Pith. sign in

REVIEW 2 major objections 6 minor 71 references

Monocular depth turns global Structure-from-Motion scale-aware and raises camera pose accuracy while still finishing with multi-view optimization.

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 ↗

T0 review · grok-4.5

2026-07-13 02:31 UTC pith:KIJTDYK4

load-bearing objection Solid systems paper that cleanly injects monocular depth into global SfM at the right stages and delivers real pose gains; depth quality is the soft spot but already stress-tested enough. the 2 major comments →

arxiv 2607.09507 v1 pith:KIJTDYK4 submitted 2026-07-10 cs.CV

DGSfM: Depth-Guided Scale-Aware Global Structure-from-Motion

Sithu Aung , Viktor Kocur , Yaqing Ding , Torsten Sattler , Zuzana Kukelova This is my paper
classification cs.CV
keywords Structure-from-MotionGlobal SfMMonocular DepthScale AveragingView-Graph FilteringCamera Pose EstimationBundle Adjustment
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.

Global Structure-from-Motion recovers camera poses and sparse 3D structure from unordered photos by solving the whole view graph at once. Its classic weakness is that pairwise epipolar geometry only recovers translation directions, not absolute scales, so noisy baselines and false image pairs can spoil the global solution. This paper shows that monocular depth maps can be injected as a metric prior at every main stage—scale-aware relative poses, view-graph and match filtering, global scale averaging, and depth-lifted initialization—so that the final multi-view optimization starts from cleaner constraints and more stable poses. On ETH3D and IMC2021 the resulting pipeline consistently improves pose accuracy over strong global baselines for both sparse and dense matchers, without replacing classical geometry by a pure feed-forward model. The practical message is that modern single-image depth is most useful as a guide that strengthens, rather than replaces, explicit multi-view reconstruction.

Core claim

DGSfM shows that monocular depth, used through depth-aware relative pose estimation, scale-consistency filtering, global scale averaging, and depth-guided camera and point initialization, converts scale-ambiguous global SfM into a scale-aware pipeline that yields substantially higher camera pose accuracy than strong baselines across sparse and dense matching front-ends, while the final reconstruction remains fixed by multi-view positioning and bundle adjustment.

What carries the argument

Depth-aware relative pose solving that returns a scaled translation and relative depth-scale for each image pair, followed by robust global scale averaging that aligns all monocular depth maps into one reconstruction scale and supplies depth-lifted cameras and points for GLOMAP-style global positioning.

Load-bearing premise

Monocular depth maps must be accurate enough that, after pairwise solving, robust scale averaging, and depth-consistency pruning, they improve rather than corrupt the global solution.

What would settle it

Replace the monocular depth maps on ETH3D with pure noise or with depths that carry a large systematic bias; if pose AUC then falls below the pure GLOMAP baseline with the same matches, the claim that depth priors help fails.

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

If this is right

  • Global SfM can absorb monocular depth as a prior without redesigning the final multi-view optimization stages.
  • False edges and matches that survive ordinary epipolar checks can be rejected by scaled depth consistency before positioning.
  • Pose accuracy gains appear across sparse (SIFT, ALIKED, LoMa) and dense (RoMa) matchers on ETH3D and IMC2021.
  • When calibration is unknown, focal lengths can be recovered by median averaging of pairwise depth-aware estimates.
  • Depth-guided initialization removes the run-to-run variation that random starts introduce in global positioning.

Where Pith is reading between the lines

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

  • As monocular depth models improve, the same injection pattern should produce further automatic gains without changing the SfM code.
  • The filtering and scale-averaging steps could transfer to incremental SfM or hybrid feed-forward-plus-optimization systems that currently lack robust metric alignment.
  • Closing the loop by refining the depth maps after bundle adjustment, as the paper suggests for future work, would turn the pipeline into a source of denser metric reconstructions.
  • Scenes where monocular depth is systematically wrong remain the natural failure mode and the natural target for uncertainty-aware relative-pose solvers.

