Pith. sign in

REVIEW 3 major objections 6 minor 61 references

SparseRecon: Neural Implicit Surface Reconstruction from Sparse Views with Feature and Depth Consistencies

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read SparseRecon reconstructs high-quality 3D surfaces from as few as three small-overlap images.

desk verdict Solid empirical advance in sparse-view surface reconstruction with a clear ablation story, but the central feature-consistency mechanism is under-validated and the paper needs a few details before publication. read the letter →

arxiv 2508.00366 v1 pith:LZQO2LIH submitted 2025-08-01 cs.CV

Liang Han , Xu Zhang , Haichuan Song , Kanle Shi , Yu-Shen Liu , Zhizhong Han This is my paper · ORCID
classification cs.CV
keywords sparse-view3Dreconstructionneuralimplicitsurfacessigneddistancefunctionfeatureconsistencyvolumerenderinguncertainty-guideddepthpriormulti-viewstereoChamfer
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

SparseRecon is a method for reconstructing a 3D surface from a very small number of RGB images whose fields of view barely overlap. It diagnoses the two standard strategies in this regime as complementary failures: generalization-based methods break on viewpoints they did not train on, while overfitting-based methods lack enough geometric evidence to fit clean geometry. The paper's proposed fix is a neural signed distance field rendered by volume rendering and constrained by a feature consistency loss that compares multi-view features at every sampled point along a ray, plus a depth prior applied only where rendered depth is uncertain. On the DTU benchmark with three small-overlap views the method reaches a mean Chamfer Distance of 1.11, the best among the 13 methods compared, and on BlendedMVS it reconstructs complete meshes where several baselines fail.

What carries the argument

The load-bearing component is the volume-rendered feature consistency loss: instead of comparing image features only at a recovered surface intersection point, it compares features at all sample points along each reference ray and weights each comparison by the volume-rendering opacity. This converts an easily mis-converged surface prior into dense scene-level supervision. The second component is an uncertainty-guided depth constraint: a monocular depth prior is calibrated with sparse 3D points from structure from motion, then masked by a confidence map built from forward-backward projection errors so that it supervises only low-confidence regions. The two operate inside a neural signed distance function rendered through volume rendering, and they are wrapped with pixel and patch color-consistency losses that refine fine detail.

What would settle it

Take a sparse small-overlap pair with near-repeating texture and compare optimization with and without the feature consistency loss: if the mean rendered-surface cosine similarity approaches zero yet the Chamfer Distance barely changes, the feature loss is not doing the claimed work. A sharper check is to swap the pretrained feature extractor for one not trained on dense multi-view stereo; the reported DTU margin should disappear if the features are the load-bearing ingredient.

Watch

Extended reading notes

Core claim

The central claim is that the ambiguity of sparse-view, small-overlap reconstruction is best removed by making feature consistency a volumetric supervision signal rather than a surface-only one. For each ray from the reference image, features are taken from a pretrained multi-view stereo network at all sampled 3D points, projected into source views, and compared by cosine similarity, with the per-point comparisons accumulated using the same volume-rendering weights as color and depth. This gives the signed distance field dense, smooth gradients even where no reliable surface point is known. In occluded or featureless regions, where that signal weakens, a monocular depth prior is calibrated against a structure-from-motion sparse point cloud and then enforced only where forward-backward projection of the rendered depth shows low confidence. Pixel and patch color warping are added for texture-level detail, and the paper reports that the combination beats both generalization-based and overfitting-based sparse-view baselines, with the largest gains in the small-overlap case.

Load-bearing premise

The whole method leans on the pretrained multi-view stereo feature extractor continuing to produce matching features across images whose viewpoints barely overlap; if those features stop corresponding, the feature-consistency loss stops constraining geometry.

Editorial extensions

If this is right

  • With three small-overlap images, the method reports the lowest mean Chamfer Distance, 1.11, among the 13 methods on the DTU benchmark.
  • Ablations show both proposed terms earn their keep: removing either the feature consistency loss or the uncertainty-guided depth loss raises the mean Chamfer Distance, and cosine feature similarity beats L1 and L2 feature distances.
  • Generalization-based sparse-view methods, including ones fine-tuned on the test scenes, fail to produce usable meshes on the BlendedMVS small-overlap scenes, while the proposed method completes them.
  • Because the feature loss supervises every sampled point along a ray, it does not depend on a sparse point cloud covering the surface, which is why it continues to work where surface-point-based feature consistency methods produce artifacts.
  • The final mesh accuracy is bounded by the quality of the pretrained feature extractor and by the supplied camera poses, a limitation the paper states explicitly.

