Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Boosting Zero-shot Stereo Matching using Large-scale Mixed Images Sources in the Real World

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read BooSTer pre-trains stereo matching on synthetic, real, and single-view images, cutting zero-shot D1 error on KITTI and ETH3D by roughly a third to a half.

desk verdict A promising data-generation recipe and a large new dataset, but the zero-shot claims are not yet fairly demonstrated because the comparisons conflate data scale, teacher quality, and architecture. read the letter →

arxiv 2505.08607 v1 pith:4CKLGZUQ submitted 2025-05-13 cs.CV

Yuran Wang , Yingping Liang , Ying Fu This is my paper · ORCID
classification cs.CV
keywords stereomatchingzero-shotgeneralizationmonoculardepthestimationdatagenerationdiffusioninpaintingpseudo-labelsupervisionvisionfoundationmodelsdomaintransfer
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

Stereo matching needs dense pixel-wise labels, but real-world ground truth is sparse and expensive, and models trained on synthetic scenes do not transfer well outdoors. This paper argues that the bottleneck is not the stereo architecture but the training data and features: pre-training on a mixture of synthetic SceneFlow frames, real DrivingStereo frames, and stereo pairs synthesized from ordinary single-view images can close much of that domain gap. The synthesis pipeline uses monocular depth from a foundation model to produce pseudo-disparity, forward-warps the right view, and fills occlusion holes with an edge-aware diffusion inpainter; the real frames are supervised with a dynamic scale- and shift-invariant loss over monocular pseudo-depth. With a hybrid DINOv2-VGG19 encoder, the resulting BooSTer model reports D1 of 3.04 on KITTI12, 3.22 on KITTI15, 0.70 on ETH3D, and 7.50 on Middlebury in zero-shot evaluation, well below prior SceneFlow-trained baselines. If correct, this makes large-scale unlabeled single-view imagery a viable source of stereo supervision.

What carries the argument

The machinery is a three-part training recipe plus a hybrid encoder. Monocular depth maps from Depth Anything V2 are randomly scaled ($D'_{\mathrm{mono}} = \alpha D_{\mathrm{mono}}$) and forward-warped to form the right view; an edge-aware inpainting module, which thresholds the horizontal gradient of the disparity map to build an edge mask, then uses Stable Diffusion to fill occlusion holes without blending foreground and background. Real DrivingStereo frames are supervised with a dynamic scale- and shift-invariant (DSSI) loss: a least-squares fit $(a,b)$ aligns the stereo prediction to the monocular pseudo-depth, outliers beyond a quintile threshold are masked, and the fit is recomputed before an MSE loss. The hybrid feature encoder concatenates DINOv2's high-level semantic features at one-thirty-second scale with VGG19's multi-scale pyramid features, feeding a StereoBase-style cost volume and refinement network. Each component is ablated in the paper.

What would settle it

Compute zero-shot error on a dense-ground-truth real dataset separately for pixels where the monocular depth teacher is accurate and where it is wrong (e.g., relative error above 10%): if the stereo model's D1 on teacher-wrong pixels stays as high as the teacher's error, the gains are inherited from the teacher rather than from stereo geometry.

Watch

Extended reading notes

Core claim

The paper's central claim is that zero-shot stereo matching can be made substantially more accurate by controlling what data a model pre-trains on and how its encoder is built. BooSTer pre-trains on a 5:6:1 mix of SceneFlow synthetic frames, generated DiffMFS pairs from 597,727 single-view images, and real DrivingStereo frames. DiffMFS is produced by taking monocular relative depth from Depth Anything V2, scaling it randomly, forward-warping the right image, and filling holes with an edge-aware Stable Diffusion inpainter. The real-world branch adds monocular pseudo-labels through a dynamic scale- and shift-invariant loss, and the encoder combines DINOv2 with VGG19. On held-out KITTI12, KITTI15, ETH3D, and Middlebury, the trained model reduces D1 substantially relative to the StereoBase baseline, for example 3.04 vs 4.98 on KITTI12, 3.22 vs 5.47 on KITTI15, and 0.70 vs 1.28 on ETH3D.

Load-bearing premise

The whole approach rests on the assumption that a single-image depth model's relative depth estimates are reliable enough, after random scaling and warping, to act as true geometry for synthetic stereo pairs and as pseudo-labels for real images; where that teacher is wrong, the stereo model will be wrong too.

