Pith. sign in

REVIEW 3 major objections 3 minor 67 references

PVNet makes a single diffusion model densify entire LiDAR scenes at arbitrary rates, trained without dense ground-truth scans.

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 ↗

PVNet uses diffusion models with point-voxel interaction to upsample LiDAR scenes at arbitrary rates without dense supervision.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection A plausible diffusion-based LiDAR upsampling pipeline with a load-bearing 'no dense supervision' claim, but the supplied full text is unreadable, so the real evidence is out of reach. the 3 major comments →

arxiv 2508.17050 v1 pith:ZUDY4QGY submitted 2025-08-23 cs.CV

PVNet: Point-Voxel Interaction LiDAR Scene Upsampling Via Diffusion Models

Xianjing Cheng , Lintai Wu , Zuowen Wang , Junhui Hou , Jie Wen , Yong Xu This is my paper
classification cs.CV
keywords LiDAR point cloud upsamplingdiffusion modelsscene-level point cloud generationclassifier-free guidancepoint-voxel interactionself-supervised 3D learningarbitrary upsampling ratesoutdoor scene perception
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

PVNet tries to show that one diffusion model can upsample a sparse outdoor LiDAR scan into a dense point cloud at any requested factor, and that this can be learned without collecting dense ground-truth LiDAR. The training signal comes from the sequence itself: the model is asked to generate a dense pseudo-cloud built by merging nearby frames, while the actual sparse scan is fed in as a conditioning signal. If that works, practitioners would no longer need expensive high-beam LiDAR or manual dense annotations to get scene-scale dense geometry. The paper also introduces a voxel completion module and a point-voxel interaction module so that every generated point is aware of both fine detail and larger scene structure, and it reports that the method outperforms existing upsamplers on standard benchmarks. The central claim is that this combination is the first scene-level upsampling method that supports arbitrary upsampling rates.

Core claim

PVNet's central claim is that a classifier-free guided diffusion model can turn one sparse LiDAR scan of an outdoor scene into a dense point cloud at any requested resolution, using no dense ground truth at training time. The training signal comes from the sequence itself: the dense target is synthesized by fusing nearby frames, and the sparse scan is fed to the network as the conditioning signal. To support scene-level geometry far beyond the object-centric inputs of earlier upsamplers, the network first voxelizes the cloud, completes coarse voxel features (voxel completion module), then lets point and voxel branches exchange information (point-voxel interaction module). The authors argue t

What carries the argument

The load-bearing mechanism is a classifier-free guided denoising diffusion model: at inference the denoiser receives the sparse LiDAR scan as a conditioning context, and at training it is taught to reverse noise added to a dense pseudo-cloud derived from nearby frames. The point-voxel interaction module is the architectural innovation that lets per-point features be refined by voxelized context features before the final coordinate regression, while the voxel completion module repairs the coarse voxel features that carry the scene-level structure.

Load-bearing premise

The dense training targets are fake clouds made by merging nearby LiDAR frames, and the method assumes those merged clouds are faithful enough to stand in for real dense ground truth even when the scene contains moving objects, misalignment, or sensor noise.

What would settle it

Run PVNet on a LiDAR sequence containing a known moving vehicle, then upsample a single frame from that sequence; if the generated cloud contains duplicate ghost points near the vehicle's positions in adjacent frames, the near-frame supervision is teaching the model to hallucinate dynamic geometry. A clean result would show no such ghosts at those locations.

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

If this is right

  • Any upsampling factor (e.g., 4x, 8x, 16x) can be handled by a single model, because the sparse cloud is a conditioning signal rather than a fixed output-size constraint.
  • Training requires only sequential LiDAR scans; dense pseudo-clouds merged from nearby frames replace expensive dense ground-truth capture.
  • The recovered geometric detail should carry over to downstream perception tasks that currently suffer from LiDAR sparsity, such as 3D detection and segmentation.
  • The point-voxel interaction design makes each upsampled point sensitive to both fine detail and large-scale scene structure, which is what allows scene-level rather than object-level upsampling.

