Pith. sign in

REVIEW 3 major objections 5 minor 65 references

Leveraging Sparse LiDAR for RAFT-Stereo: A Depth Pre-Fill Perspective

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that pre-filling sparse LiDAR disparity with interpolation restores the benefit of LiDAR guidance in RAFT-Stereo, enabling state-of-the-art stereo depth with only a few hundred points.

desk verdict Pre-filling sparse LiDAR disparities fixes RAFT-Stereo's failure mode, and the analysis is the real contribution; the main open question is whether the simulated sparse sensors transfer to real hardware. read the letter →

arxiv 2507.19738 v1 pith:ZMZBPW4Z submitted 2025-07-26 cs.CV

Jinsu Yoo , Sooyoung Jeon , Zanming Huang , Tai-Yu Pan , Wei-Lun Chao This is my paper · ORCID
classification cs.CV
keywords RAFT-StereoLiDAR-guidedstereosparseDARdepthpre-fillingdisparityinitializationiterativerefinementcompletionmatching
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

RAFT-Stereo, an iterative stereo matcher, can exploit very sparse LiDAR depth if the sparse disparity initialization is first densified by interpolation. Naive injection of a few hundred LiDAR points fails, the paper argues, because the cost-volume retrieval is dominated by zero-initialized pixels and the recurrent convolutions treat the sparse accurate values as high-frequency noise. Pre-filling the missing disparities makes the guidance smooth enough to survive retrieval, so 300 pre-filled points outperform 64-beam guidance without pre-filling. Combining this late-fusion pre-fill with an early-fusion branch that appends 3D coordinates to image features yields GRAFT-Stereo, which the paper shows beats prior LiDAR-guided methods on KITTI, VKITTI2, and MS2 under sparse conditions.

What carries the argument

The load-bearing mechanism is the iterative cost-volume lookup of RAFT-Stereo, $S(h,w,k) = C(h,w, w - D(h,w) + k)$, which indexes the correlation volume by the current disparity estimate. With sparse guidance, most pixels index at zero disparity, so their retrieved features dominate and the sparse accurate retrievals become high-frequency spikes that low-pass-filtering convolutions suppress. Pre-filling the initial disparity map (e.g., with IP-Basic interpolation) replaces zero disparities with plausible values, converting sparse guidance into a smooth, low-frequency initialization. A separate early-fusion path backprojects the pre-filled depth into 3D coordinates and concatenates them with RGB to reinforce stereo correspondence.

What would settle it

Collect a stereo-LiDAR dataset from an actual low-beam sensor (e.g., 4- or 8-beam) in driving scenes, retrain RAFT-Stereo with naive guidance and with pre-filling, and compare; if pre-filling no longer beats naive injection on real sparse scans, the paper's mechanism and remedy do not transfer beyond subsampled dense LiDAR.

Watch

Extended reading notes

Core claim

The central claim is that pre-filling the sparse initial disparity map is what makes LiDAR guidance work inside RAFT-Stereo, for a specific mechanism. RAFT-Stereo retrieves a local slab of the correlation volume indexed by the current disparity estimate; when most pixels are zero-initialized, they dominate the retrieved features, and the sparse LiDAR-guided values appear as high-frequency outliers that the recurrent 2D convolutions attenuate. Densifying the initial map removes these discontinuities, letting the informative retrieval survive and propagate. For early fusion, the paper claims pre-filling also helps, but for a different reason — it must supply accurate correspondences — so a neural depth-completion with confidence-based top-1k subsampling is needed rather than coarse interpolation.

Load-bearing premise

The paper's sparse LiDAR is simulated by uniformly subsampling or beam-synthesizing dense 64-beam ground-truth scans, and it assumes this faithfully represents real low-cost sensors' noise, coverage, and beam patterns.

Editorial extensions

If this is right

  • With as few as 300 LiDAR points per frame, pre-filled RAFT-Stereo achieves lower disparity error than RAFT-Stereo guided by full 64-beam LiDAR without pre-filling.
  • GRAFT-Stereo reports lower RMSE, MAE, Bad1, and average disparity error than EG-Depth and SDG-Depth under uniform-sampled and beam-sampled sparsity on KITTI, and on VKITTI2 and MS2 with 300 points.
  • The two pre-fill strategies are not interchangeable: coarse interpolation helps late fusion but hurts early fusion, where a confidence-subsampled neural completion is required.
  • Because RAFT-Stereo supports anytime prediction, the late-fusion pre-fill variant provides a fast CPU-compatible path (IP-Basic) while the full model retains the accuracy–iteration trade-off.

