Pith. sign in

REVIEW 4 major objections 3 minor 1 cited by

IDCNet: Guided Video Diffusion for Metric-Consistent RGBD Scene Generation with Precise Camera Control

T0 review · 4 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The abstract and the submitted full text describe different systems: one proposes camera-conditioned joint RGB-D video generation, the other zero-shot instance segmentation.

desk verdict The submission is a shell: title/abstract advertise IDCNet RGB-D video generation, but the entire body is an unrelated OC-DiT zero-shot segmentation paper, so no claim in the abstract is checkable. read the letter →

arxiv 2508.04147 v1 pith:Y4VIVJNU submitted 2025-08-06 cs.CV

Lijuan Liu , Wenfa Li , Dongbo Zhang , Shuo Wang , Shaohui Jiao This is my paper
classification cs.CV
keywords RGB-Dvideogenerationdiffusionmodelcameratrajectorycontrolgeometry-awaretransformerzero-shotinstancesegmentationObject-Conditionedmanuscriptmismatch3Dscenereconstruction
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 as submitted presents two irreconcilable documents. Its abstract claims IDC-Net, a geometry-aware diffusion model that jointly produces RGB images and depth maps under explicit camera-trajectory control, so that generated sequences can feed directly into 3D scene reconstruction without post-processing. If that claim were true, it would remove a major bottleneck in controllable scene generation, since current pipelines typically estimate depth separately or need clean-up before reconstruction. The full-text body, however, is a different manuscript about an Object-Conditioned Diffusion Transformer for zero-shot instance segmentation, and it never mentions RGB-D video, camera conditioning, or the IDC-Net name. The abstract's contribution is therefore stated but unsupported in the submitted document, and a sympathetic reader cannot verify the promised benefits from the text alone.

What carries the argument

The abstract's proposed machinery is an Image-Depth Consistency Network (IDC-Net): a unified geometry-aware diffusion model that jointly generates RGB frames and depth maps, conditioned on camera poses, with a geometry-aware transformer block to inject fine-grained camera control into the denoising process. The body text's actual machinery, if read on its own, is the Object-Conditioned Diffusion Transformer (OC-DiT), which conditions latent diffusion on template and image features via cross-attention and self-attention blocks with adaptive layer norm. The two mechanisms are disjoint; the submitted document provides no description of the former.

What would settle it

Open the manuscript and search the body for 'IDC-Net', 'depth', 'camera trajectory', or 'RGB-D' beyond the abstract; the body is a different paper on zero-shot instance segmentation and contains none of these components. That observation, verifiable by any reader, is enough to show the abstract's central claim is not established by the submitted document.

Watch

Extended reading notes

Core claim

On the abstract's own terms, the central claim is that RGB and depth can be synthesized jointly in a single diffusion model conditioned on camera trajectory, yielding metric-consistent RGB-D video sequences that need no post-processing before downstream 3D reconstruction. The accompanying mechanism is a geometry-aware transformer block for fine-grained camera control, supported by a camera-image-depth dataset with metric-aligned poses. The submitted full text, however, describes a different system: a transformer-based latent diffusion model that generates instance segmentation masks conditioned on object templates and image features, trained on large synthetic data and evaluated on standard

Load-bearing premise

The load-bearing assumption is that the submitted full text is the implementation and support for the abstract's IDC-Net claims; in this manuscript that assumption fails, so the abstract's assertions stand without backing.

Editorial extensions

If this is right

  • If IDC-Net works as claimed, generated RGB-D sequences should be usable for 3D scene reconstruction without post-processing, because depth and RGB are jointly aligned during generation.
  • Joint generation should improve inter-frame geometric consistency relative to generating RGB and depth separately, since a single model enforces shared geometric reasoning.
  • Explicit camera-trajectory conditioning should allow fine-grained control over viewpoint changes in the generated video, enabling controllable scene fly-throughs.
  • A metric-aligned dataset of RGB videos, depth maps, and camera poses would be a new resource for training and evaluating camera-conditioned generative models.