Where Pith is reading between the lines

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

  • A strong testable implication is that a model trained on one rate (say 4x) should still produce coherent output at far higher rates (e.g., 16x); the paper does not explicitly verify extrapolation beyond its trained distribution.
  • The pseudo-dense supervision assumes a mostly static scene across neighboring frames; in urban traffic, moving vehicles will appear at multiple positions in the synthesized target, so a natural extension would mask or down-weight dynamic regions to prevent the model from learning ghost artifacts.
  • The same self-supervised recipe should transfer to other sparse depth sensors (solid-state LiDAR, stereo depth) whenever temporal neighbors are available, a consequence of the formulation that the paper does not discuss.
  • Ablating the voxel completion module would isolate whether the reported gains come from the coarse-to-fine conditioning or from the diffusion framework itself, a check the paper 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, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 3 minor

Summary. The paper proposes PVNet, a diffusion-based framework for scene-level LiDAR point cloud upsampling that avoids dense supervision. The method uses a sparse point cloud as a conditioning signal and synthesized point clouds aggregated from nearby frames as the training target/input. It introduces a voxel completion module and a point-voxel interaction module. The abstract claims state-of-the-art performance on benchmarks and that this is the first scene-level point cloud upsampling method supporting arbitrary upsampling rates. However, the full text as provided is almost entirely garbled/undecodable, so the technical derivations, experimental setup, quantitative results, and ablations cannot be inspected.

Significance. If the claims are correct, PVNet would be a practically valuable contribution: label-free densification of outdoor LiDAR scenes at arbitrary upsampling factors could benefit downstream perception without requiring dense ground-truth point clouds. The combination of classifier-free diffusion guidance with point-voxel feature interaction is a reasonable design direction. The paper also promises code release. However, because the full text is unreadable in the provided manuscript, I cannot verify any of the technical or empirical claims. The significance is therefore conditional on a complete, readable version being provided.

major comments (3)
  1. [Full text (all sections after Abstract)] The manuscript body is corrupted and unreadable: equations, tables, figures, and experimental numbers cannot be checked. The abstract's core claims — state-of-the-art performance and the first scene-level upsampling method with arbitrary rates — are therefore unsupported in the provided record. This is a blocking issue for evaluation. Please provide a complete, decodable manuscript with full experimental details, baseline comparisons, and ablations.
  2. [Abstract, training-signal description] The 'without dense supervision' claim depends entirely on synthesized point clouds derived from nearby frames as training targets. Nearby-frame LiDAR in outdoor scenes contains moving objects, ego-motion misalignment, occlusion changes, and range-dependent noise. The abstract does not state whether these targets are filtered, weighted, or corrected. If they are not, the diffusion model will learn to reproduce dynamic-object artifacts and misalignment errors, inflating benchmark scores relative to the static scene geometry that downstream perception requires. Please specify the aggregation procedure and provide ablations that test target faithfulness, e.g., dynamic-object masking/removal, registration-error analysis, or evaluation against a subset with dense ground truth.
  3. [Abstract, contributions] The claim 'first scene-level point cloud upsampling method supporting arbitrary upsampling rates' requires a clear definition of 'arbitrary upsampling rates' and a comparison with prior scene-level upsampling methods. The abstract alone does not show such a comparison. Please state how arbitrary rates are achieved (e.g., sampling variable numbers of output points), what range of rates is supported, and how the method compares with existing scene-level approaches on benchmarks.
minor comments (3)
  1. [Abstract] The roles of the sparse point cloud and the synthesized nearby-frame cloud are described as 'guiding condition' and 'input', respectively. In DDPM terminology, it is unclear which is the noisy target x_t and which is the condition c. Please clarify the exact formulation, including how classifier-free guidance is implemented.
  2. [Abstract] Please define 'arbitrary upsampling rates' precisely: integer factors, continuous density control, or any number of output points? The relationship between upsampling rate and memory/computation should also be discussed.
  3. [Full text, reproducibility] The GitHub URL is stated, but no license, checkpoints, or evaluation scripts are mentioned. For reproducibility, please include those details in the final version.

Circularity Check

0 steps flagged

No significant circularity: the upsampling target is an external temporal aggregation, not a restatement of the input or a fitted parameter.

full rationale

PVNet's central self-supervised scheme uses a sparse LiDAR frame as the diffusion condition and a dense cloud synthesized from nearby frames as the training target. This is a pseudo-ground-truth construction: the target is not defined as a function of the model output, and the prediction is not statistically forced by the input, since nearby frames contribute information beyond the sparse frame. No fitted parameter is renamed as a prediction, and no load-bearing argument is reduced to a self-citation. The claims of first scene-level arbitrary-rate upsampling and SOTA performance are empirical claims against external benchmarks; while the faithfulness of the synthesized nearby-frame targets is a legitimate correctness risk, it is not a circularity. I found no equation, definition, or citation chain in which the derived result is equivalent to the input by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

No free parameters are visible from the abstract. The axioms reflect the method's reliance on self-supervision signals and architectural design choices. No new physical entities are introduced.

axioms (2)
  • domain assumption Synthesized point clouds from nearby frames provide a valid dense supervision proxy.
    The abstract states the diffusion model uses 'synthesized point clouds derived from its nearby frames' as input, implying these are treated as reasonable targets despite potential misalignment or dynamic objects.
  • domain assumption Voxel completion refines coarse voxel features in a way that improves upsampling.
    The voxel completion module is introduced as a design choice; its benefit is asserted rather than derived from first principles.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of PVNet: Point-Voxel Interaction LiDAR Scene Upsampling Via Diffusion Models." pith.science (2026). https://pith.science/paper/ZUDY4QGY