Reading between the lines

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

  • A natural next test is whether the same volume-rendered feature consistency loss transfers to Gaussian-splatting surface reconstruction, where geometry is extracted from opacity fields rather than signed distance functions; if it does, the idea reaches a wider class of renderers.
  • The uncertainty-gating rule is a generic regularizer: depth priors are applied only where rendered depth is self-inconsistent, so it could be dropped into other depth-supervised neural rendering pipelines to reduce harm from bad priors.
  • On scenes with repeated texture or extreme baseline, pretrained feature correspondences will degrade, so the reported margin over baselines should shrink; measuring the average feature similarity on such inputs would quantify this ceiling.
  • Inaccurate camera poses would corrupt the forward-backward projection confidence map, so jointly refining poses with the feature consistency signal is a testable extension that turns the stated limitation into an optimization target.
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. The paper proposes SparseRecon, a per-scene overfitting method for neural implicit surface reconstruction from sparse views. The method adds a volume-rendering-based feature consistency loss that compares pretrained Vis-MVSNet features of points projected across views, an uncertainty-guided depth constraint that uses a MonoSDF-style monocular depth prior calibrated by COLMAP sparse points and applied only where rendered depth confidence is low, and color/patch consistency losses borrowed from NeuralWarp. Experiments on 11 DTU scenes with three small-overlapping views (22, 25, 28) report a mean Chamfer Distance of 1.11, which is lower than all 13 compared methods, and qualitative results on BlendedMVS. Ablations show that the feature consistency loss is the largest single contributor to the improvement.

Significance. If the reported results hold, SparseRecon is a practically useful step for sparse-view surface reconstruction, a setting where overfitting-based methods struggle with small-overlap views. The paper has concrete strengths: it evaluates against a broad set of dense, generalization-based, and overfitting-based baselines on DTU; it provides an ablation study isolating each loss; it reports visual comparisons and a failure case; and it includes a limitations section that acknowledges dependence on the feature extractor and on COLMAP. The central claim of state-of-the-art quality on small-overlap sparse views is supported by the DTU numbers, but the attribution of the gain to view-consistent features rests on an assumption about Vis-MVSNet features that is not directly tested. Several equations, especially the depth-confidence definition, are too underspecified to reproduce.

major comments (3)
  1. [Section 3.1, Eq. (1), Table 2] The feature consistency loss is the largest single contributor in the ablation: removing it raises mean CD from 1.11 to 1.59, and adding it to the baseline lowers CD from 3.35 to 1.76. Yet the paper does not directly verify that Vis-MVSNet features are view-consistent for the 22/25/28 small-overlap protocol. The authors should report quantitative feature-correspondence statistics, e.g., the distribution of cosine similarity between projected features of ground-truth surface points compared with wrong-depth projections, and should state explicitly whether the features come from the 2D encoder or the 3D cost volume of Vis-MVSNet. Without such a sanity check, the central claim that the improvement is due to view-consistent feature constraints remains an unverified assumption.
  2. [Section 3.2, Eqs. (2)-(5)] The depth-confidence definition is not reproducible as written. Equation (4) is garbled with missing symbols and the conditional expression is unclear: it is not specified what D_conf measures, what the forward and backward projection errors are in terms of the variables in Eqs. (2)-(3), or what happens at the boundary when the two distance terms are equal. The subsequent setting of the occlusion threshold tau to 0 is also confusing: if D_conf is nonnegative, tau=0 would make the mask include essentially all pixels, which contradicts the intent of masking occluded regions. Please rewrite Eqs. (2)-(5) with complete notation and specify the exact mask logic.
  3. [Section 4.4, Table 1, Figure 7] The claim of "significantly" better performance over UFORecon (mean CD 1.11 vs 1.39) is based on a single run per scene with no error bars, no variance, and no significance test. Since the improvement is about 20% on a mean over 11 scenes, paired per-scene differences should be reported, and ideally multiple seeds. In addition, the BlendedMVS comparison is qualitative only; please provide quantitative metrics (e.g., CD against ground-truth scans, or at least completeness and accuracy) or state clearly why they are unavailable for those scenes.
minor comments (6)
  1. [Section 3.1, Eq. (1)] The notation in Eq. (1) is incomplete: w_i is not defined (presumably the NeuS volume rendering weight), the feature maps F_ref and F_src are not formally introduced, and the occlusion mask O_ij is not defined here. Please add definitions so the loss is self-contained.
  2. [Section 4.2, Section 3.2] Please clarify the relationship between the depth confidence D_conf, the threshold tau, and the occlusion mask. The statement "the threshold tau of the occlusion mask is set to 0" should be reconciled with Eq. (4) and with the sentence defining the mask.
  3. [Table 2] The ablation table uses symbols such as L_f^L1, L_f^L2, L_f^onsurface, and L_depth^MonoSDF that are not all defined in the table caption or in the text. Please add a legend or define each row in the prose.
  4. [Section 4.4] There is a typo: "NeurSurf" should be "NeuSurf".
  5. [Section 4.4] The large-overlap results are only in the supplementary material. Please include a compact table or at least a summary sentence of the large-overlap comparison in the main paper, since this is a standard evaluation setting for DTU.
  6. [Section 5] The limitation that "the accuracy of the features determines the performance of the feature consistency constraint" is exactly the point that should be empirically quantified in the main experiments; please address this connection explicitly in the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method's losses are defined from external pre-trained networks and ground-truth colors, and the evaluation is against external benchmarks.

