Pith. sign in

REVIEW 4 major objections 5 minor 68 references

Gaussian Object Carver: Object-Compositional Gaussian Splatting with surfaces completion

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Gaussian Object Carver reconstructs indoor scenes as separable, watertight object meshes at over ten times the speed of SDF-based compositional methods, using a zero-shot completion module for occluded surfaces.

desk verdict A useful integration of object-compositional Gaussian splatting with a learned completion module, but the zero-shot generalization claim is not actually tested; efficiency win is real. read the letter →

arxiv 2412.02075 v1 pith:U37FXLYW submitted 2024-12-03 cs.CV cs.RO

Liu Liu , Xinjie Wang , Jiaxiong Qiu , Tianwei Lin , Xiaolin Zhou , Zhizhong Su This is my paper · ORCID
classification cs.CVcs.RO
keywords 3DGaussianSplattingobject-compositionalreconstructionzero-shotsurfacecompletionoccupancynetworksmonoculargeometrypriorsmulti-viewregularizationinstancesegmentationwatertightmeshextraction
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

Gaussian Object Carver (GOC) aims to show that object-compositional 3D scene reconstruction — splitting an indoor scene into individually editable objects — can be done with 3D Gaussian splatting instead of slow neural signed-distance fields. The paper's central claim is a tenfold speedup over existing compositional reconstructions while producing watertight, separable object meshes even when objects are partially occluded. The speed comes from replacing per-scene SDF optimization with differentiable Gaussian rasterization, and the watertightness comes from a separately trained zero-shot Object Surface Completion (OSC) module that fills in unobserved geometry from partial point clouds. If the claim holds, digital twins for simulation, AR/VR, and robotics could be built from real-world captures in about an hour per scene instead of a day, with objects that can be picked out, rearranged, and reinserted.

What carries the argument

The load-bearing object is the Object Surface Completion (OSC) model: a VAE-style occupancy network that encodes a partial surface point cloud (2048 farthest-point-sampled points, Fourier-position-encoded, through cross- and self-attention) into a latent set, and decodes per-query-point occupancy probabilities over a 3D grid; marching cubes then extracts a watertight mesh. In the reconstruction stage, the machinery is the alpha-blended depth and semantic-feature rendering of 3D Gaussians, with two regularization families — monocular priors (depth L1, alpha-weighted normal alignment) and multi-view geometry (photometric reprojection plus depth circular-consistency) — that keep the point cloud clean enough for the OSC module to complete.

What would settle it

Run GOC on a real ScanNet scene with masks produced by an automatic 2D segmenter and depth from a monocular estimator, then measure the fraction of completed object meshes that are actually watertight and separable; if small mask or depth errors cause object meshes to merge or leak, the claimed robustness to occlusion in real-world conditions fails.

Watch

Extended reading notes

Core claim

The paper introduces a two-stage pipeline. First, a 3D Gaussian representation of the scene is optimized with monocular depth and normal priors, photometric reprojection, and depth-consistency losses, while per-Gaussian semantic features are rendered and classified into instance masks, so the reconstructed geometry can be carved into objects. Second, the sampled point cloud of each object is fed into a zero-shot Object Surface Completion model, a point-cloud-conditioned occupancy network trained on roughly 400,000 watertight meshes, that predicts occupancy on a query grid and extracts a watertight mesh with marching cubes. The authors report that the full pipeline reconstructs a synthetic scene and completes all its objects in about 1 hour 9 minutes, roughly 5% of the time of the compared SDF-based methods, with better Chamfer distance and F-score on both object and scene metrics, and that the completion module transfers to unseen objects without fine-tuning.

Load-bearing premise

The pipeline's headline results assume that every training view has accurate per-object instance masks and trustworthy monocular depth and normal priors; if those inputs are noisy, object separation erodes and the completion module cannot rescue the geometry.

Editorial extensions

If this is right

  • A full indoor scene can be turned into separable, editable object meshes in roughly an hour on a single consumer GPU, instead of tens of hours, making per-scene digital twins practical at scale.
  • Objects observed from a sparse set of viewpoints (about 30% of a full capture) still come out with higher completeness and F-score than dense SDF-based reconstruction, because the completion module fills occluded backsides.
  • The completion module, trained on a large mixed dataset of watertight meshes, generalizes zero-shot to reconstructed point clouds from Gaussian splatting, so no per-scene or per-category fine-tuning is needed.
  • Because objects are individually separated and watertight, downstream manipulation — deleting, duplicating, rearranging objects, or reinserting them into a simulator — becomes possible.
  • The monocular depth prior is the single most important regularizer: removing it cuts scene F-score from 0.8243 to 0.3483 in the real-world ablation.