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

2 major / 6 minor

Summary. DGSfM is a global Structure-from-Motion pipeline that injects monocular depth priors into GLOMAP-style reconstruction while keeping final poses and points determined by multi-view optimization. For each verified pair it runs a depth-aware relative pose solver (RePoseD) to obtain scaled translations, relative depth scales, and pairwise focals; it then prunes the view graph with Doppelgangers++ and triplet support, and prunes matches by scaled depth consistency. Global scale averaging aligns monocular depths, after which MST chaining of scaled translations and median depth-lifted tracks initialize cameras and points for global positioning and bundle adjustment. On ETH3D, IMC2021, and LaMAR the method reports consistent pose-AUC gains over COLMAP and GLOMAP across sparse and dense matchers, with ablations of each module and sensitivity to several depth estimators.

Significance. If the reported gains hold under independent reimplementation, the paper offers a practical and timely recipe for making global SfM less sensitive to scale-ambiguous baselines and weak view graphs without abandoning explicit multi-view optimization. Strengths include open code, multi-front-end evaluation (SIFT, ALIKED, LoMa, RoMa), module-level ablations (Tab. 5), depth-model sensitivity (Tab. 4), runtime comparison, and point-cloud accuracy/completeness on ETH3D. The design is complementary to dense matchers and feed-forward reconstructors rather than a replacement, which increases likely adoption. The main scientific contribution is systems-level: showing how monocular metric geometry can be used for scale-aware pairwise constraints, filtering, and initialization inside a classical global pipeline.

major comments (2)
  1. Tab. 5 and Tab. 1 (Ours vs Ours†): removing Doppelgangers++ causes one of the largest AUC drops (e.g., ETH3D LoMa-B AUC@5° 85.43→78.78), yet Doppelgangers++ is a visual-disambiguation module independent of monocular depth. The abstract and contribution list frame the gains primarily as depth-guided scale awareness. Please restate the main claim so that non-depth graph filtering and depth-specific components (RePoseD, scale averaging, depth consistency, depth-lifted init) are clearly separated, and keep Ours† as the primary comparison whenever claiming improvement from monocular geometry alone.
  2. §3.3 and Eq. (17): after depth-guided initialization, global positioning uses only multi-view ray consistency and BA; monocular depth does not enter the final residual. This is intentional, but the central claim that depth priors “improve rather than corrupt” global positioning then rests on BA being able to escape bad depth-induced init. Tab. 4 shows weaker depth models still beat non-depth baselines on ETH3D, which is helpful, but the paper should add a short failure-mode analysis (e.g., low-parallax or depth-biased scenes) quantifying when scale averaging / lifted points hurt pose AUC relative to GLOMAP with the same matches, rather than only average gains.
minor comments (6)
  1. Fig. 2 pipeline diagram is dense; label which stages consume monocular depth versus only epipolar geometry so readers can map modules to Sec. 3.1–3.3.
  2. Eq. (7): the relative depth error is normalized by s_ij D_j(x_j); briefly justify the relative form versus absolute depth error and whether τ_d=0.3 was tuned per depth model.
  3. Tab. 1: InstantSfM IMC entries are blank; either fill them or mark N/A explicitly to avoid implying missing evaluation.
  4. Appendix A1 hyperparameters (τ_DG, τ_tri, τ_π, τ_d, min inliers) should be cross-referenced from the main experimental section so the free-parameter set is visible without the appendix.
  5. Related Work comparison to MASt3R-SfM is useful; a short table row already exists, but one sentence on computational profile (feed-forward joint pose+depth vs classical RA + positioning) would help practitioners choose.
  6. Minor wording: “false edges from visually ambiguous pairs can furtherdegradereconstruction” and similar missing spaces appear in the abstract/intro; a proofread pass is needed.

Circularity Check

0 steps flagged

No significant circularity: empirical SfM pipeline uses monocular depth as prior, then refines via multi-view optimization; results are benchmark comparisons, not forced by construction.

full rationale

DGSfM is a systems paper whose central claim is empirical pose-AUC gains on ETH3D/IMC2021 from injecting monocular depth (via RePoseD pairwise solves, depth-consistency pruning, global scale averaging, and depth-lifted init) into a GLOMAP-style global SfM pipeline that still ends with multi-view ray consistency and bundle adjustment. Scale factors, relative poses, and points are estimated from data and then optimized; final poses are not definitionally equal to the monocular inputs. Self-citations (RePoseD, GLOMAP lineage) supply reusable components with independent prior evidence, not load-bearing uniqueness theorems or fitted constants later renamed as predictions. Ablations (Tabs. 4–5) and limitations treat depth quality as an empirical risk rather than a circular premise. No self-definitional loop, fitted-input-as-prediction, or ansatz-smuggling reduction appears in the derivation chain.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The method rests on standard multi-view geometry plus the practical assumption that modern monocular depth is good enough to supply usable scale and consistency signals after robust averaging and filtering. Free parameters are the usual hand-tuned geometric thresholds; no new physical entities are postulated.

