2024

    Pith/arXiv arXiv 2024

  23. [23]

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In International conference on machine learning, pages 19730--19742. PMLR, 2023 a

    work page 2023

  24. [24]

    Instant3d: Fast text-to-3d with sparse-view generation and large reconstruction model

    Jiahao Li, Hao Tan, Kai Zhang, Zexiang Xu, Fujun Luan, Yinghao Xu, Yicong Hong, Kalyan Sunkavalli, Greg Shakhnarovich, and Sai Bi. Instant3d: Fast text-to-3d with sparse-view generation and large reconstruction model. arXiv preprint arXiv:2311.06214, 2023 b

    Pith/arXiv arXiv 2023

  25. [25]

    Syncdreamer: Generating multiview-consistent images from a single-view image

    Yuan Liu, Cheng Lin, Zijiao Zeng, Xiaoxiao Long, Lingjie Liu, Taku Komura, and Wenping Wang. Syncdreamer: Generating multiview-consistent images from a single-view image. arXiv preprint arXiv:2309.03453, 2023

    Pith/arXiv arXiv 2023

  26. [26]

    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

    Pith/arXiv arXiv 2024

  27. [27]

    Wonder3d: Single image to 3d using cross-domain diffusion

    Xiaoxiao Long, Yuan-Chen Guo, Cheng Lin, Yuan Liu, Zhiyang Dou, Lingjie Liu, Yuexin Ma, Song-Hai Zhang, Marc Habermann, Christian Theobalt, et al. Wonder3d: Single image to 3d using cross-domain diffusion. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9970--9980, 2024

    work page 2024

  28. [28]

    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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11461--11471, 2022

    work page 2022

  29. [29]

    Scalable 3d captioning with pretrained models

    Tiange Luo, Chris Rockwell, Honglak Lee, and Justin Johnson. Scalable 3d captioning with pretrained models. Advances in Neural Information Processing Systems, 36: 0 75307--75337, 2023

    work page 2023

  30. [30]

    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 0 (1): 0 99--106, 2021

    2021

  31. [31]

    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 Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20669--20679, 2023

    work page 2023

  32. [32]

    Sdxl: Improving latent diffusion models for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M \"u ller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion models for high-resolution image synthesis. arXiv preprint arXiv:2307.01952, 2023

    Pith/arXiv arXiv 2023

  33. [33]

    Dreamfusion: Text-to-3d using 2d diffusion

    Ben Poole, Ajay Jain, Jonathan T Barron, and Ben Mildenhall. Dreamfusion: Text-to-3d using 2d diffusion. arXiv preprint arXiv:2209.14988, 2022

    Pith/arXiv arXiv 2022

  34. [34]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pages 8748--8763. PmLR, 2021

    2021

  35. [35]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj \"o rn 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

    2022

  36. [36]

    Photorealistic text-to-image diffusion models with deep language understanding

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photorealistic text-to-image diffusion models with deep language understanding. Advances in neural information processing systems, 35: 0 36479--36494, 2022

    work page 2022

  37. [37]

    Laion-400m: Open dataset of clip-filtered 400 million image-text pairs

    Christoph Schuhmann, Richard Vencu, Romain Beaumont, Robert Kaczmarczyk, Clayton Mullis, Aarush Katta, Theo Coombes, Jenia Jitsev, and Aran Komatsuzaki. Laion-400m: Open dataset of clip-filtered 400 million image-text pairs. arXiv preprint arXiv:2111.02114, 2021

    Pith/arXiv arXiv 2021

  38. [38]

    Laion-5b: An open large-scale dataset for training next generation image-text models

    Christoph Schuhmann, Romain Beaumont, Richard Vencu, Cade Gordon, Ross Wightman, Mehdi Cherti, Theo Coombes, Aarush Katta, Clayton Mullis, Mitchell Wortsman, et al. Laion-5b: An open large-scale dataset for training next generation image-text models. Advances in neural information processing systems, 35: 0 25278--25294, 2022

    2022

  39. [39]

    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 a

    Pith/arXiv arXiv 2023

  40. [40]

    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 b

    Pith/arXiv arXiv 2023

  41. [41]

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

    Zhihao Shi, Dong Huo, Yuhongze Zhou, Yan Min, Juwei Lu, and Xinxin Zuo. Imfine: 3d inpainting via geometry-guided multi-view refinement. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 26694--26703, 2025

    work page 2025

  42. [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. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pages 2149--2159, 2022

    work page 2022

  43. [43]

    Sv3d: Novel multi-view synthesis and 3d generation from a single image using latent video diffusion

    Vikram Voleti, Chun-Han Yao, Mark Boss, Adam Letts, David Pankratz, Dmitry Tochilkin, Christian Laforte, Robin Rombach, and Varun Jampani. Sv3d: Novel multi-view synthesis and 3d generation from a single image using latent video diffusion. In European Conference on Computer Vision, pages 439--457. Springer, 2024

    work page 2024

  44. [44]

    Wan: Open and advanced large-scale video generative models

    Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, et al. Wan: Open and advanced large-scale video generative models. arXiv preprint arXiv:2503.20314, 2025

    Pith/arXiv arXiv 2025

  45. [45]