Reading between the lines

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

  • A testable extension the authors do not run: replace their ground-truth instance-mask supervision with an automatic open-vocabulary 2D segmenter at inference time, to check whether object separation survives mask noise on real scans; their ScanNet comparison uses ground-truth instance masks, so this is the unresolved gap for fully automatic real-world use.
  • A second is to measure completion quality as a function of occlusion fraction; the paper's synthetic sparse setting fixes one viewing budget, but a sweep over viewing angles would show where the OSC prior starts to hallucinate rather than complete.
  • Because the OSC model encodes geometry only and deliberately ignores texture and appearance, a natural next step is to condition completion on multi-view appearance or language-image features — the authors themselves list this as future work — which would let the same module disambiguate symmetric or self-similar objects.
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

4 major / 5 minor

Summary. The paper proposes Gaussian Object Carver (GOC), a 3D Gaussian Splatting (3DGS) framework for object-compositional scene reconstruction. GOC combines monocular normal/depth priors, multi-view photometric and geometric consistency, and an instance-segmentation feature field to reconstruct per-object geometry. A separate Object Surface Completion (OSC) model, trained on about 400k ShapeNet and Objaverse meshes, takes partial point clouds and outputs complete occupancy fields, producing watertight object meshes. The paper reports a more than 10x training-time speedup over SDF-based compositional baselines (ObjectSDF++, RICO) on synthetic scenes, competitive or better geometric accuracy, and a zero-shot completion module that requires no fine-tuning. Experiments include five private BlenderKit synthetic scenes, ScanNet real-world scenes, and a ShapeNet test-set evaluation of OSC.

Significance. If the reported results are robust, the framework would be a meaningful practical advance: it is, to my knowledge, the first 3DGS-based object-compositional reconstruction method, and the efficiency gain over SDF-based compositional methods is large and plausible given the rendering speed of 3DGS. The paper is also commendably transparent about implementation details, and the release of code (promised) and the extensive ablation over loss terms, OSC architecture, and masking strategies are strengths. However, the central novelty claim about zero-shot, occlusion-robust object completion is not yet supported by the experimental evidence: the ShapeNet evaluation uses complete point clouds, the synthetic scenes are a private five-scene benchmark with no category-overlap audit, and the real-world ScanNet evaluation explicitly excludes OSC. The quantitative support for OSC is further muddied by an inconsistency in Table 2. These gaps concern the load-bearing claims of the paper, so revision is required.

major comments (4)
  1. [§6.2, §6.4, Supplementary Table 7] The zero-shot OSC generalizability claim (Contribution 3) is not validated under the conditions that the claim requires. The ShapeNet test-set evaluation in Supplementary Table 7 uses complete point clouds as input, so it tests surface reconstruction rather than completion from partial or occluded observations. The synthetic-scene experiments (Tables 1 and 2) are limited to about ten BlenderKit assets per scene, with no reported category split and no audit of category overlap with the ShapeNet/Objaverse training set. There is no held-out-category experiment and no test on real, noisy, partially observed point clouds. The supplementary Limitation section itself concedes the model 'may struggle with complex object models due to ambiguity,' which underscores that the generalization claim is currently an assumption. A held-out-category evaluation and/or a real partial-scan experiment is needed to support the zero-shot claim.
  2. [§6.2, Table 2] The text accompanying Table 2 contains a factual inconsistency that reverses the direction of the effect. The text states that 'our accuracy decreased slightly after applying completion (GOC w/o OSC), from 0.0073 to 0.0038,' but the table lists GOC w/o OSC Accuracy = 0.0038 and GOC Accuracy = 0.0073. Since Accuracy is a lower-is-better metric, this means applying OSC degraded accuracy by a factor of about two, not improved it. The same pattern is visible in Table 1 (0.0045 to 0.0062). These numbers should be corrected and the interpretation revised, because they directly affect whether OSC can be claimed to improve geometric fidelity or merely completeness.
  3. [§6.3, §4.3] The ScanNet experiment is missing a description of how instance-mask ground truth is obtained. Section 4.3 supervises the segmentation feature field with 'instance mask GT,' but Section 6.3 does not specify whether these masks come from the ScanNet annotations, a pretrained segmentation network, or manual labeling. The object-reconstruction metrics in Table 3 are directly conditioned on these masks, so without this information the real-world object-level results are not reproducible and their validity cannot be assessed.
  4. [§6.2, Tables 1-2] The synthetic benchmark is private (five scenes with manually configured camera paths) and all reported numbers come from a single run with no error bars or multiple seeds. Since the headline efficiency claim (5% of the time of SDF-based methods) and the accuracy comparisons rest on this benchmark, the absence of variance information makes the quantitative comparisons fragile. Reporting per-scene results, multiple random seeds, or releasing the benchmark would materially strengthen the evidence; at minimum, the current single-run table does not support strong claims of superiority over RICO and ObjectSDF++.
minor comments (5)
  1. [§4.3] The heading 'Segmentatin Regularization' contains a typo; it should read 'Segmentation Regularization.'
  2. [§5] The heading 'Obejct Surfaces Completion Model' contains a typo; it should read 'Object Surface Completion Model.'
  3. [§3] In Section 3, the phrase 'As shown in Fig. 7' appears, but the overview figure is Figure 2 in the main text; Figure 7 is in the supplementary material. The reference should be corrected.
  4. [§5.2, Eq. (11)] Equation (11) is ambiguous: the expression '0.5 · 1 − 0.5 × SDF(Qg)/Tiso' should be parenthesized as '0.5 * (1 - 0.5 * SDF(Qg) / Tiso)' to make the intended labeling-smoothing formula clear.
  5. [§6.2, Table 1] In Table 1, the GOC and GOC w/ ShapeFormer rows have dashes ('-') in the scene-reconstruction columns, while the object-reconstruction columns are filled. This makes the table visually incomplete; the scene-reconstruction metrics for these variants should be reported, or the table should clarify that scene metrics are not applicable after object completion.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the reconstruction and completion pipelines are trained and evaluated on independent data, and no claimed result reduces to a fitted parameter or to a self-citation chain.

