Pith. sign in

REVIEW 4 major objections 6 minor 68 references

Scalable and High-Quality Neural Implicit Representation for 3D Reconstruction

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Modeling a scene as overlapping local neural SDFs, registered and blended, yields higher-fidelity and scalable 3D reconstruction than a single global network.

desk verdict A solid engineering contribution: the core claim that local SDFs beat a global one survives scrutiny, but the scalability story is under-validated and the registration evaluation is partly self-referential. read the letter →

arxiv 2501.08577 v1 pith:R5KOZBAX submitted 2025-01-15 cs.CV cs.GR

Leyuan Yang , Bailin Deng , Juyong Zhang This is my paper · ORCID
classification cs.CVcs.GR
keywords signeddistancefunctionneuralimplicitsurface3DreconstructionscenegraphregistrationSDFblendingscalabledivide-and-conquer
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

The paper argues that the standard neural-SDF approach—one global multi-layer perceptron representing the whole scene—limits both accuracy and scale, and it proposes a divide-and-conquer representation instead: a graph of independent local SDFs with overlapping regions that are reconstructed separately, registered using shared images, and blended into one global SDF. The intended payoff is that each local network concentrates its capacity on a small region, so fine detail improves while the scene can be extended by adding nodes. On the Lego scene the Chamfer distance improves by 45.6% going from the single-network baseline to eight nodes, and the same pipeline reconstructs a 25-node campus scene covering roughly 1200 m by 800 m. If the representation works as claimed, it is a practical route to urban-scale neural surface reconstruction without a single global pose optimization.

What carries the argument

The signed distance function (SDF) is the function whose zero set is the surface, and each graph node stores one local SDF. Two mechanisms carry the argument. First, overlap-based registration: because shared images see the same 3D point in both local coordinate systems, the relation $P^i_h \mathcal{T}'_{ij}=P^j_h$ yields an overdetermined linear system whose least-squares solution is an initial similarity transform; this transform is then optimized by volume-rendering the registered node's color field in the reference node and minimizing the color difference over the shared-image masks. Second, softmax SDF blending: instead of taking $\min(f_i(x),f_j(x))$ in overlapping regions, the method averages with weights that decay with distance to each node's boundary, eliminating the seam while staying close to a valid SDF. The graph's minimum spanning tree is used so that each pair of adjacent nodes is registered only once and the global coordinate frame is propagated from node to node.

What would settle it

Take a pair of overlapping nodes from a large scene, apply a smooth non-rigid warp (for example, a 5% scale gradient across the overlap) to one node's structure-from-motion output, then run the full registration; if the optimized color-consistency loss converges to a visible seam or a wrong transform, the single-similarity-transform assumption is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a single global MLP is the bottleneck in neural SDF reconstruction, and that representing a scene as a Neural SDF-Graph removes that bottleneck. The graph $G=(V,E)$ has one node $v_k$ per local SDF, each reconstructed from its own subset of images and its own camera poses; an edge $e_{ij}$ marks two nodes that share images and therefore overlap in space. Adjacent nodes are aligned by an initial similarity transform $\mathcal{T}'_{ij}$ (rotation, translation, scale) obtained in closed form from the shared camera poses, then refined by minimizing a color-consistency loss through volume rendering. The aligned local SDFs are fused with softmax weights $\omega^k_{ij}(x)=e^{-\beta d_k(x)}/(e^{-\beta d_i(x)}+e^{-\beta d_j(x)})$, where $d_k(x)$ is the distance to the boundary of node $v_k$, producing a global SDF without the seam that a plain min-union creates. The quantitative evidence is that reconstruction error falls steadily as node count rises—45.6% better Chamfer distance on Lego with eight nodes than the single-network baseline—and that the same recipe handles a 25-node, roughly 1200 m by 800 m campus.

Load-bearing premise

The pipeline assumes that any two overlapping node reconstructions differ only by one global rotation, translation, and scale; if per-node camera estimation leaves non-rigid drift or locally varying scale, the registration has no model for it and the color-based refinement must silently compensate.

Editorial extensions

If this is right

  • With more local nodes, reconstruction detail improves monotonically in the reported tests: the Lego Chamfer distance drops from $5.209\times 10^{-4}$ for a single network to $2.836\times 10^{-4}$ with eight nodes.
  • Scene size is no longer bounded by one network: adding nodes along the data-collection path produced a 25-node reconstruction of a roughly 1200 m by 800 m campus from 5973 oblique aerial images.
  • Per-node camera estimation is much cheaper than global pose estimation: on the Sub-Campus scene, structure-from-motion time fell from 14 h 25 min to 5 h 8 min.
  • The independence of nodes supports downstream uses: each node's mesh and texture can be extracted and optimized separately, and a node's registration transformation can be changed to move objects within the scene.
  • Softmax-based SDF blending removes the seam at overlap boundaries while preserving reconstruction accuracy in the overlap area.

