Pith. sign in

REVIEW 3 major objections 6 minor 2 cited by

OMNI-DC: Highly Robust Depth Completion with Multiresolution Depth Integration

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A single depth-completion model, trained only on synthetic data, transfers zero-shot to real LiDAR, SfM, and VIO sparse depth and beats per-dataset-trained rivals on several benchmarks.

desk verdict Strong zero-shot depth completion with genuine contributions, but the headline ETH3D result needs a train/test overlap check before the 43% claim is trusted. read the letter →

arxiv 2411.19278 v2 pith:AQT3QA4A submitted 2024-11-28 cs.CV

Yiming Zuo , Willow Yang , Zeyu Ma , Jia Deng This is my paper · ORCID
classification cs.CV
keywords depthcompletionzero-shotgeneralizationmulti-resolutionintegrationLaplacianlossscaleequivariancesynthetictrainingdataLiDARsimulationnovelviewsynthesis
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

This paper tries to establish that depth completion — predicting a dense depth map from an RGB image plus a few known depth points — can be made to work reliably across many sensors, scene types, and sparsity levels without retraining, using a single model. The authors claim their multi-resolution depth integrator, uncertainty-aware Laplacian loss, and synthetic-only training recipe with simulated LiDAR and SfM patterns let one model outperform all zero-shot baselines on seven real-world datasets, with up to a 43 percent error reduction, and even beat models trained directly on KITTI on one metric. A reader should care because today's depth completion models typically fail on unseen sparse patterns and must be retrained per dataset, which blocks practical use in robotics, reconstruction, and view synthesis.

What carries the argument

The Multi-resolution Depth Integrator is a parameter-free layer that estimates the dense depth map as the solution of a linear least-squares problem, balancing a sparse-depth fidelity term (with a learned confidence weight) against predicted depth-gradient constraints computed at several resolutions: full resolution and successive average-pooled half-resolutions. Down-sampling the optimization target before taking finite differences makes the effective integration step from a known pixel to a far pixel shrink by a factor of $2^{R-1}$, which the paper's 1D analysis shows reduces accumulated variance from $n\sigma^2$ toward $n\sigma^2/2^{R-1}$. The same least-squares formulation, with sparse depths expressed in log space and the network input normalized by the median log-depth, yields the paper's proven guarantee that the output scales linearly with any multiplicative change in input depth.

What would settle it

Take a real sparse depth map from an unseen sensor (for instance mmWave radar on the ZJU-4DRadarCam benchmark) and compare OMNI-DC's zero-shot RMSE against the in-domain-trained RadarCam-Depth model; the paper's appendix already shows a two-fold gap. If that gap persists or widens when realistic radar patterns are added to the training simulation, the claim that the model is sensor-agnostic is falsified. A more direct test: on a controlled input with known sparse points, measure the RMSE of the predicted depth versus distance to the nearest known pixel; the Multi-resolution DDI predicts the growth slows with the number of resolutions, so a linear-in-distance growth matching the single-resolution bound would contradict the paper's mechanism.

Watch

Extended reading notes

Core claim

The paper's central claim is that the poor generalization of depth completion to new datasets and unseen sparse depth patterns is not an intrinsic limit of the task but a fixable design problem. The authors diagnose the failure mode by analyzing a simplified 1D version of the optimization-guided neural iteration: with an i.i.d. Gaussian error on predicted depth gradients $\hat{G}_i = G^{\mathrm{gt}}_i + n_i$, integrating from a single known pixel at position 0 yields $\hat{D}_n \sim N(D_0 + \sum_i G^{\mathrm{gt}}_i,\; n\,\sigma^2)$, so uncertainty grows linearly with distance to the nearest known pixel. Because real sparse depth from SfM or active sensors often leaves large holes, this error accumulation makes single-resolution integration unusable on extremely sparse inputs. The paper's remedy is a Multi-resolution Depth Integrator that solves a least-squares problem enforcing gradient constraints at several down-sampled resolutions, reducing the integration distance and the accumulated variance. Combined with a per-pixel Laplacian loss that models depth ambiguity, a log-depth scale normalization that gives guaranteed scale equivariance, and training on five synthetic datasets with simulated SIFT-based, LiDAR-line, and noise-corrupted sparse patterns, the resulting model generalizes zero-shot to LiDAR, SfM, VIO, and consumer depth across seven datasets, with the largest reported gain a 43 percent RMSE reduction on the outdoor ETH3D-SfM split and a zero-shot KITTI MAE that beats all in-domain-trained baselines. The same depth prior, injected as a loss term into 3D Gaussian Splatting training, lifts rendering PSNR from 15.64 to 20.38 on ETH3D.

Load-bearing premise

The load-bearing premise is that the five synthetic training datasets and the simulated sparse patterns (SIFT keypoints, randomized LiDAR lines, and injected noise) are representative enough of real sensors and scenes that a model trained on them alone transfers zero-shot to real LiDAR, SfM, VIO, and consumer-depth data.

