Pith. sign in

REVIEW 4 major objections 7 minor 25 references

A diffusion network jointly segments glass and estimates its depth from unpaired real masks and synthetic depth, then filters broken sensors to recover metric maps robots can navigate.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

CLIP-conditioned single-step diffusion regression jointly does glass segmentation and affine-invariant depth, then aligns metric depth by masking bad sensor returns, beating prior glass methods on Mirage 18k and public sets.

T0 review reviewed 2026-07-31 challenge →

load-bearing objection Usable robotics pipeline for glass seg+depth without paired real glass labels; planar Mirage GT softens the mutual-information mechanism claim more than the abstract admits. the 4 major comments →

arxiv 2607.24249 v1 pith:M3SKXEKD submitted 2026-07-27 cs.CV cs.RO

SILICA: Repurposing Diffusion Priors for Joint Glass Segmentation and Depth Estimation

Tarun R , Anuj Verma , Laksh Nanwani , Sourav Garg , K. Madhava Krishna This is my paper
classification cs.CV cs.RO
keywords glass segmentationmonocular depth estimationdiffusion priorstransparent surfaceszero-shot transferCLIP conditioningrobotic navigationMirage 18k
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

Standard depth cameras and LiDAR fail on glass, so robots build wrong maps and risk collisions. Collecting real paired glass depth is so hard that specialized models do not generalize to new rooms. SILICA shows that the spatial priors inside a text-to-image diffusion model are enough to fix this: one shared network is taught real-world glass masks and synthetic indoor depth at the same time, routed by simple text prompts. The two tasks reinforce each other so the model learns which surface is glass and how far away it sits, without ever seeing a real glass depth label. At deployment the predicted mask throws away the sensor’s wrong glass readings and aligns the network’s depth into metric values usable for mapping and collision avoidance, with strong zero-shot results on unseen indoor scenes.

Core claim

Repurposing a text-to-image diffusion U-Net with CLIP task prompts for single-step joint latent regression lets real glass segmentation and synthetic depth teach each other a correct foreground-background hierarchy. The resulting model produces glass masks and glass-aware affine-invariant depth that transfer zero-shot, and the mask is required to filter corrupted sensor points so metric glass depth can be recovered for 3D mapping—entirely without paired real-world glass depth annotations.

What carries the argument

CLIP-conditioned joint latent regression: the phrases “predict glass segmentation” and “predict depth map” are encoded and injected into the frozen diffusion U-Net’s cross-attention at a fixed final timestep, so one forward pass yields both latents from unpaired real and synthetic inputs and forces mutual spatial hierarchy.

Load-bearing premise

Training only on synthetic indoor depth plus real glass masks is enough for the shared network to invent correct real-world glass surface depth; real glass depth labels are unnecessary.

What would settle it

Retrain the identical joint pipeline after replacing all synthetic depth with a large corpus of real paired glass depth maps (or after removing every synthetic scene that contains glass) and measure whether Mirage and GW-Depth glass-region errors drop; a large drop would show that real glass depth labels were not dispensable.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Standard RGB-D robots can build occupancy maps that treat glass as solid obstacles by mask-filtering sensor points and aligning the network depth.
  • Glass-aware perception can be deployed in new indoor layouts without collecting fresh real glass depth labels.
  • Re-engaging the unused text-conditioning path inside diffusion depth models is sufficient to specialize them for transparent surfaces.
  • Unfiltered sensor alignment produces catastrophic glass-region error; the predicted mask is a practical prerequisite for usable metric depth.
  • A single ~5 Hz forward pass plus voxel-grid fusion is already enough for static glass in real wheelchair navigation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same CLIP task-routing pattern could couple other unpaired real/synthetic pairs (for example material masks with surface normals) without new joint labels.
  • If hierarchy learning works without glass depth labels, annotation cost for mirrors, water, and other specular surfaces may fall the same way.
  • Distilling the single-step U-Net while preserving mask–depth coupling is a direct route to higher frame-rate onboard perception.
  • Extending planar marker-based depth capture to curved or multi-layer glass would test whether the learned hierarchy still holds.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper presents SILICA, a single-step deterministic latent-regression model fine-tuned from Stable Diffusion that jointly predicts glass segmentation masks and affine-invariant depth, with task routing via CLIP text prompts injected into the U-Net's pre-trained cross-attention layers. Training pairs real-world glass segmentation data (including the authors' new Mirage 18k dataset) with synthetic Hypersim depth, deliberately avoiding any paired real-world glass depth supervision. At deployment, the predicted mask filters corrupted sensor returns on glass before a least-squares scale/shift alignment of the predicted disparity to raw sensor disparity on background pixels, yielding fused metric depth. The authors report state-of-the-art segmentation on GDD/3DRef/Trans10k/Mirage/Mall and glass-aware depth on NYUv2, GW-Depth, and Mirage, with the headline "almost 20%" gain coming from Mirage δ1 (95.34 vs. Lotus-D 79.88), plus ablations on CLIP conditioning, glass-free Hypersim training, and joint vs. single-task training, and a wheelchair navigation deployment.