Reading between the lines

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

  • Not stated in the paper: the registration step is agnostic to what the local field represents, so the same overlapping-node graph could wrap explicit representations such as 3D Gaussian splatting or mesh blocks; the softmax blending would need an SDF-like overlap analogue.
  • Not stated in the paper: because the global frame is propagated along a minimum spanning tree, any residual registration error on an early edge is inherited by all later nodes, so a loop-closure or global bundle-adjustment pass over the whole graph is the natural stress test for very large scenes.
  • Not stated in the paper: the overlap fraction is chosen by hand in the experiments (roughly 15-42% for Lego divisions, 20% for Jade), so an automatic schedule that trades registration reliability against wasted training is a testable extension.
  • Not stated in the paper: the authors note that neural SDF methods assume watertight surfaces; a non-watertight local SDF model would remove the irregular iso-surface shapes they observe near node boundaries.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes Neural SDF-Graph, a divide-and-conquer representation for neural implicit surface reconstruction. A scene is decomposed into overlapping local SDF nodes, each reconstructed independently with an existing method such as NeuS or Bakedangelo; adjacent nodes are registered using a least-squares similarity transform initialized from shared COLMAP poses and then refined by a volume-rendering color loss; finally the local SDFs are combined with a softmax-weighted blending scheme. Experiments on Lego, Jade, BlendedMVS, Actors-HQ, and two drone-captured campus datasets report improved Chamfer distance and F-score relative to single-network baselines, and demonstrate scaling to 25 nodes over a roughly 1200m x 800m area, along with texture generation and scene editing applications.

Significance. The central claim—that partitioning a scene into multiple overlapping local SDFs improves reconstruction fidelity and enables scalable reconstruction—is supported by consistent quantitative gains on external benchmarks with ground-truth meshes, e.g., Lego Chamfer distance improving from 5.209e-4 (baseline) to 2.836e-4 (8 nodes) in Table 1 and Jade from 2.507e-4 to 0.958e-4 (4 nodes) in Table 2. The paper also includes a parameter-count ablation (Table 7) and a comparison against recent 3DGS-based reconstructions, and it uses publicly available components (COLMAP, NeuS, SDFStudio), which helps reproducibility. The main open risk is the validation of the registration chain underlying the city-scale claim: only five of the twenty-four Campus edges are reported in Table 6, the registration optimization is self-supervised with no independent geometric ground truth, and the effect of error accumulation along the MST is not quantified.

major comments (4)
  1. [Sec. 3.3 and Sec. 4.4] The city-scale claim rests on composing 24 pairwise similarity transforms along the MST to obtain a global coordinate system, but the paper reports rendering metrics for only 5 of the 24 edges (Table 6) and provides no analysis of how rotation, translation, and scale residuals propagate through the composition. Please add either a global geometric validation (e.g., distances between fixed landmarks, comparison with a global COLMAP alignment, or loop-closure errors) or a quantitative residual analysis along the full chain; without this, the absence of visible seams in Figures 10–11 cannot rule out accumulated misalignment.
  2. [Sec. 3.3, Eq. (12)] The registration optimization minimizes the color difference between two renderings from the target node's color field C_j, and the Table 6 metrics are computed against those same rendered images. This is a self-supervised validation: a local minimum of Eq. (12) can leave a systematic geometric bias (e.g., a small rotation or scale error) that rendering metrics would not reveal. Please validate the final alignment with an independent geometric criterion on at least one scene, such as a global bundle adjustment or manually measured control points.
  3. [Sec. 3.3, Eqs. (4)-(9)] The initial registration in Eq. (8) assumes that the two per-node COLMAP reconstructions are related by a single similarity transform over the whole overlap region. If the per-node reconstructions contain non-rigid drift or spatially varying scale, the least-squares initial guess is biased and the color-based refinement may not compensate. The manuscript does not analyze the magnitude of such residuals or the sensitivity of the final reconstruction to them; please add an experiment that perturbs the initial transform and measures the resulting registration error, or report drift statistics over the overlap regions.
  4. [Sec. 4.2, Tables 1-3] All quantitative claims, including the headline '45.6% improvement in Lego Chamfer distance,' are based on single runs with no error bars or multiple random seeds. Given that the core assertion is a consistent quality gain from increasing the number of nodes, please provide at least three runs per configuration or otherwise report the variance, so that the significance of the gains relative to training stochasticity can be assessed.
minor comments (6)
  1. [Throughout] There are several typos and grammatical slips, e.g., 'loacl' (Sec. 1), 'ddemonstrate' (Sec. 4.3), 'basline' (Sec. 4.3), 'specfied' (Sec. 3.4), and 'T raining' (Appendix A.1.1).
  2. [Sec. 3.3, Eq. (8)] The summation index and matrix dimension use 'N12' in Eq. (8), but the text defines the number of shared images as Nij; please correct the notation.
  3. [Sec. 4.2] The reported overlap percentages for the Lego two-node division (41.67%, 24.75%, 18.38%) are stated without explaining which denominator each percentage refers to; please clarify the definitions of 'lower node range,' 'upper node range,' and 'global range.'
  4. [Table 5] The COLMAP runtimes for the four Sub-Campus nodes sum to 5h2m (1h47m + 58m + 1h33m + 44m), but the table reports 'Total: 5h8m'; please check the arithmetic.
  5. [Fig. 14 and Sec. 4.6] Figure 14 shows ablation results labeled 'Lego,' 'Campus,' and 'Human body,' but the text of Section 4.6 appears to describe only the Lego case; please clarify which dataset each quantitative pair corresponds to.
  6. [Sec. 4.4] The statement that each overlapping node pair shares 'around 100' images is vague; please provide a range or a small table of shared-image counts for the Campus nodes.

