Pith. sign in

REVIEW 2 major objections 5 minor 70 references

Prism: Semi-Supervised Multi-View Stereo with Monocular Structure Priors

T0 review · 2 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read The paper claims that transferring a monocular relative-depth teacher's structure through deep-feature and multi-scale SSIM losses lets semi-supervised MVS outperform all baselines on ScanNet++ and ARKitScenes.

desk verdict Prism is a solid, honestly-reported distillation method for semi-supervised MVS that deserves a serious referee, though its exact margins rest on a single run and a teacher that is trusted without confidence weighting. read the letter →

arxiv 2412.05771 v1 pith:K77GUJSL submitted 2024-12-08 cs.CV

classification cs.CV
keywords semi-supervisedlearningmulti-viewstereomonoculardepthpriorrelativeestimationdeepfeaturelosspyramidSSIMindoorscenereconstructionunlabeledvideotraining
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

The paper proposes a semi-supervised training framework, Prism, that lets multi-view stereo (MVS) networks learn from both unlabeled smartphone video and labeled synthetic renders. Its central claim is that a monocular relative-depth network trained on synthetic data can serve as a structure teacher for real video, provided the transfer is done through two perceptual-inspired losses rather than pixel-wise comparison: a deep feature loss and a pyramid-based multi-scale SSIM loss. If this works, MVS training no longer needs expensive real depth ground truth, and the paper reports that Prism beats unsupervised, synthetic-supervised, and simple semi-supervised baselines on every metric on both ScanNet++ and ARKitScenes.

What carries the argument

The load-bearing mechanism is the pair of monocular structure losses. The deep feature loss takes aligned depth maps, duplicates them into three-channel images, passes them through a pretrained image encoder, and minimizes the mean $\ell^2$ distance between channel-normalized deep embeddings of the teacher and student predictions. The pyramid SSIM loss replaces standard MS-SSIM's multiplicative scale combination with a normalized sum of single-scale SSIM computed on four downsampled levels, which the paper finds more stable during training. Both losses operate on the teacher output after it is aligned to the MVS prediction by the global scale and shift of Eq. 3, and the framework also keeps the usual unsupervised photometric, smoothness, and augmentation-consistency losses plus a supervised regression, gradient, and normal loss on synthetic data.

What would settle it

Train Prism with a deliberately broken teacher whose depth maps are replaced by constant values or by spatially shuffled patches, keeping every other training detail the same; if performance relative to the semi-supervised baseline does not drop substantially, the monocular losses are not carrying the structure prior the paper attributes to them.

Watch

Extended reading notes

Core claim

The central discovery is that comparing aligned depth maps in a learned feature space, together with a summed multi-scale SSIM statistic, transfers monocular structure far more effectively than pixel-wise $\ell^1$, SSIM, or MS-SSIM losses. After aligning the teacher's affine-invariant prediction to the MVS prediction with a global scale and shift (Eq. 3), the deep feature loss (Eq. 4) normalizes channel-wise deep embeddings of the duplicated depth images and penalizes their mean $\ell^2$ distance, while the pyramid SSIM loss (Eqs. 5–6) computes a normalized sum of SSIM over four downsampled levels. The two losses are complementary: the deep feature loss handles global structure, the statistical loss sharpens local detail, and together with the supervised and unsupervised terms they improve results on textureless and reflective surfaces and on thin structures. Every component of the framework contributes, and the monocular losses are responsible for most of the gap over the semi-supervised baseline.

Load-bearing premise

The load-bearing premise is that a monocular relative-depth network trained only on synthetic indoor renders produces trustworthy structure on real handheld smartphone video, and that its output differs from the multi-view prediction by no more than a global scale and shift.

Editorial extensions

If this is right

  • If Prism's claim is correct, MVS networks can be trained on large unlabeled video corpora plus small synthetic sets, with the monocular teacher supplying the structure that photometric losses miss on reflective and textureless surfaces.
  • The gains transfer: a model trained on ScanNet++ videos and Hypersim synthetics also improves on ARKitScenes without fine-tuning, suggesting the learned prior is not dataset-specific.
  • The ablation results imply that the choice of loss dominates the choice of teacher: both an out-of-the-box monocular network and one retrained on Hypersim produce similar improvements when combined with the two structure losses.
  • The paper's pyramid SSIM formulation and deep-feature comparison each independently outperform pixel-wise $\ell^1$, SSIM, and MS-SSIM at transferring monocular depth structure.

