Pith. sign in

REVIEW 2 major objections 4 minor 1 cited by

Learning Inverse Laplacian Pyramid for Progressive Depth Completion

T0 review · 2 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read LP-Net runs the Laplacian pyramid in reverse: it predicts a coarse global depth, then progressively re-adds high-frequency detail via a learned smoothness-versus-sharpness filter, and reports top KITTI accuracy with each filter applied…

desk verdict LP-Net has a real efficiency story and strong external KITTI numbers, but the Selective Depth Filtering module as written is internally inconsistent and needs fixing before the results can be trusted. read the letter →

arxiv 2502.07289 v1 pith:NJB7GKKS submitted 2025-02-11 cs.CV

Kun Wang , Zhiqiang Yan , Junkai Fan , Jun Li , Jian Yang This is my paper · ORCID
classification cs.CV
keywords depthcompletionLaplacianpyramidmulti-scalepredictionselectivefilteringsparseRGB-guidedprogressiverefinementfeature
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

LP-Net argues that depth completion should not be formulated as iterative propagation from a coarse estimate, but as the inverse of a Laplacian pyramid: first predict the low-frequency residual at 1/16 resolution to capture the whole scene, then upsample and restore the missing high-frequency structure one scale at a time. The paper introduces two modules, a Multi-path Feature Pyramid for global context and a Selective Depth Filtering module that learns both a smoothness and a sharpness filter, and shows that this progressive scheme reaches the top of the KITTI Depth Completion leaderboard and the best RMSE on TOFDC while running faster and using less memory than recent propagation-based competitors. The results matter because the dominant family of single-scale propagation methods pays a heavy computational cost for many iterations of pixel-level message passing; if the pyramid formulation is right, accurate dense depth can be obtained with a single forward pass of cheap per-scale filtering.

What carries the argument

The load-bearing object is the Laplacian pyramid, used in reverse. Instead of decomposing an image into bandpass levels and a low-frequency residual, LP-Net predicts the low-frequency residual first and then reconstructs the pyramid upward. The Multi-path Feature Pyramid (MFP) module splits the deepest feature map into p pathways, applies different numbers of stride-2 convolutions per pathway, and re-fuses them to capture global context cheaply. The Selective Depth Filtering (SDF) module is the refinement engine: it generates deformable filter weights and offsets, builds a smoothness filter whose weights are softmax-normalized to sum to 1 and a sharpness filter whose tanh-normalized, mean-subtracted weights sum to 0, applies each exactly once, and blends the two outputs with a learned spatial attention map. A confidence mechanism fuses each predicted scale with the weighted pooling of the sparse input, and a multi-scale L1 plus L2 loss supervises every level.

What would settle it

Run LP-Net inference with the smoothness and sharpness filters applied a second time at each scale on KITTI validation and compare RMSE and MAE to the single-pass model; if the second pass improves RMSE by more than a small margin, the claim that a single non-iterative filtering pass recovers the high-frequency detail is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that a dense depth map can be generated by mirroring Laplacian pyramid reconstruction. The network starts from the lowest-resolution prediction, fuses it with a confidence-weighted downsampled sparse depth, and at each higher scale bilinearly upsamples the previous estimate, fuses the sparse measurements, and applies a learned selection between a smoothing filter whose kernel weights sum to 1 and a sharpening filter whose kernel weights sum to 0, both with deformable offsets. This replaces the iterative recurrent propagation of CSPN-style methods with a strictly progressive, once-per-scale refinement. On the official KITTI DC test set LP-Net reaches RMSE 684.71 mm, MAE 186.63 mm, iRMSE 1.81, and iMAE 0.80, and it reports 63.88 ms inference and 1.76 GB GPU memory on an RTX 4090, which the paper attributes precisely to avoiding iteration.

Load-bearing premise

The whole speed and memory advantage rests on one premise: a single pass of the learned smoothness and sharpness filters at each resolution is enough to restore the missing detail; if those filters must be iterated many times to reach the reported accuracy, the efficiency claim disappears.