Circularity Check

1 steps flagged · score 4.0 of 10

Central reconstruction claims are grounded in external ground-truth benchmarks, but the registration-optimization validation is partly circular: the reported 'final' rendering metrics are the very objective minimized in Eq. (12).

  1. fitted input called prediction [Section 3.3, Eq. (12); Section 4.4, Table 6]
    "Therefore, we employ a color loss to optimize the registration matrix: ∆Tij = arg min X h ∥(cij h − cjj h ) · M ij h ∥. (12) ... We expect the optimized “final” to reach the “target” metrics, which means that the registration is aligned."

    The registration parameters are optimized by minimizing the masked color difference between renders from the transformed pose and renders from the target node's pose (Eq. 12). Table 6 then reports PSNR/SSIM/LPIPS of 'final' versus 'target' on exactly those rendered views. PSNR is monotonic in the color error minimized by Eq. (12), so 'final' approaching 'target' is forced by the optimization rather than by any independent geometric check. A transform that overfits the color field could also reduce this loss, so this validation cannot by itself establish geometric alignment.

full rationale

The paper's central claims—higher-fidelity reconstruction and scalability via multiple local SDFs—are checked against external ground-truth meshes (Lego CD/F-score, BlendedMVS, Actors-HQ) and against single-MLP baselines under matched settings, so those claims are not circular. The MST error-accumulation concern is a real correctness risk but is not a circularity. The one genuine circular element is the registration-optimization validation: the metric used to declare success (Table 6 'final' vs 'target' rendering metrics) is the same color-rendering objective being minimized in Eq. (12), so the improvement is partly by construction. No load-bearing self-citation chain, imported uniqueness theorem, or self-defined equivalence was found.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a small number of hand-set hyperparameters (blending temperature, node partitions) and on domain assumptions about SfM consistency and about the validity of rendering-based registration. No new physical entities are introduced; the SDF graph is an organizational abstraction, not a postulated entity. Existing neural implicit methods (NeuS, Bakedangelo) and COLMAP are used as black-box components.

free parameters (2)
  • softmax blending temperature beta = 10
    User-specified in Section 3.4, Eq. (16); controls the sharpness of SDF blending and is held at 10 for all experiments. The choice is not derived from first principles and is introduced to avoid seams.
  • node partition and overlap ratios = e.g., Lego overlaps 41.67%, 24.75%, 18.38% of ranges; Jade 20%
    Section 4.2 describes hand-chosen cuboid partitions and overlap amounts per dataset, based on object structure or data distribution. These choices affect registration success and reconstruction quality; they are manual hyperparameters, not learned.
assumptions (4)
  • domain assumption Independent COLMAP reconstructions of overlapping nodes are related by a single similarity transformation (rotation, translation, scale) over the whole overlap region.
    Invoked in Section 3.3, Eq. (4)-(9), where a single registration matrix T_ij^0 is solved from shared camera poses. If the SfM reconstructions have non-rigid drift, the initial registration is biased.
  • domain assumption The target node's trained radiance field provides an accurate appearance in the overlap, so the color loss in Eq. (12) is a valid proxy for geometric alignment.
    The registration refinement minimizes the difference between renderings of the source camera pose in the target field and the target's own renderings (Eq. 11-12). If the target field has appearance errors or lighting/exposure differences in the shared region, the optimization may produce a biased alignment. The authors acknowledge using renderings rather than ground truth to avoid wrong supervision directions.
  • ad hoc to paper The softmax-weighted combination of two SDFs in Eq. (15)-(16) yields a valid global SDF whose zero level set approximates the true surface.
    Section 3.4 introduces the blending scheme with a user-specified beta to remove seams; a weighted average of SDFs is not generally a signed distance function, and the zero set may be biased where local SDFs disagree. The paper relies on this for the final global SDF and mesh extraction.
  • domain assumption Existing neural implicit reconstruction methods assume watertight surfaces, which may not hold for each node's shape.
    Stated in the Discussion (Section 5). The paper notes this can cause irregular iso-surface shapes at node boundaries, mitigated only heuristically by SDF blending.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scalable and High-Quality Neural Implicit Representation for 3D Reconstruction." pith.science (2026). https://pith.science/paper/R5KOZBAX

@misc{pith2026250108577,
  author       = {Pith},
  title        = {Pith review of: Scalable and High-Quality Neural Implicit Representation for 3D Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R5KOZBAX}},
  note         = {Machine review of arXiv:2501.08577}
}
read the original abstract

Various SDF-based neural implicit surface reconstruction methods have been proposed recently, and have demonstrated remarkable modeling capabilities. However, due to the global nature and limited representation ability of a single network, existing methods still suffer from many drawbacks, such as limited accuracy and scale of the reconstruction. In this paper, we propose a versatile, scalable and high-quality neural implicit representation to address these issues. We integrate a divide-and-conquer approach into the neural SDF-based reconstruction. Specifically, we model the object or scene as a fusion of multiple independent local neural SDFs with overlapping regions. The construction of our representation involves three key steps: (1) constructing the distribution and overlap relationship of the local radiance fields based on object structure or data distribution, (2) relative pose registration for adjacent local SDFs, and (3) SDF blending. Thanks to the independent representation of each local region, our approach can not only achieve high-fidelity surface reconstruction, but also enable scalable scene reconstruction. Extensive experimental results demonstrate the effectiveness and practicality of our proposed method.