free parameters (5)
  • Doppelgangers++ confidence threshold τ_DG = 0.8
    Hand-set to 0.8; controls pairwise visual-disambiguation edge pruning (Appendix A1).
  • Triplet support threshold τ_tri = 0.6
    Hand-set to 0.6; controls triplet-guided edge pruning (Appendix A1).
  • Reprojection / depth-consistency thresholds τ_π, τ_d = 12.0 / 0.3
    Hand-set to 12.0 and 0.3 for scaled depth consistency filtering of correspondences (Appendix A1).
  • Min inliers and inlier ratio for retained pairs = 15 / 0.25
    At least 15 inliers and 0.25 ratio required after filtering (Appendix A1).
  • RePoseD RANSAC / Sampson / reprojection thresholds = 1.0 / 16.0 / 1000
    Sampson 1.0, reprojection 16.0, 1000 RANSAC iterations for depth-aware relative pose (Appendix A1).
axioms (4)
  • domain assumption Monocular depth maps (MoGe family or similar) provide geometry that is accurate enough, after relative-scale estimation and robust averaging, to convert epipolar constraints into useful scale-aware relative poses and to initialize cameras/points.
    Stated throughout Sec. 3 and tested via depth-model ablation (Tab. 4); acknowledged as a limitation in Sec. 6.
  • standard math Standard multi-view geometry: relative rotations compose, scaled translations relate camera centers via R_j(c_i - c_j), and bundle adjustment / global positioning refine under robust multi-view ray consistency.
    Used in rotation averaging, MST chaining (Eqs. 12–13), and global positioning objective (Eq. 17).
  • domain assumption Pixels are square and principal points are at image centers, so only focal lengths need estimation via median aggregation of pairwise RePoseD focals.
    Explicitly assumed in Sec. 3.3 Focal Averaging, citing common practice [15,58].
  • domain assumption A maximum spanning tree of the filtered view graph yields a sufficiently connected, low-noise path for chaining scaled translations into initial camera centers.
    Sec. 3.3 Depth-Scaled Pose-Point Initialization; standard graph-based SfM practice.

pith-pipeline@v1.1.0-grok45 · 24034 in / 3046 out tokens · 31633 ms · 2026-07-13T02:31:18.351371+00:00 · methodology

0 comments
read the original abstract

Global Structure-from-Motion (SfM) is an efficient paradigm for recovering camera poses and sparse 3D structure from unordered images. However, its reliance on scale-ambiguous epipolar geometry makes global positioning sensitive to noisy baseline estimates and weak view-graph constraints, while false edges from visually ambiguous pairs can further degrade reconstruction. We propose DGSfM, a depth-aware global SfM pipeline that uses monocular depth maps as a scalable prior while preserving explicit multi-view optimization. For each image pair, we use a depth-aware relative pose solver to convert scale-ambiguous epipolar constraints into scale-aware relative pose constraints. We further improve robustness through view-graph filtering and depth-consistency-based correspondence pruning, which suppress false edges and matches that remain plausible under epipolar geometry alone. Finally, global scale averaging and depth-guided pose-point initialization align monocular depth maps into a common reconstruction scale and provide stable initialization for global positioning and bundle adjustment. Experiments on ETH3D and IMC2021 show that DGSfM consistently improves over strong global SfM baselines across sparse and dense matching front-ends, achieving substantial gains in pose accuracy. Code is available at https://github.com/sithu31296/DGSfM.

Figures

Figures reproduced from arXiv: 2607.09507 by Sithu Aung, Torsten Sattler, Viktor Kocur, Yaqing Ding, Zuzana Kukelova.