Editorial extensions

If this is right

  • If the claims hold, iterative spatial propagation is not necessary for top-tier depth completion, so the field can shift from recurrent refinement to feed-forward multi-scale refinement.
  • The progressive steps can be truncated: earlier pyramid levels already produce usable depth, so the same network can trade accuracy for latency at runtime by stopping after fewer upsampling stages.
  • The success of MFP over a ViT block in ablations suggests global scene context in depth completion can be gathered with simple multi-scale convolutions instead of self-attention, reducing cost.
  • A learned per-pixel blend between a smoothing and a sharpening filter is enough to denoise interiors while preserving boundaries, a recipe likely to transfer to other image-guided dense prediction problems.
  • Demonstrated on a LiDAR outdoor benchmark, a Kinect indoor benchmark, and a ToF indoor benchmark, the pyramid scheme is not tied to one sensor type or depth range.

Reading between the lines

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

  • Editorial inference: the inverse-pyramid formulation is mathematically equivalent to learning a set of bandpass residuals, so the intermediate predictions D(4) through D(1) could be treated as explicit residual channels and inspected or regularized directly, a diagnostic the paper does not run.
  • Editorial inference: because the filters are applied once, the receptive field of the full-resolution stage is small; very thin or very distant structures that require global reasoning may be the first failure mode, which is testable by stress-testing on long thin objects.
  • Editorial inference: the same progressive completion schedule could be applied to other sparse-to-dense tasks such as depth super-resolution, normal completion, or sparse-view stereo, where the bottleneck is the recovery of high-frequency detail rather than coarse layout.
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

2 major / 4 minor

Summary. The paper proposes LP-Net, a depth completion network built around an inverse Laplacian pyramid: it first regresses a low-resolution depth residual, then progressively upsamples and refines the depth through four additional scales. Two modules are introduced: the Multi-path Feature Pyramid (MFP), which splits features into parallel pathways with different downsampling factors, and the Selective Depth Filtering (SDF) module, which learns a smoothness filter and a zero-sum high-pass sharpness filter, combined by a learned spatial selection map. Experiments report state-of-the-art results on KITTI DC using official leaderboard numbers, competitive results on NYUv2, and best results on TOFDC, alongside lower inference time and memory than recent propagation-based methods.

Significance. If the results are reproducible, LP-Net is a meaningful contribution: it challenges the prevailing iterative spatial-propagation paradigm, offers a strong accuracy-efficiency trade-off, and introduces a progressive multi-scale architecture for depth completion. The use of externally hosted KITTI leaderboard numbers and standard public benchmarks is a credible evaluation strategy. The main limitations are the mathematical description of the key SDF module, which appears internally inconsistent, and an overstatement of the NYUv2 rankings. These issues affect the paper's central claims and need to be resolved before the work can be fully accepted.

major comments (2)
  1. [Section III-C3 (Eqs. 8-9)] The sharpness filter weights are explicitly normalized to sum to zero by subtracting the mean (Eq. 8), so the output D'_a is a high-pass residual: it is approximately zero on smooth regions and, by construction, cannot carry absolute depth information. Eq. (9) then forms a convex combination D_hat = a*D'_m + (1-a)*D'_a. Since the text states that the selection map favors the sharpness branch (1-a high) at boundaries, the final output at boundaries would be dominated by a zero-mean high-pass signal, which cannot plausibly reconstruct the surface depth there. This is also inconsistent with the Laplacian pyramid reconstruction in Eq. (2), which is additive. Please either change Eq. (9) to an additive residual form (e.g., D_hat = D'_m + (1-a)*D'_a), or normalize the sharpness kernel to sum to one, or provide a concrete explanation and empirical evidence that the current convex combination produces valid depth values in the sharpness-selected regions.
  2. [Section IV-D2 (Table II)] The text claims LP-Net achieves 'second-best performance in RMSE and REL' on NYUv2. This is not supported by Table II: BP-Net, DySPN, BEV@DC, and PointDC all report RMSE 0.089 while LP-Net reports 0.090; for REL, LRRU and DFU report 0.011 while LP-Net reports 0.012. The claim overstates the method's ranking and should be corrected to match the table, or the table must be revised.
minor comments (4)
  1. [Table VI] In the RMSE column of Variant-iii, '721,83' uses a comma as the decimal separator; it should be '721.83'.
  2. [Section III-B and Fig. 6] The paper's SDF module is described as a key component of the inverse Laplacian pyramid, but Eq. (9) is a weighted average rather than an additive reconstruction. The figure caption mentions an 'Addition' operation; please align the figure with the equation or clarify the relationship to Eq. (2).
  3. [Section III-C3] If the sharpness filter is intended as a high-pass residual (unsharp-mask style), please cite the relevant image-enhancement literature and clearly define D'_a as a residual so that the formulation in Eq. (9) can be interpreted correctly.
  4. [Section IV-E] The ablation tables report single-run results without error bars. Given that several metric differences (e.g., MAE 185.38 vs 186.10 in Table V) are small, reporting mean and standard deviation over multiple runs would strengthen the component-wise conclusions.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: LP-Net's claims are empirical and externally benchmarked; the SDF zero-sum filter issue is an internal correctness concern, not a self-referential derivation.

