Pith. sign in

REVIEW 5 minor 52 references

Wat3R: Underwater 3D Geometry Learning without Annotations

T0 review · 0 major / 5 minor · reviewed 2026-07-10 · glm-5.2

Pith's one-line read Underwater 3D Reconstruction Without Any Underwater Annotations

desk verdict Semi-supervised adaptation of feed-forward 3D reconstruction to underwater scenes without annotations — works well, but the semi-supervised branch is fragile without careful constraints. read the letter →

arxiv 2607.08772 v1 pith:4V2BTG6G submitted 2026-07-09 cs.CV

classification cs.CV
keywords underwaterwat3rannotationsgeometryattenuationcurrentdatasetdesigned
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 claims that a feed-forward 3D geometry model trained on land can be adapted to underwater scenes without a single annotated underwater depth map, camera pose, or point cloud. The mechanism is a teacher-student semi-supervised loop: synthetic underwater degradation is applied to existing on-land 3D datasets to seed geometric priors, then a teacher network generates pseudo-labels on real unlabeled underwater video while a student network learns from both. A cross-view consistency loss reprojects depth across views and supervises only on pixels that are geometrically stable across multiple viewpoints, filtering out dynamic objects and regions where scattering has destroyed reliable structure. The authors also construct Water3D, a 42-scene underwater benchmark with pose and depth annotations, and show that their adapted model outperforms current feed-forward baselines on multi-view depth, point cloud reconstruction, camera pose, and monocular depth tasks across five public datasets plus their own.

What carries the argument

Mean Teacher semi-supervised framework with EMA-updated teacher producing pseudo-labels on unlabeled underwater video; physics-based underwater image formation model (direct transmission attenuation, backscatter, background light) applied to on-land 3D-annotated datasets to create labeled synthetic underwater training data; cross-view geometry consistency loss that backprojects teacher depth from one view, reprojects into other views, and supervises student depth only on pixels passing a multi-view static mask (depth-consistent foreground across N-2 views); sequence-level augmentation with frame shuffling and rotation to prevent pose collapse; VGGT as base feed-forward geometry backbone.

What would settle it

If the teacher network produces systematically biased geometry on real underwater video—particularly in green-water or high-turbidity conditions where the pretrained model has no prior—the student would reinforce those errors. The cross-view static mask filters per-pixel inconsistency but cannot detect systematic biases that are consistent across all views of a degraded scene. The paper acknowledges this limitation for open-water and very deep scenes, but the boundary between where the method works and where it fails is not sharply characterized.

Watch

Extended reading notes

Core claim

The central claim is that the domain gap between air-trained 3D geometry models and underwater scenes can be bridged entirely through synthetic degradation plus semi-supervised consistency training on unlabeled video, without any underwater 3D ground truth. The cross-view consistency loss is the key technical device: it compensates for information lost to attenuation and scattering in one view by pulling geometric evidence from other views of the same scene, applying supervision only where multiple views agree on depth. This mechanism, combined with a physics-based underwater image formation model for synthetic data generation, allows a large pretrained geometry transformer to generalize to

Load-bearing premise

The teacher network's depth and pose predictions on real underwater video are accurate enough to serve as training targets for the student. The entire semi-supervised branch depends on this; if the teacher is systematically wrong on turbid or low-visibility regions, the student learns to reproduce those errors. The cross-view consistency mask is designed to filter unreliable pixels, but it assumes enough views have overlapping static structure, which the authors acknowledge稀疏

Editorial extensions

If this is right

  • If the approach generalizes, any domain with abundant unlabeled video but no 3D annotations—subterranean, foggy outdoor, or endoscopic scenes—could be adapted similarly by combining synthetic degradation with cross-view consistency training.
  • The finding that underwater image enhancement as a preprocessing step does not improve 3D geometry estimation suggests the field should shift from restoration-then-reconstruction pipelines toward joint geometry-aware adaptation.
  • The Water3D benchmark with 42 scenes and both depth and pose annotations fills a gap that could recalibrate how underwater 3D methods are evaluated, since prior datasets lacked comprehensive geometric annotations.
  • The cross-view static masking strategy, which naturally sparsifies when scenes are dynamic or turbid, provides a principled mechanism for self-limiting supervision that could apply to any semi-supervised multi-view learning setting with unreliable regions.