Reading between the lines

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

  • The same deep-feature plus multi-scale-statistics recipe could be applied to transfer other affine-invariant dense predictions, such as surface normals or illumination, to a multiview student, although the paper only demonstrates depth.
  • Because the scale-shift alignment of Eq. 3 is global, a scene where the teacher and MVS disagree locally—for example, a large mirror—could pull the student toward the teacher's hallucinated geometry; a confidence-weighted or region-masked variant of the feature loss is a natural untested extension.
  • One could isolate the value of the teacher by fixing all training settings and swapping the teacher's predictions for the rendered ground-truth depth of the unlabeled frames; if Prism's advantage shrinks, the framework's gain really is the transferable structure prior rather than the extra regularizer.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper proposes Prism, a semi-supervised multi-view stereo (MVS) training framework that jointly uses unlabeled real smartphone video and labeled synthetic data. The key idea is to distill structural priors from a monocular relative-depth network (Marigold) into the MVS prediction via two novel losses: a deep feature loss inspired by LPIPS and a pyramid-SSIM statistical loss. The monocular prediction is aligned to the MVS prediction through a global least-squares scale/shift fit, and the losses are applied on the unlabeled real data. The total loss combines these monocular losses with an unsupervised loss (DIV loss) and a supervised loss on synthetic data. Experiments on ScanNet++ and ARKitScenes show consistent improvements over unsupervised, synthetic-supervised, and semi-supervised baselines, with ablations demonstrating that each component contributes positively.

Significance. If the results hold, this is a valuable contribution to semi-supervised MVS. The idea of transferring structural priors from a monocular relative-depth network using perceptual-similarity-inspired losses is novel and practically motivated. The paper provides extensive empirical validation: comparisons to multiple baselines on two datasets, thorough ablations, and a study across three different monocular teachers. The gains are large and consistent across metrics, and the analysis of LPIPS-induced artifacts shows a thoughtful design process. The paper is clearly written and the methodology is reproducible. The central claim—that the proposed losses transfer structure better than pixel-wise alternatives—is well supported by the ablation study.

major comments (2)
  1. [§3.1, Eq. (3)] The global least-squares scale/shift alignment in Eq. (3) is the only mechanism that reconciles the monocular teacher's relative depth with the MVS prediction, yet it is computed over all pixels with no confidence weighting. The paper does not analyze how erroneous MVS predictions (e.g., on textureless or reflective surfaces) bias the estimated (s,t), nor does it bound the damage from a teacher that is systematically wrong on certain structures. The only safeguard is the one-epoch delay in §3.3, which addresses initial scale but not structured teacher errors. Because Lmono is the central contribution, the absence of a failure analysis or a confidence-masked variant leaves a load-bearing assumption unvalidated.
  2. [§4.5, Fig. 6] The paper shows that a deep feature loss based on LPIPS causes hallucinations that look like sharp details and reports that switching to the SD-v2 encoder alleviates this artifact. However, it does not investigate whether the final loss combination can still impose incorrect structure when the teacher disagrees with multi-view photometric evidence. A per-scene error breakdown or an experiment with a deliberately corrupted teacher would clarify whether the improvements are robust to teacher errors, rather than merely reflecting cases where the teacher happens to be correct.
minor comments (5)
  1. [Abstract] The phrase 'This is a best-case-scenario result' is unclear; please clarify whether it refers to the experimental setup or the outcome.
  2. [§3.1] The duplication of the depth map into three channels before feature extraction is motivated only by the feature extractor's input format; a brief justification of why a single-channel input is not used would improve clarity.
  3. [§3.1] The claim that comparing only the deepest feature embedding gives the best results is not supported by a visible ablation; consider reporting this comparison or citing a supplementary.
  4. [Table 2] The rows are numbered 0-9, but the text refers to them inconsistently; consider labeling rows explicitly with names or using the same numbering consistently in the text.
  5. [Title and text] There are a few typographical errors, e.g., 'S tructure' in the title and 'confirming' instead of 'confirming', that should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the monocular-prior transfer is validated on held-out real depth, and the self-citations are not load-bearing.

full rationale

Prism is a knowledge-distillation and semi-supervised pipeline, not a derivation that reduces to its inputs. The monocular teacher is trained on the labeled synthetic set (Sec. 3.1), then used through Eq. 3-7 to define a structure-transfer loss on unlabeled real frames. The resulting MVS predictions are evaluated against held-out laser-scan ground truth on ScanNet++ and ARKitScenes validation (Sec. 4.2), data that does not enter the teacher or the losses, so the reported improvements cannot be true by construction. The claimed novelty is the pair of losses in Eqs. 4 and 5-6, and the paper supports that claim with controlled ablations (Table 2, rows 3-9) and with an independently pre-trained Omnidata teacher (Table 3), which breaks any self-citation chain. The paper does cite the authors' own DIV loss [38] and 3DVNet [37], and it uses DIV as both a baseline and as Lunsup; however, the central improvement claim is attributed to Lmono and is verified by adding Lmono to the same DIV baseline, so these citations are not load-bearing. The affine scale/shift fit of Eq. 3 aligns the teacher to the current MVS prediction, but the residual after the least-squares fit still contains the structure being learned, so the comparison is not identity-by-construction. The assumption that a Hypersim-trained teacher provides reliable structure on real indoor imagery is a genuine correctness and robustness risk, especially given the LPIPS artifact failure shown in Fig. 6, but that is an empirical limitation, not circular reasoning. No step of the form 'X is defined in terms of Y' or 'fitted parameter renamed as prediction' appears in the paper.

