pith. sign in

arxiv: 2506.13215 · v2 · submitted 2025-06-16 · 💻 cs.CV

DVP-MVS++: Synergize Depth-Normal-Edge and Harmonized Visibility Prior for Multi-View Stereo

Zhenlong Yuan , Dapeng Zhang , Zehao Li , Chengxuan Qian , Jianing Chen , Yinda Chen , Kehua Chen , Tianlu Mao
show 3 more authors
Zhaoxin Li Hao Jiang Zhaoqi Wang
This is my paper

Pith reviewed 2026-05-19 09:49 UTC · model grok-4.3

classification 💻 cs.CV
keywords multi-view stereopatch deformationdepth mapsedge alignmentvisibility priors3D reconstructiongeometry consistency
0
0 comments X

The pith

Aligning coarse depth normal and edge maps with harmonized visibility priors enables stable patch deformation for multi-view stereo.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that patch deformation in multi-view stereo can be made more reliable by tackling two main sources of error: patches skipping over edges and failures to account for visibility differences across views. It generates coarse depth and normal maps from existing estimators along with edge maps from the Roberts operator, then aligns them through erosion and dilation to create consistent boundaries. View selection is turned into visibility maps that support cross-view depth reprojection and area maximization to keep patches balanced and focused on visible regions. Geometry consistency is enforced with aggregated normals and depth differences, plus highlight correction, so that the final reconstructions avoid deviations in difficult areas.

Core claim

DVP-MVS++ produces coarse depth maps with DepthPro and Metric3Dv2, normal maps, and Roberts edge maps, then aligns these via erosion-dilation to yield fine-grained homogeneous boundaries that support robust patch deformation. View selection weights are recast as visibility maps, and an enhanced cross-view depth reprojection together with an area-maximization strategy supplies harmonized priors that restore visible areas and balance deformed patches. Aggregated normals from view selection and projection depth differences along epipolar lines establish geometry consistency, while SHIQ performs highlight correction to add highlight-aware perception during propagation and refinement.

What carries the argument

Depth-normal-edge alignment through erosion-dilation combined with reformulated visibility maps that drive cross-view depth reprojection and area maximization for visibility-aware patch deformation.

Load-bearing premise

Coarse depth maps from DepthPro and Metric3Dv2 together with Roberts edges, once aligned by erosion-dilation, produce fine-grained homogeneous boundaries that reliably prevent edge-skipping during patch deformation.

What would settle it

Re-running the method on the ETH3D or Tanks & Temples test sets and finding that edge-skipping artifacts remain visible in the output meshes or that accuracy in occluded regions does not exceed prior patch-deformation baselines would indicate the alignment and visibility steps are not delivering the claimed stability.

Figures

Figures reproduced from arXiv: 2506.13215 by Chengxuan Qian, Dapeng Zhang, Hao Jiang, Jianing Chen, Kehua Chen, Tianlu Mao, Yinda Chen, Zehao Li, Zhaoqi Wang, Zhaoxin Li, Zhenlong Yuan.

Figure 1
Figure 1. Figure 1: Comparison between traditional patch deformation-based methods [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An illustrated pipeline of DVP-MVS++. Specifically, we first introduce DepthPro, Metric3Dv2 and Roberts operator to respectively obtain depth, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Depth-Normal-Edge Aligned Prior. In (b) and (c), areas with lower color temperature show smaller depths, while higher color temperatures indicate [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Cross-View Prior. The red line in (a) separates visible and invisible [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Highlight-Aware Geometry-Driven Propagation and Refinement. In (a) and (b), the blue view cone corresponds to the reference image [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Highlight correction. By removing the detected highlight area (b) of the original image (a), we acquire the corresponding correction image (c). [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Visualized point cloud results between different methods on partial scenes of ETH3D datasets ( [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Reconstructed point clouds of our method on Tanks & Temples dataset without any fine-tuning. [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
read the original abstract