full rationale

The paper's central claim is an empirical comparison against external baselines. The feature consistency loss (Eq. 1) uses features from a fixed pre-trained Vis-MVSNet; the depth constraint uses a fixed pre-trained monocular depth network plus COLMAP sparse points; the color loss uses ground-truth colors; and the Eikonal regularization is standard. None of these quantities is defined in terms of the output mesh or fitted to the benchmark. The DTU evaluation deliberately selects scenes not used to train Vis-MVSNet, so the test is external, and the ablation study compares loss variants rather than predicting fitted parameters. Although several references share authors (e.g., NeuSurf [13], FatesGS [14], and CAP-UDF [58]), they are used as baselines or related work and are not load-bearing premises. The limitation that 'the accuracy of the features determines the performance of the feature consistency constraint' is an acknowledged assumption about the pre-trained feature extractor, not a circular definition. No equation reduces to its own input, and no fitted parameter is renamed as a prediction. The derivation chain is self-contained with respect to the claims made.

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

The method does not introduce new physics or entities; its free parameters are loss weights and a threshold. Its main external dependencies are the pre-trained feature and depth networks and COLMAP, which are treated as black boxes.

free parameters (4)
  • lambda_feat (feature consistency loss weight) = not reported
    Weight of L_feat in Eq. 7; the paper does not specify its value, only lambda_depth=0.5 and lambda_eik=0.1.
  • lambda_depth (depth loss weight) = 0.5
    Set in Section 4.2; chosen by hand, not justified by a search.
  • lambda_eik (Eikonal loss weight) = 0.1
    Set in Section 4.2, following common practice in NeuS-based methods.
  • occlusion mask threshold tau = 0
    Set in Section 4.2; the effect of this threshold on the depth constraint is not analyzed.
assumptions (4)
  • domain assumption Vis-MVSNet features are sufficiently discriminative and view-consistent for sparse-view scenes with small overlap.
    The feature consistency loss in Eq. 1 relies on features from a pre-trained MVS network; if these features fail to match across views, the loss cannot constrain geometry. The paper's limitations admit dependence on feature accuracy.
  • domain assumption Monocular depth priors, after calibration with COLMAP sparse points, are accurate enough in uncertain and occluded regions to improve geometry.
    The depth constraint in Eq. 5 assumes the calibrated monocular prior is closer to true depth than the rendered depth in low-confidence regions. The paper provides no quantitative validation of this assumption.
  • domain assumption COLMAP sparse point clouds are available and lie on the object surface for calibration.
    Depth calibration uses COLMAP keypoints; in low-texture scenes COLMAP can fail, as the paper notes in Limitations.
  • standard math NeuS volume rendering correctly models the SDF and enables optimizing geometry from color, feature, and depth losses.
    The method builds on the NeuS framework and its volume rendering equations; this is a standard established formulation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SparseRecon: Neural Implicit Surface Reconstruction from Sparse Views with Feature and Depth Consistencies." pith.science (2026). https://pith.science/paper/LZQO2LIH