Editorial extensions

If this is right

  • Monocular depth models can be repurposed as dense stereo supervisors, reducing dependence on LiDAR ground truth.
  • Large single-view image collections become usable stereo training data at negligible annotation cost.
  • Zero-shot accuracy on KITTI and ETH3D improves by roughly a third to a half in D1 relative to SceneFlow-only training, making cross-dataset deployment more practical.
  • The hybrid VFM-CNN encoder and the edge-aware inpainting each contribute, with the data mixture giving the largest gain according to the paper's ablations.

Reading between the lines

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

  • Implicit in the paper, and testable, is that the monocular teacher's accuracy sets the ceiling for the synthesized and pseudo-labeled data; corrupting the teacher should measurably degrade the stereo model on the affected surface types.
  • Because the DSSI loss aligns any stereo prediction to monocular depth up to an affine transform, the same mechanism could plausibly extend to unlabeled stereo video or self-training loops with uncertainty weighting.
  • The 5:6:1 sampling ratio and the random scaling range are choices rather than derived quantities, so per-target-domain tuning of these hyperparameters could yield further gains beyond the reported numbers.
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 / 5 minor

Summary. The paper proposes BooSTer, a stereo matching framework trained on a mixed dataset of synthetic SceneFlow data, real-world DrivingStereo data, and DiffMFS, a newly generated dataset of stereo pairs synthesized from large-scale single-view images. The synthesis pipeline uses Depth Anything V2 to produce monocular disparity, random scaling to obtain pixel disparity, forward warping to create a right view, and Stable Diffusion inpainting with an edge-aware mask to fill occlusion holes. For real stereo data with sparse LiDAR labels, the authors add a dynamic scale- and shift-invariant (DSSI) loss that aligns the network prediction to Depth Anything V2 pseudo-disparity. The architecture replaces the standard encoder with a hybrid DINOv2-VGG19 encoder. The paper reports zero-shot gains on KITTI 2012, KITTI 2015, ETH3D, and Middlebury, with KITTI12 D1 improving from 4.98 (StereoBase) to 3.04 and ETH3D D1 from 1.28 to 0.70. Ablations in Table 2 compare SceneFlow, MFS, and DiffMFS under the same network architectures, and Table 3 ablates data mixture, inpainting, and encoder choices.

Significance. If the reported gains survive a fully controlled comparison, the paper makes a useful contribution to zero-shot stereo matching: it demonstrates that large-scale single-view data can be converted into stereo training pairs at scale, that real-world pseudo-labels from a monocular teacher can supplement sparse LiDAR, and that a frozen VFM encoder can improve transfer. The cross-domain evaluation on externally acquired LiDAR ground truth means the headline result is not circular. The same-architecture comparison in Table 2 is a genuine strength and provides support for the data-generation claim, as does the ablation structure in Table 3. The main unresolved risk is that all non-SceneFlow supervision derives from a single monocular teacher, so the method's ceiling in hard regions may be set by the teacher and the paper does not quantify that ceiling.

