Pith. sign in

REVIEW 4 major objections 3 minor 43 references

Adding semantic logits to neural implicit fields cuts sparse-view 3D reconstruction error by up to 44 percent on the DTU benchmark, and by 69 percent when plugged into dense-view baselines.

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

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

A semantic-aware implicit reconstruction method claims 44% and 20% lower Chamfer distance than SparseNeuS and VolRecon, and 69%/68% error reductions as a NeuS/Neuralangelo plugin.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection The abstract describes a plausible sparse-view reconstruction method, but the submitted full text is an unrelated calibration paper, so there is nothing to referee. the 4 major comments →

arxiv 2508.18314 v1 pith:5CWQDCWR submitted 2025-08-23 cs.CV

SERES: Semantic-aware neural reconstruction from sparse views

Bo Xu , Yuhu Guo , Yuchao Wang , Wenting Wang , Yeung Yam , Charlie C.L. Wang , Xinyi Le This is my paper
classification cs.CV
keywords semantic-aware reconstructionsparse-view 3D reconstructionneural implicit surfacessigned distance fieldradiance fieldsemantic logitsgeometric primitive maskschamfer distance
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

This paper is trying to establish that adding semantic information to neural implicit reconstruction removes a specific failure mode: from very few camera views, pixel matching is ambiguous and the reconstructed surface drifts. The proposed method attaches patch-based semantic logits (per-region semantic scores) to the signed distance field and radiance field and optimizes all three together, then uses geometric primitive masks to regularize shape away from ambiguity. If the reported results hold, the approach cuts average reconstruction error on the DTU dataset by 44% compared with SparseNeuS and 20% compared with VolRecon. Used as a plugin for the dense-view methods NeuS and Neuralangelo, it reduces average error by 69% and 68%. The point matters because sparse-view 3D reconstruction is the practical setting where users cannot capture dozens of images.

Core claim

The paper's central claim is that its semantic-aware scheme, SERES, makes neural implicit reconstruction from sparse images high-fidelity where prior methods suffer from radiance ambiguity. Instead of treating semantics as a post-hoc label, the method injects patch-based semantic logits (per-region semantic scores) directly into the neural representation and optimizes them alongside the signed distance field and radiance field. A regularization term built from geometric primitive masks is added to constrain shape in ambiguous areas. The authors report that on the DTU benchmark the average chamfer distance falls 44% relative to SparseNeuS and 20% relative to VolRecon, and that when the same s

What carries the argument

The central object is a neural implicit field with three coupled components: a signed distance field (the shape), a radiance field (color and density for rendering), and a set of patch-based semantic logits (per-region class scores). The argument's load-bearing move is to optimize all three together from sparse views, so semantic consistency can disambiguate matches that radiance alone cannot, while geometric primitive masks act as a regularization that prevents the shape from drifting in ambiguous regions.

Load-bearing premise

The semantic logits extracted from the sparse views must remain consistent enough to guide the jointly optimized geometry, and this joint optimization must converge; otherwise the claimed error reductions would not materialize.

What would settle it

Run an ablation that zeros out the semantic logits while keeping all other losses identical; if the DTU chamfer distance does not degrade by the reported margins, the semantic component is not causal. A second check is to measure geometry error on target objects whose semantic boundary crosses a continuous surface; if error grows there, the logits are dragging geometry toward label edges.

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

If this is right

  • Sparse-view reconstruction can be made substantially more accurate without extra images or sensors; the extra signal is semantic consistency, not more geometry.
  • Because the semantic module works as a plugin for NeuS and Neuralangelo, dense-view pipelines can inherit the same benefit by adding a jointly optimized semantic branch.
  • Geometry and semantic labels in neural implicit reconstruction should be solved in a joint optimization rather than in separate stages.
  • Geometric primitive masks are a useful source of shape regularization, reducing ambiguity in regions where color and density alone do not disambiguate the surface.