full rationale

The paper's derivation chain is self-contained in the relevant sense. The 3D Gaussian Splatting reconstruction stage (Sec. 4) optimizes Gaussians with photometric loss, monocular depth and normal losses (Eqs. 3-4), multi-view photometric and geometry reprojection losses (Eq. 5 and Sec. 4.2), and a segmentation cross-entropy loss (Sec. 4.3); these are standard regularizers borrowed from prior work, and none of them encodes the paper's target result of watertight, occlusion-completed object meshes. The Object Surface Completion model (Sec. 5) is a separately trained VAE/occupancy network: its encoder consumes masked surface points, its decoder predicts occupancy, and it is optimized with BCE, IoU, and KL losses (Eqs. 13-16) against ground-truth SDF and occupancy values computed from ShapeNet Core v2 and Objaverse meshes (Sec. 5.3). It is then applied at inference time without fine-tuning to BlenderKit synthetic scenes (Sec. 6.2), i.e., to data outside its training set. No parameter is fitted to the test scenes and renamed a prediction. The supplementary limitation (Sec. 5) concedes that OSC 'may struggle with complex object models due to ambiguity,' but that is a generalization and robustness caveat, not an admission that a result is true by construction. The main residual risks are benchmark self-selection and the absence of a held-out-category test for the zero-shot completion claim; those are soundness concerns, not circularity. There is also no load-bearing self-citation chain: references to MonoSDF, DN-Splatter, Gaussian Grouping, ObjectSDF++, and RICO support component choices or serve as baselines, and none is used to justify the central claim by fiat.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claims rest on standard domain assumptions about monocular priors, multi-view consistency, and the transferability of large-scale 3D training data. No new physical entities are introduced. The hand-tuned hyperparameters and private benchmark choices are the main additional burden on the reader.

free parameters (7)
  • Total loss weights (lambda_d, lambda_n, lambda_pho, lambda_geo, lambda_o) = 0.3, 0.1, 0.3, 0.3, 0.1
    Section 4.4 sets these weights by hand without sensitivity analysis.
  • Occupancy isosurface threshold T_b = 0.3
    Section 5.2 uses T_b for both the IoU loss and final mesh extraction; chosen by hand.
  • SDF smoothing threshold T_iso = 1/128
    Section 5.2 maps signed distance to soft occupancy within [-T_iso, T_iso].
  • OSC loss weights (lambda_BCE, lambda_IoU, lambda_KL) = 1.0, 0.01, 0.0001
    Section 5.2 sets these to balance losses without reported tuning.
  • Surface point count M and latent dimension D = M = 2048, D = 16
    Section 5.1 fixes these architecture choices; Table 5 ablates encoder depth but not M or D.
  • Sparse observation viewpoint ratio = 50 of 170 images (30%)
    Section 6.1 defines the sparse observation dataset; this ratio is chosen to simulate partial views.
  • Per-image monocular depth scale k and offset b = fitted by least squares per image
    Supplementary Eq. 17 fits k and b to align monocular depth with rendered depth.
assumptions (6)
  • domain assumption Monocular depth and normal estimates provide reliable geometry priors for indoor scenes.
    Section 4.2 uses depth prior L_d and normal prior L_dn as core regularizers; scale ambiguity is acknowledged and handled only approximately.
  • domain assumption Multi-view photometric and geometric consistency losses improve reconstruction without introducing drift.
    Section 4.2 introduces L_pho and L_geom, which assume roughly Lambertian appearance and accurate depth for warping.
  • domain assumption Training on ShapeNet Core v2 and filtered Objaverse meshes transfers to arbitrary real-world object categories.
    Section 5.3 trains OSC on these datasets; the zero-shot generalization claim depends on distributional coverage.
  • ad hoc to paper Camera-visibility-based masking during training approximates real occlusion patterns in captured scenes.
    Section 5.1 and Table 6 select the visible-mask strategy as best; this is a training-data choice specific to the paper.
  • domain assumption Ground-truth instance masks are available for segmentation supervision in all evaluation settings.
    Section 4.3 says 'We use a cross entropy loss L_o between instance mask GT and M_o', but the real-world ScanNet mask source is not described.
  • standard math Marching Cubes with isosurface threshold T_b extracts the intended watertight surface from the occupancy field.
    Section 5.2 relies on Marching Cubes; watertightness of the output depends on the consistency of the predicted occupancy field.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Gaussian Object Carver: Object-Compositional Gaussian Splatting with surfaces completion." pith.science (2026). https://pith.science/paper/U37FXLYW