Editorial extensions

If this is right

  • A single OMNI-DC checkpoint can replace per-dataset-trained depth completion systems on any new sensor or scene whose sparse pattern resembles the simulated LiDAR, SfM, and noise patterns used in training.
  • The model's zero-shot KITTI results (MAE 0.597 on 8-line LiDAR, better than every in-domain-trained baseline) indicate that broad synthetic training with realistic pattern simulation can beat in-domain training, not merely approach it.
  • Because the output is provably scale-equivariant, users can feed SfM reconstructions with unknown metric scale directly into the model and get consistent relative depth without estimating a scale factor.
  • OMNI-DC's dense depth can serve as a geometry prior for 3D Gaussian Splatting, improving novel view synthesis on sparse-view scenes (PSNR 20.38 vs 15.64 for vanilla 3DGS).
  • The multi-resolution integrator's robustness to extremely sparse inputs (0.03% density, REL=0.034 in the synthetic-pattern suite) should carry over to downstream tasks where measurements are scarce, such as VIO with few tracked points.

Reading between the lines

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

  • An implication the authors leave implicit is that the multi-resolution gradient-integration idea should transfer to other integrating dense prediction tasks — such as surface-normal or optical-flow estimation from sparse constraints — where single-resolution propagation suffers the same variance growth with distance.
  • If the pattern simulator is the key to zero-shot transfer, then extending it to cover radar, structured-light holes, and object-removal gaps should extend the model's coverage; the paper's own radar experiments identify exactly this as the current weak spot.
  • The paper's finding that mixing real NYU data into training hurts performance — because real ground-truth depth is blurry — suggests that label sharpness, not domain realism, drives the success, a hypothesis the paper does not directly test.
  • A practical extension would be to let the model fall back to relative (monocular) depth when the sparse input is empty or nearly empty, since the current architecture explicitly does not handle the zero-sparse-point case; a smooth interpolation between the two regimes would broaden the model's applicability to monocular depth estimation.
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

3 major / 6 minor

Summary. The paper presents OMNI-DC, a depth completion model aimed at zero-shot generalization across datasets and sparse depth sensor patterns. The main components are a multi-resolution variant of the Differentiable Depth Integrator (DDI), a Laplacian-based probabilistic loss, a log-depth scale normalization scheme with a claimed scale-equivariance guarantee, and a training pipeline that mixes five synthetic datasets with synthetically generated SfM, LiDAR, and noise patterns. The method is evaluated on seven real-world benchmarks (KITTI, NYUv2, VOID, ETH3D, iBims, ARKitScenes, DIODE) and is reported to outperform zero-shot baselines consistently, including a 43% RMSE reduction on ETH3D-SfM outdoor compared with Marigold. A downstream application to 3D Gaussian Splatting view synthesis is also presented.

Significance. If the claims hold, this is a substantively useful contribution: a single model that handles a wide range of densities, noise levels, and sensor types, with a parameter-free integration layer, a scale-equivariance property, and released code/checkpoints. The synthetic-only training recipe is interesting and the view-synthesis application is a practical demonstration. The main caveat is that the headline zero-shot result on ETH3D-SfM depends on the training data being truly held out, a property the manuscript does not verify. The scale-equivariance derivation and the careful ablations on held-out validation splits are strengths, as is the breadth of the evaluation.

major comments (3)
  1. [Sec. 3.6, Table 1; Sec. 4.1, Appendix J] The training set includes BlendedMVS (115K images) and the evaluation includes ETH3D-SfM (13 scenes, 454 images). BlendedMVS is publicly known to incorporate scene content derived from ETH3D, and the manuscript does not report any check that the ETH3D evaluation scenes are disjoint from the BlendedMVS training scenes. Since the headline result (Sec. 4.5, Table 3: RMSE 1.069 vs 1.883 for Marigold, a 43% reduction) and the claim of zero-shot generalization to real SfM both rest on ETH3D being unseen, this is a load-bearing issue. Please provide a scene-level and image-level disjointness check, such as listing the scene identifiers in BlendedMVS and the ETH3D evaluation scenes, or otherwise rule out overlap.
  2. [Sec. 4.1, Appendix J] The ETH3D-SfM evaluation projects COLMAP sparse points into image space, but the manuscript does not state how the arbitrary scale of the COLMAP reconstruction is converted to metric units before computing RMSE. If the ground-truth LiDAR is used to align the COLMAP scale, that should be stated explicitly, because it affects what the zero-shot claim covers. Please specify the exact scale-alignment procedure and any use of ground-truth data for scale recovery.
  3. [Sec. 4.6, Table 2] Aggregating indoor and outdoor results by dividing outdoor RMSE by an arbitrary factor of 5.0 can change the ranking of methods and is not a standard metric. The per-dataset breakdown is relegated to Appendix N, which makes the main aggregated table difficult to interpret. Please either report the per-dataset numbers in the main table or use a scale-invariant aggregation such as REL, with the choice of the scaling factor justified.
minor comments (6)
  1. [Sec. 3.4] The statement that this work is "the first to apply probability-based losses to depth estimation or depth completion" is too strong. Heteroscedastic Laplacian and Gaussian losses have been used in prior uncertainty-aware depth regression and depth estimation works. Please temper the novelty claim accordingly.
  2. [Tables 2-5] No error bars or multi-seed results are reported for the main benchmark tables. Given that several reported margins are modest, please report run-to-run variance or results from multiple seeds for the main comparisons.
  3. [Sec. 3.2, Fig. 3] The error-accumulation analysis assumes i.i.d. Gaussian gradient noise and a single known pixel, and the statement that multiresolution integration reduces the number of integration steps from n to n/2^{R-1} is heuristic. The ablation in Table 4 supports the design empirically, but the theoretical motivation should be qualified as an illustrative model.
  4. [Eq. (10)] The loss weights 0.5 and 2.0 in the final loss are not ablated. Please include a small sensitivity study of these weights or explain how they were selected.
  5. [Table 3] The grouping "Trained on KITTI/NYU" includes methods that are evaluated zero-shot on ETH3D, which may confuse readers. Please clarify in the caption that these methods are trained on the in-domain datasets and tested on ETH3D without fine-tuning.
  6. [Sec. 4.8] The novel-view-synthesis results are reported on a single random split of 1/8 of the views. Please report variance across multiple splits or at least state whether the reported numbers are stable.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the scale-equivariance proof, loss design, and ablations are self-contained; the flagged ETH3D/BlendedMVS overlap is a data-contamination risk, not a circular step.