Significance. If the results hold, this is a practically useful contribution for robot navigation in glass-heavy environments: a modular, monocular, zero-shot pipeline with an integrated sensor-fusion step, demonstrated on a real wheelchair stack at ~5 Hz. The paper ships several concrete assets that raise confidence and reuse value: released code and project page, the Mirage 18k dataset (18,353 segmentation masks plus a 2,406-sample metric glass-depth evaluation split — a genuinely scarce resource), a VAE-reconstruction sanity check for the shared latent space (99.97% mIoU), and at least qualitative ablations of the key design choices. The central empirical result — that diffusion priors plus joint training beat both domain-specific (GW-Depth) and foundation (Lotus-D, Depth Anything V2) baselines on glass depth — is plausible and, on GW-Depth and NYUv2, supported by Tab. II. However, the strength of the evidence for the mechanistic claim (CLIP-routed mutual information teaching glass surface depth) is weaker than the strength of the language used to state it, for reasons detailed in the major comments.

major comments (4)
  1. [Sec. IV-A, Tab. II, Tab. IV] Planar ground truth is a confound for the headline depth claim. Mirage's glass depth GT is constructed by fitting a plane (in inverse depth) through corner markers on each pane; GW-Depth likewise targets planar glass walls. A deterministic latent-regression U-Net trained on Hypersim will tend to continue the surrounding frame/wall surface smoothly into ambiguous transparent regions, and when the pane is coplanar with its frame that smooth continuation *is* the GT plane — no glass-specific perception required. The largest reported gain (Mirage δ1 95.34 vs. 79.88, the source of the 'almost 20%' claim) is exactly where this confound is strongest, and the paper's own Tab. IV is consistent with the weaker explanation: removing all >10%-glass Hypersim scenes costs only 0.89 AbsRel / 0.87 δ1, i.e., glass-specific depth supervision contributes almost nothing. The authors should add a control tha
  2. [Sec. V-D, Fig. 6] The CLIP-conditioning ablation — the paper's central architectural contribution — is supported only qualitatively. Fig. 6 shows one positional-switch failure pair versus one CLIP-conditioned success, and the text says the positional variant 'quickly collapses', but no quantitative comparison (mIoU, AbsRel, δ1/δ2 on any benchmark) is reported for positional vs. CLIP task routing. Given that Eq. (2) vs. Eq. (3) is the load-bearing design choice and the text uses strong language ('empirically proves that CLIP conditioning is essential'), a small quantitative ablation table is needed; this is cheap to produce since both variants were already trained.
  3. [Tab. III, Sec. V-A-2] Tab. III's cross-method comparison is asymmetric and the derived claim is overstated. The daggered baselines (Lotus-D, Depth Anything V2) are aligned using raw sensor points *without* glass-mask filtering, while SILICA uses its predicted mask; the resulting 'catastrophic error spike of over 85%' therefore measures the filtering step, not model quality, and it is also unclear what quantity the 85% refers to (AbsRel 161.6 vs. 24.73 is ~85% relative reduction, but this should be stated). It is also ambiguous whether GW-Depth's own segmentation output was used to filter its alignment. Please either (i) report filtered and unfiltered variants for all methods, or (ii) reframe the table explicitly as an ablation of the proposed masking/alignment protocol (e.g., SILICA with and without mask filtering). Also clarify why SILICA's glass-region AbsRel here (24.73) differs from Tab. II (18.04) — pres
  4. [Abstract, Sec. IV-D, Tab. I] The 'zero-shot' framing and the 'almost 20%' claim need qualification. SILICA trains on the official train splits of Trans10k, GDD, GSD-S, GW-Depth, and Mirage, so the Tab. I results on GDD/3DRef/Trans10k are in-domain test-split results, not zero-shot; only the Mall set (segmentation) and Mirage-depth/GW-Depth (depth) are genuinely unseen. Relatedly, Tab. I mixes protocols: GDNet, GlassSemNet, and SAM3 are inference-only with pretrained weights (†), TransLab/EBLNet are retrained, and some numbers are copied from literature (*) — this should be flagged per cell, since the table is the basis for 'consistently outperforms prior baselines'. Finally, 'outperforming state-of-the-art models by almost 20%' in the abstract is a single-dataset, single-metric figure (Mirage δ1); on GW-Depth the δ1 margin over Depth Anything V2 is 0.77 points. The abstract should scope the claim accordingly.