Reading between the lines

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

  • The reliance on a static mask that becomes sparse in open-water or highly turbid scenes suggests a natural performance ceiling: the method may work well in structured underwater environments (reefs, shipwrecks, seabed with texture) but degrade precisely where underwater robotics needs it most—featureless open water with poor visibility.
  • The fact that the same framework improves both multi-view and monocular depth estimation suggests the cross-view training is teaching the model something general about separating underwater degradation from scene geometry, not just memorizing multi-view correspondences. This raises the question of whether the learned representations could transfer to other degraded-media tasks like imaging through
  • If the teacher's pseudo-labels are systematically biased in certain underwater conditions—for example, consistently overestimating depth in green-water scenes—the student would reinforce that bias. The cross-view mask filters individual pixel errors but may not catch systematic biases that are consistent across views.
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

0 major / 5 minor

Summary. The manuscript proposes Wat3R, a cross-domain semi-supervised framework for adapting feed-forward 3D reconstruction models (specifically VGGT) to underwater environments without requiring any underwater 3D annotations. The method follows a Mean Teacher paradigm: synthetic underwater degradation is applied to labeled on-land data to initialize geometric priors, while unlabeled real underwater video is used for semi-supervised refinement. A cross-view consistency loss with a static mask is introduced to mitigate the impact of degraded views by aggregating geometric cues across multiple views. The authors also construct Water3D, a new multi-view underwater dataset with 42 scenes featuring camera pose and depth annotations. The method is evaluated on multi-view depth estimation, point cloud estimation, camera pose estimation, and monocular depth estimation, demonstrating consistent improvements over VGGT and other recent feed-forward baselines.

Significance. The paper addresses a practical and significant problem: the domain gap between terrestrial and underwater 3D vision. The use of unlabeled video to bypass the annotation bottleneck is well-motivated. The authors provide public code and a new benchmark dataset (Water3D), which adds considerable value to the community. The experimental evaluation is comprehensive, spanning multiple public datasets and tasks. The ablation studies transparently show the contribution of each component, including the finding that the semi-supervised branch requires careful constraints to be net-positive. The cross-view consistency loss is a reasonable mechanism to handle view-dependent degradation.

minor comments (5)
  1. §3.3, Eq. (8): The static mask threshold is set to k=N-2. The text states this is a 'conservative setting.' It would help to briefly justify this specific choice or provide a sensitivity analysis, as the mask is load-bearing for the cross-view loss.
  2. Table 1: The shaded rows indicating two-stage pipelines (UIE + VGGT) are mentioned in the caption, but the visual distinction (shading) is not easily discernible in standard text review. Ensure the formatting is clear in the final version.
  3. §4.1: The ratio of unlabeled to labeled samples is 1:3, and the unsupervised loss weight ramps to 0.5. It would be beneficial to include a brief discussion or ablation on how sensitive the final performance is to this specific ratio and peak weight.
  4. Appendix D.2, Table A3: The comparison with COLMAP is useful. However, the 'Failure rate' for COLMAP+MINIMA is 99/152, which is very high. It would be helpful to clarify whether the reported metrics for COLMAP are computed only on the successfully reconstructed scenes, which would make the direct comparison with Wat3R (0 failure rate) less straightforward.
  5. Fig. 1: The y-axis labels of the bar charts in the top right are rotated vertically and are somewhat difficult to read. Consider adjusting the layout for clarity.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the central claim is validated against external benchmarks with independent ground truth, and the one self-citation (MINIMA) is used only for dataset construction, not as a load-bearing derivation step.