@misc{pith2026241202075,
  author       = {Pith},
  title        = {Pith review of: Gaussian Object Carver: Object-Compositional Gaussian Splatting with surfaces completion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U37FXLYW}},
  note         = {Machine review of arXiv:2412.02075}
}
read the original abstract

3D scene reconstruction is a foundational problem in computer vision. Despite recent advancements in Neural Implicit Representations (NIR), existing methods often lack editability and compositional flexibility, limiting their use in scenarios requiring high interactivity and object-level manipulation. In this paper, we introduce the Gaussian Object Carver (GOC), a novel, efficient, and scalable framework for object-compositional 3D scene reconstruction. GOC leverages 3D Gaussian Splatting (GS), enriched with monocular geometry priors and multi-view geometry regularization, to achieve high-quality and flexible reconstruction. Furthermore, we propose a zero-shot Object Surface Completion (OSC) model, which uses 3D priors from 3d object data to reconstruct unobserved surfaces, ensuring object completeness even in occluded areas. Experimental results demonstrate that GOC improves reconstruction efficiency and geometric fidelity. It holds promise for advancing the practical application of digital twins in embodied AI, AR/VR, and interactive simulation environments.

Figures

Figures reproduced from arXiv: 2412.02075 by the authors.

Figure 1
Figure 1. Invisible Surface Completion: We introduce a novel, ef [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of GOC: Given multi-view images of a scene, we optimize 3D Gaussian Splatting (3D GS) to generate scene geometry [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Illustration of OSC for surface reconstruction from [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of surface reconstruction and [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison of reconstruction quality with [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison of reconstruction quality with [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Semantic mesh results of GOC on Synthetic Scenes [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 37 canonical work pages

  1. [1]

    Blender: 3D modelling and rendering pack- age

    Blender team. Blender: 3D modelling and rendering pack- age. Available at https://www.blender.org. 6

    work page

  2. [2]

    Blenderkit

    BlenderKit Team. Blenderkit. Online; accessed 14 Novem- ber 2024. Available at https://www.blenderkit. com. 6

    work page 2024

  3. [3]

    Shapenet: An information-rich 3d model repository

    Angel X Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo Li, Silvio Savarese, Manolis Savva, Shuran Song, Hao Su, et al. Shapenet: An information-rich 3d model repository. arXiv preprint arXiv:1512.03012, 2015. 6, 7, 8, 1

    arXiv 2015

  4. [4]

    Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction

    Danpeng Chen, Hai Li, Weicai Ye, Yifan Wang, Weijian Xie, Shangjin Zhai, Nan Wang, Haomin Liu, Hujun Bao, and Guofeng Zhang. Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction. arXiv preprint arXiv:2406.06521, 2024. 2, 4

    arXiv 2024

  5. [5]

    Implicit functions in feature space for 3d shape reconstruc- tion and completion

    Julian Chibane, Thiemo Alldieck, and Gerard Pons-Moll. Implicit functions in feature space for 3d shape reconstruc- tion and completion. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 6970–6981, 2020. 7, 8, 1

    work page 2020

  6. [6]

    Diffcomplete: Diffusion-based generative 3d shape completion

    Ruihang Chu, Enze Xie, Shentong Mo, Zhenguo Li, Matthias Nießner, Chi-Wing Fu, and Jiaya Jia. Diffcomplete: Diffusion-based generative 3d shape completion. Advances in Neural Information Processing Systems, 36, 2024. 3

    work page 2024

  7. [7]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 5828–5839, 2017. 6, 8

    work page 2017

  8. [8]

    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. 6, 7

    work page 2023

Show all 68 references
  1. [9]