Recently, patch deformation-based methods have demonstrated significant effectiveness in multi-view stereo due to their incorporation of deformable and expandable perception for reconstructing textureless areas. However, these methods generally focus on identifying reliable pixel correlations to mitigate matching ambiguity of patch deformation, while neglecting the deformation instability caused by edge-skipping and visibility occlusions, which may cause potential estimation deviations. To address these issues, we propose DVP-MVS++, an innovative approach that synergizes both depth-normal-edge aligned and harmonized cross-view priors for robust and visibility-aware patch deformation. Specifically, to avoid edge-skipping, we first apply DepthPro, Metric3Dv2 and Roberts operator to generate coarse depth maps, normal maps and edge maps, respectively. These maps are then aligned via an erosion-dilation strategy to produce fine-grained homogeneous boundaries for facilitating robust patch deformation. Moreover, we reformulate view selection weights as visibility maps, and then implement both an enhanced cross-view depth reprojection and an area-maximization strategy to help reliably restore visible areas and effectively balance deformed patch, thus acquiring harmonized cross-view priors for visibility-aware patch deformation. Additionally, we obtain geometry consistency by adopting both aggregated normals via view selection and projection depth differences via epipolar lines, and then employ SHIQ for highlight correction to enable geometry consistency with highlight-aware perception, thus improving reconstruction quality during propagation and refinement stage. Evaluation results on ETH3D, Tanks & Temples and Strecha datasets exhibit the state-of-the-art performance and robust generalization capability of our proposed method.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript proposes DVP-MVS++, a patch-deformation MVS pipeline that first generates coarse depth/normal maps with DepthPro and Metric3Dv2 plus Roberts edges, aligns them via erosion-dilation to create homogeneous boundaries, reformulates view-selection weights as visibility maps, applies enhanced cross-view depth reprojection and area-maximization for harmonized priors, and uses aggregated normals, epipolar depth differences, and SHIQ highlight correction to enforce geometry consistency. The central claim is that these steps together prevent edge-skipping and visibility failures, yielding state-of-the-art results on ETH3D, Tanks & Temples, and Strecha.

Significance. If the alignment and visibility mechanisms prove reliable, the approach could strengthen patch-based MVS in textureless and occluded regions by composing existing monocular estimators with lightweight morphological and reprojection steps.

major comments (1)
  1. [Abstract and §3.1] Abstract and §3.1 (depth-normal-edge alignment): the claim that erosion-dilation of off-the-shelf DepthPro/Metric3Dv2 depths with Roberts edges produces fine-grained homogeneous boundaries that reliably block edge-skipping during patch deformation is load-bearing for the entire contribution. Because the depth estimators are applied without dataset-specific fine-tuning, systematic biases (over-smoothing in low-texture areas, scale drift) can persist; simple morphological operations cannot correct residual misalignments, leaving patches free to cross true discontinuities. This assumption requires explicit validation (e.g., boundary-error metrics or ablation removing the alignment step) before the robustness claim can be accepted.
minor comments (1)
  1. [Abstract] The abstract states that geometry consistency is obtained via “aggregated normals via view selection and projection depth differences via epipolar lines,” yet the precise aggregation rule and how these terms are weighted inside the propagation/refinement stage are not specified.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed and constructive review. We address the major comment below and have revised the manuscript to incorporate additional explicit validation of the depth-normal-edge alignment.

read point-by-point responses

  1. Referee: [Abstract and §3.1] Abstract and §3.1 (depth-normal-edge alignment): the claim that erosion-dilation of off-the-shelf DepthPro/Metric3Dv2 depths with Roberts edges produces fine-grained homogeneous boundaries that reliably block edge-skipping during patch deformation is load-bearing for the entire contribution. Because the depth estimators are applied without dataset-specific fine-tuning, systematic biases (over-smoothing in low-texture areas, scale drift) can persist; simple morphological operations cannot correct residual misalignments, leaving patches free to cross true discontinuities. This assumption requires explicit validation (e.g., boundary-error metrics or ablation removing the alignment step) before the robustness claim can be accepted.

    Authors: We acknowledge that off-the-shelf monocular estimators can exhibit biases and that morphological operations have limits. However, the erosion-dilation is applied specifically to the Roberts edge maps to enforce boundary homogeneity for patch deformation, which is a lightweight post-processing step rather than a full correction of the depth field. In the revised manuscript we have added (i) an ablation that disables the alignment step and reports the resulting increase in edge-skipping artifacts and reconstruction error on ETH3D and Tanks & Temples, and (ii) quantitative boundary-error metrics (mean boundary displacement and F-score at 1-pixel threshold) computed against available ground-truth depth on a held-out subset of ETH3D. These new results show measurable improvement attributable to the alignment and are presented in Section 4.3 and the supplementary material. revision: yes

Circularity Check

0 steps flagged

No significant circularity; method composes external tools with independent alignment steps

full rationale

The paper's derivation describes a pipeline that applies off-the-shelf models (DepthPro, Metric3Dv2, Roberts operator, SHIQ) followed by proposed erosion-dilation alignment and reformulated visibility priors. No equations, fitted parameters renamed as predictions, or self-citation chains are shown that reduce outputs to inputs by construction. The central claims rest on the composition and new strategies rather than tautological definitions, with results validated on independent external datasets (ETH3D, Tanks & Temples, Strecha). This is self-contained against benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The method rests on the domain assumption that off-the-shelf monocular depth and normal estimators plus a classical edge detector can be aligned to produce usable homogeneous boundaries; no free parameters or new entities are named in the abstract.