full rationale

The paper's central claim—that Wat3R enables feed-forward underwater 3D geometry estimation without underwater annotations—is tested against external benchmarks (Sea-thru, FLSea, SQUID, SeaThru-NeRF) with independent ground truth. The training pipeline uses VGGT (external, Ref. 40) as the base model and DA3 (external, Ref. 22) for monocular depth in synthetic rendering. The authors' own MINIMA matcher (Ref. 30, self-citation by Ren et al.) is used only for Water3D dataset construction (Appendix B), not as a load-bearing step in the method's derivation chain. The cross-view consistency loss (Eq. 10) and static mask (Eq. 8) are defined in terms of teacher predictions and geometric reprojection operators, not in terms of the evaluation targets. The semi-supervised teacher-student framework (Eq. 2-6) follows the standard Mean Teacher paradigm (Ref. 35, external). No equation in the method reduces to its own inputs by construction, and no 'prediction' is a fitted parameter renamed. The ablation (Table 5) shows the semi-supervised branch is fragile without augmentation and cross-view loss, but this is a correctness/robustness concern, not circularity. The one minor self-citation (MINIMA) is not load-bearing for the central claim and is code-reproducible, so it does not raise the score beyond 1. The derivation is self-contained against external benchmarks. No circularity found.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities
free parameters (6)
  • λ (EMA smoothing coefficient) = not explicitly stated
    Controls teacher update rate in Eq. 2. Standard Mean Teacher parameter, value not specified in main text.
  • λ_p (per-view loss weight) = not explicitly stated
    Weight for per-view consistency loss in Eq. 4. Peak unsupervised loss weight is 0.5 at step 12,800, but the split between per-view and cross-view is unclear.
  • λ_c (cross-view loss weight) = not explicitly stated
    Weight for cross-view consistency loss in Eq. 4. Combined with λ_p under the unsupervised loss weight of 0.5.
  • δ (depth consistency threshold) = not explicitly stated
    Threshold in Eq. 7 for determining visibility/depth-consistency of reprojected pixels. Controls which pixels are considered static.
  • k (static mask threshold) = N-2
    Conservative threshold in Eq. 8 requiring pixel to be consistent across N-2 of N teacher views. Chosen empirically.
  • β_D, β_B, B_∞ (underwater rendering parameters) = sampled from [0,1]
    Attenuation, backscatter, and background light parameters in Eq. 3. Randomly sampled with physical constraint that red attenuates faster than green/blue.
assumptions (4)
  • domain assumption VGGT provides strong geometric priors that can be transferred to underwater domains via fine-tuning.
    §3.1: 'we chose VGGT as the base model such that it can provide strong geometric priors when adapting to underwater environments.' This is the foundation of the entire approach.
  • domain assumption The revised underwater image formation model (Eq. 3) adequately simulates real underwater degradation for training purposes.
    §3.2: Synthetic underwater data is generated using Eq. 3 with randomly sampled parameters. The quality of this simulation directly affects how well the model initializes underwater geometric priors.
  • domain assumption Teacher pseudo-labels on real underwater video are sufficiently accurate for semi-supervised learning.
    §3.2-3.3: The per-view consistency loss (Eq. 6) treats teacher predictions as pseudo ground truth. If teacher predictions are systematically wrong on degraded regions, the student learns incorrect geometry.
  • domain assumption COLMAP-derived annotations on successfully reconstructed underwater scenes are reliable ground truth for evaluation.
    §4.1 and Appendix B: Water3D ground truth is constructed using COLMAP + MINIMA. Only 42 of ~100 scenes survived manual inspection. The paper acknowledges COLMAP failures (Fig. A3) but assumes surviving reconstructions are accurate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Wat3R: Underwater 3D Geometry Learning without Annotations." pith.science (2026). https://pith.science/paper/4V2BTG6G

@misc{pith2026260708772,
  author       = {Pith},
  title        = {Pith review of: Wat3R: Underwater 3D Geometry Learning without Annotations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4V2BTG6G}},
  note         = {Machine review of arXiv:2607.08772}
}
read the original abstract

Estimating 3D geometry in underwater environments presents unique challenges due to light attenuation, scattering, and the absence of large-scale, high-quality 3D annotations. Pioneering methods rely on massive dense annotations that are impractical in underwater settings. In this paper, we propose Wat3R, a cross-domain semi-supervised learning framework designed to adapt feed-forward 3D reconstruction models from air to underwater scenes. Uniquely, our method eliminates the need for any annotated underwater data following a teacher-student architecture, that learns robust geometry representations merely on abundant unlabeled real underwater video footage. We also design a cross-view consistency loss that leverages geometric cues from other views to compensate for the information degradation in the current view caused by water attenuation and scattering. Furthermore, considering the lack of comprehensive evaluation benchmarks, we construct Water3D, a diverse dataset covering various water bodies and underwater scenarios, designed for geometric task evaluation. Experimental results demonstrate that Wat3R outperforms current state-of-the-art methods in underwater multi-view depth estimation and point cloud reconstruction. The dataset and code are available at https://github.com/LSXI7/Wat3R .

Figures

Figures reproduced from arXiv: 2607.08772 by the authors.