major comments (3)
  1. [Section 4.3, Table 1] The headline comparison in Table 1 is not controlled: 'Ours' simultaneously changes the training data (mixed dataset vs. SceneFlow-only) and the encoder (DINOv2-VGG19 hybrid vs. the baselines' standard encoders), so the reported gains over StereoBase and IGEV conflate the data-generation contribution with the architecture contribution. The controlled comparison in Table 2 is better evidence, but it does not repair Table 1's headline claim. Please add an experiment that trains StereoBase or IGEV on the full mixed dataset with the baseline encoder, and also report a version of BooSTer trained on SceneFlow only, so the two factors can be separated.
  2. [Sections 3.2-3.3 and 4.3] The paper does not separate monocular-teacher error from stereo-model error. DiffMFS supervision is derived from D'_mono = alpha D_mono (Eq. 4), and DrivingStereo supervision uses an affine fit to the same D_mono (Eqs. 6-9); both use Depth Anything V2. Wherever that teacher is systematically wrong in textureless, transparent, or reflective regions, the generated data and pseudo-labels are wrong in exactly those regions, yet Section 4.3 claims improvements in such regions. A KITTI12 D1 of 3.04 could in principle be approached by a model that copies the aligned teacher's output; the paper does not report the aligned teacher's own D1 on KITTI/ETH3D or provide a per-region error decomposition. Please add an experiment that aligns Depth Anything V2 predictions to the LiDAR ground truth with the same scale/shift procedure and reports its D1, EPE, and per-region errors on the evaluation sets, and compare those numbers against BooSTer's errors.
  3. [Section 4.4, Table 2] The comparison between MFS Dataset and DiffMFS in Table 2 does not isolate the proposed data-generation components because the two pipelines use different monocular depth models and different inpainting strategies. MFS uses the original Mono-for-Stereo pipeline, while DiffMFS uses Depth Anything V2 plus a Stable Diffusion and edge-aware inpainting; the gains could come from the stronger teacher rather than from the edge-aware inpainting or scaling scheme. Please include an ablation that keeps the teacher and inpainting fixed and varies only the proposed component, or at least report the teacher model used for the MFS baseline and quantify its depth accuracy on a common validation set.
minor comments (5)
  1. [Section 4.2, Implementation Details] Several hyperparameters that directly control the generated data and loss are not reported, including the sampling range [dmin, dmax] for alpha in Eq. 4, the edge threshold tau in Eq. 5, the DSSI loss weight beta in Eq. 10, the definition of the 'Quintiles operation' in Eq. 8, and the number of DrivingStereo images used with the 5:6:1 sampling ratio. Please state these values for reproducibility.
  2. [Conclusion] The conclusion contains a typo: 'scenting an advancement' should be 'cementing an advancement' or 'representing an advancement'.
  3. [Table 3] The Table 3 header 'Stable-Diffsuion' is a typo for 'Stable-Diffusion'.
  4. [Section 3.1, Eq. 2] The notation for the valid mask is inconsistent: the text defines Mval as the valid mask, but the equation writes Mval ⊙ (D_GT − D_pred) and the surrounding text also uses M without a subscript. Please unify the notation.
  5. [General] The paper does not mention a code or dataset release plan. Given that the contribution is a training data pipeline and architecture, releasing the DiffMFS generation code and the trained checkpoints would substantially strengthen reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central zero-shot claim is validated on external LiDAR benchmarks, and the monocular-teacher training signal is not a fitted prediction of those benchmarks.

full rationale

The paper's headline result is cross-domain zero-shot performance on KITTI 2012, KITTI 2015, ETH3D, and Middlebury, all of which use externally measured LiDAR or depth-camera ground truth (Section 4.1). None of these benchmark labels are used to fit the model or to generate DiffMFS training pairs, so the central comparison in Table 1 does not reduce by construction. The DiffMFS pipeline builds right views by forward warping a randomly scaled monocular depth map (Eq. 4), which makes the generated training pairs geometrically self-consistent by definition, but that is a data-augmentation strategy rather than a circular validation. Similarly, the DSSI loss (Eqs. 6-9) fits a scale and shift to align the stereo prediction to monocular pseudo-depth during training; the final reported disparity maps are the raw stereo predictions, not the aligned pseudo-labels, so the loss is not being renamed as a benchmark prediction. The one self-citation, [Wang et al., 2024b], appears in a related-work list of dataset-generation approaches and is not load-bearing; it is not used to justify the central design or to exclude alternatives. The concern that systematic monocular-teacher errors can be baked into the generated data and pseudo-labels is a real correctness and robustness risk, but it is not circularity because the evaluation is externally grounded. Thus no circular step can be exhibited from the paper's own equations or citation chain.

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

The method does not introduce new physical entities. Its load-bearing assumptions are the reliability of a monocular depth teacher for generating data and pseudo-labels, the geometric validity of warped and inpainted pairs, the sufficiency of scale-shift alignment for supervision, the transferability of DINOv2 features to stereo cost volumes, and the independence of training-split evaluations. The central free parameters are alpha, tau, beta, the outlier quantile, and the data sampling ratio; none are specified with values or ablated thoroughly.