full rationale

LP-Net (arXiv:2502.07289) is an empirical depth-completion architecture paper. Its central claims are state-of-the-art accuracy on KITTI DC (Table I, sourced from the official online leaderboard), NYUv2 (Table II, standard public protocol), and TOFDC (Table III), plus efficiency measurements on a single RTX 4090 (Table IV). None of these results is derived from the paper's own definitions by construction: the multi-scale Laplacian Pyramid scheme (Eqs. 1-5) is a network-architecture strategy, not a fitted identity, and the training loss (Eq. 10) supervises all scales directly against ground-truth depth. The MFP and SDF modules are learned components whose ablations (Tables V-VI) compare against baselines rather than renaming a fitted parameter as a prediction. The paper cites several prior works from the same group (RigNet [18], TPVD [36], DCDepth [8]), but these citations are contextual or provide a dataset/benchmark; no load-bearing derivation is justified solely by a self-citation, and no uniqueness theorem or ansatz is imported from the authors' prior work. The skeptic's flagged issue with the SDF sharpness filter (Eq. 8: weights are mean-subtracted so that a constant input maps to zero, and Eq. 9 then interpolates between fm and fa) is a potential internal inconsistency or implementation detail, but it is a correctness/engineering concern, not circularity: it does not make any reported metric equivalent to the model's inputs by definition. Because the accuracy evidence against external benchmarks is self-contained and the method's components are tested against alternatives, the appropriate circularity score is 0.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

This is an empirical deep-learning paper. The 'free parameters' are training hyperparameters and architectural choices selected by ablation; there is no analytic derivation. No new physical entities are introduced. The key assumptions are standard domain assumptions about deep learning training and multi-scale fusion.

free parameters (5)
  • MFP path count p = 4
    Hand-chosen after ablation (Table V); controls the number of parallel multi-scale pathways and directly affects global-context fusion.
  • SDF kernel size k = not specified
    The deformable filter kernel size is never stated in Section III-C3, yet it sets the receptive field for smoothness and sharpness filtering.
  • DropPath maximum rate = 0.5
    Regularization schedule set in Section IV-B; affects generalization but is not derived from any principle.
  • EMA decay rate = 0.9999
    Model averaging hyperparameter in Section IV-B; standard practice, chosen by hand.
  • Number of progressive scales = 5
    The full model uses five refinement steps (Section III-B); fewer steps trade accuracy for speed (Fig. 12), so the choice of 5 is application-driven.
assumptions (3)
  • domain assumption Bilinear up/down sampling preserves enough geometric structure for depth maps at different scales
    Section III-A defines the Laplacian pyramid with up/down operators; in real scenes, downsampling can smear thin structures and depth discontinuities.
  • ad hoc to paper The learned confidence c_i (Eq. 4) can reliably choose between regressed depth and pooled sparse depth at every scale
    Eq. 5 linearly interpolates based on this confidence; the entire fusion mechanism depends on this learned scalar field being trustworthy.
  • domain assumption End-to-end training with the multi-scale L1+L2 loss (Eq. 10) is sufficient to produce accurate dense depth
    This is standard practice in depth completion; no analysis links the loss to the evaluation metrics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Inverse Laplacian Pyramid for Progressive Depth Completion." pith.science (2026). https://pith.science/paper/NJB7GKKS