full rationale

The paper's derivation chain does not reduce to its inputs. The scale-equivariance result (Sec. 3.5 and Appendix L) is a genuine mathematical consequence: normalizing the network input by log(median(O)) makes the network's gradient predictions scale-invariant, and the log-space DDI energy is linear in log(O), so scaling O by β shifts the solution by log β. This is a theorem proved from the stated equations, not a fitted parameter renamed as a prediction. The Multi-resolution DDI (Sec. 3.3) is motivated by a simplified error-accumulation analysis (Sec. 3.2) but its benefit is validated by ablations on a held-out ETH3D-SfM validation split (Sec. 4.7, Tab. 4), with no overlap with the test set. The Laplacian loss (Sec. 3.4) is a standard negative log-likelihood objective; the final loss weights are determined by ablations, not by fitting to test metrics. Self-citations to OGNI-DC [74] by the same authors are architectural lineage, not load-bearing circularity: the paper re-derives the DDI equations (Eqs. 2-3), extends them (Eqs. 6-8), and reports OGNI-DC as a baseline that it outperforms. The synthetic test patterns on monocular benchmarks (Sec. 4.1) overlap in distribution with training patterns, but that is distributional similarity rather than a construction-level equivalence, and the central zero-shot claims also rest on out-of-distribution real sensors (KITTI LiDAR, VOID VIO, ETH3D-SfM). One correctness risk deserves explicit flagging: Table 1 lists BlendedMVS in the training mixture while Sec. 4.5 and Appendix J evaluate ETH3D-SfM zero-shot, and the paper does not report a disjointness check between BlendedMVS imagery and the ETH3D evaluation scenes. If BlendedMVS contains any ETH3D content, the zero-shot ETH3D result and the 43% reduction claim would be invalid due to train/test leakage. That is a data-hygiene correctness issue, not a circular derivation, and therefore does not raise the circularity score. Appendix G also states honest limitations (transparent/reflective surfaces, sky, radar patterns), which is consistent with a non-circular empirical evaluation.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central method rests on a few hand-chosen hyperparameters (loss weights, number of resolutions, sampling ratios, alpha) and on two unproven domain assumptions: a 1D Gaussian gradient-error model motivates the multi-res design, and synthetic training distributions transfer to real sensors. No invented physical entities are introduced.

free parameters (5)
  • alpha (DDI data term weight) = not reported
    Weight balancing sparse-depth fidelity vs gradient constraints in Eq. 2; presumably tuned on validation.
  • Loss weights for Laplacian and gradient-matching terms = 0.5 and 2.0 (Eq. 10)
    Hand-chosen coefficients in L = L1 + 0.5*LLap + 2.0*Lgm; not justified by a derivation.
  • Number of DDI resolutions R = 3
    R=3 was selected for the main results; ablation shows 3 > 2 > 1, but no sweep beyond 3 is reported.
  • Synthetic pattern sampling ratio (random:SfM:LiDAR) = 2:1:1
    Empirically chosen ratio in Sec F.2; paper states performance is not sensitive to it.
  • Outdoor RMSE scaling factor in Tab 2 aggregation = 5.0
    Outdoor RMSE divided by 5.0 before averaging with indoor results to 'approximately match scales'; arbitrary constant affects headline averages (separate numbers in Appendix N).
assumptions (4)
  • domain assumption The 1D i.i.d. Gaussian error model on predicted gradients (Eqs. 4-5) captures the dominant error-accumulation behavior of DDI on sparse inputs.
    Used in Sec 3.2 to motivate multi-res DDI; real gradient errors are correlated and non-Gaussian, but the argument is illustrative.
  • domain assumption Training on five synthetic datasets with simulated SfM/LiDAR/noise patterns transfers to real sensor distributions.
    Load-bearing for zero-shot generalization; stated in Sec 3.6 and Tab 1. Empirically supported by evaluation, but not proven.
  • standard math The linear least squares in DDI has a unique solution and can be differentiated through for training.
    Relies on standard linear algebra; DDI is a parameter-free layer from OGNI-DC.
  • domain assumption Post-DDI upsampling and DySPN refinement are linear in depth, preserving scale equivariance.
    Scale equivariance proof in Appendix L covers DDI only; the full pipeline assumes SPN and upsampling do not break equivariance, plausible but not proven for the whole model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of OMNI-DC: Highly Robust Depth Completion with Multiresolution Depth Integration." pith.science (2026). https://pith.science/paper/AQT3QA4A