free parameters (5)
  • alpha sampling range [dmin, dmax] = not specified
    Eq. 4 samples alpha from U(dmin,dmax) to convert monocular relative disparity to pixel disparity. The range is a hand-chosen hyperparameter that determines all disparity magnitudes in DiffMFS; the paper never states its values.
  • tau edge threshold = not specified
    Eq. 5 binarizes horizontal disparity gradients into an object edge mask; tau controls the number of edge pixels used in edge-aware inpainting and is stated only as a threshold without value.
  • beta DSSI loss weight = not specified
    Eq. 10 combines sparse LiDAR loss and DSSI pseudo-label loss; beta is a hand-set balancing weight and no ablation across beta is reported.
  • outlier threshold quantile in Q() = not specified
    Eq. 8 uses a quintile operation Q to derive tau_l; the exact quantile is ambiguous and not specified.
  • mixed dataset sampling ratio 5:6:1 = 5:6:1
    Section 4.1: SceneFlow, DiffMFS, and DrivingStereo are sampled with frequency 5:6:1 per epoch; this ratio is chosen by hand and not ablated.
assumptions (5)
  • domain assumption Depth Anything V2 estimates are an adequate proxy for stereo disparity up to scale and shift
    Section 3.2 Eq. 4 and Section 3.3 Eqs. 6 to 9 use Depth Anything V2 disparity as ground truth for DiffMFS and as pseudo-labels for DrivingStereo; the method inherits all systematic monocular depth errors.
  • domain assumption Forward-warped right views with diffusion inpainting form valid stereo pairs
    Section 3.2; the training signal is only as good as the geometric consistency of the generated pairs, and inpainting can erase or distort objects near holes.
  • domain assumption Least-squares scale and shift alignment with quantile-based outlier removal reliably aligns stereo predictions to monocular pseudo-depth
    Section 3.3 Eqs. 6 to 9; outliers are removed with a quintile rule that is not fully specified, and a poor alignment would corrupt supervision.
  • domain assumption DINOv2 features improve stereo matching when concatenated at 1/32 resolution
    Section 3.4; no ablation isolates the encoder contribution on all four benchmarks, only a single KITTI15 comparison in Table 3.
  • domain assumption Training-set metrics on KITTI, ETH3D and Middlebury measure zero-shot generalization
    Section 4.1; if ETH3D was used as a validation set during development, its numbers are not independent.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Boosting Zero-shot Stereo Matching using Large-scale Mixed Images Sources in the Real World." pith.science (2026). https://pith.science/paper/4CKLGZUQ

@misc{pith2026250508607,
  author       = {Pith},
  title        = {Pith review of: Boosting Zero-shot Stereo Matching using Large-scale Mixed Images Sources in the Real World},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4CKLGZUQ}},
  note         = {Machine review of arXiv:2505.08607}
}
read the original abstract

Stereo matching methods rely on dense pixel-wise ground truth labels, which are laborious to obtain, especially for real-world datasets. The scarcity of labeled data and domain gaps between synthetic and real-world images also pose notable challenges. In this paper, we propose a novel framework, \textbf{BooSTer}, that leverages both vision foundation models and large-scale mixed image sources, including synthetic, real, and single-view images. First, to fully unleash the potential of large-scale single-view images, we design a data generation strategy combining monocular depth estimation and diffusion models to generate dense stereo matching data from single-view images. Second, to tackle sparse labels in real-world datasets, we transfer knowledge from monocular depth estimation models, using pseudo-mono depth labels and a dynamic scale- and shift-invariant loss for additional supervision. Furthermore, we incorporate vision foundation model as an encoder to extract robust and transferable features, boosting accuracy and generalization. Extensive experiments on benchmark datasets demonstrate the effectiveness of our approach, achieving significant improvements in accuracy over existing methods, particularly in scenarios with limited labeled data and domain shifts.

Figures

Figures reproduced from arXiv: 2505.08607 by the authors.

Figure 1
Figure 1. Comparison between Monocular Depth and Stereo Match [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall architecture of our proposed method consists of two main parts. 1. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Comparison between naive inpainting module and our [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Qualitative results of IGEV and StereoBase trained with Sceneflow and our model trained on our mixed dataset. By default, IGEV [PITH_FULL_IMAGE:figures/full_fig_p006_5.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. AquaStereo: Enabling Underwater Stereo Matching via Depth-Conditioned Diffusion and Geometry Self-Distillation

    cs.CV 2026-07 conditional novelty 6.0 of 10

    Depth-conditioned diffusion, geometry self-distillation, and perception frames yield strong zero-shot underwater stereo matching without real underwater labels.