@misc{pith2026250207289,
  author       = {Pith},
  title        = {Pith review of: Learning Inverse Laplacian Pyramid for Progressive Depth Completion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NJB7GKKS}},
  note         = {Machine review of arXiv:2502.07289}
}
read the original abstract

Depth completion endeavors to reconstruct a dense depth map from sparse depth measurements, leveraging the information provided by a corresponding color image. Existing approaches mostly hinge on single-scale propagation strategies that iteratively ameliorate initial coarse depth estimates through pixel-level message passing. Despite their commendable outcomes, these techniques are frequently hampered by computational inefficiencies and a limited grasp of scene context. To circumvent these challenges, we introduce LP-Net, an innovative framework that implements a multi-scale, progressive prediction paradigm based on Laplacian Pyramid decomposition. Diverging from propagation-based approaches, LP-Net initiates with a rudimentary, low-resolution depth prediction to encapsulate the global scene context, subsequently refining this through successive upsampling and the reinstatement of high-frequency details at incremental scales. We have developed two novel modules to bolster this strategy: 1) the Multi-path Feature Pyramid module, which segregates feature maps into discrete pathways, employing multi-scale transformations to amalgamate comprehensive spatial information, and 2) the Selective Depth Filtering module, which dynamically learns to apply both smoothness and sharpness filters to judiciously mitigate noise while accentuating intricate details. By integrating these advancements, LP-Net not only secures state-of-the-art (SOTA) performance across both outdoor and indoor benchmarks such as KITTI, NYUv2, and TOFDC, but also demonstrates superior computational efficiency. At the time of submission, LP-Net ranks 1st among all peer-reviewed methods on the official KITTI leaderboard.

Figures

Figures reproduced from arXiv: 2502.07289 by the authors.

Figure 1
Figure 1. Visual comparison with recent state-of-the-art (SOTA) methods on [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Visual comparison between existing single-scale, propagation [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overall Framework of LP-Net. MFP, RH and SDF stand for the Multi-path Feature Pyramid module, Regression Head and Selective Depth Filtering ˆˆ [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Evolution of depth completion results. We illustrate the progression of our depth completion scheme by showcasing the intermediate results on the ˆˆ [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Illustration of the Multi-path Feature Pyramid (MFP) Module. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: Illustration of the Selective Depth Filtering (SDF) Module. This [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Illustration of the selective filtering mechanism in the SDF module. In [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison on the official KITTI DC test set. From top to bottom: color images, sparse depth maps, and results from GuideNet [33], [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: Qualitative comparison on the NYUv2 test set. From left to right: color images, sparse depth maps, results from CSPN [15], NLSPN [16], CFormer [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Qualitative comparison on the TOFDC test set. From left to right: [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 12
Figure 12. Figure 12: Relationship between depth completion accuracy and inference time [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Event-Driven Dynamic Scene Depth Completion

    cs.CV 2025-05 conditional novelty 6.0 of 10

    EventDC uses event-stream-conditioned dynamic convolutions (EMA and LDF modules) to align and refine RGB and sparse-depth features, achieving state-of-the-art results on a new event-based depth completion benchmark.

Reference graph

Works this paper leans on

59 extracted references · 54 canonical work pages · cited by 1 Pith paper

  1. [1]

    Hybrid-mvs: Robust multi-view reconstruction with hybrid optimization of visual and depth cues,

    C. Li, L. Zhou, H. Jiang, Z. Zhang, X. Xiang, H. Sun, Q. Luan, H. Bao, and G. Zhang, “Hybrid-mvs: Robust multi-view reconstruction with hybrid optimization of visual and depth cues,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 33, no. 12, pp. 7630– 7644, 2023

    work page 2023

  2. [2]

    Altnerf: Learning robust neural radiance field via alternating depth-pose optimization,

    K. Wang, Z. Yan, H. Tian, Z. Zhang, X. Li, J. Li, and J. Yang, “Altnerf: Learning robust neural radiance field via alternating depth-pose optimization,” in AAAI, vol. 38, no. 6, 2024, pp. 5508–5516

    work page 2024

  3. [3]