@misc{pith2026241119278,
  author       = {Pith},
  title        = {Pith review of: OMNI-DC: Highly Robust Depth Completion with Multiresolution Depth Integration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AQT3QA4A}},
  note         = {Machine review of arXiv:2411.19278}
}
read the original abstract

Depth completion (DC) aims to predict a dense depth map from an RGB image and a sparse depth map. Existing DC methods generalize poorly to new datasets or unseen sparse depth patterns, limiting their real-world applications. We propose OMNI-DC, a highly robust DC model that generalizes well zero-shot to various datasets. The key design is a novel Multi-resolution Depth Integrator, allowing our model to deal with very sparse depth inputs. We also introduce a novel Laplacian loss to model the ambiguity in the training process. Moreover, we train OMNI-DC on a mixture of high-quality datasets with a scale normalization technique and synthetic depth patterns. Extensive experiments on 7 datasets show consistent improvements over baselines, reducing errors by as much as 43%. Codes and checkpoints are available at https://github.com/princeton-vl/OMNI-DC.

Figures

Figures reproduced from arXiv: 2411.19278 by the authors.

Figure 1
Figure 1. Given an RGB image and a sparse depth map, OMNI-DC generates high-quality dense depth maps on different types of scenes [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall pipeline of OMNI-DC. The RGB image and the normalized sparse depth (Sec. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The Multi-resolution DDI reduces the error accumulation in the depth integration. In [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: The sparse depth map from COLMAP [46] often has large holes on the textureless surfaces (b). These areas with high ambiguity dominate the L1 training error (c). As a result, the model trained without the Laplacian loss does not converge well, producing artifacts in the…
Figure 5
Figure 5. Figure 5: Rendered images and depths on test views. We train 3DGS [ [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: All numbers are benchmarked on the ETH3D [ [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. PacGDC: Label-Efficient Generalizable Depth Completion with Projection Ambiguity and Consistency

    cs.CV 2025-07 conditional novelty 7.0 of 10

    PacGDC synthesizes diverse pseudo training geometries by rescaling depth predictions from foundation models, improving zero-shot and few-shot generalization of depth completion.

  2. Depth Anything with Any Prior

    cs.CV 2025-05 conditional novelty 6.0 of 10

    Prior Depth Anything unifies depth completion, super-resolution, and inpainting by aligning a monocular prediction to partial metric priors pixel by pixel, then refining with a conditioned network.

Reference graph

Works this paper leans on

75 extracted references · 61 canonical work pages · cited by 2 Pith papers

  1. [1]

    Infrared system for 3d scanning of metallic surfaces

    Olivier Aubreton, Alban Bajard, Benjamin Verney, and Fred- eric Truchetet. Infrared system for 3d scanning of metallic surfaces. Machine vision and applications , 24:1513–1524,

    work page

  2. [2]

    Revisiting depth completion from a stereo matching perspective for cross-domain generaliza- tion

    Luca Bartolomei, Matteo Poggi, Andrea Conti, Fabio Tosi, and Stefano Mattoccia. Revisiting depth completion from a stereo matching perspective for cross-domain generaliza- tion. In International Conference on 3D Vision (3DV), pages 1360–1370, 2024. 1, 2, 12

    work page 2024

  3. [3]

    Arkitscenes: A diverse real-world dataset for 3d indoor scene understanding using mobile rgb-d data

    Gilad Baruch, Zhuoyuan Chen, Afshin Dehghan, Tal Dimry, Yuri Feigin, Peter Fu, Thomas Gebauer, Brandon Joffe, Daniel Kurz, Arik Schwartz, et al. Arkitscenes: A diverse real-world dataset for 3d indoor scene understanding using mobile rgb-d data. NeurIPS Datasets & Benchmarks , 2021. 6, 16

    work page 2021

  4. [4]

    Zoedepth: Zero-shot trans- fer by combining relative and metric depth

    Shariq Farooq Bhat, Reiner Birkl, Diana Wofk, Peter Wonka, and Matthias M ¨uller. Zoedepth: Zero-shot trans- fer by combining relative and metric depth. arXiv preprint arXiv:2302.12288, 2023. 2, 3, 8, 13

    arXiv 2023

  5. [5]

    Weight uncertainty in neural network

    Charles Blundell, Julien Cornebise, Koray Kavukcuoglu, and Daan Wierstra. Weight uncertainty in neural network. In ICML, pages 1613–1622, 2015. 5

    work page 2015

  6. [6]

    Depth pro: Sharp monocular metric depth in less than a second

    Aleksei Bochkovskii, Ama ¨el Delaunoy, Hugo Germain, Marcel Santos, Yichao Zhou, Stephan R Richter, and Vladlen Koltun. Depth pro: Sharp monocular metric depth in less than a second. arXiv preprint arXiv:2410.02073 , 2024. 3, 6, 7, 8, 16, 19, 23, 24, 25

    arXiv 2024

  7. [7]