Reading between the lines

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

  • The zero-dominance mechanism suggests that any smoothness-inducing initialization — not only depth completion — should recover most of the gain, since even nearest-neighbor fill in the toy example reduces the retrieved-feature Laplacian from 0.65 to 0.33.
  • Real low-beam LiDAR sensors add beam-specific noise and irregular coverage that uniform subsampling of dense scans does not mimic; testing on actual 4- or 8-beam hardware would determine whether the pre-fill benefit persists.
  • The pre-fill-before-injection principle may transfer to other iterative refinement networks such as RAFT optical flow when they receive sparse external guidance.
  • The confidence-based top-1k subsampling result implies that under a fixed guidance budget, where you place the completed points matters more than the raw accuracy of the completion.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper studies how to inject sparse LiDAR depth into RAFT-Stereo for disparity estimation. The authors identify a late-fusion path (initializing the disparity map) and an early-fusion path (concatenating XYZ coordinates with RGB at the feature encoder). They observe that naive injection degrades sharply as LiDAR points become sparse (300 points) and propose a depth pre-filling step: for late fusion, filling missing disparities with interpolation (IP-Basic or a learned completion network); for early fusion, retaining only the top-1k confident completed points. The resulting GRAFT-Stereo combines both fusion paths and is evaluated on KITTI Depth Completion, VKITTI2, and MS2 under uniform subsampling and simulated 4/8/16-beam LiDAR, reporting consistent improvements over EG-Depth and SDG-Depth in disparity and depth metrics.

Significance. If the reported results hold, the paper makes a useful and counterintuitive contribution: a simple interpolation-based pre-fill of the initial disparity map can restore the benefit of very sparse LiDAR guidance to a strong iterative stereo baseline, with a signal-processing rationale (sparse guidance creates high-frequency discontinuities that low-pass recurrent convolution attenuates). The paper is also transparent about its design choices: it retrains baselines with official code, reports both disparity and depth metrics, provides ablations of late-only and full models, and includes supplementary experiments on 64-beam LiDAR, iteration trade-offs, test-time point counts, and mid-fusion. The main weaknesses are the lack of error bars or significance tests (each model trained once), the absence of real low-beam LiDAR hardware validation, and the post-hoc selection of the top-1k early-fusion subsample based on validation performance.

major comments (3)
  1. [Sec. 4.2, Tables 4–6, Figs. S4–S5] The paper's practical claim is that pre-filling enables sparse LiDAR guidance to work in affordable, low-cost settings, but all sparse-guidance experiments use either uniform subsampling of the semi-dense aggregated ground-truth depth map (noiseless, spatially uniform seeds) or beam sampling from a 64-beam Velodyne scan (which retains the accuracy and calibration of a high-end sensor). Real low-beam sensors exhibit larger range noise, mixed pixels, irregular coverage, and synchronization errors relative to the stereo pair. Since pre-filling propagates seed values to large unmeasured regions, biased or noisy seeds could amplify errors rather than help. The supplementary Limitations (S1.1) does not acknowledge this sensor-realism gap. I would recommend either adding experiments with synthetically injected LiDAR noise/calibration error or explicitly reframing the contribution as simulation-based and discussing the transfer risk.
  2. [Sec. 5.3, Table 5, S3] The early-fusion pre-filling approach relies on a post-hoc selection of the top-1k depth-completed points based on validation performance. The paper reports that retaining top-1k works, but does not report the sensitivity to the number of retained points (e.g., 500, 2k, 5k) or to the confidence measure used. Since this is a free parameter tuned on the validation set, a short sensitivity analysis would strengthen the claim that the chosen value is not overfit to the validation split. As reported, the difference between row ➅ (top-1k, 3.33 Bad1) and row ➆ (dense, 3.44 Bad1) is small, and without variance estimates it is unclear whether the improvement is significant.
  3. [Sec. 5.1–5.2, Figs. 4–5, Tables 2–4] The signal-processing explanation (sparse guidance creates high-frequency discontinuities that are attenuated by low-pass filtering in RAFT-Stereo) is supported by a toy example and a qualitative FF, but the connection between the Laplacian of the retrieved feature map and the final disparity error is not quantitatively established on real data. The paper shows in Table 2 that adding noise to a zero disparity map degrades accuracy, and in Table 4 that pre-filling improves accuracy, but it does not directly measure the 'feature domination' quantity E (L2 distance to ground-truth retrieval) on KITTI before and after pre-filling. A small experiment reporting E or a similar retrieval-quality metric on real validation frames would make the proposed mechanism more than a plausible narrative.
minor comments (5)
  1. [Abstract & Sec. 1] The abstract claims that GRAFT-Stereo 'significantly outperforms' existing methods, but no statistical significance tests are reported; consider softening to 'consistently outperforms' or adding significance measures.
  2. [Sec. 5.2, Table 3] The table reports IP-Basic and neural-net pre-filling alone with Bad1 of 52.51% and 17.59%, but the text does not state how these maps are evaluated (e.g., on the semi-dense ground truth or the full image); please clarify.
  3. [Sec. 5.3, Fig. 7] The text says 'For pixels lacking a projected LiDAR point, we concatenate zero values', but it is unclear whether this is done before or after depth pre-filling; please clarify the preprocessing order.
  4. [Sec. 6, Table 6 and Table S2] The main paper reports RMSE/MAE for depth while Table S1 reports disparity metrics; it would be helpful to report both in the main table or clearly state that disparity metrics are in the supplement.
  5. [Supplementary, Sec. S2] The beam sampling follows [63] and samples 'more LiDAR lines from the lower part of the scene'; this detail is important for reproducibility and should be briefly mentioned in the main text or at least in the caption of Fig. 1b.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the main comparisons are held-out empirical evaluations, with only a minor validation-selected hyperparameter and a non-load-bearing citation overlap.