Figures

Figures reproduced from arXiv: 2501.08577 by the authors.

Figure 1
Figure 1. Pipeline. At the bottom, we present the main construction processes of our method: Constructing the distribution and overlap relationship of the local radiance fields based on object structure or data distribution, Adjacent Nodes Registration, and SDF Blending. At the top, we take adjacent nodes vi and vj as an example to explain in detail how to use volume rendering for registration and then blending by the softmax… view at source ↗
Figure 2
Figure 2. Mesh Visualization for Optimization of Registration. The left image shows misalignment caused by the initial registration, while the right displays the optimized result. matrix, which denotes the transformation from the world coordinate system into the camera coordinate system: xc = Pxw, (3) where xw ∈ R 3 is a point in the world coordinate system, and xc ∈ R 3 is the corresponding point in the camera coordinate sys… view at source ↗
Figure 3
Figure 3. Blending. The top images depict simplified SDF contours within a vertical y-axis section. The left image shows a visible seam caused by directly taking the minimum value, while the right displays the smoothed result achieved through Softmax weighting. Note that we use the rendering results rather than the ground-truth images as supervision. This is because various factors may hinder accurate learning in certain regi… view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Lego Nodes Division. From left to right are the node divisions of 2-nodes, 4-nodes and 8-nodes. Ground-Truth Baseline 2-Nodes 4-Nodes 8-Nodes [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Visualization of High-Quality Reconstruction Results for Lego. The right side is the comparison, demonstrating that our representation progressively reconstructs finer details. In addition, for the evaluation of scalable large scene re￾construction, we constructed a la…
Figure 6
Figure 6. Figure 6: Visualization of High-Quality Reconstruction of Jade. It can be seen that as the number of nodes increases, the carvings and hollowing out of the jade are gradually reconstructed and refined. TABLE 2 Quantitative Evaluation on the Jade Dataset. The optimal result is in…
Figure 7
Figure 7. Figure 7: Visualization of High-Quality Reconstruction of BlendedMVS Datasets. By magnifying and comparing local details, our method achieves higher-quality reconstruction and recovers more high-fidelity details. TABLE 3 Quantitative Evaluation on the BlendedMVS Datasets. For F-…
Figure 8
Figure 8. Figure 8: Visualization of High-Quality Reconstruction of Human Body. By the magnified comparison of local areas, our method recovers more details on clothes and shoes than the overall reconstruction. reconstructed and delineated with increasing detail. There￾fore, both quantita…
Figure 9
Figure 9. Figure 9: Visualization of High-Quality Reconstruction of Sub-Campus with 4 Nodes. On the left side, we present an overview of the object-based scene Sub-Campus reconstructed using our method, featuring the four objects utilized for node partitioning. A detailed comparison of lo…
Figure 10
Figure 10. Figure 10: Process of data collection and Scalable Scene Reconstruction of Campus Dataset. The image above illustrates the process of reconstructing the scalable large scene, Campus, based on the direction and extent of data collection. The Reconstructed Complete Mesh Rendering …
Figure 11
Figure 11. Figure 11: Visualization Results of Scalable Scene Reconstruction on Campus Dataset. A comprehensive view of meshes and textures is presented above. Below is a partial close-up comparison. The top row displays the ground-truth image, the middle row showcases the rendered image u…
Figure 12
Figure 12. Figure 12: Rendering Evaluation for Optimization of Registration. We present rendering visualizations and metrics for the registered node, initial registration, optimized registration, and a colormap indicating errors relative to supervision (registered node). The partial magnif…
Figure 13
Figure 13. Figure 13: Editing Results. From left to right: RGB images, original mesh, edited mesh examples 1 and 2. numerical and visual results demonstrates the necessity and effectiveness of our registration optimization method. After registration optimization, we conduct blending for al…
Figure 14
Figure 14. Figure 14: Ablation Study of SDF Blending. Quantitatively evaluate the precision of the mesh after minimum-based and softmax-based SDF blending in the common area respectively. Clearly, our proposed softmax-based blending not only eliminates the seam problem but also maintains t…
Figure 15
Figure 15. Figure 15: Visualization Results of Sub-Campus Dataset Compared with Other Representation. We compare our method with some novel and state-of-the-art geometric surface reconstruction methods based on 3DGS [62], including Gaussian Opacity Fields (GOF) [63], PGSR [64], and 2D Gaus…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 58 canonical work pages

  1. [1]

    Accurate, dense, and robust multi- view stereopsis,

    Y. Furukawa and J. Ponce, “Accurate, dense, and robust multi- view stereopsis,” IEEE T ransactions on Pattern Analysis and Machine Intelligence, vol. 32, no. 8, pp. 1362–1376, 2010. 1

    work page 2010

  2. [2]