axioms (1)
  • domain assumption Coarse depth and normal maps from DepthPro and Metric3Dv2 plus Roberts edges can be aligned via erosion-dilation to yield fine-grained homogeneous boundaries suitable for patch deformation.
    Invoked in the depth-normal-edge alignment paragraph of the abstract.

pith-pipeline@v0.9.0 · 5849 in / 1236 out tokens · 26356 ms · 2026-05-19T09:49:33.201754+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

86 extracted references · 86 canonical work pages · 1 internal anchor

  1. [1]

    Noise-Transfer2Clean: denoising cryo-EM images based on noise modeling and transfer,

    H. Li, H. Zhang, X. Wan, Z. Yang, C. Li, J. Li, R. Han, P. Zhu, and F. Zhang, “Noise-Transfer2Clean: denoising cryo-EM images based on noise modeling and transfer,” Bioinformatics, vol. 38, no. 7, pp. 2022– 2029, 02 2022

    work page 2022

  2. [2]

    Self- supervised noise modeling and sparsity guided electron tomography volumetric image denoising,

    Z. Yang, D. Zang, H. Li, Z. Zhang, F. Zhang, and R. Han, “Self- supervised noise modeling and sparsity guided electron tomography volumetric image denoising,” Ultramicroscopy, vol. 255, p. 113860, 2024

    work page 2024

  3. [3]

    Self-supervised cryo-electron tomogra- phy volumetric image restoration from single noisy volume with sparsity constraint,

    Z. Yang, F. Zhang, and R. Han, “Self-supervised cryo-electron tomogra- phy volumetric image restoration from single noisy volume with sparsity constraint,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , October 2021, pp. 4056–4065

    work page 2021

  4. [4]

    Ehss: An efficient hybrid-supervised symmetric stereo matching network,

    D. Zhang, P. Zhi, B. Yong, J.-Q. Wang, Y . Hou, L. Guo, Q. Zhou, and R. Zhou, “Ehss: An efficient hybrid-supervised symmetric stereo matching network,” 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC) , pp. 1044–1051, 2023. [Online]. Available: https://api.semanticscholar.org/CorpusID:267661311

    work page 2023

  5. [5]

    Mapexpert: Online hd map construction with simple and efficient sparse map element expert,

    D. Zhang, D. Chen, P. Zhi, Y . Chen, Z. Yuan, C. Li, Sunjing, R. Zhou, and Q. Zhou, “Mapexpert: Online hd map construction with simple and efficient sparse map element expert,” 2024. [Online]. Available: https://arxiv.org/abs/2412.12704

    work page arXiv 2024

  6. [6]

    Xvtp3d: cross-view trajectory prediction using shared 3d queries for autonomous driving,

    Z. Song, H. Bi, R. Zhang, T. Mao, and Z. Wang, “Xvtp3d: cross-view trajectory prediction using shared 3d queries for autonomous driving,” arXiv preprint arXiv:2308.08764 , 2023

    work page arXiv 2023

  7. [7]

    Mmgdreamer: Mixed-modality graph for geometry-controllable 3d indoor scene generation,

    Z. Yang, K. Lu, C. Zhang, J. Qi, H. Jiang, R. Ma, S. Yin, Y . Xu, M. Xing, Z. Xiao et al. , “Mmgdreamer: Mixed-modality graph for geometry-controllable 3d indoor scene generation,” arXiv preprint arXiv:2502.05874, 2025

    work page arXiv 2025

  8. [8]

    Audio-driven emotion-aware 3d talking face generation from single image,

    C.-S. Qiu, F.-L. Liu, H. Fu, F. Zhang, Y .-P. Cao, Y .-K. Lai, and L. Gao, “Audio-driven emotion-aware 3d talking face generation from single image,” in IEEE International Conference on Multimedia and Expo, ICME 2025. IEEE, 2025

    work page 2025

  9. [9]

    Myportrait: Mor- phable prior-guided personalized portrait generation,

    B. Ding, Z. Fan, S. Yang, and S. Xia, “Myportrait: Mor- phable prior-guided personalized portrait generation,” arXiv preprint arXiv:2312.02703, 2023

    work page arXiv 2023

  10. [10]

    D2gv: Deformable 2d gaussian splatting for video representation in 400fps,

    M. Liu, Q. Yang, M. Zhao, H. Huang, L. Yang, Z. Li, and Y . Xu, “D2gv: Deformable 2d gaussian splatting for video representation in 400fps,” arXiv preprint arXiv:2503.05600 , 2025

    work page arXiv 2025

  11. [11]

    Light4gs: Lightweight compact 4d gaussian splatting generation via context model,

    M. Liu, Q. Yang, H. Huang, W. Huang, Z. Yuan, Z. Li, and Y . Xu, “Light4gs: Lightweight compact 4d gaussian splatting generation via context model,” arXiv preprint arXiv:2503.13948 , 2025

    work page arXiv 2025

  12. [12]