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

The framework introduces no new physical or ontological entities; its contributions are loss functions and a training scheme. The free parameters are loss weights and normalization constants set by hand without validation-based tuning, and the load-bearing domain assumptions concern the transferability of monocular priors and the effectiveness of DIV losses on real video.

free parameters (5)
  • Alpha (monocular loss balance) = 1.0
    Hand-set weight in Eq. 7 balancing the deep feature loss and pyramid SSIM loss; not tuned on a validation set.
  • Pyramid levels L = 4
    Number of scales in Eq. 6, chosen by hand.
  • Total loss weights lambda1, lambda2, lambda3 = 10, 1, 10
    Hand-set in Sec. 3.3 to balance the three loss terms; the authors state that better values may exist with validation-set tuning.
  • DIV loss weights alpha1, alpha2, alpha3, alpha4 = 12.0, 6.0, 18.0, 1.0 (with alpha3 and alpha4 scaled by 100)
    Inherited from the prior DIV loss paper [38] and hand-adjusted to account for dataset scale differences (Sec. 3.2).
  • Quantile normalization q2, q98 = 2nd and 98th percentiles
    Robust normalization in Eq. 2, chosen by hand to map monocular depth to the [0,1] range.
assumptions (5)
  • domain assumption Camera parameters (intrinsics and poses) are known for both labeled and unlabeled training data.
    Stated in Sec. 3: 'Each is assumed to have known camera parameters.' The photometric, supervised, and monocular loss machinery all depend on this.
  • domain assumption A monocular depth network trained on synthetic data yields a structure prior that transfers to real indoor scenes.
    The central enabling premise of Sec. 3.1; tested indirectly through results and Table 3, but not proven in general.
  • domain assumption The DIV unsupervised losses (photometric, smoothness, augmentation consistency) provide effective supervision on real smartphone video.
    Adopted from Rich et al. [38] in Eq. 8 without modification; it is load-bearing for the unsupervised part of the training.
  • standard math The least-squares scale and shift alignment (Eq. 3) has a unique analytic solution and adequately resolves the monocular-to-MVS ambiguity for structure transfer.
    Standard closed-form linear regression; the adequacy for depth-map pairs is assumed rather than derived.
  • domain assumption SSIM and pyramid SSIM formulations apply meaningfully to depth maps as image-like signals.
    The paper's key perceptual-metric analogy, validated by ablations but assumed in the loss design.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Prism: Semi-Supervised Multi-View Stereo with Monocular Structure Priors." pith.science (2026). https://pith.science/paper/K77GUJSL

@misc{pith2026241205771,
  author       = {Pith},
  title        = {Pith review of: Prism: Semi-Supervised Multi-View Stereo with Monocular Structure Priors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/K77GUJSL}},
  note         = {Machine review of arXiv:2412.05771}
}
read the original abstract

The promise of unsupervised multi-view-stereo (MVS) is to leverage large unlabeled datasets, yet current methods underperform when training on difficult data, such as handheld smartphone videos of indoor scenes. Meanwhile, high-quality synthetic datasets are available but MVS networks trained on these datasets fail to generalize to real-world examples. To bridge this gap, we propose a semi-supervised learning framework that allows us to train on real and rendered images jointly, capturing structural priors from synthetic data while ensuring parity with the real-world domain. Central to our framework is a novel set of losses that leverages powerful existing monocular relative-depth estimators trained on the synthetic dataset, transferring the rich structure of this relative depth to the MVS predictions on unlabeled data. Inspired by perceptual image metrics, we compare the MVS and monocular predictions via a deep feature loss and a multi-scale statistical loss. Our full framework, which we call Prism, achieves large quantitative and qualitative improvements over current unsupervised and synthetic-supervised MVS networks. This is a best-case-scenario result, opening the door to using both unlabeled smartphone videos and photorealistic synthetic datasets for training MVS networks.

Figures

Figures reproduced from arXiv: 2412.05771 by the authors.