    Photorealistic scene reconstruction by voxel coloring,

    S. Seitz and C. Dyer, “Photorealistic scene reconstruction by voxel coloring,” in Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition , 1997, pp. 1067–1073. 1, 2

    work page 1997

  3. [3]

    Efficient large-scale multi-view stereo for ultra high-resolution image sets,

    E. Tola, C. Strecha, and P . Fua, “Efficient large-scale multi-view stereo for ultra high-resolution image sets,” Mach. Vision Appl. , vol. 23, no. 5, p. 903–920, sep 2012. [Online]. Available: https://doi.org/10.1007/s00138-011-0346-8 1

    work page doi:10.1007/s00138-011-0346-8 2012

  4. [4]

    Nerf: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P . P . Srinivasan, M. Tancik, J. T. Barron, R. Ra- mamoorthi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,” in ECCV, 2020. 1, 6, 18

    work page 2020

  5. [5]

    Volume rendering of neural implicit surfaces,

    L. Yariv, J. Gu, Y. Kasten, and Y. Lipman, “Volume rendering of neural implicit surfaces,” in Thirty-Fifth Conference on Neural Information Processing Systems , 2021. 1, 2, 4

    work page 2021

  6. [6]

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

    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,” in Proc. Advances in Neural Information Processing Systems (NeurIPS) , vol. 34, 2021, pp. 27 171–27 183. 1, 2, 4, 8

    work page 2021

  7. [7]

    Neus2: Fast learning of neural implicit surfaces for multi- view reconstruction,

    Y. Wang, Q. Han, M. Habermann, K. Daniilidis, C. Theobalt, and L. Liu, “Neus2: Fast learning of neural implicit surfaces for multi- view reconstruction,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , 2023. 1, 2, 4

    work page 2023

  8. [8]

    Neuralangelo: High-fidelity neural surface reconstruction,

    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. 1, 2, 4, 10, 18

    work page 2023

Show all 68 references
  1. [9]

    Structure-from-motion re- visited,

    J. L. Sch ¨onberger and J.-M. Frahm, “Structure-from-motion re- visited,” in Conference on Computer Vision and Pattern Recognition (CVPR), 2016. 1, 5, 8

    2016

  2. [10]

    Pixelwise view selection for unstructured multi-view stereo,

    J. L. Sch ¨onberger, E. Zheng, M. Pollefeys, and J.-M. Frahm, “Pixelwise view selection for unstructured multi-view stereo,” in European Conference on Computer Vision (ECCV) , 2016. 1, 8

    2016

  3. [11]

    From coarse to fine: Robust hierarchical localization at large scale,

    P .-E. Sarlin, C. Cadena, R. Siegwart, and M. Dymczyk, “From coarse to fine: Robust hierarchical localization at large scale,” in CVPR, 2019. 1

    2019

  4. [12]

    Block-nerf: Scalable large scene neural view synthesis,

    M. Tancik, V . Casser, X. Yan, S. Pradhan, B. Mildenhall, P . P . Srinivasan, J. T. Barron, and H. Kretzschmar, “Block-nerf: Scalable large scene neural view synthesis,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2022, pp...

    2022

  5. [13]

    Mega-nerf: Scal- able construction of large-scale nerfs for virtual fly-throughs,

    H. Turki, D. Ramanan, and M. Satyanarayanan, “Mega-nerf: Scal- able construction of large-scale nerfs for virtual fly-throughs,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2022, pp. 12 922–12 931. 1, 2

    2022

  6. [14]

    A probabilistic framework for space carving,

    A. Broadhurst, T. Drummond, and R. Cipolla, “A probabilistic framework for space carving,” in Proceedings Eighth IEEE Inter- national Conference on Computer Vision. ICCV 2001 , vol. 1, 2001, pp. 388–393 vol.1. 2

    2001

  7. [15]

    Three-dimensional reconstruction of complex shapes based on the Delaunay triangulation,

    J.-D. Boissonnat and B. Geiger, “Three-dimensional reconstruction of complex shapes based on the Delaunay triangulation,” in Biomedical Image Processing and Biomedical Visualization , R. S. Acharya and D. B. Goldgof, Eds., vol. 1905, International Society for Optics and Photon...

    1905 doi

  8. [16]

    Coarse-to-fine surface reconstruction from silhouettes and range data using mesh deformation,

    Y. Sahillio ˘glu and Y. Yemez, “Coarse-to-fine surface reconstruction from silhouettes and range data using mesh deformation,” Computer Vision and Image Understanding , vol. 114, no. 3, pp. 334–348, 2010. [Online]. Available: https://www.sciencedirect. com/science/article/pii/...