@misc{pith2026250800366,
  author       = {Pith},
  title        = {Pith review of: SparseRecon: Neural Implicit Surface Reconstruction from Sparse Views with Feature and Depth Consistencies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LZQO2LIH}},
  note         = {Machine review of arXiv:2508.00366}
}
read the original abstract

Surface reconstruction from sparse views aims to reconstruct a 3D shape or scene from few RGB images. The latest methods are either generalization-based or overfitting-based. However, the generalization-based methods do not generalize well on views that were unseen during training, while the reconstruction quality of overfitting-based methods is still limited by the limited geometry clues. To address this issue, we propose SparseRecon, a novel neural implicit reconstruction method for sparse views with volume rendering-based feature consistency and uncertainty-guided depth constraint. Firstly, we introduce a feature consistency loss across views to constrain the neural implicit field. This design alleviates the ambiguity caused by insufficient consistency information of views and ensures completeness and smoothness in the reconstruction results. Secondly, we employ an uncertainty-guided depth constraint to back up the feature consistency loss in areas with occlusion and insignificant features, which recovers geometry details for better reconstruction quality. Experimental results demonstrate that our method outperforms the state-of-the-art methods, which can produce high-quality geometry with sparse-view input, especially in the scenarios with small overlapping views. Project page: https://hanl2010.github.io/SparseRecon/.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 57 canonical work pages

  1. [1]

    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. IEEE Transactions on Visualization and Computer Graphics , PP: 1–12, 2024. 3

    work page 2024

  2. [2]

    NeuSG: Neu- ral implicit surface reconstruction with 3D gaussian splatting guidance

    Hanlin Chen, Chen Li, and Gim Hee Lee. NeuSG: Neu- ral implicit surface reconstruction with 3D gaussian splatting guidance. arXiv preprint arXiv:2312.00846, 2023. 3

    arXiv 2023

  3. [3]

    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 European Conference on Computer Vision, pages 370–386. Springer, 2025. 3

    work page 2025

  4. [4]

    High-quality surface recon- struction using gaussian surfels

    Pinxuan Dai, Jiamin Xu, Wenxiang Xie, Xinguo Liu, Huamin Wang, and Weiwei Xu. High-quality surface recon- struction using gaussian surfels. In ACM SIGGRAPH 2024 Conference Proceedings, 2024. 3

    work page 2024

  5. [5]

    Improving neural im- plicit surfaces geometry with patch warping

    Franc ¸ois Darmon, B´en´edicte Bascle, Jean-Cl ´ement Devaux, Pascal Monasse, and Mathieu Aubry. Improving neural im- plicit surfaces geometry with patch warping. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 6260–6269, 2022. 1, 2, 5, 6

    work page 2022

  6. [6]

    Depth-supervised NeRF: Fewer views and faster training for free

    Kangle Deng, Andrew Liu, Jun-Yan Zhu, and Deva Ra- manan. Depth-supervised NeRF: Fewer views and faster training for free. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 12882–12891, 2022. 3

    work page 2022

  7. [7]

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

    Qiancheng Fu, Qingshan Xu, Yew Soon Ong, and Wen- bing Tao. Geo-Neus: Geometry-consistent neural implicit surfaces learning for multi-view reconstruction. Advances in Neural Information Processing Systems , 35:3403–3416,

    work page

  8. [8]

    Implicit geometric regularization for learning shapes

    Amos Gropp, Lior Yariv, Niv Haim, Matan Atzmon, and Yaron Lipman. Implicit geometric regularization for learning shapes. In Proceedings of Machine Learning and Systems 2020, pages 3569–3579. 2020. 5

    work page 2020

Show all 61 references
  1. [9]

    Cascade cost volume for high-resolution multi-view stereo and stereo matching

    Xiaodong Gu, Zhiwen Fan, Siyu Zhu, Zuozhuo Dai, Feitong Tan, and Ping Tan. Cascade cost volume for high-resolution multi-view stereo and stereo matching. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2495–2504, 2020. 2