    Geo-neus: Geometry-consistent neural implicit surfaces learning for multi-view reconstruction

    Qiancheng Fu, Qingshan Xu, Yew-Soon Ong, and Wenbing Tao. Geo-neus: Geometry-consistent neural implicit surfaces learning for multi-view reconstruction. Advances in Neural Information Processing Systems (NeurIPS), 2022. 2, 4

    2022

  2. [10]

    Bros- tow

    Cl ´ement Godard, Oisin Mac Aodha, and Gabriel J. Bros- tow. Unsupervised monocular depth estimation with left- right consistency. In CVPR, 2017. 4

    2017

  3. [11]

    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. 2

    2024

  4. [12]

    Objectcarver: Semi-automatic segmentation, re- construction and separation of 3d objects

    Gemmechu Hassena, Jonathan Moon, Ryan Fujii, Andrew Yuen, Noah Snavely, Steve Marschner, and Bharath Har- iharan. Objectcarver: Semi-automatic segmentation, re- construction and separation of 3d objects. arXiv preprint arXiv:2407.19108, 2024. 2

    2024 arXiv

  5. [13]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll´ar, and Ross Girshick. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 16000– 16009, 2022. 5, 8

    2022

  6. [14]

    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. 2

    2024

  7. [15]

    Neural kernel surface re- construction

    Jiahui Huang, Zan Gojcic, Matan Atzmon, Or Litany, Sanja Fidler, and Francis Williams. Neural kernel surface re- construction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 4369– 4379, 2023. 3

    2023

  8. [16]

    Poisson surface reconstruction

    Michael Kazhdan, Matthew Bolitho, and Hugues Hoppe. Poisson surface reconstruction. In Proceedings of the fourth Eurographics symposium on Geometry processing, 2006. 3

    2006

  9. [17]

    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. [18]

    3d gaussian splatting as markov chain monte carlo

    Shakiba Kheradmand, Daniel Rebain, Gopal Sharma, Wei- wei Sun, Jeff Tseng, Hossam Isack, Abhishek Kar, An- drea Tagliasacchi, and Kwang Moo Yi. 3d gaussian splatting as markov chain monte carlo. arXiv preprint arXiv:2404.09591, 2024. 5, 1

    2024 arXiv

  11. [19]

    Kingma and Max Welling

    Diederik P. Kingma and Max Welling. Auto-encoding vari- ational bayes. CoRR, abs/1312.6114, 2013. 5

    2013 arXiv

  12. [20]

    Rico: Regularizing the unobservable for indoor compositional reconstruction

    Zizhang Li, Xiaoyang Lyu, Yuanyuan Ding, Mengmeng Wang, Yiyi Liao, and Yong Liu. Rico: Regularizing the unobservable for indoor compositional reconstruction. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 17761–17771, 2023. 1, 2, 7, 8

    2023

  13. [21]

    One-2-3-45++: Fast single im- age to 3d objects with consistent multi-view generation and 3d diffusion

    Minghua Liu, Ruoxi Shi, Linghao Chen, Zhuoyang Zhang, Chao Xu, Xinyue Wei, Hansheng Chen, Chong Zeng, Ji- ayuan Gu, and Hao Su. One-2-3-45++: Fast single im- age to 3d objects with consistent multi-view generation and 3d diffusion. In Proceedings of the IEEE/CVF Conference on ...

    2024

  14. [22]

    One-2-3-45: Any single image to 3d mesh in 45 seconds without per-shape optimiza- tion

    Minghua Liu, Chao Xu, Haian Jin, Linghao Chen, Mukund Varma T, Zexiang Xu, and Hao Su. One-2-3-45: Any single image to 3d mesh in 45 seconds without per-shape optimiza- tion. Advances in Neural Information Processing Systems , 36, 2024. 3

    2024

  15. [23]

    Marching cubes: A high resolution 3d surface construction algorithm

    William E Lorensen and Harvey E Cline. Marching cubes: A high resolution 3d surface construction algorithm. InSem- inal graphics: pioneering efforts that shaped the field, pages 347–353. 1998. 1

    1998

  16. [24]

    Occupancy networks: Learning 3d reconstruction in function space

    Lars Mescheder, Michael Oechsle, Michael Niemeyer, Se- bastian Nowozin, and Andreas Geiger. Occupancy networks: Learning 3d reconstruction in function space. InProceedings 9 of the IEEE/CVF conference on computer vision and pattern recognition, pages 4460–4470, 2019. 2, 3, 5, 6, 1

    2019

  17. [25]

    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. 2, 4, 5

    2021

  18. [26]

    When does label smoothing help? Advances in neural in- formation processing systems, 32, 2019

    Rafael M ¨uller, Simon Kornblith, and Geoffrey E Hinton. When does label smoothing help? Advances in neural in- formation processing systems, 32, 2019. 6

    2019

  19. [27]

    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 ...