Reference graph

Works this paper leans on

39 extracted references · 34 canonical work pages · cited by 1 Pith paper

  1. [1]

    Correlate-and-excite: Real-time stereo matching via guided cost volume excitation

    [Bangunharcana et al., 2021] Antyanta Bangunharcana, Jae Won Cho, Seokju Lee, In So Kweon, Kyung-Soo Kim, and Soohyun Kim. Correlate-and-excite: Real-time stereo matching via guided cost volume excitation. In 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 3542–3548,

    work page 2021

  2. [5]

    Vision transformers need registers,

    [Darcet et al., 2023] Timoth´ee Darcet, Maxime Oquab, Julien Mairal, and Piotr Bojanowski. Vision transformers need registers,

    work page 2023

  3. [10]

    Lidar-based person re-identification

    [Guo et al., 2024] Wenxuan Guo, Zhiyu Pan, Yingping Liang, Ziheng Xi, Zhicheng Zhong, Jianjiang Feng, and Jie Zhou. Lidar-based person re-identification. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17437–17447,

    work page 2024

  4. [13]

    Microsoft coco: Com- mon objects in context

    [Lin et al., 2014] Tsung-Yi Lin, Michael Maire, Serge Be- longie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Com- mon objects in context. In The European Conference on Computer Vision (ECCV), pages 740–755,

    work page 2014

  5. [15]

    A large dataset to train con- volutional networks for disparity, optical flow, and scene flow estimation

    [Mayer et al., 2016] Nikolaus Mayer, Eddy Ilg, Philip Hausser, Philipp Fischer, Daniel Cremers, Alexey Doso- vitskiy, and Thomas Brox. A large dataset to train con- volutional networks for disparity, optical flow, and scene flow estimation. In the IEEE conference on Computer Vi- sion and Pattern Recognition (CVPR) , pages 4040–4048,

    work page 2016

  6. [16]

    Object scene flow for autonomous vehicles

    [Menze and Geiger, 2015] Moritz Menze and Andreas Geiger. Object scene flow for autonomous vehicles. In the IEEE conference on Computer Vision and Pattern Recognition (CVPR), pages 3061–3070,

    work page 2015

  7. [18]

    [Oquab et al., 2023] Maxime Oquab, Timoth´ee Darcet, Theo Moutakanni, Huy V . V o, Marc Szafraniec, Vasil Khali- dov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, Russell Howes, Po-Yao Huang, Hu Xu, Vasu Sharma, Shang-Wen Li, Wojciech Galuba, Mike Rabbat, Mido Assran, Nicolas Ballas, Gabriel Syn- naeve, Ishan Misra, Herve Jegou, J...

    work page 2023

  8. [19]

    Slam2: an open- source slam system for monocular, stereo, and rgb-d cam- eras

    [Orb, 2017] Mur-Artal R Tard ´os JD Orb. Slam2: an open- source slam system for monocular, stereo, and rgb-d cam- eras. IEEE Transactions on Robotics , 33(5):1255–1262,

    work page 2017

Show all 39 references
  1. [20]

    Learning transferable visual models from nat- ural language supervision

    [Radford et al., 2021] Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agar- wal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from nat- ural language supervision. In International confere...

    2021

  2. [21]

    Sam 2: Seg- ment anything in images and videos

    [Ravi et al., 2024] Nikhila Ravi, Valentin Gabeur, Yuan- Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman R ¨adle, Chloe Rolland, Laura Gustafson, Eric Mintun, Junting Pan, Kalyan Vasudev Al- wala, Nicolas Carion, Chao-Yuan Wu, Ross Girshick, Pi- otr Doll...

    2024 arXiv

  3. [22]

    High-resolution stereo datasets with subpixel-accurate ground truth

    [Scharstein et al., 2014] Daniel Scharstein, Heiko Hirschm¨uller, York Kitajima, Greg Krathwohl, Nera Neˇsi´c, Xi Wang, and Porter Westling. High-resolution stereo datasets with subpixel-accurate ground truth. In Pattern Recognition: 36th German Conference, GCPR 2014, M ¨unste...

    2014

  4. [23]

    A multi-view stereo benchmark with high-resolution images and multi-camera videos

    [Schops et al., 2017] Thomas Schops, Johannes L Schon- berger, Silvano Galliani, Torsten Sattler, Konrad Schindler, Marc Pollefeys, and Andreas Geiger. A multi-view stereo benchmark with high-resolution images and multi-camera videos. In the IEEE conference on Computer Vision ...