@misc{pith2026250817050,
  author       = {Pith},
  title        = {Pith review of: PVNet: Point-Voxel Interaction LiDAR Scene Upsampling Via Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZUDY4QGY}},
  note         = {Machine review of arXiv:2508.17050}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Accurate 3D scene understanding in outdoor environments heavily relies on high-quality point clouds. However, LiDAR-scanned data often suffer from extreme sparsity, severely hindering downstream 3D perception tasks. Existing point cloud upsampling methods primarily focus on individual objects, thus demonstrating limited generalization capability for complex outdoor scenes. To address this issue, we propose PVNet, a diffusion model-based point-voxel interaction framework to perform LiDAR point cloud upsampling without dense supervision. Specifically, we adopt the classifier-free guidance-based DDPMs to guide the generation, in which we employ a sparse point cloud as the guiding condition and the synthesized point clouds derived from its nearby frames as the input. Moreover, we design a voxel completion module to refine and complete the coarse voxel features for enriching the feature representation. In addition, we propose a point-voxel interaction module to integrate features from both points and voxels, which efficiently improves the environmental perception capability of each upsampled point. To the best of our knowledge, our approach is the first scene-level point cloud upsampling method supporting arbitrary upsampling rates. Extensive experiments on various benchmarks demonstrate that our method achieves state-of-the-art performance. The source code will be available at https://github.com/chengxianjing/PVNet.

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

67 extracted references · 62 canonical work pages

  1. [1]

    M. Wei, Z. Wei, H. Zhou, F. Hu, H. Si, Z. Chen, Z. Zhu, J. Qiu, X. Yan, Y. Guo et al., ``Agconv: Adaptive graph convolution on 3d point clouds,'' IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 8, pp. 9374--9392, 2023

    work page 2023

  2. [2]

    L. Wu, Q. Zhang, J. Hou, and Y. Xu, ``Leveraging single-view images for unsupervised 3d point cloud completion,'' IEEE Transactions on Multimedia, vol. 27, pp. 940--953, 2025

    work page 2025

  3. [3]

    Y. Guo, Y. Li, D. Ren, X. Zhang, J. Li, L. Pu, C. Ma, X. Zhan, J. Guo, M. Wei et al., ``Lidar-net: A real-scanned 3d point cloud dataset for indoor scenes,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2024, pp. 21\,989--21\,999

    work page 2024

  4. [4]

    L. Wu, J. Hou, L. Song, and Y. Xu, ``3d shape completion on unseen categories: A weakly-supervised approach,'' IEEE Transactions on Visualization and Computer Graphics, vol. 31, no. 9, pp. 5547--5561, 2025

    work page 2025

  5. [5]

    C. Ma, Y. Yang, J. Guo, M. Wei, C. Wang, Y. Guo, and W. Wang, ``Collaborative completion and segmentation for partial point clouds with outliers,'' IEEE Transactions on Visualization and Computer Graphics, vol. 30, no. 9, pp. 6118--6129, 2023

    work page 2023

  6. [6]

    J. Han, K. Liu, W. Li, G. Chen, W. Wang, and F. Zhang, ``A large-scale network construction and lightweighting method for point cloud semantic segmentation,'' IEEE Transactions on Image Processing, 2024

    work page 2024

  7. [7]

    L. Wu, X. Cheng, Y. Xu, H. Zeng, and J. Hou, ``Unsupervised 3d point cloud completion via multi-view adversarial learning,'' IEEE Transactions on Visualization and Computer Graphics, 2025

    work page 2025

  8. [8]

    F. Yang, X. Li, and J. Shen, ``Nested architecture search for point cloud semantic segmentation,'' IEEE Transactions on Image Processing, vol. 32, pp. 2889--2900, 2022

    work page 2022

  9. [9]

    Y. Zeng, J. Hou, Q. Zhang, S. Ren, and W. Wang, ``Dynamic 3d point cloud sequences as 2d videos,'' IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 46, no. 12, pp. 9371--9386, 2024

    work page 2024

  10. [10]

    S. Ren, J. Hou, X. Chen, H. Xiong, and W. Wang, ``Ddm: A metric for comparing 3d shapes using directional distance fields,'' IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 8, pp. 6631--6646, 2025

    work page 2025

  11. [11]

    Y. Guo, H. Wang, Q. Hu, H. Liu, L. Liu, and M. Bennamoun, ``Deep learning for 3d point clouds: A survey,'' IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 12, pp. 4338--4364, 2020

    work page 2020

  12. [12]