    A low-cost and scalable framework to build large-scale localization benchmark for augmented reality,

    H. Liu, L. Zhao, Z. Peng, W. Xie, M. Jiang, H. Zha, H. Bao, and G. Zhang, “A low-cost and scalable framework to build large-scale localization benchmark for augmented reality,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 34, no. 4, pp. 2274– 2288, 2024

    work page 2024

  4. [4]

    Designing for depth perceptions in augmented reality,

    C. Diaz, M. Walker, D. A. Szafir, and D. Szafir, “Designing for depth perceptions in augmented reality,” in ISMAR. IEEE, 2017, pp. 111–122

    work page 2017

  5. [5]

    Digital video stabilization method based on periodic jitters of airborne vision of large flapping wing robots,

    J. Ye, E. Pan, and W. Xu, “Digital video stabilization method based on periodic jitters of airborne vision of large flapping wing robots,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 34, no. 4, pp. 2591–2603, 2024. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 11

    work page 2024

  6. [6]

    Towards real-time monocular depth estimation for robotics: A survey,

    X. Dong, M. A. Garratt, S. G. Anavatti, and H. A. Abbass, “Towards real-time monocular depth estimation for robotics: A survey,” IEEE Transactions on Intelligent Transportation Systems , vol. 23, no. 10, pp. 16 940–16 961, 2022

    work page 2022

  7. [7]

    Sparse-to- dense depth estimation in videos via high-dimensional tensor voting,

    B. Wang, J. Zou, Y . Li, K. Ju, H. Xiong, and Y . F. Zheng, “Sparse-to- dense depth estimation in videos via high-dimensional tensor voting,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 29, no. 1, pp. 68–79, 2019

    work page 2019

  8. [8]

    Dcdepth: Progressive monocular depth estimation in discrete cosine domain,

    K. Wang, Z. Yan, J. Fan, W. Zhu, X. Li, J. Li, and J. Yang, “Dcdepth: Progressive monocular depth estimation in discrete cosine domain,” in NeurIPS, 2024

    work page 2024

Show all 59 references
  1. [9]

    Regularizing nighttime weirdness: Efficient self-supervised monocular depth estimation in the dark,

    K. Wang, Z. Zhang, Z. Yan, X. Li, B. Xu, J. Li, and J. Yang, “Regularizing nighttime weirdness: Efficient self-supervised monocular depth estimation in the dark,” in ICCV, 2021, pp. 16 055–16 064

    2021

  2. [10]

    Depth- centric dehazing and depth-estimation from real-world hazy driving video,

    J. Fan, K. Wang, Z. Yan, X. Chen, S. Gao, J. Li, and J. Yang, “Depth- centric dehazing and depth-estimation from real-world hazy driving video,” in AAAI, 2025

    2025

  3. [11]

    Sgnet: Structure guided network via gradient-frequency awareness for depth map super-resolution,

    Z. Wang, Z. Yan, and J. Yang, “Sgnet: Structure guided network via gradient-frequency awareness for depth map super-resolution,” in AAAI, 2024, pp. 5823–5831

    2024

  4. [14]

    Deep depth completion of a single rgb-d image,

    Y . Zhang and T. Funkhouser, “Deep depth completion of a single rgb-d image,” in CVPR, 2018, pp. 175–185

    2018

  5. [15]

    Learning depth with convolutional spatial propagation network,

    X. Cheng, P. Wang, and R. Yang, “Learning depth with convolutional spatial propagation network,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 42, no. 10, pp. 2361–2379, 2019

    2019

  6. [16]

    Non-local spatial propagation network for depth completion,

    J. Park, K. Joo, Z. Hu, C.-K. Liu, and I. So Kweon, “Non-local spatial propagation network for depth completion,” in ECCV. Springer, 2020, pp. 120–136

    2020

  7. [18]

    Rignet: Repetitive image guided network for depth completion,

    Z. Yan, K. Wang, X. Li, Z. Zhang, J. Li, and J. Yang, “Rignet: Repetitive image guided network for depth completion,” in ECCV. Springer, 2022, pp. 214–230