    Haif-gs: Hierarchical and induced flow-guided gaussian splatting for dynamic scene,

    J. Chen, Z. Li, Y . Cai, H. Jiang, C. Qian, J. Kang, S. Gao, H. Zhao, T. Mao, and Y . Zhang, “Haif-gs: Hierarchical and induced flow-guided gaussian splatting for dynamic scene,” 2025. [Online]. Available: https://arxiv.org/abs/2506.09518

    work page arXiv 2025

  13. [13]

    Stdr: Spatio-temporal decoupling for real-time dynamic scene rendering,

    Z. Li, H. Jiang, Y . Cai, J. Chen, B. Bi, S. Gao, H. Zhao, Y . Wang, T. Mao, and Z. Wang, “Stdr: Spatio-temporal decoupling for real-time dynamic scene rendering,” 2025. [Online]. Available: https://arxiv.org/abs/2505.22400

    work page arXiv 2025

  14. [14]

    Gradiseg: Gradient-guided gaussian segmentation with enhanced 3d boundary precision,

    Z. Li, W. Han, Y . Cai, H. Jiang, B. Bi, S. Gao, H. Zhao, and Z. Wang, “Gradiseg: Gradient-guided gaussian segmentation with enhanced 3d boundary precision,” 2024. [Online]. Available: https://arxiv.org/abs/2412.00392

    work page arXiv 2024

  15. [15]

    Learning multi-view stereo with geometry-aware prior,

    K. Chen, Z. Yuan, H. Xiao, T. Mao, and Z. Wang, “Learning multi-view stereo with geometry-aware prior,” publisher: IEEE

    work page

  16. [16]

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

    T. Schops, J. L. Schonberger, S. Galliani, T. Sattler, K. Schindler, M. Pollefeys, and A. Geiger, “A multi-view stereo benchmark with high- resolution images and multi-camera videos,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR) , July 2017

    work page 2017

  17. [17]

    Tanks and temples: Benchmarking large-scale scene reconstruction,

    A. Knapitsch, J. Park, Q.-Y . Zhou, and V . Koltun, “Tanks and temples: Benchmarking large-scale scene reconstruction,” 2017. 13

    work page 2017

  18. [18]

    On benchmarking camera calibration and multi-view stereo for high resolution imagery,

    C. Strecha, W. von Hansen, L. Van Gool, P. Fua, and U. Thoennessen, “On benchmarking camera calibration and multi-view stereo for high resolution imagery,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2008, pp. 1–8

    work page 2008

  19. [19]

    Blendedmvs: A large-scale dataset for generalized multi-view stereo networks,

    Y . Yao, Z. Luo, S. Li, J. Zhang, Y . Ren, L. Zhou, T. Fang, and L. Quan, “Blendedmvs: A large-scale dataset for generalized multi-view stereo networks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 1790–1799

    work page 2020

  20. [20]

    Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision,

    L. Ling, Y . Sheng, Z. Tu, W. Zhao, C. Xin, K. Wan, L. Yu, Q. Guo, Z. Yu, and Y . Lu, “Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 22 160–22 169

    work page 2024

  21. [21]

    Dual-level precision edges guided multi-view stereo with accurate planarization,

    K. Chen, Z. Yuan, T. Mao, and Z. Wang, “Dual-level precision edges guided multi-view stereo with accurate planarization,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 39, pp. 2105–2113

    work page

  22. [22]

    NeRF-based polarimetric multi-view stereo,

    J. Cao, Z. Yuan, T. Mao, Z. Wang, and Z. Li, “NeRF-based polarimetric multi-view stereo,” vol. 158, p. 111036, publisher: Elsevier

    work page

  23. [23]

    Topology-aware 3d gaussian splatting: Leveraging persistent homology for optimized struc- tural integrity,

    T. Shen, S. Liu, J. Feng, Z. Ma, and N. An, “Topology-aware 3d gaussian splatting: Leveraging persistent homology for optimized struc- tural integrity,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 7, 2025, pp. 6823–6832

    work page 2025

  24. [24]

    Di-mvs: learning efficient multi- view stereo with depth-aware iterations,

    J. Jiang, M. Cao, J. Yi, and C. Li, “Di-mvs: learning efficient multi- view stereo with depth-aware iterations,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024, pp. 3180–3184

    work page 2024

  25. [25]

    Rrt-mvs: Recurrent regularization transformer for multi-view stereo,

    J. Jiang, L. Wang, H. Yu, T. Hu, J. Chen, and H. Ma, “Rrt-mvs: Recurrent regularization transformer for multi-view stereo,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 39, no. 4, 2025, pp. 3994–4002

    work page 2025

  26. [26]