Figure 1
Figure 1. Using structure priors from a monocular relative- [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of Prism framework for semi-supervised MVS. We leverage both real unlabeled smartphone data and labeled synthetic data to train MVS networks. Our central idea is to (1) train an existing monocular relative-depth prediction network on the synthetic set in order to capture high-quality structure priors and (2) teach the MVS network to use these structural priors on the unlabeled set via losses inspired by per… view at source ↗
Figure 3
Figure 3. Visual Ablation Study. Each component of Prism interacts constructively, helping bring out fine detail (top row) and global structure (bottom row). The monocular structure prior significantly helps in the case of textureless/reflective surfaces. The supervised loss helps add a final smoothness to flat surfaces and sharpness to object boundaries. See Sec. 4.5 for details and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative Results. Prism outperforms all baselines, producing sharp and accurate depth maps with excellent global structure and fine-grained local detail. With the arrows, we indicate hard cases where Prism performs well: textureless and reflective surfaces (rows 1, …
Figure 5
Figure 5. Figure 5: Visual Ablation Study. We isolate the effect of various monocular losses. Only our multi-scale loss ℓssim and our deep feature loss ℓfeat can handle confusing geometry (top row) and textureless surfaces (bottom row). See [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: LPIPS causes artifacts. We initially tried LPIPS [69], and found that it causes spurious artifacts. Though these look like sharp details, they are actually hallucinations. Our solution allevi￾ates this problem. See Sec 4.5 for details and discussion. unsup. Omni- Marig…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 67 canonical work pages

  1. [1]

    SceneScript: Reconstructing scenes with an autoregr es- sive structured language model

    Armen Avetisyan, Christopher Xie, Henry Howard-Jenkin s, Tsun-Yi Yang, Samir Aroudj, Suvam Patra, Fuyang Zhang, Duncan Frost, Luke Holland, Campbell Orme, Jakob En- gel, Edward Miller, Richard Newcombe, and V asileios Bal- ntas. SceneScript: Reconstructing scenes with an autoregr es- sive structured language model. In European Conference on Computer Visio...

  2. [2]

    ARK- itscenes - a diverse real-world dataset for 3D indoor scene understanding using mobile RGB-D data

    Gilad Baruch, Zhuoyuan Chen, Afshin Dehghan, Tal Dimry, Y uri Feigin, Peter Fu, Thomas Gebauer, Brandon Joffe, Daniel Kurz, Arik Schwartz, and Elad Shulman. ARK- itscenes - a diverse real-world dataset for 3D indoor scene understanding using mobile RGB-D data. In Advances in Neural Information Processing Systems, 2021. 2, 5, 7

  3. [3]

    Richter, and Vladlen Koltun

    Aleksei Bochkovskii, Ama¨ el Delaunoy, Hugo Germain, Marcel Santos, Yichao Zhou, Stephan R. Richter, and Vladlen Koltun. Depth Pro: Sharp monocular metric depth in less than a second. arXiv, 2024. 2

  4. [4]

    Vir- tual kitti 2

    Y ohann Cabon, Naila Murray, and Martin Humenberger. Vir- tual kitti 2. arXiv, 2020. 8

  5. [5]

    MVSFormer: Multi-view stereo by learning robust image features and temperature-based depth

    Chenjie Cao, Xinlin Ren, and Yanwei Fu. MVSFormer: Multi-view stereo by learning robust image features and temperature-based depth. Transactions on Machine Learn- ing Research, 2022. 1, 2

  6. [6]

    RC- MVSNet: Unsupervised multi-view stereo with neural ren- dering

    Di Chang, Aljaˇ z Boˇ ziˇ c, Tong Zhang, Qingsong Yan, Ying- cong Chen, Sabine S¨ usstrunk, and Matthias Nießner. RC- MVSNet: Unsupervised multi-view stereo with neural ren- dering. In European Conference on Computer Vision , 2022. 1, 2, 5, 7

  7. [7]

    VCR-GauS: View consistent depth-normal regularizer for gaussian surface reconstruction

    Hanlin Chen, Fangyin Wei, Chen Li, Tianxin Huang, Y un- song Wang, and Gim Hee Lee. VCR-GauS: View consistent depth-normal regularizer for gaussian surface reconstruction. Advances in Neural Information Processing Systems , 2024. 2

  8. [8]

    Depth-regularized optimization for 3D gaussian splatting in few-shot images

    Jaeyoung Chung, Jeongtaek Oh, and Kyoung Mu Lee. Depth-regularized optimization for 3D gaussian splatting in few-shot images. In Conference on Computer Vision and Pattern Recognition W orkshops, 2024. 2

Show all 70 references
  1. [9]

    Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner

    Angela Dai, Angel X. Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. ScanNet: Richly-annotated 3D reconstructions of indoor scenes. In Conference on Computer Vision and Pattern Recognition , 2017. 1

  2. [10]