    2020

  2. [10]

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

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

    2024

  3. [11]

    Binocular-guided 3D gaussian splatting with view con- sistency for sparse view synthesis

    Liang Han, Junsheng Zhou, Yu-Shen Liu, and Zhizhong Han. Binocular-guided 3D gaussian splatting with view con- sistency for sparse view synthesis. In Advances in Neural Information Processing Systems, 2024. 3

    2024

  4. [12]

    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 Proceedings, 2024. 3

    2024

  5. [13]

    NeuSurf: On-surface priors for neural surface reconstruction from sparse input views

    Han Huang, Yulun Wu, Junsheng Zhou, Ge Gao, Ming Gu, and Yu-Shen Liu. NeuSurf: On-surface priors for neural surface reconstruction from sparse input views. In Proceed- ings of the AAAI Conference on Artificial Intelligence, pages 2312–2320, 2024. 1, 2, 4, 6, 7, 8

    2024

  6. [14]

    FatesGS: Fast and accurate sparse-view surface reconstruction using gaussian splatting with depth- feature consistency

    Han Huang, Yulun Wu, Chao Deng, Ge Gao, Ming Gu, and Yu-Shen Liu. FatesGS: Fast and accurate sparse-view surface reconstruction using gaussian splatting with depth- feature consistency. In Proceedings of the AAAI Conference on Artificial Intelligence, 2025. 2, 3, 4, 6, 7

    2025

  7. [15]

    Putting NeRF on a Diet: Semantically consistent few-shot view synthesis implementation

    Ajay Jain, Matthew Tancik, and Pieter Abbeel. Putting NeRF on a Diet: Semantically consistent few-shot view synthesis implementation. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision , pages 5885–5894,

  8. [16]

    Large scale multi-view stereopsis evalu- ation

    Rasmus Jensen, Anders Dahl, George V ogiatzis, Engin Tola, and Henrik Aanæs. Large scale multi-view stereopsis evalu- ation. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 406–413, 2014. 2, 5, 7, 8

    2014

  9. [17]

    Transformers are rnns: Fast autoregressive transformers with linear attention

    Angelos Katharopoulos, Apoorv Vyas, Nikolaos Pappas, and Franc ¸ois Fleuret. Transformers are rnns: Fast autoregressive transformers with linear attention. In International Confer- ence on Machine Learning, pages 5156–5165, 2020. 2

    2020

  10. [18]

    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 Transactions on Graphics, 42 (4):1–14, 2023. 3

    2023

  11. [19]

    DNGaussian: Optimizing sparse-view 3D gaussian radiance fields with global-local depth normaliza- tion

    Jiahe Li, Jiawei Zhang, Xiao Bai, Jin Zheng, Xin Ning, Jun Zhou, and Lin Gu. DNGaussian: Optimizing sparse-view 3D gaussian radiance fields with global-local depth normaliza- tion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 2...

  12. [20]

    GaussianUDF: Inferring unsigned distance functions through 3D gaussian splatting

    Shujuan Li, Yu-Shen Liu, and Zhizhong Han. GaussianUDF: Inferring unsigned distance functions through 3D gaussian splatting. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 27113–27123, 2025. 3

    2025

  13. [21]

    Neuralangelo: High-fidelity neural surface reconstruction

    Zhaoshuo Li, Thomas M ¨uller, Alex Evans, Russell H Tay- lor, Mathias Unberath, Ming-Yu Liu, and Chen-Hsuan Lin. Neuralangelo: High-fidelity neural surface reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 8456–8465, 2023. 2

    2023

  14. [22]

    ReTR: Modeling rendering via transformer for generalizable neural surface re- construction

    Yixun Liang, Hao He, and Yingcong Chen. ReTR: Modeling rendering via transformer for generalizable neural surface re- construction. Advances in Neural Information Processing Systems, 36, 2024. 1, 2, 6

    2024

  15. [23]

    Reconstructing com- plex shaped clothing from a single image with feature stable unsigned distance fields

    Xinqi Liu, Jituo Li, and Guodong Lu. Reconstructing com- plex shaped clothing from a single image with feature stable unsigned distance fields. IEEE Transactions on Visualization and Computer Graphics, 2024. 1

    2024

  16. [24]