    2017

  5. [24]

    Cfnet: Cascade and fused cost volume for robust stereo matching

    [Shen et al., 2021] Zhelun Shen, Yuchao Dai, and Zhibo Rao. Cfnet: Cascade and fused cost volume for robust stereo matching. In the IEEE conference on Computer Vision and Pattern Recognition (CVPR) , pages 13906– 13915,

    2021

  6. [25]

    Simonyan and A

    [Simonyan and Zisserman, 2015] K. Simonyan and A. Zis- serman. Very deep convolutional networks for large-scale image recognition. In International Conference on Learn- ing Representations (ICLR), May

    2015

  7. [26]

    Diode: A dense indoor and outdoor depth dataset

    [Vasiljevic et al., 2019] Igor Vasiljevic, Nick Kolkin, Shanyi Zhang, Ruotian Luo, Haochen Wang, Falcon Z Dai, An- drea F Daniele, Mohammadreza Mostajabi, Steven Basart, Matthew R Walter, et al. Diode: A dense indoor and outdoor depth dataset. arXiv preprint arXiv:1908.00463 ,

    2019 arXiv

  8. [27]

    Fadnet++: Real-time and accurate disparity estimation with config- urable networks

    [Wang et al., 2021a] Qiang Wang, Shaohuai Shi, Shizhen Zheng, Kaiyong Zhao, and Xiaowen Chu. Fadnet++: Real-time and accurate disparity estimation with config- urable networks. arXiv preprint arXiv:2110.02582,

    arXiv

  9. [28]

    Mono2stereo: Monocular knowledge transfer for enhanced stereo matching

    [Wang et al., 2024b] Yuran Wang, Yingping Liang, Hesong Li, and Ying Fu. Mono2stereo: Monocular knowledge transfer for enhanced stereo matching. arXiv preprint arXiv:2411.09151,

    arXiv

  10. [29]

    Learning stereo from single images

    [Watson et al., 2020] Jamie Watson, Oisin Mac Aodha, Daniyar Turmukhambetov, Gabriel J Brostow, and Michael Firman. Learning stereo from single images. In The European Conference on Computer Vision (ECCV) , pages 722–740,

    2020

  11. [30]

    Structure-guided ranking loss for single image depth prediction

    [Xian et al., 2020] Ke Xian, Jianming Zhang, Oliver Wang, Long Mai, Zhe Lin, and Zhiguo Cao. Structure-guided ranking loss for single image depth prediction. Inthe IEEE conference on Computer Vision and Pattern Recognition (CVPR), pages 611–620,

    2020

  12. [31]

    Iterative geometry encoding volume for stereo matching

    [Xu et al., 2023] Gangwei Xu, Xianqi Wang, Xiaohuan Ding, and Xin Yang. Iterative geometry encoding volume for stereo matching. In the IEEE conference on Computer Vision and Pattern Recognition (CVPR) , pages 21919– 21928,

    2023

  13. [32]

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

    [Yang et al., 2024] Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth anything: Unleashing the power of large-scale unlabeled data. In the IEEE conference on Computer Vision and Pat- tern Recognition (CVPR), pages 10371–10381,

    2024

  14. [33]

    Building a 3-d line-based map using stereo slam

    [Zhang et al., 2015] Guoxuan Zhang, Jin Han Lee, Jong- woo Lim, and Il Hong Suh. Building a 3-d line-based map using stereo slam. IEEE Transactions on Robotics , 31(6):1364–1377,

    2015

  15. [34]

    Ga-net: Guided ag- gregation net for end-to-end stereo matching

    [Zhang et al., 2019] Feihu Zhang, Victor Prisacariu, Ruigang Yang, and Philip HS Torr. Ga-net: Guided ag- gregation net for end-to-end stereo matching. In the IEEE conference on Computer Vision and Pattern Recognition (CVPR), pages 185–194,

    2019

  16. [35]

    Learning temporal consistency for low light video enhancement from single images

    [Zhang et al., 2021] Fan Zhang, Yu Li, Shaodi You, and Ying Fu. Learning temporal consistency for low light video enhancement from single images. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 4967–4976,

    2021

  17. [36]

    Learning rain location prior for nighttime derain- ing

    [Zhang et al., 2023] Fan Zhang, Shaodi You, Yu Li, and Ying Fu. Learning rain location prior for nighttime derain- ing. In Proceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 13148–13157,

    2023

  18. [37]

    Atlantis: Enabling underwater depth estimation with stable diffusion

    [Zhang et al., 2024] Fan Zhang, Shaodi You, Yu Li, and Ying Fu. Atlantis: Enabling underwater depth estimation with stable diffusion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 11852–11861,