Reading between the lines

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

  • Because the full text does not describe IDC-Net, the first test for any reader is to resolve which document is authoritative: if the abstract is authoritative, the body's experimental claims are irrelevant; if the body is authoritative, the abstract's title and promises are unsupported.
  • If a future version supplies the missing method, a natural testable extension would be to compare joint RGB-D diffusion against a two-stage pipeline that generates RGB first and estimates depth afterward, measuring reconstruction error and camera-consistency metrics.
  • The claimed 'directly feed for downstream 3D reconstruction' could be operationalized by measuring reconstruction accuracy, such as Chamfer distance, from generated sequences versus ground-truth RGB-D sequences—a test the current document does not report.
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 / 3 minor

Summary. The submission is titled 'IDCNet: Guided Video Diffusion for Metric-Consistent RGBD Scene Generation with Precise Camera Control,' and its abstract promises a unified geometry-aware diffusion model that jointly synthesizes RGB images and depth maps under explicit camera-trajectory control, trained on a camera-image-depth consistent dataset, and producing sequences directly usable for 3D scene reconstruction. However, the full text is a completely different manuscript: 'Conditional Latent Diffusion Models for Zero-Shot Instance Segmentation' (OC-DiT). The body contains no description of IDC-Net, no RGB-D video generation, no depth-map or camera-pose conditioning, no geometry-aware transformer block for camera control, no RGB-D video dataset, and no 3D-reconstruction experiments. Every load-bearing component of the advertised central claim is therefore unsupported by the provided document.

Significance. If the claims in the abstract were backed by the appropriate technical content, the work could be significant for controllable RGB-D video synthesis and its downstream use in 3D reconstruction. However, the submitted manuscript does not contain that work. The OC-DiT content appears to be a self-contained paper on zero-shot instance segmentation, but it is not evidence for IDC-Net's claims. I cannot assess the significance of IDC-Net from this submission because the artifact needed for that assessment is absent.

major comments (4)
  1. [Abstract vs. Full Text] The front matter advertises IDC-Net for joint RGB-depth video generation with camera control, but the full text is a different paper on zero-shot instance segmentation (OC-DiT). The body contains no occurrence of IDC-Net, no depth maps, no camera poses, and no RGB-D video experiments. The central claim is therefore unsupported by any technical derivation, experiment, or dataset in the submitted document.
  2. [Section 3] The method section describes the OC-DiT architecture for generating instance segmentation masks, not a unified RGB-depth diffusion model. Equations (1)-(7) are standard diffusion/EDM formulations and do not model camera trajectories or geometric consistency between RGB and depth. No 'geometry-aware transformer block' for fine-grained camera control is defined or analyzed.
  3. [Section 4] The experimental section evaluates zero-shot instance segmentation on BOP benchmarks (Tables 1-4) and reports AP metrics. There are no experiments on RGB-D video generation, no camera-trajectory control metrics, no evaluation of metric-consistent depth, and no downstream 3D reconstruction results. The abstract's claim that generated sequences 'can be directly feed for downstream 3D Scene reconstruction tasks' is unverified in this document.
  4. [Overall] This is not a case of a defensible claim with incomplete evidence; the submitted document contains no evidence for the claimed method at all. The mismatch between the front matter and the body is internal and cannot be repaired by local revisions. The appropriate action is to reject and, if this was a packaging error, resubmit the correct manuscript.
minor comments (3)
  1. [Abstract] The phrase 'can be directly feed' should be 'can be directly fed.'
  2. [Throughout] If the intended submission was the OC-DiT paper, the title and abstract should be corrected accordingly; as written, the pronoun 'our method' refers to different works in the abstract and the body.
  3. [Full Text] The project URL https://idcnet-scene.github.io appears only in the abstract and is never referenced or supported by the body text.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity identified; the IDC-Net abstract has no supporting technical body, so circularity is unevaluable rather than present.