Figure 1
Figure 1. Wat3R reconstructs from the open-domain underwater images in a feed￾forward manner without requiring any underwater 3D annotations. Our Wat3R achieves significant enhancement in both single-view and multi-view tasks. Statistic results also reveal the superior performance of our Wat3R against the SOTA. In recent years, 3D vision has witnessed a paradigm shift from classical multi￾view geometry pipelines to feed-forwa… view at source ↗
Figure 2
Figure 2. Overview of our constructed Water3D dataset. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Overview of our Wat3R framework. Our pipeline follows a Mean Teacher semi-supervised paradigm, where the teacher network produces pseudo-labels for depth, camera parameters, and point maps to supervise the student network. Training lever￾ages labeled synthetic underwater data together with unlabeled real underwater videos, enabling adaptation without underwater 3D annotations. Additional per-view and cross-view cons… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: 3D reconstruction results on our constructed Water3D [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison on in-the-wild images [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Monocular depth estimation using Wat3R, DA3 [ [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Robustness and failure case analysis [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 52 canonical work pages

  1. [1]

    CACM54(10), 105–112 (2011)

    Agarwal, S., Furukawa, Y., Snavely, N., Simon, I., Curless, B., Seitz, S.M., Szeliski, R.: Building rome in a day. CACM54(10), 105–112 (2011)

  2. [2]

    In: CVPR

    Akkaynak, D., Treibitz, T.: A revised underwater image formation model. In: CVPR. pp. 6723–6732 (2018)

  3. [3]

    In: CVPR

    Akkaynak, D., Treibitz, T.: Sea-thru: A method for removing water from under- water images. In: CVPR. pp. 1682–1691 (2019)

  4. [4]

    IEEE TPAMI (2020)

    Berman, D., Levy, D., Avidan, S., Treibitz, T.: Underwater single image color restoration using haze-lines and a new quantitative dataset. IEEE TPAMI (2020)

  5. [5]

    In: BMVC

    Bleyer, M., Rhemann, C., Rother, C.: Patchmatch stereo-stereo matching with slanted support windows. In: BMVC. vol. 11, pp. 1–11 (2011)

  6. [6]

    In: CVPR

    Cabon, Y., Stoffl, L., Antsfeld, L., Csurka, G., Chidlovskii, B., Revaud, J., Leroy, V.: Must3r: Multi-view network for stereo 3d reconstruction. In: CVPR. pp. 1050– 1060 (2025)

  7. [7]

    In: ICCV

    Chen, D., Liu, Z., Yang, C., Wang, D., Yan, Y., Xu, Y., Ji, X.: Conformalsam: Un- locking the potential of foundational segmentation models in semi-supervised se- mantic segmentation with conformal prediction. In: ICCV. pp. 24045–24055 (2025)

  8. [8]

    In: CVPR

    Chen, S., Guo, H., Zhu, S., Zhang, F., Huang, Z., Feng, J., Kang, B.: Video depth anything: Consistent depth estimation for super-long videos. In: CVPR. pp. 22831– 22840 (2025)

Show all 52 references
  1. [9]

    In: CVPR

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

  2. [10]

    IEEE TPAMI35(12), 2841–2853 (2012)

    Crandall, D.J., Owens, A., Snavely, N., Huttenlocher, D.P.: Sfm with mrfs: Discrete-continuous optimization for large-scale structure from motion. IEEE TPAMI35(12), 2841–2853 (2012)

  3. [11]

    In: CVPR

    Cui, H., Gao, X., Shen, S., Hu, Z.: Hsfm: Hybrid structure-from-motion. In: CVPR. pp. 1212–1221 (2017)

  4. [12]

    IEEE TIM (2025)

    Ding, Y., Li, K., Mei, H., Liu, S., Hou, G.: Watermono: Teacher-guided anomaly masking and enhancement boosting for robust underwater self-supervised monoc- ular depth estimation. IEEE TIM (2025)

  5. [13]

    In: ICRA

    Ebner, L., Billings, G., Williams, S.: Metrically scaled monocular depth estimation through sparse priors for underwater robots. In: ICRA. pp. 3751–3757 (2024)

  6. [14]

    Cambridge university press (2003)

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

  7. [15]

    In: CVPR

    Huang,S.,Wang,K.,Liu,H.,Chen,J.,Li,Y.:Contrastivesemi-supervisedlearning for underwater image restoration via reliable bank. In: CVPR. pp. 18145–18155 (2023)

  8. [16]