Where Pith is reading between the lines

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

  • A natural stress test not reported in the abstract: scenes where object semantic boundaries cut across a continuous geometric surface (e.g., a person leaning against a textured wall). If the joint optimization biases geometry toward label edges, reconstruction error should rise there; this would be a falsifiable extension of the method's assumption.
  • The reported gains on DTU may not transfer to object categories or scene types absent from that benchmark, since semantic priors are learned from the training classes. Testing on out-of-distribution scenes would show how much of the gain is semantic prior versus generic regularization.
  • The plugin result on dense baselines suggests the semantic signal is orthogonal to multi-view photometric consistency, so applying the same co-optimization to monocular reconstruction, where ambiguity is worst, is a possible follow-through.
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 manuscript, identified as arXiv:2508.18314, presents SERES, a semantic-aware neural reconstruction method for sparse-view 3D reconstruction. The abstract claims that adding patch-based semantic logits jointly optimized with a signed distance field and a radiance field, together with a regularization based on geometric primitive masks, reduces average chamfer distances on the DTU dataset by 44% for SparseNeuS, 20% for VolRecon, 69% for NeuS, and 68% for Neuralangelo. However, the full text provided is entirely a different paper, 'Does Calibration Affect Human Actions?' by Nizri et al. (arXiv:2508.18317v1), about prospect-theory-based calibration and human decision-making. The full text contains no description of SERES, no neural reconstruction method, no DTU experiments, and no equations, architecture, loss definitions, or results related to the claims in the abstract. Consequently, the central claims of the manuscript are completely unsupported by the submitted document.

Significance. If the claimed reductions are correct and reproducible, SERES would be a substantial contribution to sparse-view neural implicit reconstruction, potentially offering large improvements over strong baselines such as SparseNeuS, VolRecon, NeuS, and Neuralangelo. However, the manuscript as submitted provides no verifiable methods or results. There are no machine-checked proofs, reproducible code, derivations, or experimental tables. The four quantitative claims in the abstract cannot be checked, placed in context, or compared with prior work. The significance of the work therefore cannot be assessed from this submission.

major comments (4)
  1. [Full text (entire)] The full text is not the paper described in the abstract. It is 'Does Calibration Affect Human Actions?' (Nizri et al., arXiv:2508.18317v1), which concerns human decision-making and calibration. It contains no mention of SERES, neural reconstruction, sparse views, DTU, signed distance fields, radiance fields, semantic logits, or geometric primitive masks. This is a load-bearing defect: none of the abstract's claims are backed by any methods or experiments in the submitted document.
  2. [Abstract, second sentence] The abstract states that 'patch-based semantic logits' are 'optimized together with the signed distance field and the radiance field,' but the manuscript provides no optimization objective, no network architecture, no mathematical notation, and no convergence or sensitivity analysis. Without these details, the proposed co-optimization cannot be evaluated for stability or for the risk that semantic supervision pulls geometry toward geometrically incorrect semantic boundaries.
  3. [Abstract, final sentence] The abstract reports average chamfer distance reductions of 44%, 20%, 69%, and 68% on DTU for SparseNeuS, VolRecon, NeuS, and Neuralangelo, respectively. The full text contains no tables, no evaluation protocol, no dataset split, no error bars, no statistical significance tests, and no details about how the reductions were computed. The figures cannot be verified or even contextualized, so they cannot be treated as established results.
  4. [Abstract, third sentence] The 'novel regularization based on the geometric primitive masks' is mentioned only in the abstract. No definition of these masks, their computation, their role in the loss function, or an ablation study is provided anywhere in the manuscript. This is a central component of the proposed method, and its absence makes the method impossible to assess.
minor comments (3)
  1. [Header/footer metadata] The arXiv identifier shown in the full text is 2508.18317v1, while the manuscript under review is 2508.18314. This mismatch suggests a submission or compilation error and should be corrected if this is to be reviewed as the intended paper.
  2. [Title] The title of the full text is 'Does Calibration Affect Human Actions?' which is unrelated to the title 'SERES: Semantic-aware neural reconstruction from sparse views'. This inconsistency reinforces the completeness problem.
  3. [Abstract] The abstract says 'The performance of our approach has been verified in experimental evaluation,' but the body contains no experimental section, no figures or tables for SERES, and no references to prior reconstructions such as SparseNeuS, VolRecon, NeuS, or Neuralangelo. The internal inconsistency should be addressed.