    2022

  8. [19]

    Cspn++: Learning context and resource aware convolutional spatial propagation networks for depth completion,

    X. Cheng, P. Wang, C. Guan, and R. Yang, “Cspn++: Learning context and resource aware convolutional spatial propagation networks for depth completion,” in AAAI, vol. 34, no. 07, 2020, pp. 10 615–10 622

    2020

  9. [20]

    Dyspn: Learning dy- namic affinity for image-guided depth completion,

    Y . Lin, H. Yang, T. Cheng, W. Zhou, and Z. Yin, “Dyspn: Learning dy- namic affinity for image-guided depth completion,” IEEE Transactions on Circuits and Systems for Video Technology , 2023

    2023

  10. [21]

    Depth seeds: Recovering incomplete depth data using superpixels,

    M. Van den Bergh, D. Carton, and L. Van Gool, “Depth seeds: Recovering incomplete depth data using superpixels,” in WACV. IEEE, 2013, pp. 363–368

    2013

  11. [22]

    Seeds: Superpixels extracted via energy-driven sampling,

    M. Van den Bergh, X. Boix, G. Roig, and L. Van Gool, “Seeds: Superpixels extracted via energy-driven sampling,”International Journal of Computer Vision , vol. 111, pp. 298–314, 2015

    2015

  12. [23]

    In defense of classical image processing: Fast depth completion on the cpu,

    J. Ku, A. Harakeh, and S. L. Waslander, “In defense of classical image processing: Fast depth completion on the cpu,” in CRV. IEEE, 2018, pp. 16–22

    2018

  13. [24]

    A surface geometry model for lidar depth completion,

    Y . Zhao, L. Bai, Z. Zhang, and X. Huang, “A surface geometry model for lidar depth completion,” IEEE Robotics and Automation Letters , vol. 6, no. 3, pp. 4457–4464, 2021

    2021

  14. [25]

    Sparsity invariant cnns,

    J. Uhrig, N. Schneider, L. Schneider, U. Franke, T. Brox, and A. Geiger, “Sparsity invariant cnns,” in 3DV. IEEE, 2017, pp. 11–20

    2017

  15. [26]

    Hms- net: Hierarchical multi-scale sparsity-invariant network for sparse depth completion,

    Z. Huang, J. Fan, S. Cheng, S. Yi, X. Wang, and H. Li, “Hms- net: Hierarchical multi-scale sparsity-invariant network for sparse depth completion,” IEEE Transactions on Image Processing, vol. 29, pp. 3429– 3441, 2019

    2019

  16. [27]

    Uncertainty- aware cnns for depth completion: Uncertainty from beginning to end,

    A. Eldesokey, M. Felsberg, K. Holmquist, and M. Persson, “Uncertainty- aware cnns for depth completion: Uncertainty from beginning to end,” in CVPR, 2020, pp. 12 014–12 023

    2020

  17. [28]

    Estimat- ing depth from rgb and sparse sensing,

    Z. Chen, V . Badrinarayanan, G. Drozdov, and A. Rabinovich, “Estimat- ing depth from rgb and sparse sensing,” in ECCV, 2018, pp. 167–182

    2018

  18. [29]

    Learning steering kernels for guided depth completion,

    L. Liu, Y . Liao, Y . Wang, A. Geiger, and Y . Liu, “Learning steering kernels for guided depth completion,” IEEE Transactions on Image Processing, vol. 30, pp. 2850–2861, 2021

    2021

  19. [30]

    Bilateral propagation network for depth completion,

    J. Tang, F.-P. Tian, B. An, J. Li, and P. Tan, “Bilateral propagation network for depth completion,” in CVPR, 2024, pp. 9763–9772

    2024

  20. [31]

    Learning joint 2d- 3d representations for depth completion,

    Y . Chen, B. Yang, M. Liang, and R. Urtasun, “Learning joint 2d- 3d representations for depth completion,” in ICCV, 2019, pp. 10 023– 10 032

    2019

  21. [32]

    Deeplidar: Deep surface normal guided depth prediction for outdoor scene from sparse lidar data and single color image,