    MVS2: Deep unsupervised multi-view stereo with multi-view sym- metry

    Y uchao Dai, Zhidong Zhu, Zhibo Rao, and Bo Li. MVS2: Deep unsupervised multi-view stereo with multi-view sym- metry. In International Conference on 3D Vision , 2019. 1, 2

  3. [11]

    TransMVS- Net: Global context-aware multi-view stereo network with transformers

    Yikang Ding, Wentao Y uan, Qingtian Zhu, Haotian Zhang, Xiangyue Liu, Y uanjiang Wang, and Xiao Liu. TransMVS- Net: Global context-aware multi-view stereo network with transformers. In Conference on Computer Vision and Pat- tern Recognition, 2022. 1, 2

  4. [12]

    KD-MVS: Knowledge dis- tillation based self-supervised learning for multi-view stereo

    Yikang Ding, Qingtian Zhu, Xiangyue Liu, Wentao Y uan, Haotian Zhang, and Chi Zhang. KD-MVS: Knowledge dis- tillation based self-supervised learning for multi-view stereo. In European Conference on Computer Vision, 2022. 2

  5. [13]

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

    Arda D¨ uzc ¸eker, Silvano Galliani, Christoph V ogel, P ablo Speciale, Mihai Dusmanu, and Marc Pollefeys. Deep- VideoMVS: Multi-view stereo on video with recurrent spatio-temporal fusion. In Conference on Computer Vision and Pattern Recognition, 2021. 2

  6. [14]

    Omnidata: A scalable pipeline for making multi- task mid-level vision datasets from 3d scans

    Ainaz Eftekhar, Alexander Sax, Jitendra Malik, and Ami r Zamir. Omnidata: A scalable pipeline for making multi- task mid-level vision datasets from 3d scans. In International Conference on Computer Vision, pages 10786–10796, 2021. 2, 8

  7. [15]

    Massively parallel multiview stereopsis by surface normal diffusion

    Silvano Galliani, Katrin Lasinger, and Konrad Schindl er. Massively parallel multiview stereopsis by surface normal diffusion. In International Conference on Computer Vision , 2015. 1

  8. [16]

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

    Xiaodong Gu, Zhiwen Fan, Zuozhuo Dai, Siyu Zhu, Feitong Tan, and Ping Tan. Cascade cost volume for high-resolution multi-view stereo and stereo matching. In Conference on Computer Vision and Pattern Recognition, 2020. 1, 4, 5, 7

  9. [17]

    Lotus: Diffusion-based visual foundation model for high-quality dense prediction

    Jing He, Haodong Li, Wei Yin, Yixun Liang, Leheng Li, Kaiqiang Zhou, Hongbo Liu, Bingbing Liu, and Ying-Cong Chen. Lotus: Diffusion-based visual foundation model for high-quality dense prediction. arXiv, 2024. 2, 3

  10. [18]

    Multi-view stereo by temporal nonparametric fusion

    Y uxin Hou, Juho Kannala, and Arno Solin. Multi-view stereo by temporal nonparametric fusion. In International Conference on Computer Vision, 2019. 2

  11. [19]

    ConsistentNeRF: Enhancing neural radiance fields with 3D consistency for sparse view synthesis

    Shoukang Hu, Kaichen Zhou, Kaiyu Li, Longhui Y u, Lan- qing Hong, Tianyang Hu, Zhenguo Li, Gim Hee Lee, and Zi- wei Liu. ConsistentNeRF: Enhancing neural radiance fields with 3D consistency for sparse view synthesis. In arXiv, 2023. 2

  12. [20]

    M3VSNet: Unsupervised multi-metric multi-view stereo network

    Baichuan Huang, Hongwei Yi, Can Huang, Yijia He, Jingbi n Liu, and Xiao Liu. M3VSNet: Unsupervised multi-metric multi-view stereo network. In International Conference on Image Processing, 2021. 1, 2

  13. [21]

    DPSNet: End-to-end deep plane sweep stereo

    Sunghoon Im, Hae-Gon Jeon, Stephen Lin, and In So Kweon. DPSNet: End-to-end deep plane sweep stereo. In In- ternational Conference on Learning Representations , 2019. 2

  14. [22]

    Large scale multi-view stereopsis evalu- ation

    Rasmus Jensen, Anders Dahl, George V ogiatzis, Engil To la, and Henrik Aanæs. Large scale multi-view stereopsis evalu- ation. In Conference on Computer Vision and Pattern Recog- nition, pages 406–413. IEEE, 2014. 1

  15. [23]

    Repurpos- ing diffusion-based image generators for monocular depth estimation

    Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Met- zger, Rodrigo Caye Daudt, and Konrad Schindler. Repurpos- ing diffusion-based image generators for monocular depth estimation. In Conference on Computer Vision and Pattern Recognition, 2024. 2, 3, 8

  16. [24]

    Chang, and Manolis Savva

    Mukul Khanna*, Y ongsen Mao*, Hanxiao Jiang, Sanjay Haresh, Brennan Shacklett, Dhruv Batra, Alexander Clegg, Eric Undersander, Angel X. Chang, and Manolis Savva. Habitat Synthetic Scenes Dataset (HSSD-200): An Analy- sis of 3D Scene Scale and Realism Tradeoffs for ObjectGoal N...