    2011

  20. [28]

    Phyrecon: Physically plausible neural scene reconstruction

    Junfeng Ni, Yixin Chen, Bohan Jing, Nan Jiang, Bin Wang, Bo Dai, Yixin Zhu, Song-Chun Zhu, and Siyuan Huang. Phyrecon: Physically plausible neural scene reconstruction. arXiv preprint arXiv:2404.16666, 2024. 1, 2, 8

    2024 arXiv

  21. [29]

    Convolutional occupancy networks

    Songyou Peng, Michael Niemeyer, Lars Mescheder, Marc Pollefeys, and Andreas Geiger. Convolutional occupancy networks. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceed- ings, Part III 16, pages 523–540. Springer, 2020. 1

    2020

  22. [30]

    Gem3d: Generative medial abstractions for 3d shape synthesis

    Dmitry Petrov, Pradyumn Goyal, Vikas Thamizharasan, Vladimir Kim, Matheus Gadelha, Melinos Averkiou, Sid- dhartha Chaudhuri, and Evangelos Kalogerakis. Gem3d: Generative medial abstractions for 3d shape synthesis. In ACM SIGGRAPH 2024 Conference Papers , pages 1–11,

    2024

  23. [31]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. Advances in neural information processing systems, 30, 2017. 5

    2017

  24. [32]

    Goi: Find 3d gaussians of interest with an optimizable open-vocabulary semantic-space hyperplane

    Yansong Qu, Shaohui Dai, Xinyang Li, Jianghang Lin, Liu- juan Cao, Shengchuan Zhang, and Rongrong Ji. Goi: Find 3d gaussians of interest with an optimizable open-vocabulary semantic-space hyperplane. In Proceedings of the 32nd ACM International Conference on Multimedia, pages ...

  25. [33]

    Patchcomplete: Learning multi-resolution patch priors for 3d shape comple- tion on unseen categories

    Yuchen Rao, Yinyu Nie, and Angela Dai. Patchcomplete: Learning multi-resolution patch priors for 3d shape comple- tion on unseen categories. Advances in Neural Information Processing Systems, 35:34436–34450, 2022. 3

    2022

  26. [34]

    3d point cloud generative adversarial network based on tree structured graph convolutions

    Dong Wook Shu, Sung Woo Park, and Junseok Kwon. 3d point cloud generative adversarial network based on tree structured graph convolutions. In Proceedings of the IEEE/CVF international conference on computer vision , pages 3859–3868, 2019. 2

    2019

  27. [35]

    Neurad: Neural rendering for autonomous driving

    Adam Tonderski, Carl Lindstr ¨om, Georg Hess, William Ljungbergh, Lennart Svensson, and Christoffer Petersson. Neurad: Neural rendering for autonomous driving. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14895–14904, 2024. 1

    2024

  28. [36]

    Reconciling reality through simulation: A real-to-sim-to-real approach for robust manipulation

    Marcel Torne, Anthony Simeonov, Zechu Li, April Chan, Tao Chen, Abhishek Gupta, and Pulkit Agrawal. Reconciling reality through simulation: A real-to-sim-to-real approach for robust manipulation. arXiv preprint arXiv:2403.03949,

    arXiv

  29. [37]

    Dn-splatter: Depth and normal priors for gaussian splatting and meshing

    Matias Turkulainen, Xuqian Ren, Iaroslav Melekhov, Otto Seiskari, Esa Rahtu, and Juho Kannala. Dn-splatter: Depth and normal priors for gaussian splatting and meshing. arXiv preprint arXiv:2403.17822, 2024. 2, 4

    2024 arXiv

  30. [38]

    Neuris: Neural reconstruction of indoor scenes using normal priors

    Jiepeng Wang, Peng Wang, Xiaoxiao Long, Christian Theobalt, Taku Komura, Lingjie Liu, and Wenping Wang. Neuris: Neural reconstruction of indoor scenes using normal priors. In European Conference on Computer Vision, pages 139–155. Springer, 2022. 2

    2022

  31. [39]

    Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction. arXiv preprint arXiv:2106.10689, 2021. 2

    2021 arXiv

  32. [40]

    Marching cubes: A high resolution 3d surface construction algorithm

    LORENSEN WE. Marching cubes: A high resolution 3d surface construction algorithm. Computer graphics, 21(1): 7–12, 1987. 3, 5, 6

    1987

  33. [41]

    Neural splines: Fitting 3d surfaces with infinitely- wide neural networks

    Francis Williams, Matthew Trager, Joan Bruna, and Denis Zorin. Neural splines: Fitting 3d surfaces with infinitely- wide neural networks. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 9949–9958, 2021. 3