    SparseNeuS: Fast generalizable neural sur- face reconstruction from sparse views

    Xiaoxiao Long, Cheng Lin, Peng Wang, Taku Komura, and Wenping Wang. SparseNeuS: Fast generalizable neural sur- face reconstruction from sparse views. In European Confer- ence on Computer Vision, pages 210–227. Springer, 2022. 1, 2, 6, 7, 8

    2022

  17. [25]

    3DGSR: Implicit surface reconstruction with 3D gaussian splatting

    Xiaoyang Lyu, Yang-Tian Sun, Yi-Hua Huang, Xiuzhe Wu, Ziyi Yang, Yilun Chen, Jiangmiao Pang, and Xiaojuan Qi. 3DGSR: Implicit surface reconstruction with 3D gaussian splatting. ACM Transactions on Graphics, 43(6):1–12, 2024. 3

    2024

  18. [26]

    Learning signed distance functions from noisy 3D point clouds via noise to noise mapping

    Baorui Ma, Yu-Shen Liu, and Zhizhong Han. Learning signed distance functions from noisy 3D point clouds via noise to noise mapping. In International Conference on Ma- chine Learning, 2023. 1

    2023

  19. [27]

    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

    2021

  20. [28]

    UFORecon: Generalizable sparse-view surface reconstruction from arbitrary and unfavorable sets

    Youngju Na, Woo Jae Kim, Kyu Beom Han, Suhyeon Ha, and Sung-Eui Yoon. UFORecon: Generalizable sparse-view surface reconstruction from arbitrary and unfavorable sets. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5094–5104, 2024. 1, 2, 6, 7

    2024

  21. [29]

    Reg- NeRF: Regularizing neural radiance fields for view synthesis from sparse inputs

    Michael Niemeyer, Jonathan T Barron, Ben Mildenhall, Mehdi SM Sajjadi, Andreas Geiger, and Noha Radwan. Reg- NeRF: Regularizing neural radiance fields for view synthesis from sparse inputs. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition,...

    2022

  22. [30]

    Learning bijective surface parameterization for inferring signed distance functions from sparse point clouds with grid deformation

    Takeshi Noda, Chao Chen, Junsheng Zhou, Weiqi Zhang, Yu-Shen Liu, and Zhizhong Han. Learning bijective surface parameterization for inferring signed distance functions from sparse point clouds with grid deformation. In Proceedings of the Computer Vision and Pattern Recognition...

    2025

  23. [31]

    GenS: Generalizable neural sur- face reconstruction from multi-view images

    Rui Peng, Xiaodong Gu, Luyang Tang, Shihe Shen, Fanqi Yu, and Ronggang Wang. GenS: Generalizable neural sur- face reconstruction from multi-view images. In Advances in Neural Information Processing Systems , pages 56932– 56945, 2023. 1, 2, 6

    2023

  24. [32]

    V olRecon: V olume rendering of signed ray distance functions for generalizable multi-view reconstruction

    Yufan Ren, Tong Zhang, Marc Pollefeys, Sabine S ¨usstrunk, and Fangjinhua Wang. V olRecon: V olume rendering of signed ray distance functions for generalizable multi-view reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, page...

    2023

  25. [33]

    Pixelwise view selection for un- structured multi-view stereo

    Johannes Lutz Sch ¨onberger, Enliang Zheng, Marc Pollefeys, and Jan-Michael Frahm. Pixelwise view selection for un- structured multi-view stereo. In European Conference on Computer Vision, 2016. 2, 4, 5, 8

    2016

  26. [34]

    SuperUDF: Self-supervised udf estimation for surface reconstruction

    Hui Tian, Chenyang Zhu, Yifei Shi, and Kai Xu. SuperUDF: Self-supervised udf estimation for surface reconstruction. IEEE Transactions on Visualization and Computer Graph- ics, 2023. 1

    2023

  27. [35]

    SPARF: Neural radiance fields from sparse and noisy poses

    Prune Truong, Marie-Julie Rakotosaona, Fabian Manhardt, and Federico Tombari. SPARF: Neural radiance fields from sparse and noisy poses. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 4190–4200, 2023. 3

    2023

  28. [36]