  17. [25]

    Learning unsupervised multi-view stereopsis via robust 9 photometric consistency

    Tejas Khot, Shubham Agrawal, Shubham Tulsiani, Christoph Mertz, Simon Lucey, and Martial Hebert. Learning unsupervised multi-view stereopsis via robust 9 photometric consistency. arXiv preprint arXiv:1905.02706 ,

  18. [26]

    Just a few points are all you need for multi-view stereo: A novel semi-supervised learning metho d for multi-view stereo

    Taekyung Kim, Jaehoon Choi, Seokeon Choi, Dongki Jung, and Changick Kim. Just a few points are all you need for multi-view stereo: A novel semi-supervised learning metho d for multi-view stereo. In International Conference on Com- puter Vision, 2021. 2

  19. [27]

    Kingma and Jimmy Ba

    Diederik P . Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In International Conference on Learning Representations, 2015. 5

  20. [28]

    DS-MVSNet: Unsupervised multi-view stereo via depth synthesis

    Jingliang Li, Zhengda Lu, Yiqun Wang, Ying Wang, and Jun Xiao. DS-MVSNet: Unsupervised multi-view stereo via depth synthesis. In ACM International Conference on Multi- media, 2022. 1, 2

  21. [29]

    MegaDepth: Learning singl e- view depth prediction from internet photos

    Zhengqi Li and Noah Snavely. MegaDepth: Learning singl e- view depth prediction from internet photos. In Conference on Computer Vision and Pattern Recognition, 2018. 2

  22. [30]

    WT-MVSNet: Window-based transformers for multi-view stereo

    Jinli Liao, Yikang Ding, Y oli Shavit, Dihe Huang, Shiha o Ren, Jia Guo, Wensen Feng, and Kai Zhang. WT-MVSNet: Window-based transformers for multi-view stereo. In Ad- vances in Neural Information Processing Systems , 2022. 1, 2

  23. [31]

    P-MVSNet: Learning patch-wise matching confidence aggregation for multi-view stereo

    Keyang Luo, Tao Guan, Lili Ju, Haipeng Huang, and Yawei Luo. P-MVSNet: Learning patch-wise matching confidence aggregation for multi-view stereo. In International Confer- ence on Computer Vision, 2019. 2

  24. [32]

    Generalized binary search network for highly-efficient multi-view stereo

    Zhenxing Mi, Chang Di, and Dan Xu. Generalized binary search network for highly-efficient multi-view stereo. In Conference on Computer Vision and Pattern Recognition , 2022. 1

  25. [33]

    Atlas: End- to-end 3D scene reconstruction from posed images

    Zak Murez, Tarrence van As, James Bartolozzi, Ayan Sinh a, Vijay Badrinarayanan, and Andrew Rabinovich. Atlas: End- to-end 3D scene reconstruction from posed images. In Con- ference on Computer Vision and Pattern Recognition , 2020. 5

  26. [34]

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

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Rai- son, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, L...

  27. [35]

    Self- supervised multi-view stereo via inter and intra network pseudo depth

    Ke Qiu, Yawen Lai, Shiyi Liu, and Ronggang Wang. Self- supervised multi-view stereo via inter and intra network pseudo depth. In International Conference on Multimedia , 2022. 1, 2

  28. [36]

    Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer

    Ren´ e Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, and Vladlen Koltun. Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer. IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, 44(3), 2022. 2, 3

  29. [37]

    3DVNet: Multi-view depth prediction and volu- metric refinement

    Alexander Rich, Noah Stier, Pradeep Sen, and Tobias H¨ ollerer. 3DVNet: Multi-view depth prediction and volu- metric refinement. In International Conference on 3D Vision, 2021. 1, 2

  30. [38]

    Smoothness, synthesis, and sampling: Re-thinking unsuper - vised multi-view stereo with DIV loss

    Alex Rich, Noah Stier, Pradeep Sen, and Tobias H¨ ollere r. Smoothness, synthesis, and sampling: Re-thinking unsuper - vised multi-view stereo with DIV loss. In European Confer- ence on Computer Vision, 2024. 1, 2, 3, 4, 5, 7

  31. [39]

    Susskind

    Mike Roberts, Jason Ramapuram, Anurag Ranjan, Atulit Kumar, Miguel Angel Bautista, Nathan Paczan, Russ Webb, and Joshua M. Susskind. Hypersim: A photorealistic syn- thetic dataset for holistic indoor scene understanding. In In- ternational Conference on Computer Vision , 2021....