    2021

  34. [42]

    Neural fields as learnable kernels for 3d reconstruction

    Francis Williams, Zan Gojcic, Sameh Khamis, Denis Zorin, Joan Bruna, Sanja Fidler, and Or Litany. Neural fields as learnable kernels for 3d reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18500–18510, 2022. 3

    2022

  35. [43]

    Unique3d: High-quality and efficient 3d mesh generation from a single image

    Kailu Wu, Fangfu Liu, Zhihan Cai, Runjie Yan, Hanyang Wang, Yating Hu, Yueqi Duan, and Kaisheng Ma. Unique3d: High-quality and efficient 3d mesh generation from a single image. arXiv preprint arXiv:2405.20343, 2024. 3

    2024 arXiv

  36. [44]

    Object- compositional neural implicit surfaces

    Qianyi Wu, Xian Liu, Yuedong Chen, Kejie Li, Chuanxia Zheng, Jianfei Cai, and Jianmin Zheng. Object- compositional neural implicit surfaces. In European Con- ference on Computer Vision, pages 197–213. Springer, 2022. 1, 2

    2022

  37. [45]

    Objectsdf++: Improved object-compositional neural implicit surfaces

    Qianyi Wu, Kaisiyuan Wang, Kejie Li, Jianmin Zheng, and Jianfei Cai. Objectsdf++: Improved object-compositional neural implicit surfaces. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 21764– 21774, 2023. 1, 2, 7, 8

    2023

  38. [46]

    Clusteringsdf: Self-organized neural implicit surfaces for 3d decomposition

    Tianhao Wu, Chuanxia Zheng, Qianyi Wu, and Tat-Jen Cham. Clusteringsdf: Self-organized neural implicit surfaces for 3d decomposition. In European Conference on Computer Vision, pages 255–272. Springer, 2025. 2

    2025

  39. [47]

    Opengaussian: Towards point-level 3d gaussian-based open vocabulary understanding

    Yanmin Wu, Jiarui Meng, Haijie Li, Chenming Wu, Yahao Shi, Xinhua Cheng, Chen Zhao, Haocheng Feng, Errui Ding, Jingdong Wang, et al. Opengaussian: Towards point-level 3d gaussian-based open vocabulary understanding. arXiv preprint arXiv:2406.02058, 2024. 2

    2024 arXiv

  40. [48]

    Grnet: Gridding resid- ual network for dense point cloud completion

    Haozhe Xie, Hongxun Yao, Shangchen Zhou, Jiageng Mao, Shengping Zhang, and Wenxiu Sun. Grnet: Gridding resid- ual network for dense point cloud completion. In European conference on computer vision , pages 365–381. Springer,

  41. [49]

    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 ,

    arXiv

  42. [50]

    Shapeformer: Transformer-based shape completion via sparse representa- tion

    Xingguang Yan, Liqiang Lin, Niloy J Mitra, Dani Lischin- ski, Daniel Cohen-Or, and Hui Huang. Shapeformer: Transformer-based shape completion via sparse representa- tion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 6239–6249,

  43. [51]

    Unisim: A neural closed-loop sensor simulator

    Ze Yang, Yun Chen, Jingkang Wang, Sivabalan Mani- vasagam, Wei-Chiu Ma, Anqi Joyce Yang, and Raquel Ur- tasun. Unisim: A neural closed-loop sensor simulator. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 1389–1399, 2023. 1

    2023

  44. [52]

    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. arXiv preprint arXiv:2312.00732, 2023. 2

    2023 arXiv

  45. [53]

    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, 2024. 2, 4

    2024

  46. [54]

    gsplat: An open-source library for Gaussian splatting

    Vickie Ye, Ruilong Li, Justin Kerr, Matias Turkulainen, Brent Yi, Zhuoyang Pan, Otto Seiskari, Jianbo Ye, Jeffrey Hu, Matthew Tancik, and Angjoo Kanazawa. gsplat: An open-source library for Gaussian splatting. arXiv preprint arXiv:2409.06765, 2024. 1

    2024 arXiv

  47. [55]

    Monosdf: Exploring monocu- lar geometric cues for neural implicit surface reconstruc- tion

    Zehao Yu, Songyou Peng, Michael Niemeyer, Torsten Sat- tler, and Andreas Geiger. Monosdf: Exploring monocu- lar geometric cues for neural implicit surface reconstruc- tion. Advances in neural information processing systems , 35:25018–25032, 2022. 2, 4, 8

    2022

  48. [56]

    Pcn: Point completion network

    Wentao Yuan, Tejas Khot, David Held, Christoph Mertz, and Martial Hebert. Pcn: Point completion network. In 2018 in- ternational conference on 3D vision (3DV), pages 728–737. IEEE, 2018. 2

