Pith. sign in

REVIEW 4 major objections 5 minor 34 references

LiteMVS: Efficient Multi-View Stereo with Foundation Distillation and Expert Aggregation

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read LiteMVS claims that injecting semantic descriptors and foundation-model pseudo-labels into a plane-sweep cost volume yields more accurate depth and 3D reconstruction at roughly 75 ms per frame.

desk verdict LiteMVS is a sensible architecture with a real training-data confound in its SOTA comparison; worth review, not yet trustworthy. read the letter →

arxiv 2608.03851 v1 pith:37AYUKRS submitted 2026-08-04 cs.CV

Tianbao Zhang , Zeyu Liu , Shuyu Wu , Fanxing Li , Zhaoxin Fan , Wenjun Wu , Danping Zou This is my paper
classification cs.CV
keywords multi-viewstereodepthestimationplane-sweepcostvolumeMixture-of-Expertsknowledgedistillationsemanticpriors3Dreconstructionembodiedmanipulation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper presents LiteMVS, a multi-view stereo method that estimates depth from multiple posed images while keeping computation low. Its central claim is that lightweight plane-sweep matching, which often fails in textureless or repetitive regions, can be made accurate by enriching the cost volume with semantic descriptors and by training on pseudo-labels distilled from monocular foundation models. On ScanNetv2 and 7-Scenes, the paper reports lower depth errors than SimpleRecon and DoubleTake, with reconstruction quality comparable to or better than heavier volumetric methods and per-frame update latency around 75 ms. The same representation also transfers to a robotic-manipulation setup, approaching a much larger 3D-aware backbone at roughly one-third of its inference cost. The paper's own Limitations section concedes that reflective surfaces, transparent objects, and low-texture regions remain difficult.

What carries the argument

The load-bearing object is the semantic-augmented 4D cost volume: per-pixel semantic descriptors from a lightweight segmentation encoder (MobileSAM lineage) are broadcast along the depth dimension and concatenated with warped multi-view features and geometric metadata, so every depth hypothesis carries both matching evidence and monocular boundary/structure cues. A Mixture-of-Experts MLP then aggregates the volume, with a gating network letting experts specialize across depth regimes, and pseudo-label losses transfer foundation-model priors at no extra inference cost.

What would settle it

Re-run SimpleRecon, DoubleTake, and LiteMVS under one shared keyframe schedule, mesh mask, and GPU, then recompute the depth and reconstruction tables; if the re-run baselines match or exceed LiteMVS's reported numbers, the central accuracy claim is an artifact of the comparison protocol.

Watch

Extended reading notes

Core claim

LiteMVS claims that an efficient multi-view stereo network does not have to choose between geometric consistency and monocular priors. It augments a standard plane-sweep feature volume with semantic descriptors from a lightweight segmentation encoder, replaces the shared cost-aggregation MLP with a Mixture-of-Experts (MoE) formulation so different experts specialize across depth ranges, and adds output-level distillation: relative-depth pseudo-labels from Depth Anything V2 and surface-normal pseudo-labels from StableNormal supervise training without changing inference. The combined system yields sharper boundaries and lower depth error than SimpleRecon and DoubleTake on ScanNetv2 and 7-Scene

Load-bearing premise

The load-bearing premise is that all methods were compared under the same protocol—same keyframes, mesh masks, hardware, and input views—since Table 1 reports some previous results from [12] or from each method's own keyframes; if the protocol is not identical, the reported accuracy advantages are not attributable to the architecture.

Editorial extensions

If this is right

  • Lightweight 2D-CNN-based MVS can reach reconstruction quality previously associated with heavier volumetric refinement, at real-time update rates.
  • Distilling relative depth and surface-normal pseudo-labels from foundation models is a low-cost way to import monocular priors into task-specific networks.
  • Letting multiple expert MLPs specialize across depth ranges improves accuracy over a single shared MLP, with gains saturating around three experts.
  • Geometry-aware MVS features can serve as efficient visual targets for embodied manipulation policies, matching heavier 3D backbones on LIBERO at about one-third the inference cost.
  • Using segmentation cues as auxiliary guidance rather than a trained head keeps the semantic prior optional at inference time.