full rationale

I find no significant circularity. The central claims are empirical: sparse LiDAR guidance degrades RAFT-Stereo, pre-filling the initial disparity map restores it, and GRAFT-Stereo outperforms retrained baselines on KITTI, VKITTI2, and MS2. The models are trained on the training splits and evaluated on held-out validation splits, and the baselines are retrained with their official code, so no reported number is an identity or a fitted parameter renamed as a prediction. There is no equation in which an output is defined as an input. Two mild caveats prevent a clean zero. First, the beam-sampling protocol cites [61] (Pseudo-LiDAR++, which includes co-author Wei-Lun Chao), though the implementation actually follows the external [63] and the citation is not load-bearing. Second, the top-1k subsample in early fusion (Sec. 5.3) was chosen after observing validation performance, while all reported comparisons use the same validation set, creating a small optimistic bias. Additionally, S1.1's limitations do not acknowledge that sparse LiDAR is simulated by subsampling or synthesizing beams from dense 64-beam scans rather than measured with real low-beam sensors; that is an external-validity concern, not circularity. Overall the derivation chain is self-contained and externally evaluated.

Assumptions & free parameters 7 free parameters · 3 assumptions · 0 invented entities

This is an empirical deep learning paper. The free parameters are standard training hyperparameters and model-selection choices. No new physical entities are introduced. The central claim does not rest on a formal derivation, so the ledger focuses on the assumptions behind the experimental setup and the proposed mechanism.

free parameters (7)
  • top_k_subset_for_early_fusion = 1000
    In early fusion, only the top 1,000 neural-net pre-filled points are retained based on confidence. This value appears selected via validation performance (Table 5).
  • loss_weight_alpha = 0.9
    Sequential L1 loss weight, taken from RAFT-Stereo's original objective; not tuned by the authors.
  • training_refinement_steps = 22
    Number of recurrent refinement steps during training, fixed for the study.
  • test_refinement_steps = 32
    Number of recurrent refinement steps during inference.
  • crop_size_late_fusion = 336x1120
    Random crop size for the late-fusion experiments, chosen due to training considerations.
  • crop_size_early_fusion = 224x784
    Reduced crop size in early fusion experiments due to GPU memory constraints.
  • max_disparity = 192
    Maximum disparity range used for evaluation and cost volume construction.
assumptions (3)
  • domain assumption RAFT-Stereo's iterative refinement and cost volume retrieval operate as described in the paper.
    The analysis in Section 5.1 relies on the specific mechanics of RAFT-Stereo: features are retrieved from the cost volume based on the current disparity estimate.
  • domain assumption Subsampling dense 64-beam LiDAR ground truth accurately simulates real sparse LiDAR sensors.
    All sparse LiDAR conditions are created by uniform subsampling or beam synthesis from dense scans, not by using actual low-beam sensor data. This is the weakest load-bearing assumption for real-world claims.
  • ad hoc to paper Sparse guidance creates high-frequency discontinuities that are attenuated by low-pass filtering in RAFT-Stereo.
    This is the paper's novel explanation, supported by a toy example and FFT analysis, but not proven theoretically. It motivates the pre-filling solution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging Sparse LiDAR for RAFT-Stereo: A Depth Pre-Fill Perspective." pith.science (2026). https://pith.science/paper/ZMZBPW4Z