    SparseNeRF: Distilling depth ranking for few-shot novel view synthesis

    Guangcong Wang, Zhaoxi Chen, Chen Change Loy, and Zi- wei Liu. SparseNeRF: Distilling depth ranking for few-shot novel view synthesis. InProceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 9065–9076,

  29. [37]

    NeuS: Learning neural im- plicit surfaces by volume rendering for multi-view recon- struction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. NeuS: Learning neural im- plicit surfaces by volume rendering for multi-view recon- struction. Advances in Neural Information Processing Sys- tems, 2021. 1, 2, 3, 5, 6

    2021

  30. [38]

    NeuS2: Fast learning of neural implicit surfaces for multi-view recon- struction

    Yiming Wang, Qin Han, Marc Habermann, Kostas Dani- ilidis, Christian Theobalt, and Lingjie Liu. NeuS2: Fast learning of neural implicit surfaces for multi-view recon- struction. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 3295–3306, 2023. 1

    2023

  31. [39]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Process- ing, 13(4):600–612, 2004. 5

    2004

  32. [40]

    S- V olSDF: Sparse multi-view stereo regularization of neural implicit surfaces

    Haoyu Wu, Alexandros Graikos, and Dimitris Samaras. S- V olSDF: Sparse multi-view stereo regularization of neural implicit surfaces. International Conference on Computer Vi- sion, 2023. 1, 2, 6, 7, 8

    2023

  33. [41]

    V oxurf: V oxel-based efficient and accurate neural surface reconstruction

    Tong Wu, Jiaqi Wang, Xingang Pan, Xudong Xu, Christian Theobalt, Ziwei Liu, and Dahua Lin. V oxurf: V oxel-based efficient and accurate neural surface reconstruction. Inter- national Conference on Learning Representations, 2022. 1, 2

    2022

  34. [42]

    C2F2NeUS: Cascade cost frustum fusion for high fidelity and generalizable neu- ral surface reconstruction

    Luoyuan Xu, Tao Guan, Yuesong Wang, Wenkai Liu, Zhao- jie Zeng, Junle Wang, and Wei Yang. C2F2NeUS: Cascade cost frustum fusion for high fidelity and generalizable neu- ral surface reconstruction. In Proceedings of the IEEE/CVF International Conference on Computer Vision, page...

    2023

  35. [43]

    FreeNeRF: Im- proving few-shot neural rendering with free frequency reg- ularization

    Jiawei Yang, Marco Pavone, and Yue Wang. FreeNeRF: Im- proving few-shot neural rendering with free frequency reg- ularization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 8254– 8263, 2023. 3

    2023

  36. [44]

    Blended- MVS: A large-scale dataset for generalized multi-view stereo networks

    Yao Yao, Zixin Luo, Shiwei Li, Jingyang Zhang, Yufan Ren, Lei Zhou, Tian Fang, and Long Quan. Blended- MVS: A large-scale dataset for generalized multi-view stereo networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 1790– 1799...

    2020

  37. [45]

    V ol- ume rendering of neural implicit surfaces

    Lior Yariv, Jiatao Gu, Yoni Kasten, and Yaron Lipman. V ol- ume rendering of neural implicit surfaces. Advances in Neu- ral Information Processing Systems, 34:4805–4815, 2021. 1, 2

    2021

  38. [46]

    Spar- seCraft: Few-shot neural reconstruction through stereopsis guided geometric linearization

    Mae Younes, Amine Ouasfi, and Adnane Boukhayma. Spar- seCraft: Few-shot neural reconstruction through stereopsis guided geometric linearization. In European Conference on Computer Vision, pages 37–56. Springer, 2024. 2, 6

    2024

  39. [47]

    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, 6, 7, 8

    2022

  40. [48]

    Gaussian Opacity Fields: Efficient adaptive surface reconstruction in unbounded scenes

    Zehao Yu, Torsten Sattler, and Andreas Geiger. Gaussian Opacity Fields: Efficient adaptive surface reconstruction in unbounded scenes. ACM Transactions on Graphics, 43(6): 1–13, 2024. 3

    2024

  41. [49]

    Neural radiance fields from sparse rgb-d images for high-quality view synthesis

    Yu-Jie Yuan, Yu-Kun Lai, Yi-Hua Huang, Leif Kobbelt, and Lin Gao. Neural radiance fields from sparse rgb-d images for high-quality view synthesis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(7):8713–8728, 2022. 3