Reading between the lines

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

  • The authors' own limitation list — reflective surfaces, transparent objects, low-texture regions — maps cleanly onto the failure modes that appearance matching and monocular priors share; if semantic cues are doing the claimed work, per-region error analysis should show the largest gains exactly there. That is a testable prediction the paper does not make.
  • The MoE saturation at three experts suggests the mechanism may be interpreted as a soft split of the depth range rather than a capacity increase; plotting gate weights against true depth would confirm the specialization directly.
  • The same pseudo-label distillation recipe could extend to other geometry outputs, such as optical flow or 4D tracking, since the paper demonstrates it only for static-scene depth and normals.
  • Because Table 1 mixes previously reported numbers with per-method keyframes, the cleanest way to test the accuracy claim is a single-keyframe re-run of all baselines on identical hardware.
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

4 major / 5 minor

Summary. The paper introduces LiteMVS, a lightweight multi-view stereo depth estimator that augments a plane-sweep cost volume with semantic descriptors from MobileSAM, aggregates cost features with a Mixture-of-Experts (MoE) MLP, and distills relative-depth and surface-normal pseudo-labels from Depth Anything V2 and StableNormal during training. The authors report lower depth errors than SimpleRecon and DoubleTake on ScanNetv2 and 7-Scenes, improved 3D reconstruction F-score, and competitive per-frame latency, as well as downstream manipulation results on LIBERO and RoboTwin 2.0. The central claim is that combining monocular semantic/structural priors with efficient MVS improves the accuracy-efficiency trade-off without adding inference cost.

Significance. If the results hold, the paper offers a practical recipe for injecting foundation-model priors into lightweight MVS while keeping inference cost unchanged, since the distilled pseudo-labels are used only as training supervision. The architecture is clearly motivated, and the evaluation spans depth, reconstruction, and embodied downstream tasks. I agree with the reader that there is no circularity: pseudo-labels are training signals, and all reported metrics are against ground-truth depth/meshes. The benchmark claims are falsifiable, and the ablations that are actually run generally support the stated components. However, the headline comparative claims are currently conditional on an uncontrolled training-data protocol and on a missing distillation ablation, so the significance is not yet established at the level claimed.

major comments (4)
  1. [Section 4, Tables 1–2] The headline comparisons are confounded by training data. Section 4 states LiteMVS is trained on 'a combination of ScanNetv2 and ScanNet++', while Table 1's caption says prior results were 'taken from [12], or evaluated for each method using their keyframes'; there is no statement that SimpleRecon or DoubleTake were retrained on the same ScanNet++-augmented split. Since ScanNet++ is larger and higher-fidelity, the reported margins (e.g., Abs Rel 0.0311 vs 0.0369 for DoubleTake; F-score 0.715 vs 0.714) may reflect a data advantage rather than an architectural one. Please retrain all baselines on the identical data split, or report a ScanNet-only comparison, and state the full evaluation protocol. Reporting multiple-seed error bars would also help for the near-tie F-score difference.
  2. [Section 5.3, Table 3] The claimed contribution of pseudo-label distillation is not ablated. Section 5.3 is titled 'Effect of Vision Foundation Model Distillation', but its text and Table 4 vary the number of MoE experts; the Table 3 row 'Ours w/oLM onoKD' is malformed and unexplained. A clean experiment removing L_rel and L_norm (Eqs. 3–5) is required to support the paper's third contribution and the claim that distillation adds 'no extra inference cost'. Please fix the row label and add the missing distillation ablation.
  3. [Section 3.5, Eq. (5)] The primary supervision term L_simple is not defined in the paper; the text defers it to a supplementary that is not included. The values of lambda_rel, lambda_norm, and lambda_grad are also omitted. Without the full SimpleRecon loss and hyperparameters, Eq. (5) is not reproducible, and the contribution of the new distillation losses cannot be quantitatively assessed. Please include the full loss formulation and all hyperparameter values in the main text or an appendix.
  4. [Section 4.2, Table 2, Fig. 1] Fig. 1(b) claims LiteMVS achieves the 'fastest inference speed', but Table 2 reports update latency 72 ms for SimpleRecon vs 75 ms for LiteMVS. This is a factual inconsistency in a headline claim. Please correct the figure/caption to say 'competitive' latency, or revise the comparison to be consistent with the reported numbers.
minor comments (5)
  1. [Table 3] Typos: 'Sematic' should be 'Semantic'; 'w/oLM onoKD' is likely intended as 'w/o KD' (without knowledge distillation).
  2. [Sections 5.3 and 5.4] Both sections contain nearly identical sentences: 'We further ablate the proposed MoE-based cost aggregation module by varying the number of experts.' This duplication should be removed, and the distillation ablation placed in Section 5.3.
  3. [Introduction, references] The citation '[9, 25? ? ? ?]' contains placeholder question marks and should be fixed. Also, 'Simplerecon' in Table 1 should be 'SimpleRecon'.
  4. [Table 2 caption] The abbreviations 'Comp', 'Acc', and 'Recall' are not defined in the caption. Please define them or refer to the TransformerFusion protocol explicitly.
  5. [Figure 1] 'Poindcloud' should be 'Point cloud'; the caption and text should avoid overstating the speed comparison (see major comment 4).