    In: CVPR

    Jensen, R., Dahl, A., Vogiatzis, G., Tola, E., Aanæs, H.: Large scale multi-view stereopsis evaluation. In: CVPR. pp. 406–413 (2014)

  9. [17]

    In: 3DV (2026)

    Keetha, N., Müller, N., Schönberger, J., Porzi, L., Zhang, Y., Fischer, T., Knapitsch, A., Zauss, D., Weber, E., Antunes, N., et al.: Mapanything: Univer- sal feed-forward metric 3d reconstruction. In: 3DV (2026)

  10. [18]

    In: IROS

    Kim, A., Eustice, R.: Pose-graph visual slam with geometric model selection for autonomous underwater ship hull inspection. In: IROS. pp. 1559–1565 (2009)

  11. [19]

    In: ECCV

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

  12. [20]

    In: CVPR

    Levy, D., Peleg, A., Pearl, N., Rosenbaum, D., Akkaynak, D., Korman, S., Treibitz, T.: Seathru-nerf: Neural radiance fields in scattering media. In: CVPR. pp. 56–65 (2023) Wat3R 17

  13. [21]

    Li, H., Song, W., Xu, T., Elsig, A., Kulhanek, J.: Watersplatting: Fast underwater 3d scene reconstruction using gaussian splatting. In: 3DV. pp. 969–978 (2025)

  14. [22]

    In: ICLR (2026)

    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. In: ICLR (2026)

  15. [23]

    In: CVPR

    Lin, H., Liang, D., Qi, Z., Bai, X.: A unified image-dense annotation generation model for underwater scenes. In: CVPR. pp. 961–970 (2025)

  16. [24]

    In: ICCV

    Lindenberger, P., Sarlin, P.E., Pollefeys, M.: Lightglue: Local feature matching at light speed. In: ICCV. pp. 17627–17638 (2023)

  17. [25]

    In: ICCV

    Liu, R., Fan, S., Wang, W., Yang, Y.: Underwater visual slam with depth uncer- tainty and medium modeling. In: ICCV. pp. 970–980 (2025)

  18. [26]

    IEEE TIP (2025)

    Liu, Y., Jiang, Q., Li, X., Luo, T., Ren, W.: Toward better than pseudo-reference in underwater image enhancement. IEEE TIP (2025)

  19. [27]

    IEEE TCSVT (2025)

    Lv, Q., Dong, J., Li, Y., Chen, S., Yu, H., Zhang, S., Wang, W.: Uwstereo: A large synthetic dataset for underwater stereo matching. IEEE TCSVT (2025)

  20. [28]

    In: NeurIPS (2025)

    Maggio, D., Lim, H., Carlone, L.: VGGT-SLAM: Dense rgb slam optimized on the sl (4) manifold. In: NeurIPS (2025)

  21. [29]

    Master’s thesis, University of Haifa (Israel) (2023)

    Randall, Y.: Flsea: Underwater visual-inertial and stereo-vision forward-looking datasets. Master’s thesis, University of Haifa (Israel) (2023)

  22. [30]

    In: CVPR

    Ren, J., Jiang, X., Li, Z., Liang, D., Zhou, X., Bai, X.: Minima: Modality invariant image matching. In: CVPR. pp. 23059–23068 (2025)

  23. [31]

    In: CVPR

    Sarlin, P.E., Cadena, C., Siegwart, R., Dymczyk, M.: From coarse to fine: Robust hierarchical localization at large scale. In: CVPR. pp. 12716–12725 (2019)

  24. [32]

    In: CVPR

    Schonberger, J.L., Frahm, J.M.: Structure-from-motion revisited. In: CVPR. pp. 4104–4113 (2016)

  25. [33]

    ACM TOG pp

    Snavely, N., Seitz, S.M., Szeliski, R.: Photo tourism: exploring photo collections in 3d. ACM TOG pp. 835–846 (2006)

  26. [34]

    In: CVPR

    Tang, Y., Zhu, C., Wan, R., Xu, C., Shi, B.: Neural underwater scene representa- tion. In: CVPR. pp. 11780–11789 (2024)

  27. [35]

    In: NeurIPS

    Tarvainen, A., Valpola, H.: Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results. In: NeurIPS. vol. 30 (2017)

  28. [36]

    IEEE TPAMI (1991)

    Umeyama, S.: Least-squares estimation of transformation parameters between two point patterns. IEEE TPAMI (1991)