    J. Qiu, Z. Cui, Y . Zhang, X. Zhang, S. Liu, B. Zeng, and M. Pollefeys, “Deeplidar: Deep surface normal guided depth prediction for outdoor scene from sparse lidar data and single color image,” in CVPR, 2019, pp. 3313–3322

    2019

  22. [33]

    Learning guided convolutional network for depth completion,

    J. Tang, F.-P. Tian, W. Feng, J. Li, and P. Tan, “Learning guided convolutional network for depth completion,” IEEE Transactions on Image Processing, vol. 30, pp. 1116–1129, 2020

    2020

  23. [34]

    Guideformer: Transformers for image guided depth completion,

    K. Rho, J. Ha, and Y . Kim, “Guideformer: Transformers for image guided depth completion,” in CVPR, 2022, pp. 6250–6259

    2022

  24. [35]

    Bev@dc: Bird’s-eye view assisted training for depth completion,

    W. Zhou, X. Yan, Y . Liao, Y . Lin, J. Huang, G. Zhao, S. Cui, and Z. Li, “Bev@dc: Bird’s-eye view assisted training for depth completion,” in CVPR, 2023, pp. 9233–9242

    2023

  25. [36]

    Tri-perspective view decomposition for geometry-aware depth completion,

    Z. Yan, Y . Lin, K. Wang, Y . Zheng, Y . Wang, Z. Zhang, J. Li, and J. Yang, “Tri-perspective view decomposition for geometry-aware depth completion,” in CVPR, 2024, pp. 4874–4884

    2024

  26. [37]

    Dynamic spatial propagation network for depth completion,

    Y . Lin, T. Cheng, Q. Zhong, W. Zhou, and H. Yang, “Dynamic spatial propagation network for depth completion,” in AAAI, vol. 36, no. 2, 2022, pp. 1638–1646

    2022

  27. [38]

    Graphcspn: Geometry- aware depth completion via dynamic gcns,

    X. Liu, X. Shao, B. Wang, Y . Li, and S. Wang, “Graphcspn: Geometry- aware depth completion via dynamic gcns,” in ECCV. Springer, 2022, pp. 90–107

    2022

  28. [39]

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

    Y . Wang, B. Li, G. Zhang, Q. Liu, T. Gao, and Y . Dai, “Lrru: Long- short range recurrent updating networks for depth completion,” in ICCV, 2023, pp. 9422–9432

    2023

  29. [40]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in MICCAI. Springer, 2015, pp. 234–241

    2015

  30. [41]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in CVPR, 2016, pp. 770–778

    2016

  31. [42]

    Improving depth completion via depth feature upsampling,

    Y . Wang, G. Zhang, S. Wang, B. Li, Q. Liu, L. Hui, and Y . Dai, “Improving depth completion via depth feature upsampling,” in CVPR, 2024, pp. 21 104–21 113

    2024

  32. [43]

    Deformable convolutional networks,

    J. Dai, H. Qi, Y . Xiong, Y . Li, G. Zhang, H. Hu, and Y . Wei, “Deformable convolutional networks,” in ICCV, 2017, pp. 764–773

    2017

  33. [44]

    Deformable convnets v2: More deformable, better results,

    X. Zhu, H. Hu, S. Lin, and J. Dai, “Deformable convnets v2: More deformable, better results,” in CVPR, 2019, pp. 9308–9316

    2019

  34. [45]

    Sparse-to-dense: Depth prediction from sparse depth samples and a single image,

    F. Ma and S. Karaman, “Sparse-to-dense: Depth prediction from sparse depth samples and a single image,” in ICRA, 2018

    2018

  35. [46]

    Confidence propagation through cnns for guided sparse depth regression,

    A. Eldesokey, M. Felsberg, and F. S. Khan, “Confidence propagation through cnns for guided sparse depth regression,” IEEE transactions on pattern analysis and machine intelligence , vol. 42, no. 10, pp. 2423– 2436, 2019

    2019

  36. [47]

    Depth completion with twin surface extrapolation at occlusion boundaries,

    S. Imran, X. Liu, and D. Morris, “Depth completion with twin surface extrapolation at occlusion boundaries,” in CVPR, 2021, pp. 2583–2592