@misc{pith2026250719738,
  author       = {Pith},
  title        = {Pith review of: Leveraging Sparse LiDAR for RAFT-Stereo: A Depth Pre-Fill Perspective},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZMZBPW4Z}},
  note         = {Machine review of arXiv:2507.19738}
}
read the original abstract

We investigate LiDAR guidance within the RAFT-Stereo framework, aiming to improve stereo matching accuracy by injecting precise LiDAR depth into the initial disparity map. We find that the effectiveness of LiDAR guidance drastically degrades when the LiDAR points become sparse (e.g., a few hundred points per frame), and we offer a novel explanation from a signal processing perspective. This insight leads to a surprisingly simple solution that enables LiDAR-guided RAFT-Stereo to thrive: pre-filling the sparse initial disparity map with interpolation. Interestingly, we find that pre-filling is also effective when injecting LiDAR depth into image features via early fusion, but for a fundamentally different reason, necessitating a distinct pre-filling approach. By combining both solutions, the proposed Guided RAFT-Stereo (GRAFT-Stereo) significantly outperforms existing LiDAR-guided methods under sparse LiDAR conditions across various datasets. We hope this study inspires more effective LiDAR-guided stereo methods.

Figures

Figures reproduced from arXiv: 2507.19738 by the authors.

Figure 1
Figure 1. Overview of GRAFT-Stereo. (a): GRAFT-Stereo effectively integrates very sparse LiDAR points with minimal modifications to RAFT-Stereo [1]. (b): GRAFT-Stereo achieves a substantial reduction in disparity error on the KITTI benchmark [14] across various LiDAR sparsity conditions, surpassing the SotA LiDAR￾guided stereo methods, EG-Depth [11] and SDG-Depth [12]. gd: naive LiDAR-guidance; Avg. Err: average disparity err… view at source ↗
Figure 2
Figure 2. Illustration of the proposed GRAFT-Stereo. Given extremely sparse LiDAR points (e.g., 300 points), we systematically identify two complementary strategies to incorporate the LiDAR information into RAFT-Stereo: early fusion and late fusion. Our analysis reveals that depth pre-filling is the key to leveraging extremely sparse LiDAR guidance for effective stereo matching. a recurrent 2D CNN to predict a residual dispar… view at source ↗
Figure 3
Figure 3. Performance of RAFT-Stereo with different numbers of LiDAR points. The dis￾parity error increases drastically as the number of LiDAR guidance points reduces. 4.3 Preliminary results Where to inject LiDAR points? The modular design of RAFT-Stereo allows LiDAR information to be integrated at different stages. In addition to early fusion in image space and intermediate fusion in feature space, RAFT-Stereo offers a uniq… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Illustration of cost volume retrieval in RAFT-Stereo [1]. We compare four different initial disparity maps for feature retrieval: ground-truth disparity, default zero initialization, sparse ground truth mimicking sparse LiDAR, and a nearest-neighbor-filled version of t…
Figure 5
Figure 5. Figure 5: Analysis of cost volume features retrieved with sparse guidance. (a)-(d): The retrieved features using sparse disparity guidance (as in (d), top) can be viewed as a sampled combination of those obtained using ground-truth and zero-initialized maps (i.e., (a)-(c)). When…
Figure 6
Figure 6. Figure 6: The neural network architecture for depth pre-filling, using non-local spatial propagation [55] for the refinement block. To effectively leverage sparse LiDAR guidance without modifying the core architecture of RAFT-Stereo, we pro￾pose a pre-densification step to fill …
Figure 7
Figure 7. Figure 7: Early fusion for enhanced cost volume. Given the pre-filled depth map, we backproject pix￾els to 3D and reproject them into both images. Each 3D point’s XYZ is concatenated with RGB from the left/right image. This fused tensor—combining appear￾ance and geometry—is pass…
Figure 8
Figure 8. Figure 8: Qualitative comparison of predicted depth maps. GRAFT-Stereo produces more accurate depth than the baseline methods. Error maps are computed with respect to ground-truth. gd: naive LiDAR-guidance. 6 Additional Experiments and Analysis Comparison to SotA LiDAR-guided me…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 60 canonical work pages