minor comments (7)
  1. [Eq. (3), Sec. III-B] Eq. (3) contains stray commas ('c_s,)', 'c_d,)') and still includes the positional switch variables s_s, s_d, although the text presents CLIP conditioning as replacing the positional switch. Please clarify whether the final model uses both, and clean up the notation.
  2. [Sec. III-C-1] 'In a single forward pass' is misleading: two task prompts (c_s, c_d) imply two U-Net evaluations per image. Please state the actual per-image compute, which also matters for the 5 Hz deployment figure in Sec. V-C.
  3. [Sec. III-C-1 / Sec. IV] The binary segmentation threshold value is never given, and training hyperparameters (learning rate, steps, batch size, image resolution) are absent. Code release mitigates this, but the key numbers should be in the paper for reproducibility.
  4. [Sec. V, Sec. V-B] Overclaiming/typo passages: 'This proves our initial intuition', 'paints the importance', 'empirically prove that paired glass depth annotation is not required', 'out model predicts accurate surface depth'. Please tone down the proof-language and fix the typos.
  5. [Fig. 3, Fig. 5] Fig. 3's x-axis labels are garbled ('windows clear door'); consider merging with the caption. In Fig. 5, overlaying the predicted glass mask contours on the depth maps (or adding per-pixel error maps) would make the depth comparisons substantially easier to read.
  6. [Tab. II] The GW-Depth footnote ('trained specifically on NYUv2') is confusing for the GW-Depth-dataset columns — clarify the source of those numbers (own run vs. literature) and the alignment protocol applied to baselines.
  7. [Sec. IV-A] Please state the Mirage release terms (license, split files, whether the 2,406-sample depth split scene IDs are disjoint from the segmentation train scenes, not just from the segmentation test scenes). A per-glass-type breakdown of depth metrics (using the Fig. 3 categories) would also strengthen the evaluation and partially address the planar-GT concern.

Circularity Check

0 steps flagged

No derivation-chain circularity: unpaired multi-task training and external benchmarks; metric alignment is ordinary least squares, not a tautology.

full rationale

SILICA is an empirical methods paper. Its load-bearing claim—that CLIP-conditioned joint latent regression on real glass masks plus synthetic Hypersim depth yields zero-shot glass segmentation and glass-aware depth without paired real glass-depth labels—is supported by training on external public segmentation corpora and Hypersim, then evaluating on held-out Mirage scenes, public glass sets (GDD, 3DRef, Trans10k), GW-Depth, NYUv2, and an unseen mall set. Affine metric recovery (Eq. 4) is standard least-squares scale/shift on non-glass sensor points using the predicted mask as a filter; it does not redefine the depth prediction by construction. Ablations (CLIP vs positional switcher; glass-free Hypersim; single-task baselines) are comparative, not self-definitional. Mirage serving as both contribution and primary glass-depth benchmark is ordinary dataset introduction, not a circular step: no equation reduces the reported gains to a fitted constant, no uniqueness theorem is imported from the authors, and no ansatz is smuggled in via self-citation. Concerns that planar GT construction may confound mechanism attribution are evaluation-validity issues, not circularity of the claimed derivation.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 2 invented entities

The result rests on transfer of pretrained LDM spatial priors, adequacy of frozen SD VAE for masks and depth, effectiveness of CLIP task strings as routers across sim/real modalities, planarity for Mirage GT, and standard affine depth alignment. Free choices are prompts, fixed timestep, mask threshold, and balanced dataset sizes. No new physical entities; SILICA and Mirage are engineered constructs.