    Imagedream: Image-prompt multi-view diffusion for 3d generation

    Peng Wang and Yichun Shi. Imagedream: Image-prompt multi-view diffusion for 3d generation. arXiv preprint arXiv:2312.02201, 2023

    Pith/arXiv arXiv 2023

  46. [46]

    Zero-shot video editing using off-the-shelf image diffusion models

    Wen Wang, Yan Jiang, Kangyang Xie, Zide Liu, Hao Chen, Yue Cao, Xinlong Wang, and Chunhua Shen. Zero-shot video editing using off-the-shelf image diffusion models. arXiv preprint arXiv:2303.17599, 2023 a

    Pith/arXiv arXiv 2023

  47. [47]

    Videofactory: Swap attention in spatiotemporal diffusions for text-to-video generation

    Wenjing Wang, Huan Yang, Zixi Tuo, Huiguo He, Junchen Zhu, Jianlong Fu, and Jiaying Liu. Videofactory: Swap attention in spatiotemporal diffusions for text-to-video generation. arXiv preprint arXiv:2305.10874, 2023 b

    Pith/arXiv arXiv 2023

  48. [48]

    View-consistent 3d editing with gaussian splatting

    Yuxuan Wang, Xuanyu Yi, Zike Wu, Na Zhao, Long Chen, and Hanwang Zhang. View-consistent 3d editing with gaussian splatting. In European conference on computer vision, pages 404--420. Springer, 2024

    work page 2024

  49. [49]

    Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distillation

    Zhengyi Wang, Cheng Lu, Yikai Wang, Fan Bao, Chongxuan Li, Hang Su, and Jun Zhu. Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distillation. Advances in neural information processing systems, 36: 0 8406--8441, 2023 c

    work page 2023

  50. [50]

    Nerfiller: Completing scenes via generative 3d inpainting

    Ethan Weber, Aleksander Holynski, Varun Jampani, Saurabh Saxena, Noah Snavely, Abhishek Kar, and Angjoo Kanazawa. Nerfiller: Completing scenes via generative 3d inpainting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20731--20741, 2024

    work page 2024

  51. [51]

    Meshlrm: Large reconstruction model for high-quality meshes

    Xinyue Wei, Kai Zhang, Sai Bi, Hao Tan, Fujun Luan, Valentin Deschaintre, Kalyan Sunkavalli, Hao Su, and Zexiang Xu. Meshlrm: Large reconstruction model for high-quality meshes. arXiv preprint arXiv:2404.12385, 2024

    Pith/arXiv arXiv 2024

  52. [52]

    Aurafusion360: Augmented unseen region alignment for reference-based 360deg 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, et al. Aurafusion360: Augmented unseen region alignment for reference-based 360deg unbounded scene inpainting. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 16366--16376, 2025

    work page 2025

  53. [53]

    Structured 3d latents for scalable and versatile 3d generation

    Jianfeng Xiang, Zelong Lv, Sicheng Xu, Yu Deng, Ruicheng Wang, Bowen Zhang, Dong Chen, Xin Tong, and Jiaolong Yang. Structured 3d latents for scalable and versatile 3d generation. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 21469--21480, 2025

    work page 2025

  54. [54]

    Instantmesh: Efficient 3d mesh generation from a single image with sparse-view large reconstruction models

    Jiale Xu, Weihao Cheng, Yiming Gao, Xintao Wang, Shenghua Gao, and Ying Shan. Instantmesh: Efficient 3d mesh generation from a single image with sparse-view large reconstruction models. arXiv preprint arXiv:2404.07191, 2024

    Pith/arXiv arXiv 2024

  55. [55]

    Deep flow-guided video inpainting

    Rui Xu, Xiaoxiao Li, Bolei Zhou, and Chen Change Loy. Deep flow-guided video inpainting. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3723--3732, 2019

    work page 2019

  56. [56]

    Advancing high-resolution video-language representation with large-scale video transcriptions

    Hongwei Xue, Tiankai Hang, Yanhong Zeng, Yuchong Sun, Bei Liu, Huan Yang, Jianlong Fu, and Baining Guo. Advancing high-resolution video-language representation with large-scale video transcriptions. In International Conference on Computer Vision and Pattern Recognition (CVPR), 2022

    2022

  57. [57]

    Videogpt: Video generation using vq-vae and transformers

    Wilson Yan, Yunzhi Zhang, Pieter Abbeel, and Aravind Srinivas. Videogpt: Video generation using vq-vae and transformers. arXiv preprint arXiv:2104.10157, 2021

    Pith/arXiv arXiv 2021

  58. [58]

    Efros, Eli Shechtman, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018

    work page 2018

  59. [59]

    Unipc: A unified predictor-corrector framework for fast sampling of diffusion models

    Wenliang Zhao, Lujia Bai, Yongming Rao, Jie Zhou, and Jiwen Lu. Unipc: A unified predictor-corrector framework for fast sampling of diffusion models. Advances in Neural Information Processing Systems, 36: 0 49842--49869, 2023

    work page 2023

  60. [60]

    Propainter: Improving propagation and transformer for video inpainting

    Shangchen Zhou, Chongyi Li, Kelvin CK Chan, and Chen Change Loy. Propainter: Improving propagation and transformer for video inpainting. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10477--10486, 2023

    work page 2023

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