    2010

  9. [17]

    Multiview neural surface reconstruction by disentan- gling geometry and appearance,

    L. Yariv, Y. Kasten, D. Moran, M. Galun, M. Atzmon, B. Ronen, and Y. Lipman, “Multiview neural surface reconstruction by disentan- gling geometry and appearance,” Advances in Neural Information Processing Systems, vol. 33, 2020. 2

    2020

  10. [18]

    Monosdf: Exploring monocular geometric cues for neural implicit surface reconstruction,

    Z. Yu, S. Peng, M. Niemeyer, T. Sattler, and A. Geiger, “Monosdf: Exploring monocular geometric cues for neural implicit surface reconstruction,” Advances in Neural Information Processing Systems (NeurIPS), 2022. 2, 4

    2022

  11. [19]

    Geo-neus: Geometry- consistent neural implicit surfaces learning for multi-view re- construction,

    Q. Fu, Q. Xu, Y.-S. Ong, and W. Tao, “Geo-neus: Geometry- consistent neural implicit surfaces learning for multi-view re- construction,” Advances in Neural Information Processing Systems (NeurIPS), 2022. 2, 4

    2022

  12. [20]

    Instant neural graphics primitives with a multiresolution hash encoding,

    T. M ¨uller, A. Evans, C. Schied, and A. Keller, “Instant neural graphics primitives with a multiresolution hash encoding,” ACM T rans. Graph., vol. 41, no. 4, pp. 102:1–102:15, Jul. 2022. [Online]. Available: https://doi.org/10.1145/3528223.3530127 2

    2022

  13. [21]

    Large scale multi-view stereopsis evaluation,

    R. Jensen, A. Dahl, G. Vogiatzis, E. Tola, and H. Aanæs, “Large scale multi-view stereopsis evaluation,” in 2014 IEEE Conference on Computer Vision and Pattern Recognition , 2014, pp. 406–413. 2, 18

    2014

  14. [22]

    Building rome in a day,

    S. Agarwal, N. Snavely, I. Simon, S. M. Seitz, and R. Szeliski, “Building rome in a day,” in2009 IEEE 12th International Conference on Computer Vision, 2009, pp. 72–79. 2

    2009

  15. [23]

    An automated method for large-scale, ground-based city model acquisition,

    C. Fr ¨uh and A. Zakhor, “An automated method for large-scale, ground-based city model acquisition,” International Journal of Com- puter Vision, vol. 60, pp. 5–24, 10 2004. 2

    2004

  16. [24]

    Modeling and recognition of landmark image collections using iconic scene graphs,

    R. Raguram, C. Wu, J. M. Frahm, and S. Lazebnik, “Modeling and recognition of landmark image collections using iconic scene graphs,” International Journal of Computer Vision , vol. 95, no. 3, pp. 213–239, 2011. 2

    2011

  17. [25]

    Detailed real-time urban 3d reconstruction from video,

    M. Pollefeys, D. Nist ´er, J. M. Frahm, A. Akbarzadeh, P . Mordohai, B. Clipp, C. Engels, D. Gallup, S. J. Kim, P . Merrell, C. Salmi, S. Sinha, B. Talton, L. Wang, Q. Yang, H. Stew ´enius, R. Yang, G. Welch, and H. Towles, “Detailed real-time urban 3d reconstruction from vide...

    2008 doi

  18. [26]

    Structure-from-motion re- visited,

    J. L. Sch ¨onberger and J.-M. Frahm, “Structure-from-motion re- visited,” in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 4104–4113. 2 16

    2016

  19. [27]

    Photo tourism: Exploring photo collections in 3d,

    N. Snavely, S. M. Seitz, and R. Szeliski, “Photo tourism: Exploring photo collections in 3d,” vol. 25, no. 3, 2006. [Online]. Available: https://doi.org/10.1145/1141911.1141964 2

    2006 arXiv

  20. [28]

    Very large-scale global sfm by distributed motion averaging,

    S. Zhu, R. Zhang, L. Zhou, T. Shen, T. Fang, P . Tan, and L. Quan, “Very large-scale global sfm by distributed motion averaging,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, 2018, pp. 4568–4577. 2

    2018

  21. [29]

    Fast training of neural lumigraph representations using meta learning,

    A. Bergman, P . Kellnhofer, and G. Wetzstein, “Fast training of neural lumigraph representations using meta learning,” in Advances in Neural Information Processing Systems , M. Ranzato, A. Beygelzimer, Y. Dauphin, P . Liang, and J. W. Vaughan, Eds., vol. 34. Curran Associates,...

    2021

  22. [30]

    Learned initializations for optimizing coordinate-based neural representations,

    M. Tancik, B. Mildenhall, T. Wang, D. Schmidt, P . P . Srinivasan, J. T. Barron, and R. Ng, “Learned initializations for optimizing coordinate-based neural representations,” CoRR, vol. abs/2012.02189, 2020. [Online]. Available: https://arxiv.org/abs/ 2012.02189 2

    2012 arXiv

  23. [32]

    Ibrnet: Learning multi-view image-based rendering,

    Q. Wang, Z. Wang, K. Genova, P . P . Srinivasan, H. Zhou, J. T. Barron, R. Martin-Brualla, N. Snavely, and T. A. Funkhouser, “Ibrnet: Learning multi-view image-based rendering,” CoRR, vol. abs/2102.13090, 2021. [Online]. Available: https://arxiv.org/abs/ 2102.13090 2

    2021 arXiv

  24. [34]

    Nerfusion: Fusing radiance fields for large-scale scene reconstruction,

    X. Zhang, S. Bi, K. Sunkavalli, H. Su, and Z. Xu, “Nerfusion: Fusing radiance fields for large-scale scene reconstruction,”CVPR,

  25. [35]

    Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction,

    C. Sun, M. Sun, and H. Chen, “Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction,” CoRR, vol. abs/2111.11215, 2021. [Online]. Available: https: //arxiv.org/abs/2111.11215 2