Circularity Check

0 steps flagged

No circularity can be identified: the provided body is a different paper, so there is no SERES derivation chain to audit.

full rationale

The only in-scope text describing SERES is the abstract, which states that patch-based semantic logits are optimized jointly with the signed distance field and radiance field, and that DTU chamfer distances are reduced by 44%/20% for sparse-view baselines and 69%/68% for dense-view baselines. However, the supplied full text is a completely different arXiv paper, 'Does Calibration Affect Human Actions?' (Nizri et al., arXiv:2508.18317v1), about prospect-theory-based calibration and human decision-making. It contains no SERES methods, equations, loss definitions, model architecture, DTU experiments, or hyperparameters. There is therefore no derivation chain to walk, no fitted parameter renamed as a prediction, no self-citation chain, and no ansatz smuggled in via citation. The mismatch is an evidence-integrity problem — the quantitative claims cannot be verified or replicated from the provided text — but it is not a circularity problem. Per the hard rules, circularity requires quoting the paper and exhibiting a specific reduction where an output equals an input by construction or where a load-bearing argument rests on a self-citation. No such reduction can be exhibited here. The reader's concern that the DTU reductions could in principle result from test-set tuning is speculation about hidden fitting, not demonstrated circularity, and therefore does not raise the score. The honest finding is 'no significant circularity' with score 0.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Only the abstract is available for review. The paper introduces no new physical entities; the semantic logits and primitive masks are methodological constructs, not invented entities in the particle/force sense. The one domain assumption listed is directly from the abstract's description of the optimization.

axioms (1)
  • domain assumption Semantic logits derived from sparse views provide stable, geometry-relevant supervision when optimized jointly with SDF and radiance.
    The abstract states semantic logits are optimized together with the signed distance field and radiance field, but no evidence is given for convergence or for the consistency of semantics across sparse views.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of SERES: Semantic-aware neural reconstruction from sparse views." pith.science (2026). https://pith.science/paper/5CWQDCWR

@misc{pith2026250818314,
  author       = {Pith},
  title        = {Pith review of: SERES: Semantic-aware neural reconstruction from sparse views},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5CWQDCWR}},
  note         = {Machine review of arXiv:2508.18314}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We propose a semantic-aware neural reconstruction method to generate 3D high-fidelity models from sparse images. To tackle the challenge of severe radiance ambiguity caused by mismatched features in sparse input, we enrich neural implicit representations by adding patch-based semantic logits that are optimized together with the signed distance field and the radiance field. A novel regularization based on the geometric primitive masks is introduced to mitigate shape ambiguity. The performance of our approach has been verified in experimental evaluation. The average chamfer distances of our reconstruction on the DTU dataset can be reduced by 44% for SparseNeuS and 20% for VolRecon. When working as a plugin for those dense reconstruction baselines such as NeuS and Neuralangelo, the average error on the DTU dataset can be reduced by 69% and 68% respectively.

discussion (0)

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

Reference graph

Works this paper leans on

43 extracted references · 36 canonical work pages

  1. [1]

    J. Wang, P. Wang, X. Long, C. Theobalt, T. Komura, L. Liu, and W. Wang, ``Neuris: Neural reconstruction of indoor scenes using normal priors,'' in Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XXXII. 1em plus 0.5em minus 0.4em Berlin, Heidelberg: Springer-Verlag, 2022, p. 139–155. [Online]....