    2022

  42. [50]

    Visibility-aware multi-view stereo network

    Jingyang Zhang, Yao Yao, Shiwei Li, Zixin Luo, and Tian Fang. Visibility-aware multi-view stereo network. The British Machine Vision Conference, 2020. 2, 3, 6

    2020

  43. [51]

    Learning signed distance field for multi-view surface reconstruction

    Jingyang Zhang, Yao Yao, and Long Quan. Learning signed distance field for multi-view surface reconstruction. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 6525–6534, 2021. 4, 6, 8

    2021

  44. [52]

    CoR-GS: sparse-view 3D gaussian splatting via co-regularization

    Jiawei Zhang, Jiahe Li, Xiaohan Yu, Lei Huang, Lin Gu, Jin Zheng, and Xiao Bai. CoR-GS: sparse-view 3D gaussian splatting via co-regularization. In European Conference on Computer Vision, pages 335–352. Springer, 2024. 3

    2024

  45. [53]

    Fast learning radiance fields by shooting much fewer rays

    Wenyuan Zhang, Ruofan Xing, Yunfan Zeng, Yu-Shen Liu, Kanle Shi, and Zhizhong Han. Fast learning radiance fields by shooting much fewer rays. IEEE Transactions on Image Processing, 2023. 1

    2023

  46. [54]

    Neu- ral signed distance function inference through splatting 3D gaussians pulled on zero-level set

    Wenyuan Zhang, Yu-Shen Liu, and Zhizhong Han. Neu- ral signed distance function inference through splatting 3D gaussians pulled on zero-level set. In Advances in Neural Information Processing Systems, 2024. 3

    2024

  47. [55]

    NeRFPrior: Learning neural radiance field as a prior for indoor scene re- construction

    Wenyuan Zhang, Emily Yue-ting Jia, Junsheng Zhou, Baorui Ma, Kanle Shi, Yu-Shen Liu, and Zhizhong Han. NeRFPrior: Learning neural radiance field as a prior for indoor scene re- construction. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 11317–...

    2025

  48. [56]

    MonoInstance: Enhancing monocular priors via multi-view instance align- ment for neural rendering and reconstruction

    Wenyuan Zhang, Yixiao Yang, Han Huang, Liang Han, Kanle Shi, Yu-Shen Liu, and Zhizhong Han. MonoInstance: Enhancing monocular priors via multi-view instance align- ment for neural rendering and reconstruction. InProceedings of the Computer Vision and Pattern Recognition Confer...

    2025

  49. [57]

    GAP: Gaussianize any point clouds with text guidance

    Weiqi Zhang, Junsheng Zhou, Haotian Geng, Wenyuan Zhang, and Yu-Shen Liu. GAP: Gaussianize any point clouds with text guidance. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision, 2025. 3

    2025

  50. [58]

    CAP-UDF: Learning unsigned distance functions progressively from raw point clouds with consistency-aware field optimization

    Junsheng Zhou, Baorui Ma, Shujuan Li, Yu-Shen Liu, Yi Fang, and Zhizhong Han. CAP-UDF: Learning unsigned distance functions progressively from raw point clouds with consistency-aware field optimization. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 1, 2

    2024

  51. [59]

    Fast learn- ing of signed distance functions from noisy point clouds via noise to noise mapping

    Junsheng Zhou, Baorui Ma, and Liu Yu-Shen. Fast learn- ing of signed distance functions from noisy point clouds via noise to noise mapping. IEEE Transactions on Pattern Anal- ysis and Machine Intelligence, 2024. 1

    2024

  52. [60]

    HDhuman: High-quality human novel-view rendering from sparse views

    Tiansong Zhou, Jing Huang, Tao Yu, Ruizhi Shao, and Kun Li. HDhuman: High-quality human novel-view rendering from sparse views. IEEE Transactions on Visualization and Computer Graphics, 2023. 1

    2023

  53. [61]

    FSGS: Real-time few-shot view synthesis using gaussian splatting

    Zehao Zhu, Zhiwen Fan, Yifan Jiang, and Zhangyang Wang. FSGS: Real-time few-shot view synthesis using gaussian splatting. In European Conference on Computer Vision , pages 145–163. Springer, 2025. 3

    2025

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.