    2021 arXiv

  26. [36]

    Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction,

    C. Sun, M. Sun, and H.-T. Chen, “Direct voxel grid optimization: Super-fast convergence for radiance fields reconstruction,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 5449–5459. 2

    2022

  27. [37]

    Plenoxels: Radiance fields without neural net- works,

    S. Fridovich-Keil, A. Yu, M. Tancik, Q. Chen, B. Recht, and A. Kanazawa, “Plenoxels: Radiance fields without neural net- works,” in CVPR, 2022. 2

    2022

  28. [38]

    Grid-guided neural radiance fields for large urban scenes,

    L. Xu, Y. Xiangli, S. Peng, X. Pan, N. Zhao, C. Theobalt, B. Dai, and D. Lin, “Grid-guided neural radiance fields for large urban scenes,” 2023. 2

    2023

  29. [39]

    Switch-nerf: Learning scene decomposition with mixture of experts for large-scale neural radiance fields,

    Z. Mi and D. Xu, “Switch-nerf: Learning scene decomposition with mixture of experts for large-scale neural radiance fields,” in International Conference on Learning Representations (ICLR) ,

  30. [40]

    nerf2nerf: Pairwise registration of neural radiance fields,

    L. Goli, D. Rebain, S. Sabour, A. Garg, and A. Tagliasacchi, “nerf2nerf: Pairwise registration of neural radiance fields,” 2022. 2

    2022

  31. [41]

    Barf: Bundle- adjusting neural radiance fields,

    C.-H. Lin, W.-C. Ma, A. Torralba, and S. Lucey, “Barf: Bundle- adjusting neural radiance fields,” 2021. 2

    2021

  32. [42]

    Local-to-global registration for bundle-adjusting neural radiance fields,

    Y. Chen, X. Chen, X. Wang, Q. Zhang, Y. Guo, Y. Shan, and F. Wang, “Local-to-global registration for bundle-adjusting neural radiance fields,” 2023. 2

    2023

  33. [43]

    Visual-inertial odometry priors for bundle-adjusting neural radiance fields,

    H. Kim, M. Song, D. Lee, and P . Kim, “Visual-inertial odometry priors for bundle-adjusting neural radiance fields,” in 2022 22nd International Conference on Control, Automation and Systems (ICCAS), 2022, pp. 1131–1136. 2

    2022

  34. [44]

    Zero nerf: Registra- tion with zero overlap,

    C. Peat, O. Batchelor, R. Green, and J. Atlas, “Zero nerf: Registra- tion with zero overlap,” 2022. 2

    2022

  35. [45]

    iNeRF: Inverting neural radiance fields for pose estima- tion,

    L. Yen-Chen, P . Florence, J. T. Barron, A. Rodriguez, P . Isola, and T.-Y. Lin, “iNeRF: Inverting neural radiance fields for pose estima- tion,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021. 5

    2021

  36. [46]

    Parallel inversion of neural radiance fields for robust pose estimation,

    Y. Lin, T. M ¨uller, J. Tremblay, B. Wen, S. Tyree, A. Evans, P . A. Vela, and S. Birchfield, “Parallel inversion of neural radiance fields for robust pose estimation,” 2023. 5

    2023

  37. [47]

    Adam: A method for stochastic optimiza- tion,

    D. P . Kingma and J. Ba, “Adam: A method for stochastic optimiza- tion,” 2017. 5

    2017

  38. [48]

    Designing with distance fields,

    S. F. Frisken and R. N. Perry, “Designing with distance fields,” ACM SIGGRAPH 2006 Courses , pp. 60–66, 2006. 6

    2006

  39. [49]

    Object-compositional neural implicit surfaces,

    Q. Wu, X. Liu, Y. Chen, K. Li, C. Zheng, J. Cai, and J. Zheng, “Object-compositional neural implicit surfaces,” in European Con- ference on Computer Vision . Springer, 2022, pp. 197–213. 6

    2022

  40. [50]

    Objectsdf++: Im- proved object-compositional neural implicit surfaces,

    Q. Wu, K. Wang, K. Li, J. Zheng, and J. Cai, “Objectsdf++: Im- proved object-compositional neural implicit surfaces,” 2023. 6

    2023

  41. [51]

    Marching cubes: A high resolution 3d surface construction algorithm,

    W. E. Lorensen and H. E. Cline, “Marching cubes: A high resolution 3d surface construction algorithm,” ser. SIGGRAPH ’87. New York, NY, USA: Association for Computing Machinery, 1987, p. 163–169. [Online]. Available: https://doi.org/10.1145/ 37401.37422 6

    1987

  42. [52]

    MeshLab: an Open-Source Mesh Processing Tool,

    P . Cignoni, M. Callieri, M. Corsini, M. Dellepiane, F. Ganovelli, and G. Ranzuglia, “MeshLab: an Open-Source Mesh Processing Tool,” in Eurographics Italian Chapter Conference , V . Scarano, R. D. Chiara, and U. Erra, Eds. The Eurographics Association, 2008. 6

    2008

  43. [53]

    Blendedmvs: A large-scale dataset for generalized multi-view stereo networks,

    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,” CoRR, vol. abs/1911.10127, 2019. [Online]. Available: http://arxiv.org/abs/1911.10127 6, 7

    1911 arXiv

  44. [54]

    Humanrf: High-fidelity neural radiance fields for humans in motion,

    M. Is ¸ık, M. R ¨unz, M. Georgopoulos, T. Khakhulin, J. Starck, L. Agapito, and M. Nießner, “Humanrf: High-fidelity neural radiance fields for humans in motion,” ACM T ransactions on Graphics (TOG), vol. 42, no. 4, pp. 1–12, 2023. [Online]. Available: https://doi.org/10.1145/3...