    work page doi:10.1007/978-3-031-19824-3_9 2022

  2. [2]

    P. Wang, L. Liu, Y. Liu, C. Theobalt, T. Komura, and W. Wang, ``Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction,'' NeurIPS, 2021

    work page 2021

  3. [3]

    Zhang, G

    K. Zhang, G. Riegler, N. Snavely, and V. Koltun, ``Nerf++: Analyzing and improving neural radiance fields,'' arXiv:2010.07492, 2020

    Pith/arXiv arXiv 2010

  4. [4]

    Y. Wei, S. Liu, Y. Rao, W. Zhao, J. Lu, and J. Zhou, ``Nerfingmvs: Guided optimization of neural radiance fields for indoor multi-view stereo,'' in ICCV, 2021

    work page 2021

  5. [5]

    B. Zhu, Y. Yang, X. Wang, Y. Zheng, and L. Guibas, ``Vdn-nerf: Resolving shape-radiance ambiguity via view-dependence normalization,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 35--45

    work page 2023

  6. [6]

    A. Yu, V. Ye, M. Tancik, and A. Kanazawa, `` pixelNeRF : Neural radiance fields from one or few images,'' in CVPR, 2021

    work page 2021

  7. [7]

    X. Long, C. Lin, P. Wang, T. Komura, and W. Wang, ``Sparseneus: Fast generalizable neural surface reconstruction from sparse views,'' ECCV, 2022

    work page 2022

  8. [8]

    A. Chen, Z. Xu, F. Zhao, X. Zhang, F. Xiang, J. Yu, and H. Su, ``Mvsnerf: Fast generalizable radiance field reconstruction from multi-view stereo,'' in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 14\,124--14\,133

    work page 2021

  9. [9]

    Z. Yu, S. Peng, M. Niemeyer, T. Sattler, and A. Geiger, ``Mono SDF : Exploring monocular geometric cues for neural implicit surface reconstruction,'' in Advances in Neural Information Processing Systems, A. H. Oh, A. Agarwal, D. Belgrave, and K. Cho, Eds., 2022. [Online]. Available: https://openreview.net/forum?id=dMK7EwoTYp

    work page 2022

  10. [10]

    Guangcong, Z. Chen, C. C. Loy, and Z. Liu, ``Sparsenerf: Distilling depth ranking for few-shot novel view synthesis,'' in ICCV, 2023

    work page 2023

  11. [11]

    Y. Ren, F. Wang, T. Zhang, M. Pollefeys, and S. S\"usstrunk, ``Volrecon: Volume rendering of signed ray distance functions for generalizable multi-view reconstruction,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 16\,685--16\,695

    work page 2023

  12. [12]

    Zhang, P

    X. Zhang, P. P. Srinivasan, B. Deng, P. Debevec, W. T. Freeman, and J. T. Barron, ``Nerfactor: Neural factorization of shape and reflectance under an unknown illumination,'' ACM Trans. Graph., vol. 40, no. 6, dec 2021. [Online]. Available: https://doi.org/10.1145/3478513.3480496

    work page arXiv 2021

  13. [13]

    K. N. Kutulakos and S. M. Seitz, ``A theory of shape by space carving,'' International Journal of Computer Vision, vol. 38, no. 3, pp. 199--218, 2000. [Online]. Available: https://doi.org/10.1023/a:1008191222954

    work page doi:10.1023/a:1008191222954 2000

  14. [14]

    Niemeyer, J

    M. Niemeyer, J. T. Barron, B. Mildenhall, M. S. M. Sajjadi, A. Geiger, and N. Radwan, ``Regnerf: Regularizing neural radiance fields for view synthesis from sparse inputs,'' in Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2022

    work page 2022

  15. [15]

    Sinha, J

    S. Sinha, J. Y. Zhang, A. Tagliasacchi, I. Gilitschenski, and D. B. Lindell, `` SparsePose : Sparse-view camera pose regression and refinement,'' in Computer Vision and Pattern Recognition (CVPR), 2023

    work page 2023

  16. [16]