full rationale

The submitted full text is a different manuscript, 'Conditional Latent Diffusion Models for Zero-Shot Instance Segmentation' (OC-DiT), and contains no derivation of IDC-Net's claimed RGB-D video generation, camera-trajectory conditioning, geometry-aware transformer, metric-consistent depth supervision, or 3D-reconstruction readiness. Circularity analysis requires exhibiting a specific reduction of a claimed result to its own inputs, fitted parameters, or self-citations. No such reduction can be exhibited because the claimed IDC-Net derivation is absent from the document: there are no equations for joint RGB-depth synthesis, no loss function tying depth to poses, and no experiments validating the abstract's central claims. The mismatch between front matter and body is a completeness/coherence failure, not a circularity failure. Within the actual OC-DiT content, the derivation is a standard conditional latent diffusion formulation with externally evaluated benchmarks and no load-bearing self-citation chain that forces the conclusions; accordingly, no circular steps are found. Score 0 is therefore appropriate: the paper is unevaluable for circularity as submitted, but no circular dependency is demonstrated.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

Only the abstract is available; the body is unrelated. We list the domain assumption explicitly stated in the abstract.

assumptions (1)
  • domain assumption The camera-image-depth dataset provides accurate metric-aligned depth and poses as geometric supervision.
    Stated in the abstract as the basis for training; without it, metric consistency is not defined.

how reviews work

0 comments
Cite this review

Pith. "Pith review of IDCNet: Guided Video Diffusion for Metric-Consistent RGBD Scene Generation with Precise Camera Control." pith.science (2026). https://pith.science/paper/Y4VIVJNU

@misc{pith2026250804147,
  author       = {Pith},
  title        = {Pith review of: IDCNet: Guided Video Diffusion for Metric-Consistent RGBD Scene Generation with Precise Camera Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y4VIVJNU}},
  note         = {Machine review of arXiv:2508.04147}
}
read the original abstract

We present IDC-Net (Image-Depth Consistency Network), a novel framework designed to generate RGB-D video sequences under explicit camera trajectory control. Unlike approaches that treat RGB and depth generation separately, IDC-Net jointly synthesizes both RGB images and corresponding depth maps within a unified geometry-aware diffusion model. The joint learning framework strengthens spatial and geometric alignment across frames, enabling more precise camera control in the generated sequences. To support the training of this camera-conditioned model and ensure high geometric fidelity, we construct a camera-image-depth consistent dataset with metric-aligned RGB videos, depth maps, and accurate camera poses, which provides precise geometric supervision with notably improved inter-frame geometric consistency. Moreover, we introduce a geometry-aware transformer block that enables fine-grained camera control, enhancing control over the generated sequences. Extensive experiments show that IDC-Net achieves improvements over state-of-the-art approaches in both visual quality and geometric consistency of generated scene sequences. Notably, the generated RGB-D sequences can be directly feed for downstream 3D Scene reconstruction tasks without extra post-processing steps, showcasing the practical benefits of our joint learning framework. See more at https://idcnet-scene.github.io.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. GeoWorld: Providing Full-frame Geometry Features to Facilitate 3D Scene Generation

    cs.CV 2025-11 conditional novelty 6.0 of 10

    GeoWorld improves image-to-3D scene generation by conditioning a video-diffusion model on full-frame geometry features extracted by a multi-view geometry model, yielding higher PSNR/SSIM/LPIPS than prior methods.

Reference graph

Works this paper leans on

50 extracted references · 38 canonical work pages · cited by 1 Pith paper

  1. [1]

    T. Amit, T. Shaharbany, E. Nachmani, and L. Wolf. Segdiff: Image segmentation with diffusion probabilistic models. arXiv preprint arXiv:2112.00390, 2021. 2

    arXiv 2021

  2. [2]

    Brachmann, A

    E. Brachmann, A. Krull, F. Michel, S. Gumhold, J. Shotton, and C. Rother. Learning 6d object pose estimation using 3d object coordinates. In ECCV, pages 536–551. Springer,

    work page

  3. [3]

    Carion, F

    N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko. End-to-end object detection with trans- formers. In ECCV, pages 213–229. Springer International Publishing, 2020. 1

    work page 2020

  4. [4]