    Patch- match: A randomized correspondence algorithm for structural image editing,

    C. Barnes, E. Shechtman, A. Finkelstein, and D. B. Goldman, “Patch- match: A randomized correspondence algorithm for structural image editing,” ACM Trans. Graph., p. 24, 2009

    work page 2009

  27. [27]

    SED-MVS: Segmentation-Driven and Edge-Aligned Deformation Multi-View Stereo with Depth Restoration and Occlusion Constraint,

    Z. Yuan, Z. Yang, Y . Cai, K. Wu, M. Liu, D. Zhang, H. Jiang, Z. Li, and Z. Wang, “SED-MVS: Segmentation-Driven and Edge-Aligned Deformation Multi-View Stereo with Depth Restoration and Occlusion Constraint,” Mar. 2025

    work page 2025

  28. [28]

    MSP-MVS: Multi-Granularity Segmentation Prior Guided Multi-View Stereo,

    Z. Yuan, C. Liu, F. Shen, Z. Li, J. Luo, T. Mao, and Z. Wang, “MSP-MVS: Multi-Granularity Segmentation Prior Guided Multi-View Stereo,” Dec. 2024

    work page 2024

  29. [29]

    SD-MVS: segmentation- driven deformation multi-view stereo with spherical refinement and EM optimization,

    Z. Yuan, J. Cao, Z. Li, H. Jiang, and Z. Wang, “SD-MVS: segmentation- driven deformation multi-view stereo with spherical refinement and EM optimization,” CoRR, vol. abs/2401.06385, 2024

    work page arXiv 2024

  30. [30]

    Tsar-mvs: Textureless-aware segmentation and correlative refinement guided multi-view stereo,

    Z. Yuan, J. Cao, Z. Wang, and Z. Li, “Tsar-mvs: Textureless-aware segmentation and correlative refinement guided multi-view stereo,” Pattern Recognition, p. 110565, 2024

    work page 2024

  31. [31]

    Multi-Scale Geometric Consistency Guided and Planar Prior Assisted Multi-View Stereo,

    Q. Xu, W. Kong, W. Tao, and M. Pollefeys, “Multi-Scale Geometric Consistency Guided and Planar Prior Assisted Multi-View Stereo,”IEEE Trans. Pattern Anal. Mach. Intell. , pp. 1–18, 2022

    work page 2022

  32. [32]

    Hierarchical prior mining for non-local multi-view stereo,

    C. Ren, Q. Xu, S. Zhang, and J. Yang, “Hierarchical prior mining for non-local multi-view stereo,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 3611–3620

    work page 2023

  33. [33]

    Phi-mvs: Plane hypothesis inference multi-view stereo for large-scale scene reconstruction,

    S. Sun, Y . Zheng, X. Shi, Z. Xu, and Y . Liu, “Phi-mvs: Plane hypothesis inference multi-view stereo for large-scale scene reconstruction,” arXiv preprint arXiv:2104.06165, 2021

    work page arXiv 2021

  34. [34]

    Adaptive patch deformation for textureless-resilient multi- view stereo,

    Y . Wang, Z. Zeng, T. Guan, W. Yang, Z. Chen, W. Liu, L. Xu, and Y . Luo, “Adaptive patch deformation for textureless-resilient multi- view stereo,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2023, pp. 1621–1630

    work page 2023

  35. [35]

    Depth Anything V2,

    L. Yang, B. Kang, Z. Huang, Z. Zhao, X. Xu, J. Feng, and H. Zhao, “Depth Anything V2,” Jun. 2024

    work page 2024

  36. [36]

    Metric3d v2: A versatile monocular geometric foundation model for zero-shot metric depth and surface normal estimation,

    M. Hu, W. Yin, C. Zhang, Z. Cai, X. Long, H. Chen, K. Wang, G. Yu, C. Shen, and S. Shen, “Metric3d v2: A versatile monocular geometric foundation model for zero-shot metric depth and surface normal estimation,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    work page 2024

  37. [37]

    DVP-MVS: Synergize Depth-Edge and Visibility Prior for Multi-View Stereo,

    Z. Yuan, J. Luo, F. Shen, Z. Li, C. Liu, T. Mao, and Z. Wang, “DVP-MVS: Synergize Depth-Edge and Visibility Prior for Multi-View Stereo,” Dec. 2024

    work page 2024

  38. [38]

    Patchmatch stereo - stereo matching with slanted support windows,

    M. Bleyer, C. Rhemann, and C. Rother, “Patchmatch stereo - stereo matching with slanted support windows,” in British Mach. Vis. Conf. (BMVC), J. Hoey, S. J. McKenna, and E. Trucco, Eds., September 2011, pp. 1–11

    work page 2011

  39. [39]