    Huang, Z

    S.-S. Huang, Z. Zou, Y. Zhang, Y.-P. Cao, and Y. Shan, ``Sc-neus: Consistent neural surface reconstruction from sparse and noisy views,'' Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 3, pp. 2357--2365, Mar. 2024. [Online]. Available: https://ojs.aaai.org/index.php/AAAI/article/view/28010

    work page 2024

  17. [17]

    Mildenhall, P

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng, ``Nerf: Representing scenes as neural radiance fields for view synthesis,'' in ECCV, 2020

    work page 2020

  18. [18]

    Zhang, Y

    J. Zhang, Y. Yao, and L. Quan, ``Learning signed distance field for multi-view surface reconstruction,'' International Conference on Computer Vision (ICCV), 2021

    work page 2021

  19. [19]

    Yariv, J

    L. Yariv, J. Gu, Y. Kasten, and Y. Lipman, ``Volume rendering of neural implicit surfaces,'' Advances in Neural Information Processing Systems, vol. 34, pp. 4805--4815, 2021

    work page 2021

  20. [20]

    Darmon, B

    F. Darmon, B. Bascle, J.-C. Devaux, P. Monasse, and M. Aubry, ``Improving neural implicit surfaces geometry with patch warping,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 6260--6269

    work page 2022

  21. [21]

    Huang, Y

    H. Huang, Y. Wu, J. Zhou, G. Gao, M. Gu, and Y.-S. Liu, ``Neusurf: On-surface priors for neural surface reconstruction from sparse input views,'' in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 3, 2024, pp. 2312--2320

    work page 2024

  22. [22]

    Yariv, Y

    L. Yariv, Y. Kasten, D. Moran, M. Galun, M. Atzmon, B. Ronen, and Y. Lipman, ``Multiview neural surface reconstruction by disentangling geometry and appearance,'' Advances in Neural Information Processing Systems, vol. 33, 2020

    work page 2020

  23. [23]

    Niemeyer, L

    M. Niemeyer, L. Mescheder, M. Oechsle, and A. Geiger, ``Differentiable volumetric rendering: Learning implicit 3d representations without 3d supervision,'' in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 1em plus 0.5em minus 0.4em Los Alamitos, CA, USA: IEEE Computer Society, jun 2020, pp. 3501--3512. [Online]. Available: ht...

    arXiv 2020

  24. [24]

    Verbin, P

    D. Verbin, P. Hedman, B. Mildenhall, T. Zickler, J. T. Barron, and P. P. Srinivasan, `` Ref-NeRF : Structured view-dependent appearance for neural radiance fields,'' CVPR, 2022

    work page 2022

  25. [25]

    W. Bian, Z. Wang, K. Li, J.-W. Bian, and V. A. Prisacariu, ``Nope-nerf: Optimising neural radiance field with no pose prior,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 4160--4169

    work page 2023

  26. [26]

    Y. Na, W. J. Kim, K. B. Han, S. Ha, and S.-E. 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 (CVPR), June 2024, pp. 5094--5104

    work page 2024

  27. [27]

    Q. Fu, Q. Xu, Y.-S. Ong, and W. Tao, ``Geo-neus: Geometry-consistent neural implicit surfaces learning for multi-view reconstruction,'' in Advances in Neural Information Processing Systems, A. H. Oh, A. Agarwal, D. Belgrave, and K. Cho, Eds., 2022. [Online]. Available: https://openreview.net/forum?id=JvIFpZOjLF4

    work page 2022

  28. [28]

    Kirillov, E

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y. Lo, P. Dollar, and R. Girshick, ``Segment anything,'' in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October 2023, pp. 4015--4026

    work page 2023

  29. [29]