    2023 doi

  45. [55]

    Pointnet: Deep learning on point sets for 3d classification and segmentation,

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,” CoRR, vol. abs/1612.00593, 2016. [Online]. Available: http: //arxiv.org/abs/1612.00593 7

    2016 arXiv

  46. [56]

    Tanks and temples: Benchmarking large-scale scene reconstruction,

    A. Knapitsch, J. Park, Q.-Y. Zhou, and V . Koltun, “Tanks and temples: Benchmarking large-scale scene reconstruction,” ACM T ransactions on Graphics, vol. 36, no. 4, 2017. 7

    2017

  47. [57]

    Multiscale structural similarity for image quality assessment,

    Z. Wang, E. P . Simoncelli, and A. C. Bovik, “Multiscale structural similarity for image quality assessment,” in The Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003 , vol. 2. IEEE, 2003, pp. 1398–1402. 8

    2003

  48. [58]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P . Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 586–595. 8

    2018

  49. [59]

    Sdfstudio: A unified framework for surface reconstruction,

    Z. Yu, A. Chen, B. Antic, S. Peng, A. Bhattacharyya, M. Niemeyer, S. Tang, T. Sattler, and A. Geiger, “Sdfstudio: A unified framework for surface reconstruction,” 2022. [Online]. Available: https://github.com/autonomousvision/sdfstudio 8

    2022

  50. [60]

    Bakedsdf: Meshing neural sdfs for real-time view synthesis,

    L. Yariv, P . Hedman, C. Reiser, D. Verbin, P . P . Srinivasan, R. Szeliski, J. T. Barron, and B. Mildenhall, “Bakedsdf: Meshing neural sdfs for real-time view synthesis,” arXiv, 2023. 10

    2023

  51. [61]

    Accelerating 3d deep learning with py- torch3d,

    N. Ravi, J. Reizenstein, D. Novotny, T. Gordon, W.-Y. Lo, J. John- son, and G. Gkioxari, “Accelerating 3d deep learning with py- torch3d,” arXiv:2007.08501, 2020. 13

    2007 arXiv

  52. [62]

    3d gaussian splatting for real-time radiance field rendering,

    B. Kerbl, G. Kopanas, T. Leimk ¨uhler, and G. Drettakis, “3d gaussian splatting for real-time radiance field rendering,” ACM T ransactions on Graphics, vol. 42, no. 4, July 2023. [Online]. Avail- able: https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/ 14

    2023

  53. [63]

    Gaussian opacity fields: Effi- cient adaptive surface reconstruction in unbounded scenes,

    Z. Yu, T. Sattler, and A. Geiger, “Gaussian opacity fields: Effi- cient adaptive surface reconstruction in unbounded scenes,” ACM T ransactions on Graphics, 2024. 14

    2024

  54. [64]

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

    D. Chen, H. Li, W. Ye, Y. Wang, W. Xie, S. Zhai, N. Wang, H. Liu, H. Bao, and G. Zhang, “Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction,” arXiv preprint arXiv:2406.06521, 2024. 14

    2024 arXiv

  55. [65]

    2d gaus- sian splatting for geometrically accurate radiance fields,

    B. Huang, Z. Yu, A. Chen, A. Geiger, and S. Gao, “2d gaus- sian splatting for geometrically accurate radiance fields,” in SIG- GRAPH 2024 Conference Papers . Association for Computing Machinery, 2024. 14

    2024

  56. [66]

    Mip-nerf 360: Unbounded anti-aliased neural radiance fields,

    J. T. Barron, B. Mildenhall, D. Verbin, P . P . Srinivasan, and P . Hed- man, “Mip-nerf 360: Unbounded anti-aliased neural radiance fields,” CVPR, 2022. 18

    2022

  57. [67]

    Zip-nerf: Anti-aliased grid-based neural radiance fields,

    ——, “Zip-nerf: Anti-aliased grid-based neural radiance fields,” ICCV, 2023. 18 17 Leyuan Yang is currently working as a master student in the School of Mathematical Science, University of Science and Technology of China. His research interests include computer vision, 3D visio...

    2005

  58. [2021]

    Available: https://arxiv.org/abs/2107.12512 2

    [Online]. Available: https://arxiv.org/abs/2107.12512 2

    arXiv

  59. [2022]

    Available: https://arxiv.org/abs/2201.12204 2

    [Online]. Available: https://arxiv.org/abs/2201.12204 2

    arXiv

  60. [2023]

    Available: https://openreview.net/forum?id= PQ2zoIZqvm 2

    [Online]. Available: https://openreview.net/forum?id= PQ2zoIZqvm 2

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.