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found; LiteMVS's derivation is self-contained and validated against external ground truth with component-level ablations.

full rationale

LiteMVS's central claim is that injecting semantic descriptors (MobileSAM), using MoE aggregation, and distilling relative depth/normal pseudo-labels from foundation models (Depth Anything V2, StableNormal) improves depth estimation and 3D reconstruction. The pseudo-labels are used only as auxiliary training supervision (Eqs. 3–5); all reported metrics in Tables 1 and 2 are computed against ScanNetv2 and 7-Scenes ground truth using standard depth and mesh-reconstruction protocols, not against the distilled pseudo-labels. Therefore, no 'prediction' reduces to a fitted input or to a self-defined target. The ablation study (Table 3) explicitly isolates each component, including 'Ours w/o LMonoKD' for distillation, showing measurable, non-tautological contributions. The loss builds on SimpleRecon [20] as an external baseline, and all cited foundation models and datasets are independent of the authors; there is no load-bearing self-citation. The only notable concern is that Table 1 takes baseline numbers from prior papers without retraining on the ScanNet++-augmented split, but this is a fairness/protocol issue rather than circularity. No equation-level equivalence, ansatz-smuggling, or uniqueness import is present.

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

All free parameters are standard training hyperparameters and architecture choices; the paper does not fit constants to the evaluation set. However, the exact values are absent, and the missing distillation ablation makes the loss balance a genuine hand-tuned knob that cannot be audited.

free parameters (4)
  • lambda_rel, lambda_norm = not stated
    Balancing weights in Eq. 5 for relative-depth and normal distillation losses; values are hand-tuned and directly trade off distillation against the SimpleRecon multi-view objective.
  • lambda_grad = not stated
    Weight on the gradient term in the normal loss (Eq. 4); chosen by hand, no value or sensitivity analysis reported.
  • K (number of MoE experts) = 3
    Selected by validation ablation in Table 4; 3 experts give best ScanNetv2 numbers, so the architecture choice is tuned on the evaluation domain.
  • depth hypothesis count and range = not stated
    The plane-sweep volume in Eq. 1 depends on a chosen set {d_k}; resolution and range affect accuracy and are not specified in the paper.
assumptions (5)
  • domain assumption Camera intrinsics, relative poses, and depth-hypothesis grids are known and correct.
    Section 3.1 assumes known K and (R,t); plane-sweep warping Eq. 1 depends on this. If poses are noisy, the cost volume is corrupted.
  • domain assumption The depth hypothesis range and density cover scene depths in ScanNetv2 and 7-Scenes.
    Depth is selected from the discrete set {d_k}; true depths outside the range or between planes are not representable, yet the paper does not state the grid.
  • domain assumption MobileSAM semantic descriptors provide structure cues beyond appearance matching.
    This is the central methodological premise behind semantic feature injection; the paper only ablates the full model versus removing semantics, without isolating whether the gain comes from semantic information itself or from extra capacity.
  • domain assumption Depth Anything V2 and StableNormal pseudo-labels are accurate enough to supervise the student on indoor scans.
    The distillation losses in Eqs. 3 and 4 assume these external pseudo-labels improve, rather than corrupt, training; no analysis of pseudo-label noise or failure regions is given.
  • standard math Differentiable homography warping and the depth-to-normal operator N() are standard differentiable operations.
    The method relies on these operations without derivation; they are standard in the cited MVS and depth literature.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LiteMVS: Efficient Multi-View Stereo with Foundation Distillation and Expert Aggregation." pith.science (2026). https://pith.science/paper/37AYUKRS