Figure 1
Figure 1. Figure 1: Reconstruction results of DGSfM using RoMa [14] matches on the full image sets from the IMC2021 [20] dataset. DGSfM integrates monocular depth into global SfM to improve scale estimation, view-graph robustness, and pose-point initialization, producing globally consistent camera poses and sparse/semi-dense 3D structure. where visually plausible but incorrect image pairs can survive geometric veri￾fication. … view at source ↗
Figure 2
Figure 2. Figure 2: Overview of DGSfM. Given unordered images, we extract sparse or dense matches, monocular depth, and compute depth-aware relative poses and pairwise focals to construct an initial view graph, which is refined through pairwise and triplet-guided edge pruning, and scaled depth consistency checks. Mapping is done with rotation av￾eraging, focal and scale averaging, global positioning with depth-scaled initiali… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison on ETH3D [48]. Compared with GLOMAP [35], our method produces more stable camera pose estimates with LoMa [34] matches, while using dense RoMa v1 [14] matches leads to more complete 3D structure. MoGe2 [57] to extract the metric depth maps. Our rotation averaging module follows the implementation of InstantSfM [70], while GPU-accelerated sparse optimization is based on BAE [67]. We r… view at source ↗

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

71 extracted references · 13 linked inside Pith

  1. [1]

    Communications of the ACM54(10), 105–112 (2011)

    Agarwal, S., Furukawa, Y., Snavely, N., Simon, I., Curless, B., Seitz, S.M., Szeliski, R.: Building rome in a day. Communications of the ACM54(10), 105–112 (2011)

    2011

  2. [2]

    arXiv preprint arXiv:2311.18801 (2023)

    Baid, A., Lambert, J., Driver, T., Krishnan, A., Stepanyan, H., Dellaert, F.: Distributed global structure-from-motion with a deep front-end. arXiv preprint arXiv:2311.18801 (2023)

    Pith/arXiv arXiv 2023

  3. [3]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops

    Berton, G., Masone, C.: Megaloc: One retrieval to place them all. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. pp. 2886– 2892 (June 2025)

    2025

  4. [4]

    arXiv preprint arXiv:2307.14460 (2023)

    Birkl, R., Wofk, D., Müller, M.: Midas v3.1 – a model zoo for robust monocular relative depth estimation. arXiv preprint arXiv:2307.14460 (2023)

    Pith/arXiv arXiv 2023

  5. [5]

    In: International Conference on Learning Representations (2025),https://arxiv

    Bochkovskii, A., Delaunoy, A., Germain, H., Santos, M., Zhou, Y., Richter, S.R., Koltun, V.: Depth pro: Sharp monocular metric depth in less than a second. In: International Conference on Learning Representations (2025),https://arxiv. org/abs/2410.02073

    Pith/arXiv arXiv 2025

  6. [6]

    In: Proceedings of the IEEE international conference on computer vision

    Chatterjee, A., Govindu, V.M.: Efficient and robust large-scale rotation averaging. In: Proceedings of the IEEE international conference on computer vision. pp. 521– 528 (2013)

    2013

  7. [7]

    In: CVPR 2011

    Crandall, D., Owens, A., Snavely, N., Huttenlocher, D.: Discrete-continuous op- timization for large-scale structure from motion. In: CVPR 2011. pp. 3001–3008. IEEE (2011)

    2011

  8. [8]

    In:Proceed- ings of the IEEE International Conference on Computer Vision (ICCV) (December 2015)

    Cui,Z., Tan, P.:Global structure-from-motionbysimilarityaveraging. In:Proceed- ings of the IEEE International Conference on Computer Vision (ICCV) (December 2015)

    2015

  9. [9]

    arXiv preprint arXiv:2507.16443 (2025)

    Deng, K., Ti, Z., Xu, J., Yang, J., Xie, J.: Vggt-long: Chunk it, loop it, align it–pushing vggt’s limits on kilometer-scale long rgb sequences. arXiv preprint arXiv:2507.16443 (2025)

    Pith/arXiv arXiv 2025

  10. [10]

    In: Proceedings of the IEEE conference on com- puter vision and pattern recognition workshops

    DeTone, D., Malisiewicz, T., Rabinovich, A.: Superpoint: Self-supervised interest point detection and description. In: Proceedings of the IEEE conference on com- puter vision and pattern recognition workshops. pp. 224–236 (2018)

    2018

  11. [11]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Ding, Y., Kocur, V., Vávra, V., Haladová, Z.B., Yang, J., Sattler, T., Kukelova, Z.: Reposed: Efficient relative pose estimation with known depth information. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 14876–14886 (2025)

    2025

  12. [12]

    In: 2025 International Conference on 3D Vision (3DV)