    Mvsformer++: Revealing the devil in transformer’s details for multi-view stereo

    Chenjie Cao, Xinlin Ren, and Yanwei Fu. Mvsformer++: Revealing the devil in transformer’s details for multi-view stereo. arXiv preprint arXiv:2401.11673, 2024. 13

    arXiv 2024

  8. [8]

    Object detection using depth completion and camera-lidar fusion for autonomous driving

    Manuel Carranza-Garc ´ıa, F Javier Gal ´an-Sales, Jos ´e Mar´ıa Luna-Romera, and Jos´e C Riquelme. Object detection using depth completion and camera-lidar fusion for autonomous driving. Integrated Computer-Aided Engineering , 29(3): 241–258, 2022. 1

    work page 2022

Show all 75 references
  1. [9]

    Learning depth with convolutional spatial propagation network

    Xinjing Cheng, Peng Wang, and Ruigang Yang. Learning depth with convolutional spatial propagation network. IEEE TPAMI, 2019. 2

    2019

  2. [10]

    Cspn++: Learning context and resource aware convo- lutional spatial propagation networks for depth completion

    Xinjing Cheng, Peng Wang, Chenye Guan, and Ruigang Yang. Cspn++: Learning context and resource aware convo- lutional spatial propagation networks for depth completion. In AAAI, 2020. 2

    2020

  3. [11]

    Depth-regularized optimization for 3d gaussian splatting in few-shot images

    Jaeyoung Chung, Jeongtaek Oh, and Kyoung Mu Lee. Depth-regularized optimization for 3d gaussian splatting in few-shot images. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition , pages 811–820, 2024. 13

    2024

  4. [12]

    Unsupervised confidence for lidar depth maps and applications

    Andrea Conti, Matteo Poggi, Filippo Aleotti, and Stefano Mattoccia. Unsupervised confidence for lidar depth maps and applications. In IROS, pages 8352–8359, 2022. 6

    2022

  5. [13]

    Spar- sity agnostic depth completion

    Andrea Conti, Matteo Poggi, and Stefano Mattoccia. Spar- sity agnostic depth completion. In WACV, pages 5871–5880,

  6. [14]

    Depth-supervised nerf: Fewer views and faster train- ing for free

    Kangle Deng, Andrew Liu, Jun-Yan Zhu, and Deva Ra- manan. Depth-supervised nerf: Fewer views and faster train- ing for free. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 12882– 12891, 2022. 13

    2022

  7. [15]

    Geowiz- ard: Unleashing the diffusion priors for 3d geometry estima- tion from a single image

    Xiao Fu, Wei Yin, Mu Hu, Kaixuan Wang, Yuexin Ma, Ping Tan, Shaojie Shen, Dahua Lin, and Xiaoxiao Long. Geowiz- ard: Unleashing the diffusion priors for 3d geometry estima- tion from a single image. In ECCV, pages 241–258, 2024. 3

    2024

  8. [16]

    Virtual worlds as proxy for multi-object tracking anal- ysis

    Adrien Gaidon, Qiao Wang, Yohann Cabon, and Eleonora Vig. Virtual worlds as proxy for multi-object tracking anal- ysis. In CVPR, pages 4340–4349, 2016. 6

    2016

  9. [17]

    3d visual perception for self-driving cars using a multi-camera system: Calibration, mapping, localization, and obstacle de- tection

    Christian H ¨ane, Lionel Heng, Gim Hee Lee, Friedrich Fraun- dorfer, Paul Furgale, Torsten Sattler, and Marc Pollefeys. 3d visual perception for self-driving cars using a multi-camera system: Calibration, mapping, localization, and obstacle de- tection. Image and Vision Comput...

    2017

  10. [18]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. arxiv e-prints. arXiv preprint arXiv:1512.03385, 10, 2015. 14

    2015 arXiv

  11. [19]

    Depth com- pletion with twin surface extrapolation at occlusion bound- aries

    Saif Imran, Xiaoming Liu, and Daniel Morris. Depth com- pletion with twin surface extrapolation at occlusion bound- aries. In CVPR, pages 2583–2592, 2021. 6, 16

    2021

  12. [20]

    On the importance of accurate geometry data for dense 3d vision tasks

    HyunJun Jung, Patrick Ruhkamp, Guangyao Zhai, Nikolas Brasch, Yitong Li, Yannick Verdie, Jifei Song, Yiren Zhou, Anil Armagan, Slobodan Ilic, et al. On the importance of accurate geometry data for dense 3d vision tasks. In Pro- ceedings of the IEEE/CVF Conference on Computer V...

    2023

  13. [21]

    Repurpos- ing diffusion-based image generators for monocular depth estimation

    Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Met- zger, Rodrigo Caye Daudt, and Konrad Schindler. Repurpos- ing diffusion-based image generators for monocular depth estimation. In CVPR, pages 9492–9502, 2024. 2, 3, 6, 7, 8, 16, 19, 23, 24, 25

    2024

  14. [22]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph., 42(4):139–1,

  15. [23]

    Adam: A method for stochastic opti- mization

    Diederik P Kingma. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980, 2014. 15

    2014 arXiv

  16. [24]

    Evaluation of cnn-based single-image depth estimation methods

    Tobias Koch, Lukas Liebel, Friedrich Fraundorfer, and Marco Korner. Evaluation of cnn-based single-image depth estimation methods. In ECCV Workshops, pages 0–0, 2018. 6, 12, 16