  1. [1]

    Raft-stereo: Multilevel recurrent field transforms for stereo matching

    Lahav Lipson, Zachary Teed, and Jia Deng. Raft-stereo: Multilevel recurrent field transforms for stereo matching. In 3DV, 2021

    work page 2021

  2. [2]

    Iterative geometry encoding volume for stereo matching

    Gangwei Xu, Xianqi Wang, Xiaohuan Ding, and Xin Yang. Iterative geometry encoding volume for stereo matching. In CVPR, 2023

    work page 2023

  3. [3]

    Selective-stereo: Adaptive frequency information selection for stereo matching

    Xianqi Wang, Gangwei Xu, Hao Jia, and Xin Yang. Selective-stereo: Adaptive frequency information selection for stereo matching. In CVPR, 2024

    work page 2024

  4. [4]

    Practical stereo matching via cascaded recurrent network with adaptive correlation

    Jiankun Li, Peisen Wang, Pengfei Xiong, Tao Cai, Ziwei Yan, Lei Yang, Jiangyu Liu, Haoqiang Fan, and Shuaicheng Liu. Practical stereo matching via cascaded recurrent network with adaptive correlation. In CVPR, 2022

    work page 2022

  5. [5]

    Pyramid stereo matching network

    Jia-Ren Chang and Yong-Sheng Chen. Pyramid stereo matching network. In CVPR, 2018

    work page 2018

  6. [6]

    End-to-end learning of geometry and context for deep stereo regression

    Alex Kendall, Hayk Martirosyan, Saumitro Dasgupta, Peter Henry, Ryan Kennedy, Abraham Bachrach, and Adam Bry. End-to-end learning of geometry and context for deep stereo regression. In ICCV, 2017

    work page 2017

  7. [7]

    Hierarchical deep stereo matching on high-resolution images

    Gengshan Yang, Joshua Manela, Michael Happold, and Deva Ramanan. Hierarchical deep stereo matching on high-resolution images. In CVPR, 2019

    work page 2019

  8. [8]

    Revisiting stereo depth estimation from a sequence-to-sequence perspective with transformers

    Zhaoshuo Li, Xingtong Liu, Nathan Drenkow, Andy Ding, Francis X Creighton, Russell H Taylor, and Mathias Unberath. Revisiting stereo depth estimation from a sequence-to-sequence perspective with transformers. In ICCV, 2021

    work page 2021

Show all 65 references
  1. [9]

    Context-enhanced stereo transformer

    Weiyu Guo, Zhaoshuo Li, Yongkui Yang, Zheng Wang, Russell H Taylor, Mathias Unberath, Alan Yuille, and Yingwei Li. Context-enhanced stereo transformer. In ECCV, 2022

    2022

  2. [10]

    V olumetric propagation network: Stereo-lidar fusion for long-range depth estimation

    Jaesung Choe, Kyungdon Joo, Tooba Imtiaz, and In So Kweon. V olumetric propagation network: Stereo-lidar fusion for long-range depth estimation. RAL, 2021

    2021

  3. [11]

    Expanding sparse lidar depth and guiding stereo matching for robust dense depth estimation

    Zhenyu Xu, Yuehua Li, Shiqiang Zhu, and Yuxiang Sun. Expanding sparse lidar depth and guiding stereo matching for robust dense depth estimation. RAL, 2023

    2023

  4. [12]

    Stereo-lidar depth estimation with deformable propagation and learned disparity-depth conversion

    Ang Li, Anning Hu, Wei Xi, Wenxian Yu, and Danping Zou. Stereo-lidar depth estimation with deformable propagation and learned disparity-depth conversion. In ICRA, 2024

    2024

  5. [13]

    Sparse lidar assisted self-supervised stereo disparity estimation

    Xiaoming Zhao, Weihai Chen, Xingming Wu, Peter CY Chen, and Zhengguo Li. Sparse lidar assisted self-supervised stereo disparity estimation. In CCDC, 2022

    2022

  6. [14]

    Sparsity invariant cnns

    Jonas Uhrig, Nick Schneider, Lukas Schneider, Uwe Franke, Thomas Brox, and Andreas Geiger. Sparsity invariant cnns. In 3DV, 2017