    Duisterhof, B.P., Zust, L., Weinzaepfel, P., Leroy, V., Cabon, Y., Revaud, J.: Mast3r-sfm: a fully-integrated solution for unconstrained structure-from-motion. In: 2025 International Conference on 3D Vision (3DV). pp. 1–10. IEEE (2025)

    2025

  13. [13]

    arXiv preprint arXiv:2511.15706 (2025)

    Edstedt, J., Nordström, D., Zhang, Y., Bökman, G., Astermark, J., Larsson, V., Heyden, A., Kahl, F., Wadenbäck, M., Felsberg, M.: Roma v2: Harder better faster denser feature matching. arXiv preprint arXiv:2511.15706 (2025)

    Pith/arXiv arXiv 2025

  14. [14]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Edstedt, J., Sun, Q., Bökman, G., Wadenbäck, M., Felsberg, M.: Roma: Robust dense feature matching. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 19790–19800 (2024)

    2024

  15. [15]

    IEEE transactions on pattern analysis and machine intelligence 34(12), 2303–2314 (2012)

    Hartley, R., Li, H.: An efficient hidden variable approach to minimal-case camera motion estimation. IEEE transactions on pattern analysis and machine intelligence 34(12), 2303–2314 (2012)

    2012

  16. [16]

    Cambridge university press (2003) DGSfM 23

    Hartley, R., Zisserman, A.: Multiple view geometry in computer vision. Cambridge university press (2003) DGSfM 23

    2003

  17. [17]

    Computer vision and image understanding 68(2), 146–157 (1997)

    Hartley, R.I., Sturm, P.: Triangulation. Computer vision and image understanding 68(2), 146–157 (1997)

    1997

  18. [18]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    He, X., Sun, J., Wang, Y., Peng, S., Huang, Q., Bao, H., Zhou, X.: Detector-free structure from motion. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 21594–21603 (2024)

    2024

  19. [19]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    Hu, M., Yin, W., Zhang, C., Cai, Z., Long, X., Chen, H., Wang, K., Yu, G., Shen, C., Shen, S.: Metric3d v2: A versatile monocular geometric foundation model for zero-shot metric depth and surface normal estimation. IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    2024

  20. [20]

    International Journal of Computer Vision129(2), 517–547 (2021)

    Jin, Y., Mishkin, D., Mishchuk, A., Matas, J., Fua, P., Yi, K.M., Trulls, E.: Image matching across wide baselines: From paper to practice. International Journal of Computer Vision129(2), 517–547 (2021)

    2021

  21. [21]

    Keetha, N., Müller, N., Schönberger, J., Porzi, L., Zhang, Y., Fischer, T., Knapitsch, A., Zauss, D., Weber, E., Antunes, N., et al.: Mapanything: Universal feed-forward metric 3d reconstruction; map-anything. github. io. In: 2026 Interna- tional Conference on 3D Vision (3DV). pp. 499–509. IEEE (2026)

    2026

  22. [22]

    Proceedings of the American Mathematical society7(1), 48–50 (1956)

    Kruskal, J.B.: On the shortest spanning subtree of a graph and the traveling salesman problem. Proceedings of the American Mathematical society7(1), 48–50 (1956)

    1956

  23. [23]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Lee, J., Kang, S., Yoo, S.: Mv-roma: From pairwise matching into multi-view track reconstruction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 7446–7456 (June 2026)

    2026

  24. [24]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Lee, J., Yoo, S.: Dense-sfm: Structure from motion with dense consistent matching. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 6404–6414 (2025)

    2025

  25. [25]

    In: European conference on computer vision

    Leroy, V., Cabon, Y., Revaud, J.: Grounding image matching in 3d with mast3r. In: European conference on computer vision. pp. 71–91. Springer (2024)

    2024

  26. [26]

    arXiv preprint arXiv:2511.10647 (2025)

    Lin, H., Chen, S., Liew, J., Chen, D.Y., Li, Z., Shi, G., Feng, J., Kang, B.: Depth anything 3: Recovering the visual space from any views. arXiv preprint arXiv:2511.10647 (2025)

    Pith/arXiv arXiv 2025

  27. [27]

    In: Proceedings of the IEEE/CVF in- ternational conference on computer vision

    Lindenberger, P., Sarlin, P.E., Larsson, V., Pollefeys, M.: Pixel-perfect structure- from-motion with featuremetric refinement. In: Proceedings of the IEEE/CVF in- ternational conference on computer vision. pp. 5987–5997 (2021)