    2018

  17. [25]

    Radarcam-depth: Radar-camera fusion for depth estimation with learned metric scale

    Han Li, Yukai Ma, Yaqing Gu, Kewei Hu, Yong Liu, and Xingxing Zuo. Radarcam-depth: Radar-camera fusion for depth estimation with learned metric scale. In Inter- national Conference on Robotics and Automation (ICRA) , pages 10665–10672. IEEE, 2024. 12, 13

    2024

  18. [26]

    Human pose regression with 9 residual log-likelihood estimation

    Jiefeng Li, Siyuan Bian, Ailing Zeng, Can Wang, Bo Pang, Wentao Liu, and Cewu Lu. Human pose regression with 9 residual log-likelihood estimation. In ICCV, pages 11025– 11034, 2021. 5

    2021

  19. [27]

    Megadepth: Learning single- view depth prediction from internet photos

    Zhengqi Li and Noah Snavely. Megadepth: Learning single- view depth prediction from internet photos. In CVPR, pages 2041–2050, 2018. 5

    2018

  20. [28]

    Dynamic spatial propagation network for depth com- pletion

    Yuankai Lin, Tao Cheng, Qi Zhong, Wending Zhou, and Hua Yang. Dynamic spatial propagation network for depth com- pletion. In AAAI, pages 1638–1646, 2022. 1, 2, 3, 6, 14

    2022

  21. [29]

    Learning affinity via spa- tial propagation networks

    Sifei Liu, Shalini De Mello, Jinwei Gu, Guangyu Zhong, Ming-Hsuan Yang, and Jan Kautz. Learning affinity via spa- tial propagation networks. In NeurIPS, 2017. 2

    2017

  22. [30]

    Graphcspn: Geometry-aware depth completion via dynamic gcns

    Xin Liu, Xiaofei Shao, Bo Wang, Yali Li, and Shengjin Wang. Graphcspn: Geometry-aware depth completion via dynamic gcns. In ECCV, 2022. 2

    2022

  23. [31]

    Depth estimation from monocular images and sparse radar using deep ordinal regression network

    Chen-Chou Lo and Patrick Vandewalle. Depth estimation from monocular images and sparse radar using deep ordinal regression network. In ICIP, pages 3343–3347. IEEE, 2021. 12

    2021

  24. [32]

    Distinctive image features from scale- invariant keypoints

    David G Lowe. Distinctive image features from scale- invariant keypoints. International journal of computer vision (IJCV), 60:91–110, 2004. 6, 7, 15, 23, 24

    2004

  25. [33]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. Communications of the ACM , 65(1):99–106, 2021. 13

    2021

  26. [34]

    Indoor segmentation and support inference from rgbd images

    Pushmeet Kohli Nathan Silberman, Derek Hoiem and Rob Fergus. Indoor segmentation and support inference from rgbd images. In ECCV, 2012. 1, 2, 6

    2012

  27. [35]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193, 2023. 15

    2023 arXiv

  28. [36]

    Test-time adaptation for depth completion

    Hyoungseob Park, Anjali Gupta, and Alex Wong. Test-time adaptation for depth completion. In CVPR, pages 20519– 20529, 2024. 3, 12

    2024

  29. [37]

    Non-local spatial propagation network for depth completion

    Jinsun Park, Kyungdon Joo, Zhe Hu, Chi-Kuei Liu, and In So Kweon. Non-local spatial propagation network for depth completion. In ECCV, pages 120–136, 2020. 1, 2

    2020

  30. [38]

    A simple yet universal framework for depth completion

    Jin-Hwi Park and Hae-Gon Jeon. A simple yet universal framework for depth completion. Advances in Neural In- formation Processing Systems , 37:23577–23602, 2025. 3, 12

    2025

  31. [39]

    Depth prompting for sensor-agnostic depth estimation

    Jin-Hwi Park, Chanhwi Jeong, Junoh Lee, and Hae-Gon Jeon. Depth prompting for sensor-agnostic depth estimation. In CVPR, pages 9859–9869, 2024. 2, 6, 7, 8, 12, 16, 19, 25

    2024

  32. [40]

    Deepli- dar: Deep surface normal guided depth prediction for out- door scene from sparse lidar data and single color image

    Jiaxiong Qiu, Zhaopeng Cui, Yinda Zhang, Xingdi Zhang, Shuaicheng Liu, Bing Zeng, and Marc Pollefeys. Deepli- dar: Deep surface normal guided depth prediction for out- door scene from sparse lidar data and single color image. In CVPR, pages 3313–3322, 2019. 6

    2019

  33. [41]

    Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer

    Ren ´e Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, and Vladlen Koltun. Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer. IEEE TPAMI, 44(3), 2022. 2, 3, 5, 15

    2022

  34. [42]

    Hypersim: A photorealistic syn- thetic dataset for holistic indoor scene understanding

    Mike Roberts, Jason Ramapuram, Anurag Ranjan, Atulit Kumar, Miguel Angel Bautista, Nathan Paczan, Russ Webb, and Joshua M Susskind. Hypersim: A photorealistic syn- thetic dataset for holistic indoor scene understanding. In ICCV, pages 10912–10922, 2021. 6

    2021

  35. [43]