  29. [37]

    IEEE TGRS62, 1–16 (2024)

    Wang, C., Xu, H., Jiang, G., Yu, M., Luo, T., Chen, Y.: Underwater monocular depth estimation based on physical-guided transformer. IEEE TGRS62, 1–16 (2024)

  30. [38]

    In: WACV

    Wang, H., Anantrasirichai, N., Zhang, F., Bull, D.: Uw-gs: Distractor-aware 3d gaussian splatting for enhanced underwater scene reconstruction. In: WACV. pp. 3280–3289 (2025)

  31. [39]

    Wang, H., Agapito, L.: 3d reconstruction with spatial memory. In: 3DV. pp. 78–89 (2025)

  32. [40]

    In: CVPR

    Wang, J., Chen, M., Karaev, N., Vedaldi, A., Rupprecht, C., Novotny, D.: Vggt: Visual geometry grounded transformer. In: CVPR. pp. 5294–5306 (2025)

  33. [41]

    In: CVPR

    Wang, S., Leroy, V., Cabon, Y., Chidlovskii, B., Revaud, J.: Dust3r: Geometric 3d vision made easy. In: CVPR. pp. 20697–20709 (2024)

  34. [42]

    In: ICLR (2026)

    Wang, Y., Zhou, J., Zhu, H., Chang, W., Zhou, Y., Li, Z., Chen, J., Pang, J., Shen, C., He, T.:π3: Permutation-equivariant visual geometry learning. In: ICLR (2026)

  35. [43]

    In: ICRA (2026) 18 J

    Wu, Z., Wang, Y., Wen, Y., Zhang, Z., Wu, B., Tang, H.: Stereoadapter: Adapting stereo depth estimation to underwater scenes. In: ICRA (2026) 18 J. Ren et al

  36. [44]

    In: CVPR

    Xie, Y., Kong, L., Chen, K., Zheng, Z., Yu, X., Yu, Z., Zheng, B.: Uveb: A large- scale benchmark and baseline towards real-world underwater video enhancement. In: CVPR. pp. 22358–22367 (2024)

  37. [45]

    In: NeurIPS (2025)

    Xu, W., Wang, C., Liang, D., Zhao, Z., Jiang, X., Zhang, P., Bai, X.: Nautilus: A large multimodal model for underwater scene understanding. In: NeurIPS (2025)

  38. [46]

    In: ICRA

    Yang, D., Leonard, J.J., Girdhar, Y.: Seasplat: Representing underwater scenes with 3d gaussian splatting and a physically grounded image formation model. In: ICRA. pp. 7632–7638 (2025)

  39. [47]

    In: CVPR

    Yang, J., Sax, A., Liang, K.J., Henaff, M., Tang, H., Cao, A., Chai, J., Meier, F., Feiszli, M.: Fast3r: Towards 3d reconstruction of 1000+ images in one forward pass. In: CVPR. pp. 21924–21935 (2025)

  40. [48]

    In: NeurIPS

    Yang, L., Kang, B., Huang, Z., Zhao, Z., Xu, X., Feng, J., Zhao, H.: Depth anything v2. In: NeurIPS. vol. 37, pp. 21875–21911 (2024)

  41. [49]

    In: ICRA (2023)

    Yu, B., Wu, J., Islam, M.J.: Udepth: Fast monocular depth estimation for visually- guided underwater robots. In: ICRA (2023)

  42. [50]

    In: CVPR

    Zhang, F., You, S., Li, Y., Fu, Y.: Atlantis: Enabling underwater depth estimation with stable diffusion. In: CVPR. pp. 11852–11861 (2024)

  43. [51]

    In: CVPR

    Zhao, Z., Yang, L., Long, S., Pi, J., Zhou, L., Wang, J.: Augmentation mat- ters: A simple-yet-effective approach to semi-supervised semantic segmentation. In: CVPR. pp. 11350–11359 (2023)

  44. [52]

    Wat3R: Underwater 3D Geometry Learning without Annotations

    Zhou, J., Liang, T., Zhang, D., Liu, S., Wang, J., Wu, E.Q.: Waterhe-nerf: Water- ray matching neural radiance fields for underwater scene reconstruction. Informa- tion Fusion115, 102770 (2025) Wat3R 19 Supplementary Material for “Wat3R: Underwater 3D Geometry Learning without...

Pith tools

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