    2021

  28. [28]

    In: Proceedings of the IEEE/CVF international conference on com- puter vision

    Lindenberger, P., Sarlin, P.E., Pollefeys, M.: Lightglue: Local feature matching at light speed. In: Proceedings of the IEEE/CVF international conference on com- puter vision. pp. 17627–17638 (2023)

    2023

  29. [29]

    Interna- tional journal of computer vision60(2), 91–110 (2004)

    Lowe, D.G.: Distinctive image features from scale-invariant keypoints. Interna- tional journal of computer vision60(2), 91–110 (2004)

    2004

  30. [30]

    Advances in Neural Information Processing Systems38, 129839–129867 (2026)

    Maggio, D., Lim, H., Carlone, L.: Vggt-slam: Dense rgb slam optimized on the sl (4) manifold. Advances in Neural Information Processing Systems38, 129839–129867 (2026)

    2026

  31. [31]

    Manam, L., Govindu, V.M.: Leveraging camera triplets for efficient and accurate structure-from-motion.In:ProceedingsoftheIEEE/CVFConferenceonComputer Vision and Pattern Recognition. pp. 4959–4968 (2024)

    2024

  32. [32]

    Meza, J., Oswald, M.R., Sattler, T.: Benchmarking efficient & effective camera poseestimationstrategiesfornovelviewsynthesis.arXivpreprintarXiv:2603.20428 (2026)

    arXiv 2026

  33. [33]

    In: International Workshop on Reproducible Research in Pattern Recog- nition

    Moulon, P., Monasse, P., Perrot, R., Marlet, R.: Openmvg: Open multiple view geometry. In: International Workshop on Reproducible Research in Pattern Recog- nition. pp. 60–74. Springer (2016) 24 S. Aung et al

    2016

  34. [34]

    arXiv preprint arXiv:2604.04931 (2026)

    Nordström, D., Edstedt, J., Bökman, G., Astermark, J., Heyden, A., Larsson, V., Wadenbäck, M., Felsberg, M., Kahl, F.: Loma: Local feature matching revisited. arXiv preprint arXiv:2604.04931 (2026)

    Pith/arXiv arXiv 2026

  35. [35]

    In: European Conference on Computer Vision

    Pan, L., Baráth, D., Pollefeys, M., Schönberger, J.L.: Global structure-from-motion revisited. In: European Conference on Computer Vision. pp. 58–77. Springer (2024)

    2024

  36. [36]

    In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition

    Pan, L., Schönberger, J., Pollefeys, M.: Global structure-from-motion meets feed- forward reconstruction. In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition. pp. 21880–21890 (2026)

    2026

  37. [37]

    In: Proceedings of the Computer Vi- sion and Pattern Recognition Conference

    Pataki, Z., Sarlin, P.E., Schönberger, J.L., Pollefeys, M.: Mp-sfm: Monocular sur- face priors for robust structure-from-motion. In: Proceedings of the Computer Vi- sion and Pattern Recognition Conference. pp. 21891–21901 (2025)

    2025

  38. [38]

    In: IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) (2025)

    Piccinelli, L., Sakaridis, C., Segu, M., Yang, Y.H., Li, S., Abbeloos, W., Van Gool, L.: UniK3D: Universal camera monocular 3d estimation. In: IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) (2025)

    2025

  39. [39]

    Piccinelli, L., Sakaridis, C., Yang, Y.H., Segu, M., Li, S., Abbeloos, W., Gool, L.V.: UniDepthV2: Universal monocular metric depth estimation made simpler (2025), https://arxiv.org/abs/2502.20110

    Pith/arXiv arXiv 2025

  40. [40]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Piccinelli, L., Yang, Y.H., Sakaridis, C., Segu, M., Li, S., Van Gool, L., Yu, F.: UniDepth: Universal monocular metric depth estimation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  41. [41]

    ICCV (2021)

    Ranftl, R., Bochkovskiy, A., Koltun, V.: Vision transformers for dense prediction. ICCV (2021)

    2021

  42. [42]

    IEEE Transactions on Pattern Analysis and Machine Intelligence44(3) (2022)

    Ranftl, R., Lasinger, K., Hafner, D., Schindler, K., Koltun, V.: Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer. IEEE Transactions on Pattern Analysis and Machine Intelligence44(3) (2022)