    Dense depth pri- ors for neural radiance fields from sparse input views

    Barbara Roessle, Jonathan T Barron, Ben Mildenhall, Pratul P Srinivasan, and Matthias Nießner. Dense depth pri- ors for neural radiance fields from sparse input views. In CVPR, pages 12892–12901, 2022. 1, 5

    2022

  36. [44]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In MICCAI, pages 234–241, 2015. 14

    2015

  37. [45]

    Orb: An efficient alternative to sift or surf

    Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary Bradski. Orb: An efficient alternative to sift or surf. InICCV, pages 2564–2571, 2011. 6, 7, 23, 24

    2011

  38. [46]

    Structure-from-motion revisited

    Johannes Lutz Sch ¨onberger and Jan-Michael Frahm. Structure-from-motion revisited. In CVPR, 2016. 5, 16

    2016

  39. [47]

    Sch¨onberger, Silvano Galliani, Torsten Sattler, Konrad Schindler, Marc Pollefeys, and An- dreas Geiger

    Thomas Sch ¨ops, Johannes L. Sch¨onberger, Silvano Galliani, Torsten Sattler, Konrad Schindler, Marc Pollefeys, and An- dreas Geiger. A multi-view stereo benchmark with high- resolution images and multi-camera videos. In CVPR, 2017. 2, 6, 8, 13, 16

    2017

  40. [48]

    Depth estimation from camera image and mmwave radar point cloud

    Akash Deep Singh, Yunhao Ba, Ankur Sarker, Howard Zhang, Achuta Kadambi, Stefano Soatto, Mani Srivastava, and Alex Wong. Depth estimation from camera image and mmwave radar point cloud. In CVPR, pages 9275–9285,

  41. [49]

    Loftr: Detector-free local feature matching with transformers

    Jiaming Sun, Zehong Shen, Yuang Wang, Hujun Bao, and Xiaowei Zhou. Loftr: Detector-free local feature matching with transformers. In CVPR, pages 8922–8931, 2021. 5

    2021

  42. [50]

    Bi- lateral propagation network for depth completion

    Jie Tang, Fei-Peng Tian, Boshi An, Jian Li, and Ping Tan. Bi- lateral propagation network for depth completion. In CVPR, pages 9763–9772, 2024. 1, 2, 6, 7, 8, 16, 19, 23, 24, 25

    2024

  43. [51]

    3d lidar re- construction with probabilistic depth completion for robotic navigation

    Yifu Tao, Marija Popovi ´c, Yiduo Wang, Sundara Tejaswi Digumarti, Nived Chebrolu, and Maurice Fallon. 3d lidar re- construction with probabilistic depth completion for robotic navigation. In IROS, pages 5339–5346, 2022. 1

    2022

  44. [52]

    Raft: Recurrent all-pairs field transforms for optical flow

    Zachary Teed and Jia Deng. Raft: Recurrent all-pairs field transforms for optical flow. In ECCV, pages 402–419, 2020. 6

    2020

  45. [53]

    Pdc-net+: Enhanced probabilistic dense cor- respondence network

    Prune Truong, Martin Danelljan, Radu Timofte, and Luc Van Gool. Pdc-net+: Enhanced probabilistic dense cor- respondence network. IEEE TPAMI , 45(8):10247–10266,

  46. [54]

    Dn-splatter: Depth and normal priors for gaussian splatting and meshing

    Matias Turkulainen, Xuqian Ren, Iaroslav Melekhov, Otto Seiskari, Esa Rahtu, and Juho Kannala. Dn-splatter: Depth and normal priors for gaussian splatting and meshing. WACV, 2025. 1, 2, 8, 13

    2025

  47. [55]

    Sparsity invariant cnns

    Jonas Uhrig, Nick Schneider, Lukas Schneider, Uwe Franke, Thomas Brox, and Andreas Geiger. Sparsity invariant cnns. In International Conference on 3D Vision (3DV), 2017. 1, 2, 6, 16

    2017

  48. [56]

    Diode: A dense indoor and outdoor depth dataset

    Igor Vasiljevic, Nick Kolkin, Shanyi Zhang, Ruotian Luo, Haochen Wang, Falcon Z Dai, Andrea F Daniele, Moham- madreza Mostajabi, Steven Basart, Matthew R Walter, et al. Diode: A dense indoor and outdoor depth dataset. arXiv preprint arXiv:1908.00463, 2019. 6, 16 10

    1908 arXiv

  49. [57]

    G2- monodepth: A general framework of generalized depth in- ference from monocular rgb+ x data

    Haotian Wang, Meng Yang, and Nanning Zheng. G2- monodepth: A general framework of generalized depth in- ference from monocular rgb+ x data. IEEE TPAMI, 2023. 2, 3, 6, 7, 8, 12, 13, 16, 19, 23, 24, 25

    2023

  50. [58]

    Irs: A large naturalis- tic indoor robotics stereo dataset to train deep models for disparity and surface normal estimation

    Qiang Wang, Shizhen Zheng, Qingsong Yan, Fei Deng, Kaiyong Zhao, and Xiaowen Chu. Irs: A large naturalis- tic indoor robotics stereo dataset to train deep models for disparity and surface normal estimation. In IEEE Interna- tional Conference on Multimedia and Expo (ICME) , pag...