@misc{pith2026260803851,
  author       = {Pith},
  title        = {Pith review of: LiteMVS: Efficient Multi-View Stereo with Foundation Distillation and Expert Aggregation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/37AYUKRS}},
  note         = {Machine review of arXiv:2608.03851}
}
read the original abstract

Real-time 3D perception is crucial for robotics, augmented reality, and embodied intelligence applications. Existing multi-view stereo (MVS) methods primarily rely on geometric correspondences, which often fail in textureless or repetitive regions, while monocular depth models leverage strong image-level priors but lack robust multi-view geometric constraints. More importantly, in robotics and embodied manipulation scenarios, high-quality 3D geometry is not only essential for static reconstruction, but also serves as a critical foundation for learning temporally consistent 4D representations. To obtain visual representations with stronger structural awareness and greater potential for spatiotemporal extension, we present LiteMVS, a lightweight multi-view depth estimation model that integrates plane-sweep geometric reasoning with strong monocular semantic and structural priors. The central idea of LiteMVS is to efficiently inject high-level monocular knowledge, obtained from lightweight segmentation models and large-scale vision foundation models, into a multi-view stereo framework. In particular, LiteMVS enriches the cost volume with semantic descriptors and employs a Mixture-of-Experts (MoE) formulation to enable adaptive geometric aggregation across depth hypotheses. Moreover, geometric priors distilled from vision foundation models further strengthen monocular guidance without increasing inference cost. Through this design, LiteMVS not only improves depth estimation and 3D reconstruction quality in static scenes, but also provides a more reliable geometric foundation for subsequent temporal modeling and 4D representation learning. Experiments on ScanNetv2 and 7-Scenes demonstrate that LiteMVS achieves high-quality depth prediction and 3D reconstruction while maintaining competitive efficiency.

Figures

Figures reproduced from arXiv: 2608.03851 by the authors.

Figure 1
Figure 1. (a) Qualitative visualization results of LiteMVS, including the input images, predicted depth maps, and reconstructed point clouds [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Framework of LiteMVS. The pipeline consists of three components: (1) monocular vision foundation model distillation that generates pseudo depth and normal labels for training supervision; (2) semantic feature injection via MobileSAM to enhance the cost volume with scene-level context; and (3) semantic-aware multi-view stereo with a Mixture-of-Experts MLP for adaptive feature aggregation and robust depth prediction. … view at source ↗
Figure 3
Figure 3. Qualitative results of LiteMVS. LiteMVS produces accurate depth predictions with sharp object boundaries and consistent surface geometry, enabling high-quality 3D reconstruction across diverse indoor scenes. Compared to existing methods, fine-grained geometric details and boundary regions are better preserved, as highlighted in the red boxes. Beyond reconstruction quality, we evaluate computa￾tional efficiency in te… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Effectiveness of Pseudo-Label Distillation. Com￾pared with the model trained without pseudo-label supervision, LiteMVS produces sharper object boundaries and more coherent depth transitions (highlighted by red boxes) [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Comparisons with SpatialForcing [16] on RoboTwin 2.0 benchmark. As shown in [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 23 canonical work pages

  1. [12]