  32. [40]

    Barron, Ben Mildenhall, Pratul P

    Barbara Roessle, Jonathan T. Barron, Ben Mildenhall, Pratul P . Srinivasan, and Matthias Nießner. Dense depth pri- ors for neural radiance fields from sparse input views. In Conference on Computer Vision and Pattern Recognition , 2022. 2

  33. [41]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨ orn Ommer. High-resolution image syn- thesis with latent diffusion models. In Conference on Com- puter Vision and Pattern Recognition, 2022. 2, 4

  34. [42]

    Sim- pleRecon: 3D reconstruction without 3D convolutions

    Mohamed Sayed, John Gibson, Jamie Watson, Victor Prisacariu, Michael Firman, and Cl´ ement Godard. Sim- pleRecon: 3D reconstruction without 3D convolutions. In European Conference on Computer Vision, 2022. 2, 4

  35. [43]

    Pixelwise view selection for un- structured multi-view stereo

    Johannes Lutz Sch¨ onberger, Enliang Zheng, Marc Pollefeys, and Jan-Michael Frahm. Pixelwise view selection for un- structured multi-view stereo. In European Conference on Computer Vision, 2016. 1

  36. [44]

    V ery deep convo- lutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. V ery deep convo- lutional networks for large-scale image recognition. In In- ternational Conference on Learning Representations , 2015. 8

  37. [45]

    Efficien t large scale multi-view stereo for ultra high resolution ima ge sets

    Engin Tola, Christoph Strecha, and Pascal Fua. Efficien t large scale multi-view stereo for ultra high resolution ima ge sets. Machine Vision and Applications, 23, 2011. 1

  38. [46]

    SparseNeRF: Distilling depth ranking for few-shot novel view synthesis

    Guangcong Wang, Zhaoxi Chen, Chen Change Loy, and Zi- wei Liu. SparseNeRF: Distilling depth ranking for few-shot novel view synthesis. In International Conference on Com- puter Vision, 2023. 2

  39. [47]

    NeuRIS: Neural reconstruction of indoor scenes using nor- mal priors

    Jiepeng Wang, Peng Wang, Xiaoxiao Long, Christian Theobalt, Taku Komura, Lingjie Liu, and Wenping Wang. NeuRIS: Neural reconstruction of indoor scenes using nor- mal priors. In European Conference on Computer Vision , 2022. 2

  40. [48]

    Zhou Wang and A.C. Bovik. A universal image quality in- dex. IEEE Signal Processing Letters , 9(3):81–84, 2002. 4

  41. [49]

    Multi- scale structural similarity for image quality assessment

    Zhou Wang, Eero P Simoncelli, and Alan C Bovik. Multi- scale structural similarity for image quality assessment. In Asilomar Conference on Signals, Systems, & Computers ,

  42. [50]

    Bovik, H.R

    Zhou Wang, A.C. Bovik, H.R. Sheikh, and E.P . Simoncelli . Image quality assessment: from error visibility to structu ral similarity. IEEE Transactions on Image Processing , 13(4): 600–612, 2004. 4

  43. [51]

    Toward practical monocular indoor depth estimation

    Cho-Ying Wu, Jialiang Wang, Michael Hall, Ulrich Neu- mann, and Shuochen Su. Toward practical monocular indoor depth estimation. In Conference on Computer Vision and Pattern Recognition, 2022. 2 10

  44. [52]

    RayMVSNet: Learning ray-based 1D implicit fields for ac- curate multi-view stereo

    Junhua Xi, Yifei Shi, Yijie Wang, Y ulan Guo, and Kai Xu. RayMVSNet: Learning ray-based 1D implicit fields for ac- curate multi-view stereo. In Conference on Computer Vision and Pattern Recognition, 2022. 2

  45. [53]

    DebSDF: Delving into the details and bias of neural indoor scene reconstruction

    Y uting Xiao, Jingwei Xu, Zehao Y u, and Shenghua Gao. DebSDF: Delving into the details and bias of neural indoor scene reconstruction. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 2

  46. [54]

    CL-MVSNet: Unsupervised multi-view stereo with dual-level contrasti ve learning

    Kaiqiang Xiong, Rui Peng, Zhe Zhang, Tianxing Feng, Jianbo Jiao, Feng Gao, and Ronggang Wang. CL-MVSNet: Unsupervised multi-view stereo with dual-level contrasti ve learning. In International Conference on Computer Vision , 2023. 1, 2, 5, 7