    Zeropose: Cad-model-based zero-shot pose estimation

    Jianqiu Chen, Mingshan Sun, Tianpeng Bao, Rui Zhao, Li- wei Wu, and Zhenyu He. Zeropose: Cad-model-based zero-shot pose estimation. arXiv preprint arXiv:2305.17934,

    arXiv

  5. [5]

    Deitke, D

    M. Deitke, D. Schwenk, J. Salvador, L. Weihs, O. Michel, E. VanderBilt, L. Schmidt, K. Ehsani, A. Kembhavi, and A. Farhadi. Objaverse: A universe of annotated 3d objects. In CVPR, pages 13142–13153, 2023. 6, 11, 12, 13

    work page 2023

  6. [6]

    Denninger, M

    M. Denninger, M. Sundermeyer, D. Winkelbauer, Y . Zidan, D. Olefir, M. Elbadrawy, A. Lodhi, and H. Katam. Blender- proc. arXiv preprint arXiv:1911.01911, 2019. 6, 11, 13

    arXiv 1911

  7. [7]

    Dhariwal and A

    P. Dhariwal and A. Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021. 2

    work page 2021

  8. [8]

    Dosovitskiy

    A. Dosovitskiy. An image is worth 16x16 words: Trans- formers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. 4, 5

    arXiv 2010

Show all 50 references
  1. [9]

    Downs, A

    L. Downs, A. Francis, N. Koenig, B. Kinman, R. Hick- man, K. Reymann, T. McHugh, and V . Vanhoucke. Google scanned objects: A high-quality dataset of 3d scanned house- hold items. In ICRA, pages 2553–2560, 2022. 6, 13

    2022

  2. [10]

    Y . Du, Y . Xiao, and V . Lepetit. Learning to better segment objects from unseen classes with unlabeled videos. In ICCV, pages 3375–3384, 2021. 2

    2021

  3. [11]

    Durner, W

    M. Durner, W. Boerdijk, M. Sundermeyer, W. Friedl, Z.-C. M´arton, and R. Triebel. Unknown object segmentation from stereo images. In IROS, pages 4823–4830. IEEE, 2021. 1, 2

    2021

  4. [12]

    K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick. Mask R- CNN. In ICCV, pages 2980–2988, 2017. 1

    2017

  5. [13]

    Hendrycks and K

    D. Hendrycks and K. Gimpel. Gaussian error linear units (gelus). arXiv preprint arXiv:1606.08415, 2016. 5

    2016 arXiv

  6. [14]

    J. Ho, A. Jain, and P. Abbeel. Denoising diffusion proba- bilistic models. NeurIPS, 33:6840–6851, 2020. 2

    2020

  7. [15]

    Hoda ˇn, M

    T. Hoda ˇn, M. Sundermeyer, Y . Labb ´e, V . N. Nguyen, G. Wang, E. Brachmann, B. Drost, V . Lepetit, C. Rother, and J. Matas. BOP Challenge 2023 on Detection, Segmenta- tion and Pose Estimation of Seen and Unseen Rigid Objects. CVPRW, 2024. 1, 2, 6

    2023

  8. [16]

    M. Humt, D. Winkelbauer, and U. Hillenbrand. Shape Com- pletion with Prediction of Uncertain Regions. InIROS, 2023. 1

    2023

  9. [17]

    Hyv ¨arinen and P

    A. Hyv ¨arinen and P. Dayan. Estimation of non-normalized statistical models by score matching. JMLR, 6(4), 2005. 3

    2005

  10. [18]

    Karras, M

    T. Karras, M. Aittala, T. Aila, and S. Laine. Elucidating the design space of diffusion-based generative models.NeurIPS, 35:26565–26577, 2022. 2, 3, 4, 5, 6, 7

    2022

  11. [19]

    Karras, M

    T. Karras, M. Aittala, J. Lehtinen, J. Hellsten, T. Aila, and S. Laine. Analyzing and improving the training dynamics of diffusion models. In CVPR, pages 24174–24184, 2024. 3, 4, 5, 6