    Dpsnet: End-to-end deep plane sweep stereo.arXiv preprint arXiv:1905.00538, 2019

    Sunghoon Im, Hae-Gon Jeon, Stephen Lin, and In So Kweon. Dpsnet: End-to-end deep plane sweep stereo.arXiv preprint arXiv:1905.00538, 2019. 6

    arXiv 1905

  2. [1]

    Midas v3

    Reiner Birkl, Diana Wofk, and Matthias M ¨uller. Midas v3. 1–a model zoo for robust monocular relative depth estima- tion.arXiv preprint arXiv:2307.14460, 2023. 3

    arXiv 2023

  3. [2]

    Transformerfusion: Monocular rgb scene reconstruction using transformers.Advances in Neural In- formation Processing Systems, 34:1403–1414, 2021

    Aljaz Bozic, Pablo Palafox, Justus Thies, Angela Dai, and Matthias Nießner. Transformerfusion: Monocular rgb scene reconstruction using transformers.Advances in Neural In- formation Processing Systems, 34:1403–1414, 2021. 5, 7

    work page 2021

  4. [3]

    Mvsformer++: Revealing the devil in transformer’s details for multi-view stereo.arXiv preprint arXiv:2401.11673, 2024

    Chenjie Cao, Xinlin Ren, and Yanwei Fu. Mvsformer++: Revealing the devil in transformer’s details for multi-view stereo.arXiv preprint arXiv:2401.11673, 2024. 2, 3

    arXiv 2024

  5. [4]

    Robotwin 2.0: A scalable data gen- erator and benchmark with strong domain randomization for robust bimanual robotic manipulation.arXiv preprint arXiv:2506.18088, 2025

    Tianxing Chen, Zanxin Chen, Baijun Chen, Zijian Cai, Yibin Liu, Zixuan Li, Qiwei Liang, Xianliang Lin, Yiheng Ge, Zhenyu Gu, et al. Robotwin 2.0: A scalable data gen- erator and benchmark with strong domain randomization for robust bimanual robotic manipulation.arXiv preprint arXiv:2506.18088, 2025. 8

    arXiv 2025

  6. [5]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 5828–5839, 2017. 5

    2017

  7. [6]

    Deep- videomvs: Multi-view stereo on video with recurrent spatio- temporal fusion

    Arda Duzceker, Silvano Galliani, Christoph V ogel, Pablo Speciale, Mihai Dusmanu, and Marc Pollefeys. Deep- videomvs: Multi-view stereo on video with recurrent spatio- temporal fusion. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pages 15324–15333, 2021. 5, 6, 7

    work page 2021

  8. [7]

    Predicting depth, surface nor- mals and semantic labels with a common multi-scale con- volutional architecture

    David Eigen and Rob Fergus. Predicting depth, surface nor- mals and semantic labels with a common multi-scale con- volutional architecture. InProceedings of the IEEE inter- national conference on computer vision, pages 2650–2658,

    work page

Show all 34 references
  1. [8]

    Depth map prediction from a single image using a multi-scale deep net- work.Advances in neural information processing systems, 27, 2014

    David Eigen, Christian Puhrsch, and Rob Fergus. Depth map prediction from a single image using a multi-scale deep net- work.Advances in neural information processing systems, 27, 2014. 5

    2014

  2. [9]

    Rpr-net: A point cloud-based rotation-aware large scale place recognition network

    Zhaoxin Fan, Zhenbo Song, Wenping Zhang, Hongyan Liu, Jun He, and Xiaoyong Du. Rpr-net: A point cloud-based rotation-aware large scale place recognition network. In European Conference on Computer Vision, pages 709–725. Springer, 2022. 2

    2022

  3. [10]

    Multi-view stereo: A tutorial.Foundations and trends® in Computer Graphics and Vision, 9(1-2):1–148, 2015

    Yasutaka Furukawa, Carlos Hern ´andez, et al. Multi-view stereo: A tutorial.Foundations and trends® in Computer Graphics and Vision, 9(1-2):1–148, 2015. 2, 3

    2015

  4. [11]