    2022

  43. [43]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Sarlin, P.E., DeTone, D., Malisiewicz, T., Rabinovich, A.: Superglue: Learning feature matching with graph neural networks. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 4938–4947 (2020)

    2020

  44. [44]

    In: ECCV (2022)

    Sarlin, P.E., Dusmanu, M., Schönberger, J.L., Speciale, P., Gruber, L., Larsson, V., Miksik, O., Pollefeys, M.: LaMAR: Benchmarking Localization and Mapping for Augmented Reality. In: ECCV (2022)

    2022

  45. [45]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Schonberger, J.L., Frahm, J.M.: Structure-from-motion revisited. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 4104–4113 (2016)

    2016

  46. [46]

    In: European conference on computer vision

    Schonberger, J.L., Zheng, E., Frahm, J.M., Pollefeys, M.: Pixelwise view selection for unstructured multi-view stereo. In: European conference on computer vision. pp. 501–518. Springer (2016)

    2016

  47. [47]

    In: European Conference on Computer Vision (ECCV) (2016)

    Schönberger, J.L., Zheng, E., Pollefeys, M., Frahm, J.M.: Pixelwise view selection for unstructured multi-view stereo. In: European Conference on Computer Vision (ECCV) (2016)

    2016

  48. [48]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Schops, T., Schonberger, J.L., Galliani, S., Sattler, T., Schindler, K., Pollefeys, M., Geiger, A.: A multi-view stereo benchmark with high-resolution images and multi-camera videos. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 3260–3269 (2017)

    2017

  49. [49]

    In: ACM siggraph 2006 papers, pp

    Snavely, N., Seitz, S.M., Szeliski, R.: Photo tourism: exploring photo collections in 3d. In: ACM siggraph 2006 papers, pp. 835–846. Association for Computing Machinery (2006)

    2006

  50. [50]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Sun, J., Shen, Z., Wang, Y., Bao, H., Zhou, X.: Loftr: Detector-free local fea- ture matching with transformers. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 8922–8931 (2021) DGSfM 25

    2021

  51. [51]

    In: International workshop on vision algorithms

    Triggs, B., McLauchlan, P.F., Hartley, R.I., Fitzgibbon, A.W.: Bundle adjust- ment—a modern synthesis. In: International workshop on vision algorithms. pp. 298–372. Springer (1999)

    1999

  52. [52]

    In: 2025 Interna- tional Conference on 3D Vision (3DV)

    Wang, H., Agapito, L.: 3d reconstruction with spatial memory. In: 2025 Interna- tional Conference on 3D Vision (3DV). pp. 78–89. IEEE (2025)

    2025

  53. [53]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Wang, J., Chen, M., Karaev, N., Vedaldi, A., Rupprecht, C., Novotny, D.: Vggt: Visual geometry grounded transformer. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 5294–5306 (2025)

    2025

  54. [54]

    arXiv preprint arXiv:2605.15195 (2026)

    Wang, J., Chen, M., Zhang, S., Karaev, N., Schönberger, J., Labatut, P., Bo- janowski, P., Novotny, D., Vedaldi, A., Rupprecht, C.: Vggt-omage. arXiv preprint arXiv:2605.15195 (2026)

    Pith/arXiv arXiv 2026

  55. [55]

    In: Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition

    Wang, J., Karaev, N., Rupprecht, C., Novotny, D.: Vggsfm: Visual geometry grounded deep structure from motion. In: Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition. pp. 21686–21697 (2024)

    2024

  56. [56]

    In: Proceedings of the Computer Vision and Pattern Recog- nition Conference

    Wang, R., Xu, S., Dai, C., Xiang, J., Deng, Y., Tong, X., Yang, J.: Moge: Unlock- ing accurate monocular geometry estimation for open-domain images with optimal training supervision. In: Proceedings of the Computer Vision and Pattern Recog- nition Conference. pp. 5261–5271 (2025)

    2025

  57. [57]

    Wang,R., Xu,S., Dong, Y.,Deng, Y.,Xiang, J.,Lv, Z.,Sun, G.,Tong,X., Yang, J.: Moge-2: Accurate monocular geometry with metric scale and sharp details (2025), https://arxiv.org/abs/2507.02546