free parameters (5)
  • CLIP task prompt strings = fixed English phrases
    ys = “predict glass segmentation”, yd = “predict depth map” chosen by authors to drive cross-attention routing; not derived.
  • fixed diffusion timestep t=T and noise-free x0 regression = t=T, no added noise
    Deterministic single-step setup selected for robotics latency/consistency rather than sampled from a schedule search reported in detail.
  • binary segmentation threshold after decoder = unspecified fixed threshold
    Fixed threshold converts continuous decoder output to mask used both for metrics and for sensor filtering.
  • balanced multi-task sample count 24811 = 24811 per task
    Hypersim randomly subsampled to exactly match aggregated real segmentation count to avoid task imbalance.
  • least-squares scale s and shift t on background disparity = per-image LS solution
    Per-image affine fit on non-glass sensor points; standard but central to claimed metric glass depth.
axioms (6)
  • domain assumption Pretrained Stable Diffusion U-Net priors encode transferable spatial hierarchy useful for glass surfaces after light fine-tuning.
    Inherited from Marigold/GeoWizard-style literature; invoked throughout Sec. I–III.
  • domain assumption Frozen SD VAE encoder/decoder can represent binary glass masks and affine depth with negligible error (reported mIoU 99.97% for masks).
    Sanity check in Sec. III-B; required to share one latent pathway.
  • ad hoc to paper CLIP text embeddings injected into pretrained cross-attention suffice for stable multi-task routing across real segmentation and synthetic depth domains.
    Core design choice vs positional switcher; justified by ablation Fig. 6 / Sec. V-D.
  • domain assumption Glass panes in Mirage depth GT can be modeled as 3D planes recovered from corner markers and inverse-depth least squares.
    Sec. IV-A / Fig. 4; underpins all Mirage depth metrics.
  • domain assumption Non-glass raw sensor depth is reliable enough to recover global metric scale/shift for the full prediction including glass.
    Sec. III-C metric alignment; Tab. III contrasts filtered vs unfiltered.
  • ad hoc to paper Joint multi-task training produces mutually beneficial features rather than destructive interference when properly conditioned.
    Stated hypothesis Sec. I and III-B; supported by joint-vs-single ablation summary in V-D.
invented entities (2)
  • SILICA joint latent-regression pipeline no independent evidence
    purpose: Single shared U-Net producing glass mask and glass-aware depth with CLIP task conditioning and sensor-fusion metric recovery.
    Named system combining known LDM parts; independent evidence is the empirical tables and robot demo, not an external physical prediction.
  • Mirage 18k dataset independent evidence
    purpose: Provide diverse real indoor glass masks plus a 2406-image metric glass-depth evaluation split.
    New annotated resource; falsifiable via public release and third-party re-annotation, claimed available on project page.

reviewed 2026-07-31 · how reviews work

0 comments
Cite this review

Pith. "Pith review of SILICA: Repurposing Diffusion Priors for Joint Glass Segmentation and Depth Estimation." pith.science (2026). https://pith.science/paper/M3SKXEKD

@misc{pith2026260724249,
  author       = {Pith},
  title        = {Pith review of: SILICA: Repurposing Diffusion Priors for Joint Glass Segmentation and Depth Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M3SKXEKD}},
  note         = {Machine review of arXiv:2607.24249}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Standard depth sensors systematically fail on transparent surfaces, creating corrupted 3D maps and severe navigation hazards. While specialized hardware sensors can detect glass, they lack modularity and have extensive hardware dependencies. Consequently, learning-based monocular depth estimation has emerged as a compelling alternative. However, domain-specific glass-aware monocular depth estimators struggle with unfamiliar indoor layouts; restricted by the severe scarcity of real-world glass depth annotations, they fail to generalize zero-shot to new settings. This motivates us to explore whether the extensive priors of text-to-image diffusion models can enable generalizable perception of transparent surfaces. We introduce SILICA, a unified pipeline leveraging these priors to jointly predict glass segmentation and glass-aware depth. This mutual information exchange establishes a robust spatial hierarchy, entirely eliminating the need for paired real-world glass depth annotations. Subsequently, we use the predicted segmentation mask to explicitly filter incorrect glass depth points from standard sensors, recovering accurate metric glass depth for downstream 3D mapping and autonomous collision avoidance. Supported by our novel Mirage 18k dataset, extensive experiments demonstrate that SILICA achieves remarkable zero-shot transfer across diverse, unseen environments, outperforming state-of-the-art models by almost 20% and setting a new benchmark for transparent surface perception.