    Multi-view stereo by temporal nonparametric fusion

    Yuxin Hou, Juho Kannala, and Arno Solin. Multi-view stereo by temporal nonparametric fusion. InProceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 2651–2660, 2019. 6

    2019

  5. [13]

    Mvsanywhere: Zero-shot multi-view stereo

    Sergio Izquierdo, Mohamed Sayed, Michael Firman, Guillermo Garcia-Hernando, Daniyar Turmukhambetov, Javier Civera, Oisin Mac Aodha, Gabriel Brostow, and Jamie Watson. Mvsanywhere: Zero-shot multi-view stereo. In Proceedings of the Computer Vision and Pattern Recognition Confer...

    2025

  6. [14]

    Fine-tuning vision-language-action models: Optimizing speed and suc- cess.arXiv preprint arXiv:2502.19645, 2025

    Moo Jin Kim, Chelsea Finn, and Percy Liang. Fine-tuning vision-language-action models: Optimizing speed and suc- cess.arXiv preprint arXiv:2502.19645, 2025. 7

    2025 arXiv

  7. [15]

    Segment any- thing

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment any- thing. InProceedings of the IEEE/CVF international confer- ence on computer vision, pages 4015–4026, 2023. 2

    2023

  8. [16]

    Spatial forcing: Implicit spatial representation align- ment for vision-language-action model.arXiv preprint arXiv:2510.12276, 2025

    Fuhao Li, Wenxuan Song, Han Zhao, Jingbo Wang, Pengxiang Ding, Donglin Wang, Long Zeng, and Haoang Li. Spatial forcing: Implicit spatial representation align- ment for vision-language-action model.arXiv preprint arXiv:2510.12276, 2025. 7, 8

    2025

  9. [17]

    Libero: Benchmarking knowl- edge transfer for lifelong robot learning.Advances in Neural Information Processing Systems, 36:44776–44791, 2023

    Bo Liu, Yifeng Zhu, Chongkai Gao, Yihao Feng, Qiang Liu, Yuke Zhu, and Peter Stone. Libero: Benchmarking knowl- edge transfer for lifelong robot learning.Advances in Neural Information Processing Systems, 36:44776–44791, 2023. 7

    2023

  10. [18]

    Mixture of ex- perts: a literature survey.Artificial Intelligence Review, 42 (2):275–293, 2014

    Saeed Masoudnia and Reza Ebrahimpour. Mixture of ex- perts: a literature survey.Artificial Intelligence Review, 42 (2):275–293, 2014. 2

    2014

  11. [19]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193, 2023. 7, 8

    2023 arXiv

  12. [20]

    Simplere- con: 3d reconstruction without 3d convolutions

    Mohamed Sayed, John Gibson, Jamie Watson, Victor Prisacariu, Michael Firman, and Cl´ement Godard. Simplere- con: 3d reconstruction without 3d convolutions. InEuropean Conference on Computer Vision, pages 1–19. Springer, 2022. 2, 3, 4, 5, 6, 7

    2022

  13. [21]

    Doubletake: Geometry guided depth estimation

    Mohamed Sayed, Filippo Aleotti, Jamie Watson, Zawar Qureshi, Guillermo Garcia-Hernando, Gabriel Brostow, Sara Vicente, and Michael Firman. Doubletake: Geometry guided depth estimation. InEuropean conference on computer vi- sion, pages 121–138. Springer, 2024. 2, 3, 6, 7

    2024

  14. [22]

    Scene co- ordinate regression forests for camera relocalization in rgb-d images

    Jamie Shotton, Ben Glocker, Christopher Zach, Shahram Izadi, Antonio Criminisi, and Andrew Fitzgibbon. Scene co- ordinate regression forests for camera relocalization in rgb-d images. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 2930–2...