    Pith/arXiv arXiv 2025

  58. [58]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Wang, S., Leroy, V., Cabon, Y., Chidlovskii, B., Revaud, J.: Dust3r: Geometric 3d vision made easy. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 20697–20709 (2024)

    2024

  59. [59]

    arXiv preprint arXiv:2507.13347 (2025)

    Wang, Y., Zhou, J., Zhu, H., Chang, W., Zhou, Y., Li, Z., Chen, J., Pang, J., Shen, C., He, T.: pi3: Permutation-equivariant visual geometry learning. arXiv preprint arXiv:2507.13347 (2025)

    Pith/arXiv arXiv 2025

  60. [60]

    In: European con- ference on computer vision

    Wilson, K., Snavely, N.: Robust global translations with 1dsfm. In: European con- ference on computer vision. pp. 61–75. Springer (2014)

    2014

  61. [61]

    In: 2013 Interna- tional Conference on 3D Vision-3DV 2013

    Wu, C.: Towards linear-time incremental structure from motion. In: 2013 Interna- tional Conference on 3D Vision-3DV 2013. pp. 127–134. IEEE (2013)

    2013

  62. [62]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Xiangli, Y., Cai, R., Chen, H., Byrne, J., Snavely, N.: Doppelgangers++: Improved visual disambiguation with geometric 3d features. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 27166–27175 (2025)

    2025

  63. [63]

    In: CVPR (2024)

    Yang, L., Kang, B., Huang, Z., Xu, X., Feng, J., Zhao, H.: Depth anything: Un- leashing the power of large-scale unlabeled data. In: CVPR (2024)

    2024

  64. [64]

    arXiv:2406.09414 (2024)

    Yang, L., Kang, B., Huang, Z., Zhao, Z., Xu, X., Feng, J., Zhao, H.: Depth anything v2. arXiv:2406.09414 (2024)

    Pith/arXiv arXiv 2024

  65. [65]

    In: Proceed- ings of the IEEE/CVF international conference on computer vision

    Yin, W., Zhang, C., Chen, H., Cai, Z., Yu, G., Wang, K., Chen, X., Shen, C.: Metric3d: Towards zero-shot metric 3d prediction from a single image. In: Proceed- ings of the IEEE/CVF international conference on computer vision. pp. 9043–9053 (2023)

    2023

  66. [66]

    In: Proceedings of the Com- puter Vision and Pattern Recognition Conference

    Yu, Y., Liu, S., Pautrat, R., Pollefeys, M., Larsson, V.: Relative pose estimation through affine corrections of monocular depth priors. In: Proceedings of the Com- puter Vision and Pattern Recognition Conference. pp. 16706–16716 (2025)

    2025

  67. [67]

    IEEE Transactions on Robotics (2026),https://arxiv.org/abs/ 2409.12190

    Zhan, Z., Xu, H., Fang, Z., Wei, X., Hu, Y., Wang, C.: Bundle adjustment in the eager mode. IEEE Transactions on Robotics (2026),https://arxiv.org/abs/ 2409.12190

    Pith/arXiv arXiv 2026

  68. [68]

    arXiv preprint arXiv:2506.09278 (2025) 26 S

    Zhang, Y., Keetha, N., Lyu, C., Jhamb, B., Chen, Y., Qiu, Y., Karhade, J., Jha, S., Hu, Y., Ramanan, D., et al.: Ufm: A simple path towards unified dense corre- spondence with flow. arXiv preprint arXiv:2506.09278 (2025) 26 S. Aung et al

    arXiv 2025

  69. [69]

    IEEE Transac- tions on Instrumentation and Measurement72, 1–16 (2023)

    Zhao, X., Wu, X., Chen, W., Chen, P.C., Xu, Q., Li, Z.: Aliked: A lighter keypoint and descriptor extraction network via deformable transformation. IEEE Transac- tions on Instrumentation and Measurement72, 1–16 (2023)

    2023

  70. [70]

    arXiv preprint (2025)

    Zhong, J., Zhan, Z., Gao, Q., Chen, Z., Lou, H., Mao, J., Neumann, U., Wang, Y.: Instantsfm: Fully sparse and parallel structure-from-motion. arXiv preprint (2025)

    2025

  71. [71]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    Zhuang, B., Cheong, L.F., Lee, G.H.: Baseline desensitizing in translation aver- aging. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 4539–4547 (2018)

    2018