Figures

Figures reproduced from arXiv: 2607.24249 by Anuj Verma, K. Madhava Krishna, Laksh Nanwani, Sourav Garg, Tarun R.

Figure 1
Figure 1. Figure 1: (Left) Conventional depth cameras and LiDAR sensors do not provide correct depth to glass, leading to misinformed mapping and navigation, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: SILICA Pipeline Overview. Our framework employs a shared U-Net architecture for joint glass segmentation and monocular depth estimation zˆ s ,zˆ d . During Training (top), we employ a fixed timestep t = T joint latent regression loss LT using CLIP-conditioning cs ,cd for task switching, where the latent U-Net learns to directly predict the annotation latents z s ,z d from input latents z xs ,z xd . During … view at source ↗
Figure 4
Figure 4. Figure 4: Acquisition of ground truth depth for Mirage 18k. The raw sensor depth image gives noisy returns for glass areas, as it is transparent to IR. The corrected depth image shows polygons with smoothly varying depth to glass, which is used for validating our results. diverse indoor environments with planar and curved trans￾parent structures, varying lighting conditions, and cluttered backgrounds. 2) Monocular D… view at source ↗
Figure 3
Figure 3. Figure 3: Diversity of type of glass in Mirage 18k. Mirage 18k provides training and testing data for several types of glass, often with multiple in the same image. 1,930 samples without any glass are also included to provide negative samples in the training set. We use the CVAT annotation tool to generate ground truth binary glass segmentation masks. Ground truth depth to glass is obtained by modeling each pane as … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison of glass segmentation and depth prediction. SILICA can jointly predict glass segmentation and glass-aware depth estimation precisely even in scenes with a busy background (seen in row 1), a half-open glass door (seen in row 2), and a glass window (seen in row 3). SILICA shows consistent performance in all these samples, even in the midst of saliency. The joint prediction approach als… view at source ↗
Figure 6
Figure 6. Figure 6: (Top) Without conditioning, task entanglement corrupts depth [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

25 extracted references · 3 linked inside Pith

  1. [1]

    Monocular depth estimation for glass walls with context: A new dataset and method,

    Y . Liang, B. Deng, W. Liu, J. Qin, and S. He, “Monocular depth estimation for glass walls with context: A new dataset and method,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 12, pp. 15 081–15 097, 2023

    2023

  2. [2]

    Enhanced boundary learning for glass-like object segmentation,

    H. He et al., “Enhanced boundary learning for glass-like object segmentation,” inProceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV), Oct. 2021, pp. 15 859–15 868

    2021

  3. [3]

    Segmenting transparent objects in the wild,

    E. Xie, W. Wang, W. Wang, M. Ding, C. Shen, and P. Luo, “Segmenting transparent objects in the wild,” inComputer Vision – ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XIII, Glasgow, United Kingdom: Springer-Verlag, 2020, pp. 696–711

    2020

  4. [4]

    Exploiting semantic relations for glass surface detection,

    J. Lin, Y .-H. Yeung, and R. Lau, “Exploiting semantic relations for glass surface detection,” inAdvances in Neural Information Pro- cessing Systems, S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, Eds., vol. 35, Curran Associates, Inc., 2022, pp. 22 490–22 504

    2022

  5. [5]

    Don’t hit me! glass detection in real-world scenes,

    H. Mei et al., “Don’t hit me! glass detection in real-world scenes,” in2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 3684–3693

    2020

  6. [6]

    Yang et al.,Depth anything v2, 2024

    L. Yang et al.,Depth anything v2, 2024

    2024

  7. [7]

    Hypersim: A photorealistic synthetic dataset for holistic indoor scene understanding,

    M. Roberts et al., “Hypersim: A photorealistic synthetic dataset for holistic indoor scene understanding,” inInternational Conference on Computer Vision (ICCV) 2021, 2021

    2021

  8. [8]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, High-resolution image synthesis with latent diffusion models, 2021

    2021

  9. [9]

    Learning transferable visual models from natu- ral language supervision,

    A. Radford et al., “Learning transferable visual models from natu- ral language supervision,” inInternational conference on machine learning, PmLR, 2021, pp. 8748–8763

    2021

  10. [10]