    2021

  37. [48]

    Adaptive context-aware multi- modal network for depth completion,

    S. Zhao, M. Gong, H. Fu, and D. Tao, “Adaptive context-aware multi- modal network for depth completion,” IEEE Transactions on Image Processing, vol. 30, pp. 5264–5276, 2021

    2021

  38. [49]

    Fcfr-net: Feature fusion based coarse-to-fine residual learning for depth completion,

    L. Liu, X. Song, X. Lyu, J. Diao, M. Wang, Y . Liu, and L. Zhang, “Fcfr-net: Feature fusion based coarse-to-fine residual learning for depth completion,” in AAAI, vol. 35, no. 3, 2021, pp. 2136–2144

    2021

  39. [50]

    Penet: Towards precise and efficient image guided depth completion,

    M. Hu, S. Wang, B. Li, S. Ning, L. Fan, and X. Gong, “Penet: Towards precise and efficient image guided depth completion,” in ICRA. IEEE, 2021, pp. 13 656–13 662

    2021

  40. [51]

    Completionformer: Depth completion with convolutions and vision transformers,

    Y . Zhang, X. Guo, M. Poggi, Z. Zhu, G. Huang, and S. Mattoccia, “Completionformer: Depth completion with convolutions and vision transformers,” in CVPR, 2023, pp. 18 527–18 536

    2023

  41. [52]

    Decomposed guided dynamic filters for efficient rgb-guided depth completion,

    Y . Wang, Y . Mao, Q. Liu, and Y . Dai, “Decomposed guided dynamic filters for efficient rgb-guided depth completion,” IEEE Transactions on Circuits and Systems for Video Technology , 2023

    2023

  42. [53]

    Ogni-dc: Robust depth completion with optimization-guided neural iterations,

    Y . Zuo and J. Deng, “Ogni-dc: Robust depth completion with optimization-guided neural iterations,” in ECCV. Springer, 2024, pp. 78–95

    2024

  43. [54]

    Are we ready for autonomous driving? the kitti vision benchmark suite,

    A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? the kitti vision benchmark suite,” in CVPR. IEEE, 2012, pp. 3354–3361

    2012

  44. [55]

    Indoor segmentation and support inference from rgbd images,

    N. Silberman, D. Hoiem, P. Kohli, and R. Fergus, “Indoor segmentation and support inference from rgbd images,” in ECCV. Springer, 2012, pp. 746–760

    2012

  45. [56]

    Pytorch: An imperative style, high-performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. Kopf, E. Yang, Z. DeVito, M. Raison, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala, “Pytorch: An imperative style, high-pe...

    2019

  46. [57]

    Fractalnet: Ultra-deep neural networks without residuals,

    G. Larsson, M. Maire, and G. Shakhnarovich, “Fractalnet: Ultra-deep neural networks without residuals,” arXiv preprint arXiv:1605.07648 , 2016

    2016 arXiv

  47. [58]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” in ICLR, 2019

    2019

  48. [59]

    Super-convergence: Very fast training of neural networks using large learning rates,

    L. N. Smith and N. Topin, “Super-convergence: Very fast training of neural networks using large learning rates,” in Artificial intelligence and machine learning for multi-domain operations applications , vol. 11006. SPIE, 2019, pp. 369–386

    2019

  49. [60]

    Aggregating feature point cloud for depth completion,

    Z. Yu, Z. Sheng, Z. Zhou, L. Luo, S.-Y . Cao, H. Gu, H. Zhang, and H.-L. Shen, “Aggregating feature point cloud for depth completion,” in ICCV, 2023, pp. 8732–8743

    2023

  50. [61]

    Sparse and noisy lidar completion with rgb guidance and uncertainty,

    W. Van Gansbeke, D. Neven, B. De Brabandere, and L. Van Gool, “Sparse and noisy lidar completion with rgb guidance and uncertainty,” in MVA. IEEE, 2019, pp. 1–6

    2019

  51. [62]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” in ICLR, 2021. Kun Wang received his B....

    2021

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.