    2017

  7. [15]

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

    Andreas Geiger, Philip Lenz, and Raquel Urtasun. Are we ready for autonomous driving? the kitti vision benchmark suite. In CVPR, 2012

    2012

  8. [16]

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

    Jason Ku, Ali Harakeh, and Steven L Waslander. In defense of classical image processing: Fast depth completion on the cpu. In CRV, 2018

    2018

  9. [17]

    Virtual kitti 2

    Yohann Cabon, Naila Murray, and Martin Humenberger. Virtual kitti 2. arXiv preprint arXiv:2001.10773, 2020

    2001 arXiv

  10. [18]

    Deep depth estimation from thermal image

    Ukcheol Shin, Jinsun Park, and In So Kweon. Deep depth estimation from thermal image. In CVPR, 2023

    2023

  11. [19]

    Non-parametric local transforms for computing visual corre- spondence

    Ramin Zabih and John Woodfill. Non-parametric local transforms for computing visual corre- spondence. In ECCV, 1994. 10

    1994

  12. [20]

    A constant-space belief propagation algorithm for stereo matching

    Qingxiong Yang, Liang Wang, and Narendra Ahuja. A constant-space belief propagation algorithm for stereo matching. In CVPR, 2010

    2010

  13. [21]

    Stereo correspondence by dynamic programming on a tree

    Olga Veksler. Stereo correspondence by dynamic programming on a tree. In CVPR, 2005

    2005

  14. [22]

    Stereo matching with color-weighted correlation, hierarchical belief propagation, and occlusion han- dling

    Qingxiong Yang, Liang Wang, Ruigang Yang, Henrik Stewénius, and David Nistér. Stereo matching with color-weighted correlation, hierarchical belief propagation, and occlusion han- dling. TPAMI, 2008

    2008

  15. [23]

    Stereo processing by semiglobal matching and mutual information.TPAMI, 2007

    Heiko Hirschmuller. Stereo processing by semiglobal matching and mutual information.TPAMI, 2007

    2007

  16. [24]

    Nerf-supervised deep stereo

    Fabio Tosi, Alessio Tonioni, Daniele De Gregorio, and Matteo Poggi. Nerf-supervised deep stereo. In CVPR, 2023

    2023

  17. [25]

    Domain generalized stereo matching via hierarchical visual transformation

    Tianyu Chang, Xun Yang, Tianzhu Zhang, and Meng Wang. Domain generalized stereo matching via hierarchical visual transformation. In CVPR, 2023

    2023

  18. [26]

    Croco v2: Improved cross-view completion pre-training for stereo matching and optical flow

    Philippe Weinzaepfel, Thomas Lucas, Vincent Leroy, Yohann Cabon, Vaibhav Arora, Romain Brégier, Gabriela Csurka, Leonid Antsfeld, Boris Chidlovskii, and Jérôme Revaud. Croco v2: Improved cross-view completion pre-training for stereo matching and optical flow. In ICCV, 2023

    2023

  19. [27]

    Graftnet: Towards domain generalized stereo matching with a broad-spectrum and task-oriented feature

    Biyang Liu, Huimin Yu, and Guodong Qi. Graftnet: Towards domain generalized stereo matching with a broad-spectrum and task-oriented feature. In CVPR, 2022

    2022

  20. [28]

    Uncertainty guided adaptive warping for robust and efficient stereo matching

    Junpeng Jing, Jiankun Li, Pengfei Xiong, Jiangyu Liu, Shuaicheng Liu, Yichen Guo, Xin Deng, Mai Xu, Lai Jiang, and Leonid Sigal. Uncertainty guided adaptive warping for robust and efficient stereo matching. In ICCV, 2023

    2023

  21. [29]

    Dps-net: Deep polarimetric stereo depth estimation

    Chaoran Tian, Weihong Pan, Zimo Wang, Mao Mao, Guofeng Zhang, Hujun Bao, Ping Tan, and Zhaopeng Cui. Dps-net: Deep polarimetric stereo depth estimation. In ICCV, 2023

    2023

  22. [30]

    Federated online adaptation for deep stereo

    Matteo Poggi and Fabio Tosi. Federated online adaptation for deep stereo. In CVPR, 2024

    2024

  23. [31]