    2024

  12. [20]

    Kaskman, S

    R. Kaskman, S. Zakharov, I. Shugurov, and S. Ilic. Home- breweddb: Rgb-d dataset for 6d pose estimation of 3d ob- jects. In ICCVW, pages 0–0, 2019. 6

    2019

  13. [21]

    Kendall, Y

    A. Kendall, Y . Gal, and R. Cipolla. Multi-task learning using uncertainty to weigh losses for scene geometry and seman- tics. In CVPR, pages 7482–7491, 2018. 3

    2018

  14. [22]

    Kirillov, E

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. Berg, W.-Y . Lo, et al. Segment anything. In ICCV, pages 4015–4026, 2023. 7

    2023

  15. [23]

    Labb´e, L

    Y . Labb´e, L. Manuelli, A. Mousavian, S. Tyree, S. Birchfield, J. Tremblay, J. Carpentier, M. Aubry, D. Fox, and J. Sivic. Megapose: 6d pose estimation of novel objects via render & compare. arXiv preprint arXiv:2212.06870, 2022. 6

    2022 arXiv

  16. [24]

    J. Lin, L. Liu, D. Lu, and K. Jia. Sam-6d: Segment anything model meets zero-shot 6d object pose estimation. In CVPR, pages 27906–27916, 2024. 2, 4, 6

    2024

  17. [25]

    Y . Lin, Y . Su, P. Nathan, S. Inuganti, Y . Di, M. Sunder- meyer, F. Manhardt, D. Stricker, J. Rambach, and Y . Zhang. Hipose: Hierarchical binary surface encoding and correspon- dence pruning for rgb-d 6dof object pose estimation. In CVPR, pages 10148–10158, 2024. 1

    2024

  18. [26]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. In European Conference on Computer Vision , pages 38–55. Springe...

    2024

  19. [27]

    Adapting pre-trained vision models for novel instance de- tection and segmentation

    Yangxiao Lu, Yunhui Guo, Nicholas Ruozzi, Yu Xiang, et al. Adapting pre-trained vision models for novel instance de- tection and segmentation. arXiv preprint arXiv:2405.17859,

    arXiv

  20. [28]

    Newbury, M

    R. Newbury, M. Gu, L. Chumbley, A. Mousavian, C. Epp- ner, J. Leitner, J. Bohg, A. Morales, T. Asfour, D. Kragic, et al. Deep learning approaches to grasp synthesis: A review. IEEE Transactions on Robotics, 39(5):3994–4015, 2023. 1

    2023

  21. [29]

    V . N. Nguyen, T. Groueix, G. Ponimatkin, V . Lepetit, and T. Hodan. Cnos: A strong baseline for cad-based novel object segmentation. In ICCV, pages 2134–2140, 2023. 2, 4, 6, 7

    2023

  22. [30]

    A. Q. Nichol and P. Dhariwal. Improved denoising diffusion probabilistic models. In ICML, pages 8162–8171. PMLR,

  23. [31]

    Oquab, T

    M. Oquab, T. Darcet, T. Moutakanni, H. V o, M. Szafraniec, V . Khalidov, P. Fernandez, D. Haziza, F. Massa, A. El- Nouby, et al. Dinov2: Learning robust visual features with- out supervision. arXiv preprint arXiv:2304.07193, 2023. 2, 6

    2023 arXiv

  24. [32]

    Peebles and S

    W. Peebles and S. Xie. Scalable diffusion models with trans- formers. In ICCV, pages 4195–4205, 2023. 3, 5

    2023

  25. [33]

    Perez, F

    E. Perez, F. Strub, H. De Vries, V . Dumoulin, and A. Courville. Film: Visual reasoning with a general condition- ing layer. In AAAI, 2018. 5

    2018

  26. [34]

    Rahman, J

    A. Rahman, J. M. J. Valanarasu, I. Hacihaliloglu, and V . M. Patel. Ambiguous medical image segmentation using diffu- sion models. In CVPR, pages 11536–11546, 2023. 2