    2018

  49. [57]

    3dilg: Ir- regular latent grids for 3d generative modeling

    Biao Zhang, Matthias Nießner, and Peter Wonka. 3dilg: Ir- regular latent grids for 3d generative modeling. Advances in Neural Information Processing Systems, 35:21871–21885,

  50. [58]

    3dshape2vecset: A 3d shape representation for neu- ral fields and generative diffusion models.ACM Transactions on Graphics (TOG), 42(4):1–16, 2023

    Biao Zhang, Jiapeng Tang, Matthias Niessner, and Peter Wonka. 3dshape2vecset: A 3d shape representation for neu- ral fields and generative diffusion models.ACM Transactions on Graphics (TOG), 42(4):1–16, 2023. 3, 5, 6, 7, 8, 1

    2023

  51. [59]

    Clay: A controllable large-scale generative model for creat- ing high-quality 3d assets

    Longwen Zhang, Ziyu Wang, Qixuan Zhang, Qiwei Qiu, Anqi Pang, Haoran Jiang, Wei Yang, Lan Xu, and Jingyi Yu. Clay: A controllable large-scale generative model for creat- ing high-quality 3d assets. ACM Transactions on Graphics (TOG), 43(4):1–20, 2024. 3

    2024

  52. [60]

    Open3d: A modern library for 3d data processing

    Qian-Yi Zhou, Jaesik Park, and Vladlen Koltun. Open3d: A modern library for 3d data processing. arXiv preprint arXiv:1801.09847, 2018. 6

    2018 arXiv

  53. [61]

    Feature 3dgs: Supercharging 3d gaussian splatting to enable distilled feature fields

    Shijie Zhou, Haoran Chang, Sicheng Jiang, Zhiwen Fan, Ze- hao Zhu, Dejia Xu, Pradyumna Chari, Suya You, Zhangyang Wang, and Achuta Kadambi. Feature 3dgs: Supercharging 3d gaussian splatting to enable distilled feature fields. InPro- ceedings of the IEEE/CVF Conference on Compu...

    2024

  54. [62]

    High- fidelity 3d textured shapes generation by sparse encoding and adversarial decoding

    Qi Zuo, Xiaodong Gu, Yuan Dong, Zhengyi Zhao, Weihao Yuan, Lingteng Qiu, Liefeng Bo, and Zilong Dong. High- fidelity 3d textured shapes generation by sparse encoding and adversarial decoding. In European Conference on Computer Vision, 2024. 6 11 Gaussian Object Carver: Object-...

    2024

  55. [63]

    Compared to state-of-the-art methods such as 3D2VS [58] and IF-Net [5], OSC demonstrated superior performance across all metrics

    OSC Reconstruction Quality on ShapeNet We evaluated the OSC model on the ShapeNet [3] test set to assess surface reconstruction quality using complete point- cloud inputs. Compared to state-of-the-art methods such as 3D2VS [58] and IF-Net [5], OSC demonstrated superior perform...

  56. [64]

    The results, presented in Figures 5 and 6 , reveal the signif- icant impact of these components on reconstruction qual- ity

    Additional Ablation Results for OSC Model To evaluate the effectiveness of key components in the OSC model, we performed an ablation study with additional ex- periments focusing on LIoU loss and label smoothing. The results, presented in Figures 5 and 6 , reveal the signif- ic...

  57. [65]

    The training itera- tions for all scenes are set to 30,000

    Additional 3D GS Implementation Details Implementation Details Our code is built based on gsplat [54]and training strategy are consistent with [18], because we observed that 3DGS [17] strategy is sensitive with ini- tialization and hyperparameter settings. The training itera- ...

  58. [66]

    From this field, the mesh is subsequently extracted using the Marching Cubes algorithm [23] 1 Scene1Scene2Scene3 Figure 7

    to construct the corresponding TSDF field. From this field, the mesh is subsequently extracted using the Marching Cubes algorithm [23] 1 Scene1Scene2Scene3 Figure 7. Semantic mesh results of GOC on Synthetic Scenes Depth Regularization For datasets with sensor-provided depth a...

  59. [67]

    Additional results Per-scene quantitative results of GOC on the Synthetic Scenes are reported in Fig. 7. This process yields water- tight and separable object meshes while preserving highly detailed features, enabling flexible scene rearrangement and object-level manipulation

  60. [68]

    It’s simple and efficient but may struggle with complex object models due to ambiguity

    Limitation Currently, our approach supports geometry completion based solely on reconstructed point cloud data. It’s simple and efficient but may struggle with complex object models due to ambiguity. In future work, we aim to integrate addi- tional observations into the 3D mod...

Pith tools

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