    Accurate and efficient stereo matching via attention concatenation volume

    Gangwei Xu, Yun Wang, Junda Cheng, Jinhui Tang, and Xin Yang. Accurate and efficient stereo matching via attention concatenation volume. TPAMI, 2023

    2023

  24. [32]

    Active stereo without pattern projector

    Luca Bartolomei, Matteo Poggi, Fabio Tosi, Andrea Conti, and Stefano Mattoccia. Active stereo without pattern projector. In ICCV, 2023

    2023

  25. [33]

    Neural markov random field for stereo matching

    Tongfan Guan, Chen Wang, and Yun-Hui Liu. Neural markov random field for stereo matching. In CVPR, 2024

    2024

  26. [34]

    Openstereo: A comprehensive benchmark for stereo matching and strong baseline

    Xianda Guo, Chenming Zhang, Juntao Lu, Yiqi Wang, Yiqun Duan, Tian Yang, Zheng Zhu, and Long Chen. Openstereo: A comprehensive benchmark for stereo matching and strong baseline. arXiv preprint arXiv:2312.00343, 2023

    2023 arXiv

  27. [35]

    Segstereo: Exploiting semantic information for disparity estimation

    Guorun Yang, Hengshuang Zhao, Jianping Shi, Zhidong Deng, and Jiaya Jia. Segstereo: Exploiting semantic information for disparity estimation. In ECCV, 2018

    2018

  28. [36]

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

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. In ICLR, 2021

    2021

  29. [37]

    High-frequency stereo matching network

    Haoliang Zhao, Huizhou Zhou, Yongjun Zhang, Jie Chen, Yitong Yang, and Yong Zhao. High-frequency stereo matching network. In CVPR, 2023

    2023

  30. [38]

    Eai-stereo: Error aware iterative network for stereo matching

    Haoliang Zhao, Huizhou Zhou, Yongjun Zhang, Yong Zhao, Yitong Yang, and Ting Ouyang. Eai-stereo: Error aware iterative network for stereo matching. In ACCV, 2022

    2022

  31. [39]

    Parameterized cost volume for stereo matching

    Jiaxi Zeng, Chengtang Yao, Lidong Yu, Yuwei Wu, and Yunde Jia. Parameterized cost volume for stereo matching. In ICCV, 2023. 11

    2023

  32. [40]

    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, 2020

    2020

  33. [41]

    Noise-aware unsupervised deep lidar-stereo fusion

    Xuelian Cheng, Yiran Zhong, Yuchao Dai, Pan Ji, and Hongdong Li. Noise-aware unsupervised deep lidar-stereo fusion. In CVPR, 2019

    2019

  34. [42]

    Sparse lidar and stereo fusion (sls-fusion) for depth estimation and 3d object detection

    N-A-M Mai, Pierre Duthon, Louahdi Khoudour, Alain Crouzil, and Sergio A Velastin. Sparse lidar and stereo fusion (sls-fusion) for depth estimation and 3d object detection. In ICPRS, 2021

    2021

  35. [43]

    3d lidar and stereo fusion using stereo matching network with conditional cost volume normalization

    Tsun-Hsuan Wang, Hou-Ning Hu, Chieh Hubert Lin, Yi-Hsuan Tsai, Wei-Chen Chiu, and Min Sun. 3d lidar and stereo fusion using stereo matching network with conditional cost volume normalization. In IROS, 2019

    2019

  36. [44]

    Slfnet: A stereo and lidar fusion network for depth completion

    Yongjian Zhang, Longguang Wang, Kunhong Li, Zhiheng Fu, and Yulan Guo. Slfnet: A stereo and lidar fusion network for depth completion. RAL, 2022

    2022

  37. [45]

    Expansion of visual hints for improved generalization in stereo matching

    Andrea Pilzer, Yuxin Hou, Niki Loppi, Arno Solin, and Juho Kannala. Expansion of visual hints for improved generalization in stereo matching. In WACV, 2023

    2023

  38. [46]

    Guided stereo matching

    Matteo Poggi, Davide Pallotti, Fabio Tosi, and Stefano Mattoccia. Guided stereo matching. In CVPR, 2019

    2019

  39. [47]

    S3: Learnable sparse signal superdensity for guided depth estimation

    Yu-Kai Huang, Yueh-Cheng Liu, Tsung-Han Wu, Hung-Ting Su, Yu-Cheng Chang, Tsung-Lin Tsou, Yu-An Wang, and Winston H Hsu. S3: Learnable sparse signal superdensity for guided depth estimation. In CVPR, 2021