    Carion et al.,Sam 3: Segment anything with concepts, 2025

    N. Carion et al.,Sam 3: Segment anything with concepts, 2025

    2025

  11. [11]

    Clear grasp: 3d shape estimation of transparent objects for manipulation,

    S. Sajjan et al., “Clear grasp: 3d shape estimation of transparent objects for manipulation,” in2020 IEEE International Conference on Robotics and Automation (ICRA), 2020, pp. 3634–3642

    2020

  12. [12]

    Transcg: A large-scale real-world dataset for transparent object depth completion and a grasping baseline,

    H. Fang, H.-S. Fang, S. Xu, and C. Lu, “Transcg: A large-scale real-world dataset for transparent object depth completion and a grasping baseline,”IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 7383–7390, 2022

    2022

  13. [13]

    Keypose: Multi-view 3d labeling and keypoint estimation for transparent objects,

    X. Liu, R. Jonschkowski, A. Angelova, and K. Konolige, “Keypose: Multi-view 3d labeling and keypoint estimation for transparent objects,” in2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 11 599–11 607

    2020

  14. [14]

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

    L. Yang, B. Kang, Z. Huang, X. Xu, J. Feng, and H. Zhao, “Depth anything: Unleashing the power of large-scale unlabeled data,” in2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 10 371–10 381

    2024

  15. [15]

    B. Ke, A. Obukhov, S. Huang, N. Metzger, R. C. Daudt, and K. Schindler,Repurposing diffusion-based image generators for monocular depth estimation, 2024

    2024

  16. [16]

    Fu et al.,Geowizard: Unleashing the diffusion priors for 3d geometry estimation from a single image, 2024

    X. Fu et al.,Geowizard: Unleashing the diffusion priors for 3d geometry estimation from a single image, 2024

    2024

  17. [17]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,”arXiv preprint arxiv:2006.11239, 2020

    Pith/arXiv arXiv 2006

  18. [18]

    Dynamic dual-output diffusion models,

    Y . Benny and L. Wolf, “Dynamic dual-output diffusion models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2022, pp. 11 482–11 491

    2022

  19. [19]

    Progressive distillation for fast sampling of diffusion models,

    T. Salimans and J. Ho, “Progressive distillation for fast sampling of diffusion models,”CoRR, vol. abs/2202.00512, 2022

    Pith/arXiv arXiv 2022

  20. [20]

    He et al.,Lotus: Diffusion-based visual foundation model for high- quality dense prediction, 2025

    J. He et al.,Lotus: Diffusion-based visual foundation model for high- quality dense prediction, 2025

    2025

  21. [21]

    Wonder3d: Single image to 3d using cross-domain diffusion,

    X. Long et al., “Wonder3d: Single image to 3d using cross-domain diffusion,”arXiv preprint arXiv:2310.15008, 2023

    Pith/arXiv arXiv 2023

  22. [22]

    3DRef: 3D Dataset and Benchmark for Reflection Detection in RGB and Lidar Data,

    X. Zhao and S. Schwertfeger, “ 3DRef: 3D Dataset and Benchmark for Reflection Detection in RGB and Lidar Data,” in2024 Interna- tional Conference on 3D Vision (3DV), Los Alamitos, CA, USA: IEEE Computer Society, Mar. 2024, pp. 225–234

    2024

  23. [23]

    Indoor segmenta- tion and support inference from rgbd images,

    N. Silberman, D. Hoiem, P. Kohli, and R. Fergus, “Indoor segmenta- tion and support inference from rgbd images,” inComputer Vision – ECCV 2012, A. Fitzgibbon, S. Lazebnik, P. Perona, Y . Sato, and C. Schmid, Eds., Berlin, Heidelberg: Springer Berlin Heidelberg, 2012, pp. 746–760

    2012

  24. [24]

    Nicki Skafte Detlefsen et al.,TorchMetrics - Measuring Repro- ducibility in PyTorch, Feb. 2022

    2022

  25. [25]

    Rtab-map as an open-source lidar and visual simultaneous localization and mapping library for large-scale and long-term online operation,

    M. Labb ´e and F. Michaud, “Rtab-map as an open-source lidar and visual simultaneous localization and mapping library for large-scale and long-term online operation,”Journal of field robotics, vol. 36, no. 2, pp. 416–446, 2019

    2019

This paper was first reviewed by grok-4.5 on July 31, 2026.