    2021

  51. [59]

    Tartanair: A dataset to push the limits of visual slam

    Wenshan Wang, Delong Zhu, Xiangwei Wang, Yaoyu Hu, Yuheng Qiu, Chen Wang, Yafei Hu, Ashish Kapoor, and Se- bastian Scherer. Tartanair: A dataset to push the limits of visual slam. In IROS, 2020. 6

    2020

  52. [60]

    Lrru: Long-short range recurrent updating networks for depth completion

    Yufei Wang, Bo Li, Ge Zhang, Qi Liu, Tao Gao, and Yuchao Dai. Lrru: Long-short range recurrent updating networks for depth completion. In ICCV, pages 9422–9432, 2023. 1, 2, 15

    2023

  53. [61]

    Sea-raft: Simple, efficient, accurate raft for optical flow

    Yihan Wang, Lahav Lipson, and Jia Deng. Sea-raft: Simple, efficient, accurate raft for optical flow. In ECCV, pages 36– 54, 2024. 5, 14

    2024

  54. [62]

    Improving depth completion via depth feature upsampling

    Yufei Wang, Ge Zhang, Shaoqian Wang, Bo Li, Qi Liu, Le Hui, and Yuchao Dai. Improving depth completion via depth feature upsampling. In CVPR, pages 21104–21113, 2024. 1, 2, 6, 7, 8, 16, 19, 23, 24, 25

    2024

  55. [63]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 13(4):600–612, 2004. 8

    2004

  56. [64]

    Unsupervised depth completion from visual iner- tial odometry

    Alex Wong, Xiaohan Fei, Stephanie Tsuei, and Stefano Soatto. Unsupervised depth completion from visual iner- tial odometry. IEEE Robotics and Automation Letters , 5(2): 1899–1906, 2020. 2, 6, 8

    1906

  57. [65]

    Towards domain-agnostic depth completion

    Guangkai Xu, Wei Yin, Jianming Zhang, Oliver Wang, Si- mon Niklaus, Simon Chen, and Jia-Wang Bian. Towards domain-agnostic depth completion. Machine Intelligence Research, 21(4):652–669, 2024. 2

    2024

  58. [66]

    Deformable spatial propagation networks for depth completion

    Zheyuan Xu, Hongche Yin, and Jian Yao. Deformable spatial propagation networks for depth completion. In ICIP, 2020. 2

    2020

  59. [67]

    Learnable differencing center for night- time depth perception

    Zhiqiang Yan, Yupeng Zheng, Deng-Ping Fan, Xiang Li, Jun Li, and Jian Yang. Learnable differencing center for night- time depth perception. Visual Intelligence, 2(1):15, 2024. 12, 13

    2024

  60. [68]

    Drivingstereo: A large-scale dataset for stereo matching in autonomous driving scenarios

    Guorun Yang, Xiao Song, Chaoqin Huang, Zhidong Deng, Jianping Shi, and Bolei Zhou. Drivingstereo: A large-scale dataset for stereo matching in autonomous driving scenarios. In CVPR, 2019. 12, 13

    2019

  61. [69]

    Depth anything: Unleashing the power of large-scale unlabeled data

    Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth anything: Unleashing the power of large-scale unlabeled data. In CVPR, pages 10371–10381, 2024. 3, 5

    2024

  62. [70]

    Depth any- thing v2

    Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xiao- gang Xu, Jiashi Feng, and Hengshuang Zhao. Depth any- thing v2. NeurIPS, 37:21875–21911, 2024. 2, 3, 5, 6, 7, 8, 12, 13, 16, 17, 19, 23, 24, 25

    2024

  63. [71]

    Blendedmvs: A large-scale dataset for generalized multi-view stereo net- works

    Yao Yao, Zixin Luo, Shiwei Li, Jingyang Zhang, Yufan Ren, Lei Zhou, Tian Fang, and Long Quan. Blendedmvs: A large-scale dataset for generalized multi-view stereo net- works. CVPR, 2020. 6

    2020

  64. [72]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, pages 586–595,

  65. [73]

    Completionformer: Depth completion with convolutions and vision transform- ers

    Youmin Zhang, Xianda Guo, Matteo Poggi, Zheng Zhu, Guan Huang, and Stefano Mattoccia. Completionformer: Depth completion with convolutions and vision transform- ers. In CVPR, pages 18527–18536, 2023. 1, 2, 5, 6, 7, 8, 12, 13, 14, 15, 16, 19, 23, 24, 25

    2023

  66. [74]

    generalizable

    Yiming Zuo and Jia Deng. Ogni-dc: Robust depth comple- tion with optimization-guided neural iterations. In ECCV, pages 78–95, 2024. 1, 2, 3, 4, 5, 6, 7, 8, 12, 13, 15, 16, 19, 23, 24, 25 11 OMNI-DC: Highly Robust Depth Completion with Multiresolution Depth Integration Appendix...

    2024

  67. [75]

    office” and the “courtyard

    Like other depth estimation models, our method faces challenges when predicting depth for transparent surfaces (e.g., glasses), reflective surfaces, or the sky. In Fig. d we show a few failure cases. 2) The backbone of our method takes 4 channels (RGB-D) input, which makes it ...

    1920

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.