    2023

  27. [35]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Om- mer. High-resolution image synthesis with latent diffusion models. In CVPR, pages 10684–10695, 2022. 2, 3, 4, 6

    2022

  28. [36]

    Sohl-Dickstein, E

    J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan, and S. Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. arXiv [cs.LG], 2015. 2

    2015

  29. [37]

    J. Song, C. Meng, and S. Ermon. Denoising diffusion im- plicit models. arXiv preprint arXiv:2010.02502, 2020. 3

    2010 arXiv

  30. [38]

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020. 3

    2011 arXiv

  31. [39]

    Stoiber, M

    M. Stoiber, M. Sundermeyer, and R. Triebel. Iterative Corre- sponding Geometry: Fusing Region and Depth for Highly Efficient 3D Tracking of Textureless Objects. In CVPR. IEEE, 2022. 1

    2022

  32. [40]

    Sundermeyer, A

    M. Sundermeyer, A. Mousavian, R. Triebel, and D. Fox. Contact-graspnet: Efficient 6-dof grasp generation in clut- tered scenes. In ICRA, pages 13438–13444. IEEE, 2021. 1

    2021

  33. [41]

    Sundermeyer, T

    M. Sundermeyer, T. Hoda ˇn, Y . Labbe, G. Wang, E. Brach- mann, B. Drost, C. Rother, and J. Matas. BOP Challenge 2023 on Detection, Segmentation and Pose Estimation of Specific Rigid Objects. In CVPRW, pages 2785–2794, 2023. 6

    2023

  34. [42]

    W. Tan, S. Chen, and B. Yan. Diffss: Diffusion model for few-shot semantic segmentation. arXiv preprint arXiv:2307.00773, 2023. 2

    2023 arXiv

  35. [43]

    Ulmer, M

    M. Ulmer, M. Durner, M. Sundermeyer, M. Stoiber, and R. Triebel. 6d object pose estimation from approximate 3d models for orbital robotics. In IROS, pages 10749–10756. IEEE, 2023. 1

    2023

  36. [44]

    Wolleb, R

    J. Wolleb, R. Sandk ¨uhler, F. Bieder, P. Valmaggia, and P. C. Cattin. Diffusion models for implicit image segmentation ensembles. In MIDL, pages 1336–1348. PMLR, 2022. 2

    2022

  37. [45]

    Xiang, T

    Y . Xiang, T. Schmidt, V . Narayanan, and D. Fox. Posecnn: A convolutional neural network for 6d object pose estimation in cluttered scenes. arXiv preprint arXiv:1711.00199, 2017. 6

    2017 arXiv

  38. [46]

    Xiang, C

    Y . Xiang, C. Xie, A. Mousavian, and D. Fox. Learning rgb-d feature embeddings for unseen object instance segmentation. In Conference on Robot Learning , pages 461–470. PMLR,

  39. [47]

    C. Xie, Y . Xiang, A. Mousavian, and D. Fox. Unseen ob- ject instance segmentation for robotic environments. IEEE Transactions on Robotics, 37(5):1343–1359, 2021. 1, 2

    2021

  40. [48]

    X. Yan, L. Lin, N. J. Mitra, D. Lischinski, D. Cohen-Or, and H. Huang. ShapeFormer: Transformer-based Shape Com- pletion via Sparse Representation. In CVPR, pages 6229– 6239, 2022. 1

    2022

  41. [49]

    Fast segment anything, 2023

    Xu Z., Wenchao D., Yongqi A., Yinglong D., Tao Y ., Min L., Ming T., and Jinqiao W. Fast segment anything, 2023. 2

    2023

  42. [50]

    Zheng, J

    Y . Zheng, J. Wu, Y . Qin, F. Zhang, and L. Cui. Zero-shot instance segmentation. In CVPR, pages 2593–2602, 2021. 2 A. Implementation and Training Details A.1. Object Templates Rendering We use BlenderProc[6] for template generation and use the camera intrinsics from the TUDL ...

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.