    J. Cen, Y. Wu, K. Wang, X. Li, J. Yang, Y. Pei, L. Kong, Z. Liu, and Q. Chen, `` SAD: segment any RGBD ,'' CoRR, vol. abs/2305.14207, 2023. [Online]. Available: https://doi.org/10.48550/arXiv.2305.14207

    Pith/arXiv arXiv doi:10.48550/arxiv.2305.14207 2023

  30. [30]

    Y. Liu, L. Kong, J. Cen, R. Chen, W. Zhang, L. Pan, K. Chen, and Z. Liu, ``Segment any point cloud sequences by distilling vision foundation models,'' in Advances in Neural Information Processing Systems, 2023

    work page 2023

  31. [31]

    Y. Yang, X. Wu, T. He, H. Zhao, and X. Liu, ``Sam3d: Segment anything in 3d scenes,'' ArXiv, vol. abs/2306.03908, 2023

    Pith/arXiv arXiv 2023

  32. [32]

    Zhang, D

    D. Zhang, D. Liang, H. Yang, Z. Zou, X. Ye, Z. Liu, and X. Bai, ``Sam3d: Zero-shot 3d object detection via segment anything model,'' ArXiv, vol. abs/2306.02245, 2023

    Pith/arXiv arXiv 2023

  33. [33]

    Siddiqui, L

    Y. Siddiqui, L. Porzi, S. R. Bul\`o, N. M\"uller, M. Nie ner, A. Dai, and P. Kontschieder, ``Panoptic lifting for 3d scene understanding with neural fields,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 9043--9052

    work page 2023

  34. [34]

    Q. Shen, X. Yang, and X. Wang, ``Anything-3d: Towards single-view anything reconstruction in the wild,'' ArXiv, vol. abs/2304.10261, 2023

    Pith/arXiv arXiv 2023

  35. [35]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, ``High-resolution image synthesis with latent diffusion models,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 10\,684--10\,695

    work page 2022

  36. [36]

    J. Cen, Z. Zhou, J. Fang, C. Yang, W. Shen, L. Xie, X. Zhang, and Q. Tian, ``Segment anything in 3d with nerfs,'' NeurIPS, 2023

    work page 2023

  37. [37]

    Zhang, F

    H. Zhang, F. Li, and N. Ahuja, ``Open-nerf: Towards open vocabulary nerf decomposition,'' in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), January 2024, pp. 3456--3465

    work page 2024

  38. [38]

    M. Liu, C. Xu, H. Jin, L. Chen, T. MukundVarma, Z. Xu, and H. Su, ``One-2-3-45: Any single image to 3d mesh in 45 seconds without per-shape optimization,'' NeurIPS, 2023

    work page 2023

  39. [39]

    Gropp, L

    A. Gropp, L. Yariv, N. Haim, M. Atzmon, and Y. Lipman, ``Implicit geometric regularization for learning shapes,'' Proceedings of Machine Learning and Systems 2020, 2020

    work page 2020

  40. [40]

    Jensen, A

    R. Jensen, A. Dahl, G. Vogiatzis, E. Tola, and H. Aan s, ``Large scale multi-view stereopsis evaluation,'' in Proceedings of the IEEE conference on computer vision and pattern recognition, 2014, pp. 406--413

    work page 2014

  41. [41]

    Y. Yao, Z. Luo, S. Li, J. Zhang, Y. Ren, L. Zhou, T. Fang, and L. Quan, ``Blendedmvs: A large-scale dataset for generalized multi-view stereo networks,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 1790--1799

    work page 2020

  42. [42]

    Z. Li, T. M\"uller, A. Evans, R. H. Taylor, M. Unberath, M.-Y. Liu, and C.-H. Lin, ``Neuralangelo: High-fidelity neural surface reconstruction,'' in IEEE Conference on Computer Vision and Pattern Recognition ( CVPR ) , 2023

    work page 2023

  43. [43]

    J. L. Schönberger and J.-M. Frahm, ``Structure-from-motion revisited,'' in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 4104--4113

    2016

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