    Massively parallel multiview stereopsis by surface normal diffusion,

    S. Galliani, K. Lasinger, and K. Schindler, “Massively parallel multiview stereopsis by surface normal diffusion,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV) , December 2015

    work page 2015

  40. [40]

    Multi-scale geometric consistency guided multi- view stereo,

    Q. Xu and W. Tao, “Multi-scale geometric consistency guided multi- view stereo,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), June 2019

    work page 2019

  41. [41]

    Mesh-guided multi-view stereo with pyramid architecture,

    Y . Wang, T. Guan, Z. Chen, Y . Luo, K. Luo, and L. Ju, “Mesh-guided multi-view stereo with pyramid architecture,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR) , June 2020, pp. 2036–2045

    work page 2020

  42. [42]

    Pyramid Multi-View Stereo with Local Consistency,

    J. Liao, Y . Fu, Q. Yan, and C. Xiao, “Pyramid Multi-View Stereo with Local Consistency,” Computer Graphics Forum, vol. 38, no. 7, pp. 335– 346, Oct. 2019

    work page 2019

  43. [43]

    Adaptive pixelwise inference multi-view stereo,

    S. Sun, J. Liu, Y . Li, H. Ying, Z. Zhai, and Y . Mou, “Adaptive pixelwise inference multi-view stereo,” in Thirteenth International Conference on Graphics and Image Processing (ICGIP 2021), D. Xu and L. Xiao, Eds. Kunming, China: SPIE, Feb. 2022, p. 77

    work page 2021

  44. [44]

    mmfas: Multimodal face anti-spoofing using multi-level alignment and switch-attention fusion,

    G. Chen, W. Xie, D. Lin, Y . Liu, and M. Wang, “mmfas: Multimodal face anti-spoofing using multi-level alignment and switch-attention fusion,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 1, 2025, pp. 58–66

    work page 2025

  45. [45]

    Adaptive label correction for robust medical image segmentation with noisy labels,

    C. Qian, K. Han, S. Ma, C. Lyu, Z. Yuan, J. Chen, and Z. Liu, “Adaptive label correction for robust medical image segmentation with noisy labels,” arXiv preprint arXiv:2503.12218 , 2025

    work page arXiv 2025

  46. [46]

    DecAlign: Hierarchical Cross-Modal Alignment for Decoupled Multimodal Representation Learning

    C. Qian, S. Xing, S. Li, Y . Zhao, and Z. Tu, “Decalign: Hierarchical cross-modal alignment for decoupled multimodal representation learn- ing,” arXiv preprint arXiv:2503.11892 , 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  47. [47]

    arXiv preprint arXiv:2503.06456 (2025)

    C. Qian, K. Han, J. Wang, Z. Yuan, C. Lyu, J. Chen, and Z. Liu, “Dyncim: Dynamic curriculum for imbalanced multimodal learning,” arXiv preprint arXiv:2503.06456 , 2025

    work page arXiv 2025

  48. [48]

    Tokenunify: Scalable autoregressive visual pre-training with mixture token prediction,

    Y . Chen, H. Shi*, X. Liu, T. Shi, R. Zhang, D. Liu, Z. Xiong, and F. Wu, “Tokenunify: Scalable autoregressive visual pre-training with mixture token prediction,” arXiv preprint arXiv:2405.16847 , 2024

    work page arXiv 2024

  49. [49]

    Text2reaction: Enabling reactive task planning using large language models,

    Z. Yang, L. Ning, H. Wang, T. Jiang, S. Zhang, S. Cui, H. Jiang, C. Li, S. Wang, and Z. Wang, “Text2reaction: Enabling reactive task planning using large language models,” IEEE Robotics and Automation Letters , 2024

    work page 2024

  50. [50]

    Hierarchical subgoal generation from language instruction for robot task planning,

    Z. Yang, L. Ning, H. Jiang, and Z. Wang, “Hierarchical subgoal generation from language instruction for robot task planning,” in 2022 China Automation Congress (CAC) . IEEE, 2022, pp. 5976–5980

    work page 2022

  51. [51]

    MR-IntelliAssist: A world cognition agent enabling adaptive human-AI symbiosis in industry 4.0,

    C. Liu, Z. Yuan, Y . Wang, Y . Yin, W. Luo, Z. He, and X. Liang, “MR-IntelliAssist: A world cognition agent enabling adaptive human-AI symbiosis in industry 4.0,” in Artificial Intelligence in HCI , H. Degen and S. Ntoa, Eds. Springer Nature Switzerland, vol. 15822, pp. 163– 177

    work page

  52. [52]

    Self-supervised neuron segmentation with multi-agent reinforcement learning,

    Y . Chen, W. Huang, S. Zhou, Q. Chen, and Z. Xiong, “Self-supervised neuron segmentation with multi-agent reinforcement learning,” in IJCAI 23, 2023

    work page 2023

  53. [53]