    2013

  15. [23]

    V ortx: V olumetric 3d reconstruction with trans- formers for voxelwise view selection and fusion

    Noah Stier, Alexander Rich, Pradeep Sen, and Tobias H¨ollerer. V ortx: V olumetric 3d reconstruction with trans- formers for voxelwise view selection and fusion. In2021 In- ternational Conference on 3D Vision (3DV), pages 320–330. IEEE, 2021. 7

    2021

  16. [24]

    Neuralrecon: Real-time coherent 3d re- construction from monocular video

    Jiaming Sun, Yiming Xie, Linghao Chen, Xiaowei Zhou, and Hujun Bao. Neuralrecon: Real-time coherent 3d re- construction from monocular video. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 15598–15607, 2021. 7

    2021

  17. [25]

    Depth from motion for smartphone ar.ACM Transactions on Graphics (ToG), 37(6):1–19, 2018

    Julien Valentin, Adarsh Kowdle, Jonathan T Barron, Neal Wadhwa, Max Dzitsiuk, Michael Schoenberg, Vivek Verma, Ambrus Csaszar, Eric Turner, Ivan Dryanovski, et al. Depth from motion for smartphone ar.ACM Transactions on Graphics (ToG), 37(6):1–19, 2018. 2

    2018

  18. [26]

    Vggt: Vi- sual geometry grounded transformer

    Jianyuan Wang, Minghao Chen, Nikita Karaev, Andrea Vedaldi, Christian Rupprecht, and David Novotny. Vggt: Vi- sual geometry grounded transformer. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 5294–5306, 2025. 7, 8

    2025

  19. [27]

    Mvdepthnet: Real-time multiview depth estimation neural network

    Kaixuan Wang and Shaojie Shen. Mvdepthnet: Real-time multiview depth estimation neural network. In2018 Interna- tional conference on 3d vision (3DV), pages 248–257. IEEE,

  20. [28]

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

    Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth anything: Unleashing the power of large-scale unlabeled data. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10371–10381, 2024. 3

    2024

  21. [29]

    Depth any- thing v2.Advances in Neural Information Processing Sys- tems, 37:21875–21911, 2024

    Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xiao- gang Xu, Jiashi Feng, and Hengshuang Zhao. Depth any- thing v2.Advances in Neural Information Processing Sys- tems, 37:21875–21911, 2024. 2, 3, 5

    2024

  22. [30]

    Mvsnet: Depth inference for unstructured multi-view stereo

    Yao Yao, Zixin Luo, Shiwei Li, Tian Fang, and Long Quan. Mvsnet: Depth inference for unstructured multi-view stereo. InProceedings of the European conference on computer vi- sion (ECCV), pages 767–783, 2018. 2, 3

    2018

  23. [31]

    Stablenormal: Reducing diffusion variance for stable and sharp normal.ACM Transactions on Graphics (TOG), 43(6):1–18, 2024

    Chongjie Ye, Lingteng Qiu, Xiaodong Gu, Qi Zuo, Yushuang Wu, Zilong Dong, Liefeng Bo, Yuliang Xiu, and Xiaoguang Han. Stablenormal: Reducing diffusion variance for stable and sharp normal.ACM Transactions on Graphics (TOG), 43(6):1–18, 2024. 2, 5

    2024

  24. [32]

    Scannet++: A high-fidelity dataset of 3d in- door scenes

    Chandan Yeshwanth, Yueh-Cheng Liu, Matthias Nießner, and Angela Dai. Scannet++: A high-fidelity dataset of 3d in- door scenes. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 12–22, 2023. 5

    2023

  25. [33]

    Sigmoid loss for language image pre-training

    Xiaohua Zhai, Basil Mustafa, Alexander Kolesnikov, and Lucas Beyer. Sigmoid loss for language image pre-training. InProceedings of the IEEE/CVF international conference on computer vision, pages 11975–11986, 2023. 7, 8

    2023

  26. [34]

    Mobilesamv2: Faster segment anything to everything.arXiv preprint arXiv:2312.09579, 2023

    Chaoning Zhang, Dongshen Han, Sheng Zheng, Jinwoo Choi, Tae-Ho Kim, and Choong Seon Hong. Mobilesamv2: Faster segment anything to everything.arXiv preprint arXiv:2312.09579, 2023. 2

    2023 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.