  47. [55]

    Self-supervised multi-view stereo via effectiv e co-segmentation and data-augmentation

    Hongbin Xu, Zhipeng Zhou, Y u Qiao, Wenxiong Kang, and Qiuxia Wu. Self-supervised multi-view stereo via effectiv e co-segmentation and data-augmentation. In AAAI Confer- ence on Artificial Intelligence , 2021. 2, 5

  48. [56]

    Digging into uncertainty in self-supervised multi-view stereo

    Hongbin Xu, Zhipeng Zhou, Yali Wang, Wenxiong Kang, Baigui Sun, Hao Li, and Y u Qiao. Digging into uncertainty in self-supervised multi-view stereo. In International Con- ference on Computer Vision, 2021. 1, 2

  49. [57]

    Semi-supervised deep multi-view stereo

    Hongbin Xu, Weitao Chen, Yang Liu, Zhipeng Zhou, Hai- hong Xiao, Baigui Sun, Xuansong Xie, and Wenxiong Kang. Semi-supervised deep multi-view stereo. In International Conference on Multimedia, 2023. 2

  50. [58]

    Learning inverse depth re- gression for multi-view stereo with correlation cost volum e

    Qingshan Xu and Wenbing Tao. Learning inverse depth re- gression for multi-view stereo with correlation cost volum e. In AAAI Conference on Artificial Intelligence , 2019. 5

  51. [59]

    Alvarez, and Miaomiao Liu

    Jiayu Yang, Wei Mao, Jose M. Alvarez, and Miaomiao Liu. Cost volume pyramid based depth inference for multi-view stereo. In Conference on Computer Vision and Pattern Recognition, 2020. 1

  52. [60]

    Alvarez, and Miaomiao Liu

    Jiayu Yang, Jose M. Alvarez, and Miaomiao Liu. Self- supervised learning of depth inference for multi-view ster eo. In Conference on Computer Vision and Pattern Recognition,

  53. [61]

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

    Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Jias hi Feng, and Hengshuang Zhao. Depth anything: Unleashing the power of large-scale unlabeled data. In Conference on Computer Vision and Pattern Recognition, 2024. 2

  54. [62]

    Depth any- thing v2

    Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xiao- gang Xu, Jiashi Feng, and Hengshuang Zhao. Depth any- thing v2. arXiv:2406.09414, 2024. 2

  55. [63]

    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. In European Conference on Computer Vision , 2018. 1, 2

  56. [64]

    ScanNet++: A high-fidelity dataset of 3d indoor scenes

    Chandan Yeshwanth, Y ueh-Cheng Liu, Matthias Nießner, and Angela Dai. ScanNet++: A high-fidelity dataset of 3d indoor scenes. In International Conference on Computer Vi- sion, 2023. 2, 5, 7

  57. [65]

    Pyramid multi-view stereo net with self-adaptive view aggregation

    Hongwei Yi, Zizhuang Wei, Mingyu Ding, Runze Zhang, Yisong Chen, Guoping Wang, and Y u-Wing Tai. Pyramid multi-view stereo net with self-adaptive view aggregation . In European Conference on Computer Vision, 2020. 2

  58. [66]

    Monosdf: Exploring monocu- lar geometric cues for neural implicit surface reconstruct ion

    Zehao Y u, Songyou Peng, Michael Niemeyer, Torsten Sat- tler, and Andreas Geiger. Monosdf: Exploring monocu- lar geometric cues for neural implicit surface reconstruct ion. Advances in Neural Information Processing Systems , 2022. 2

  59. [67]

    A Semi-Supervised Method for PatchMatch Multi-View Stereo with Sparse Points

    Weida Zhan, Keliang Cao, Yichun Jiang, Y u Chen, Jiale Wang, and Yang Hong. A Semi-Supervised Method for PatchMatch Multi-View Stereo with Sparse Points. Photon- ics, 9(12):983, 2022. 2

  60. [68]

    ElasticMVS: Learning elastic part rep- resentation for self-supervised multi-view stereopsis

    Jinzhi Zhang, Ruofan Tang, Zheng Cao, Jing Xiao, Ruqi Huang, and Lu Fang. ElasticMVS: Learning elastic part rep- resentation for self-supervised multi-view stereopsis. I n Ad- vances in Neural Information Processing Systems , 2022. 1

  61. [69]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shech tman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In Conference on Computer Vision and Pattern Recognition, 2018. 2, 3, 4, 8

  62. [70]

    Open3D: A modern library for 3D data processing

    Qian-Yi Zhou, Jaesik Park, and Vladlen Koltun. Open3D: A modern library for 3D data processing. arXiv:1801.09847,

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.