    Mask- factory: Towards high-quality synthetic data generation for dichotomous image segmentation,

    H. Qian*, Y . Chen*, S. Lou, F. S. Khan, X. Jin, and D.-P. Fan, “Mask- factory: Towards high-quality synthetic data generation for dichotomous image segmentation,” NeurIPS 24, 2024

    work page 2024

  54. [54]

    Generative text-guided 3d vision-language pretraining for unified medical image segmentation,

    Y . Chen, C. Liu*, W. Huang, X. Liu, S. Cheng, R. Arcucci, and Z. Xiong, “Generative text-guided 3d vision-language pretraining for unified medical image segmentation,” arXiv preprint arXiv:2306.04811, 2023

    work page arXiv 2023

  55. [55]

    Structure-adaptive multi-view graph clustering for remote sensing data,

    R. Guan, W. Tu, S. Wang, J. Liu, D. Hu, C. Tang, Y . Feng, J. Li, B. Xiao, and X. Liu, “Structure-adaptive multi-view graph clustering for remote sensing data,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, 2025, pp. 16 933–16 941

    work page 2025

  56. [56]

    Contrastive multiview subspace clustering of hyperspectral images based on graph convolutional networks,

    R. Guan, Z. Li, W. Tu, J. Wang, Y . Liu, X. Li, C. Tang, and R. Feng, “Contrastive multiview subspace clustering of hyperspectral images based on graph convolutional networks,” IEEE Transactions on Geoscience and Remote Sensing , vol. 62, pp. 1–14, 2024

    work page 2024

  57. [57]

    Spatial-spectral graph contrastive clustering with hard sample mining for hyperspectral images,

    R. Guan, W. Tu, Z. Li, H. Yu, D. Hu, Y . Chen, C. Tang, Q. Yuan, and X. Liu, “Spatial-spectral graph contrastive clustering with hard sample mining for hyperspectral images,”IEEE Transactions on Geoscience and Remote Sensing, pp. 1–16, 2024

    work page 2024

  58. [58]

    Program: Prototype graph model based pseudo-label learning for test-time adaptation,

    H. Sun, L. Xu, S. Jin, P. Luo, C. Qian, and W. Liu, “Program: Prototype graph model based pseudo-label learning for test-time adaptation,” in The Twelfth International Conference on Learning Representations

    work page

  59. [59]

    Unsupervised continual domain shift learning with multi- prototype modeling,

    H. Sun, Y . Zhang, L. Xu, S. Jin, P. Luo, C. Qian, W. Liu, and Y . Chen, “Unsupervised continual domain shift learning with multi- prototype modeling,” in Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) , June 2025, pp. 10 131–10 141

    work page 2025

  60. [60]

    Text2earth: Unlocking text-driven remote sensing image generation with a global-scale dataset and a foundation model,

    C. Liu, K. Chen, R. Zhao, Z. Zou, and Z. Shi, “Text2earth: Unlocking text-driven remote sensing image generation with a global-scale dataset and a foundation model,” IEEE Geoscience and Remote Sensing Mag- azine, pp. 2–23, 2025

    work page 2025

  61. [61]

    Rscama: Remote sensing image change captioning with state space model,

    C. Liu, K. Chen, B. Chen, H. Zhang, Z. Zou, and Z. Shi, “Rscama: Remote sensing image change captioning with state space model,” IEEE Geoscience and Remote Sensing Letters , vol. 21, pp. 1–5, 2024. 14

    work page 2024

  62. [62]

    Change-agent: Toward interactive comprehensive remote sensing change interpretation and analysis,

    C. Liu, K. Chen, H. Zhang, Z. Qi, Z. Zou, and Z. Shi, “Change-agent: Toward interactive comprehensive remote sensing change interpretation and analysis,” IEEE Transactions on Geoscience and Remote Sensing , vol. 62, pp. 1–16, 2024

    work page 2024

  63. [63]

    Remote sensing image change captioning with dual-branch transformers: A new method and a large scale dataset,

    C. Liu, R. Zhao, H. Chen, Z. Zou, and Z. Shi, “Remote sensing image change captioning with dual-branch transformers: A new method and a large scale dataset,” IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1–20, 2022

    work page 2022

  64. [64]

    Remote sensing spatio-temporal vision-language models: A comprehensive survey,

    C. Liu, J. Zhang, K. Chen, M. Wang, Z. Zou, and Z. Shi, “Remote sensing spatio-temporal vision-language models: A comprehensive survey,” 2025. [Online]. Available: https://arxiv.org/abs/2412.02573

    work page arXiv 2025

  65. [65]