    2024

  19. [38]

    Unaligned rgb guided hyperspectral image super-resolution with spatial- spectral concordance,

    [Zhang et al., 2025] Yingkai Zhang, Zeqiang Lai, Tao Zhang, Ying Fu, and Chenghu Zhou. Unaligned rgb guided hyperspectral image super-resolution with spatial- spectral concordance,

    2025

  20. [2012]

    Cascade cost volume for high-resolution multi-view stereo and stereo matching

    [Gu et al., 2020] Xiaodong Gu, Zhiwen Fan, Siyu Zhu, Zuozhuo Dai, Feitong Tan, and Ping Tan. Cascade cost volume for high-resolution multi-view stereo and stereo matching. In the IEEE conference on Computer Vision and Pattern Recognition (CVPR), June

    2020

  21. [2014]

    Raft-stereo: Multilevel recurrent field transforms for stereo matching

    [Lipson et al., 2021] Lahav Lipson, Zachary Teed, and Jia Deng. Raft-stereo: Multilevel recurrent field transforms for stereo matching. In 2021 International Conference on 3D Vision (3DV), pages 218–227. IEEE,

    2021

  22. [2015]

    The mapillary vistas dataset for semantic understanding of street scenes

    [Neuhold et al., 2017] Gerhard Neuhold, Tobias Ollmann, Samuel Rota Bulo, and Peter Kontschieder. The mapillary vistas dataset for semantic understanding of street scenes. In the IEEE International Conference on Computer Vision (ICCV), pages 4990–4999,

    2017

  23. [2016]

    Diml/cvl rgb-d dataset: 2m rgb-d images of natural indoor and outdoor scenes

    [Cho et al., 2021] Jaehoon Cho, Dongbo Min, Youngjung Kim, and Kwanghoon Sohn. Diml/cvl rgb-d dataset: 2m rgb-d images of natural indoor and outdoor scenes. arXiv preprint arXiv:2110.11590,

    2021 arXiv

  24. [2017]

    Mpi-flow: Learning realistic opti- cal flow with multiplane images

    [Liang et al., 2023] Yingping Liang, Jiaming Liu, Debing Zhang, and Ying Fu. Mpi-flow: Learning realistic opti- cal flow with multiplane images. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 13857–13868,

    2023

  25. [2018]

    Single-image depth perception in the wild

    [Chen et al., 2016] Weifeng Chen, Zhao Fu, Dawei Yang, and Jia Deng. Single-image depth perception in the wild. Advances in Neural Information Processing Systems (NeurIPS), 29,

    2016

  26. [2019]

    Openstereo: A comprehensive benchmark for stereo matching and strong baseline

    [Guo et al., 2023] Xianda Guo, Juntao Lu, Chenming Zhang, 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 arXiv

  27. [2020]

    Group-wise correla- tion stereo network

    [Guo et al., 2019] Xiaoyang Guo, Kai Yang, Wukui Yang, Xiaogang Wang, and Hongsheng Li. Group-wise correla- tion stereo network. In the IEEE conference on Computer Vision and Pattern Recognition (CVPR) , pages 3273– 3282,

    2019

  28. [2021]

    Pyramid stereo matching network

    [Chang and Chen, 2018] Jia-Ren Chang and Yong-Sheng Chen. Pyramid stereo matching network. In the IEEE conference on Computer Vision and Pattern Recognition (CVPR), pages 5410–5418,

    2018

  29. [2023]

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

    [Geiger et al., 2012] Andreas Geiger, Philip Lenz, and Raquel Urtasun. Are we ready for autonomous driving? the kitti vision benchmark suite. Inthe IEEE conference on Computer Vision and Pattern Recognition (CVPR) , pages 3354–3361,

    2012

  30. [2024]

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

    [Kendall et al., 2017] Alex Kendall, Hayk Martirosyan, Saumitro Dasgupta, Peter Henry, Ryan Kennedy, Abraham Bachrach, and Adam Bry. End-to-end learning of geom- etry and context for deep stereo regression. In the IEEE conference on Computer Vision and Pattern Recognition (CVP...

    2017

  31. [2025]

    Scene parsing through ade20k dataset

    [Zhou et al., 2017] Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene parsing through ade20k dataset. Inthe IEEE conference on Computer Vision and Pattern Recognition (CVPR) , pages 633–641, 2017

    2017

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.