    2021

  40. [48]

    High-precision depth estimation with the 3d lidar and stereo fusion

    Kihong Park, Seungryong Kim, and Kwanghoon Sohn. High-precision depth estimation with the 3d lidar and stereo fusion. In ICRA, 2018

    2018

  41. [49]

    Listereo: Generate dense depth maps from lidar and stereo imagery

    Junming Zhang, Manikandasriram Srinivasan Ramanagopal, Ram Vasudevan, and Matthew Johnson-Roberson. Listereo: Generate dense depth maps from lidar and stereo imagery. In ICRA, 2020

    2020

  42. [50]

    Dfusenet: Deep fusion of rgb and sparse depth information for image guided dense depth completion

    Shreyas S Shivakumar, Ty Nguyen, Ian D Miller, Steven W Chen, Vijay Kumar, and Camillo J Taylor. Dfusenet: Deep fusion of rgb and sparse depth information for image guided dense depth completion. In ITSC, 2019

    2019

  43. [51]

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

    Jiaxiong Qiu, Zhaopeng Cui, Yinda Zhang, Xingdi Zhang, Shuaicheng Liu, Bing Zeng, and Marc Pollefeys. Deeplidar: Deep surface normal guided depth prediction for outdoor scene from sparse lidar data and single color image. In CVPR, 2019

    2019

  44. [52]

    Penet: Towards precise and efficient image guided depth completion

    Mu Hu, Shuling Wang, Bin Li, Shiyu Ning, Li Fan, and Xiaojin Gong. Penet: Towards precise and efficient image guided depth completion. In ICRA, 2021

    2021

  45. [53]

    Learning guided convolutional network for depth completion

    Jie Tang, Fei-Peng Tian, Wei Feng, Jian Li, and Ping Tan. Learning guided convolutional network for depth completion. TIP, 2020

    2020

  46. [54]

    Mff-net: Towards efficient monocular depth completion with multi-modal feature fusion

    Lina Liu, Xibin Song, Jiadai Sun, Xiaoyang Lyu, Lin Li, Yong Liu, and Liangjun Zhang. Mff-net: Towards efficient monocular depth completion with multi-modal feature fusion. RAL, 2023

    2023

  47. [55]

    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, 2020

    2020

  48. [56]

    Learning affinity via spatial propagation networks

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

    2017

  49. [57]

    Depth estimation via affinity learned with convolutional spatial propagation network

    Xinjing Cheng, Peng Wang, and Ruigang Yang. Depth estimation via affinity learned with convolutional spatial propagation network. In ECCV, 2018

    2018

  50. [58]

    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, 2023. 12

    2023

  51. [59]

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

    Lina Liu, Xibin Song, Xiaoyang Lyu, Junwei Diao, Mengmeng Wang, Yong Liu, and Liangjun Zhang. Fcfr-net: Feature fusion based coarse-to-fine residual learning for depth completion. In AAAI, 2021

    2021

  52. [60]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR, 2016

    2016

  53. [61]

    Pseudo-lidar++: Accurate depth for 3d object detection in autonomous driving

    Yurong You, Yan Wang, Wei-Lun Chao, Divyansh Garg, Geoff Pleiss, Bharath Hariharan, Mark Campbell, and Kilian Q Weinberger. Pseudo-lidar++: Accurate depth for 3d object detection in autonomous driving. ICLR, 2020

    2020

  54. [62]

    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, 2024

    2024

  55. [63]

    A surface geometry model for lidar depth completion

    Yiming Zhao, Lin Bai, Ziming Zhang, and Xinming Huang. A surface geometry model for lidar depth completion. RAL, 2021

    2021

  56. [64]

    Fpga accelerated real-time recurrent all-pairs field transforms for optical flow

    Yingxiang Li, Yingke Gao, Zhiwen Su, Shitao Chen, and Longjun Liu. Fpga accelerated real-time recurrent all-pairs field transforms for optical flow. In CAC, 2022

    2022

  57. [65]

    Scene01

    Ye Liu, Shuang Hao, Kun Huang, Minghui Yang, Zili Huang, Xiuyuan Qi, Yiting Li, Liang Zhou, Yu Long, and Jun Zhou. An fpga-based ultra-high performance and scalable optical flow hardware accelerator for autonomous driving. In ISCAS, 2024. 13 Supplementary Material for Leveragi...

    2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.