    Mvsnet: Depth inference for unstructured multi-view stereo,

    Y . Yao, Z. Luo, S. Li, T. Fang, and L. Quan, “Mvsnet: Depth inference for unstructured multi-view stereo,” in Proc. Eur. Conf. Comput. Vis. (ECCV), September 2018

    work page 2018

  66. [66]

    Recurrent mvsnet for high-resolution multi-view stereo depth inference,

    Y . Yao, Z. Luo, S. Li, T. Shen, T. Fang, and L. Quan, “Recurrent mvsnet for high-resolution multi-view stereo depth inference,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 5525–5534

    work page 2019

  67. [67]

    Itermvs: Itera- tive probability estimation for efficient multi-view stereo,

    F. Wang, S. Galliani, C. V ogel, and M. Pollefeys, “Itermvs: Itera- tive probability estimation for efficient multi-view stereo,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR) , 2022, pp. 8606–8615

    work page 2022

  68. [68]

    Cost volume pyramid based depth inference for multi-view stereo,

    J. Yang, W. Mao, J. M. Alvarez, and M. Liu, “Cost volume pyramid based depth inference for multi-view stereo,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR) , 2020, pp. 4876–4885

    work page 2020

  69. [69]

    Patch- matchnet: Learned multi-view patchmatch stereo,

    F. Wang, S. Galliani, C. V ogel, P. Speciale, and M. Pollefeys, “Patch- matchnet: Learned multi-view patchmatch stereo,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2021, pp. 14 194–14 203

    work page 2021

  70. [70]

    MVSTER: Epipolar transformer for efficient multi-view stereo,

    X. Wang, Z. Zhu, G. Huang, F. Qin, Y . Ye, Y . He, X. Chi, and X. Wang, “MVSTER: Epipolar transformer for efficient multi-view stereo,” in European Conference on Computer Vision . Springer, 2022, pp. 573– 591

    work page 2022

  71. [71]

    Epp-mvsnet: Epipolar-assembling based depth prediction for multi-view stereo,

    X. Ma, Y . Gong, Q. Wang, J. Huang, L. Chen, and F. Yu, “Epp-mvsnet: Epipolar-assembling based depth prediction for multi-view stereo,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV) , 2021, pp. 5712–5720

    work page 2021

  72. [72]

    GeoMVSNet: Learning Multi-View Stereo With Geometry Perception,

    Z. Zhang, R. Peng, Y . Hu, and R. Wang, “GeoMVSNet: Learning Multi-View Stereo With Geometry Perception,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 21 508–21 518

    work page 2023

  73. [73]

    Multi-View Stereo Representation Revist: Region-Aware MVSNet,

    Y . Zhang, J. Zhu, and L. Lin, “Multi-View Stereo Representation Revist: Region-Aware MVSNet,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 17 376–17 385

    work page 2023

  74. [74]

    Pixelwise view selection for unstructured multi-view stereo,

    J. L. Sch ¨onberger, E. Zheng, J.-M. Frahm, and M. Pollefeys, “Pixelwise view selection for unstructured multi-view stereo,” in Proc. Eur. Conf. Comput. Vis. (ECCV) , 2016, pp. 501–518

    work page 2016

  75. [75]

    Accurate multiple view 3d reconstruction using patch-based stereo for large-scale scenes,

    S. Shen, “Accurate multiple view 3d reconstruction using patch-based stereo for large-scale scenes,”IEEE Trans. Image Process., vol. 22, no. 5, pp. 1901–1914, 2013

    work page 1901

  76. [76]

    Depth Pro: Sharp Monocular Metric Depth in Less Than a Second,

    A. Bochkovskii, A. Delaunoy, H. Germain, M. Santos, Y . Zhou, S. R. Richter, and V . Koltun, “Depth Pro: Sharp Monocular Metric Depth in Less Than a Second,” Oct. 2024

    work page 2024

  77. [77]

    Repurposing diffusion-based image generators for monoc- ular depth estimation,

    B. Ke, A. Obukhov, S. Huang, N. Metzger, R. C. Daudt, and K. Schindler, “Repurposing diffusion-based image generators for monoc- ular depth estimation,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition, n CVPR 2024, Seattle, WA, USA, June 16-22,

    work page 2024

  78. [78]

    9492–9502

    IEEE, 2024, pp. 9492–9502

    work page 2024

  79. [79]

    Nddepth: Normal- distance assisted monocular depth estimation and completion,

    S. Shao, Z. Pei, W. Chen, P. C. Y . Chen, and Z. Li, “Nddepth: Normal- distance assisted monocular depth estimation and completion,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 12, pp. 8883–8899, 2024

    work page 2024

  80. [80]

    Efficient edge-preserving multi-view stereo network for depth estimation,

    W. Su and W. Tao, “Efficient edge-preserving multi-view stereo network for depth estimation,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, 2023, pp. 2348–2356

    work page 2023

Showing first 80 references.