    Vizzo, B

    I. Vizzo, B. Mersch, R. Marcuzzi, L. Wiesmann, J. Behley, and C. Stachniss, ``Make it dense: Self-supervised geometric scan completion of sparse 3d lidar scans in large outdoor environments,'' IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 8534--8541, 2022

    work page 2022

  13. [13]

    L. Zhou, Y. Du, and J. Wu, ``3d shape generation and completion through point-voxel diffusion,'' in Proceedings of IEEE International Conference on Computer Vision, 2021, pp. 5826--5835

    work page 2021

  14. [14]

    Nunes, R

    L. Nunes, R. Marcuzzi, B. Mersch, J. Behley, and C. Stachniss, ``Scaling diffusion models to real-world 3d lidar scene completion,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2024, pp. 14\,770--14\,780

    work page 2024

  15. [15]

    Y. Jiao, Z. Jie, S. Chen, J. Chen, L. Ma, and Y.-G. Jiang, ``Msmdfusion: Fusing lidar and camera at multiple scales with multi-depth seeds for 3d object detection,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2023, pp. 21\,643--21\,652

    work page 2023

  16. [16]

    J. S. Hu, T. Kuai, and S. L. Waslander, ``Point density-aware voxels for lidar 3d object detection,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2022, pp. 8469--8478

    work page 2022

  17. [17]

    Y. Chen, J. Liu, X. Zhang, X. Qi, and J. Jia, ``Voxelnext: Fully sparse voxelnet for 3d object detection and tracking,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2023, pp. 21\,674--21\,683

    work page 2023

  18. [18]

    L. Yu, X. Li, C.-W. Fu, D. Cohen-Or, and P.-A. Heng, ``Pu-net: Point cloud upsampling network,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 2790--2799

    work page 2018

  19. [19]

    R. Li, X. Li, C.-W. Fu, D. Cohen-Or, and P.-A. Heng, ``Pu-gan: a point cloud upsampling adversarial network,'' in Proceedings of IEEE International Conference on Computer Vision, 2019, pp. 7203--7212

    work page 2019

  20. [20]

    G. Qian, A. Abualshour, G. Li, A. Thabet, and B. Ghanem, ``Pu-gcn: Point cloud upsampling using graph convolutional networks,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2021, pp. 11\,683--11\,692

    work page 2021

  21. [21]

    C. Long, W. Zhang, R. Li, H. Wang, Z. Dong, and B. Yang, ``Pc2-pu: Patch correlation and point correlation for effective point cloud upsampling,'' in Proceedings of the 30th ACM International Conference on Multimedia, 2022, pp. 2191--2201

    work page 2022

  22. [22]

    Y. Qian, J. Hou, S. Kwong, and Y. He, ``Deep magnification-flexible upsampling over 3d point clouds,'' IEEE Transactions on Image Processing, vol. 30, pp. 8354--8367, 2021

    work page 2021

  23. [23]

    A. Mao, Z. Du, J. Hou, Y. Duan, Y.-j. Liu, and Y. He, ``Pu-flow: A point cloud upsampling network with normalizing flows,'' IEEE Transactions on Visualization and Computer Graphics, 2022

    work page 2022

  24. [24]

    H. Liu, H. Yuan, J. Hou, R. Hamzaoui, and W. Gao, ``Pufa-gan: A frequency-aware generative adversarial network for 3d point cloud upsampling,'' IEEE Transactions on Image Processing, vol. 31, pp. 7389--7402, 2022

    work page 2022

  25. [25]

    R. Li, X. Li, P.-A. Heng, and C.-W. Fu, ``Point cloud upsampling via disentangled refinement,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2021, pp. 344--353

    work page 2021

  26. [26]

    Y. Qian, J. Hou, S. Kwong, and Y. He, ``Pugeo-net: A geometry-centric network for 3d point cloud upsampling,'' in Proceedings of European Conference on Computer Vision. 1em plus 0.5em minus 0.4em Springer, 2020, pp. 752--769

    work page 2020

  27. [27]

    X. Liu, X. Liu, Y.-S. Liu, and Z. Han, ``Spu-net: Self-supervised point cloud upsampling by coarse-to-fine reconstruction with self-projection optimization,'' IEEE Transactions on Image Processing, vol. 31, pp. 4213--4226, 2022

    work page 2022

  28. [28]

    W. Qu, Y. Shao, L. Meng, X. Huang, and L. Xiao, ``A conditional denoising diffusion probabilistic model for point cloud upsampling,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2024, pp. 20\,786--20\,795

    work page 2024

  29. [29]

    Zhang, J

    Q. Zhang, J. Hou, Y. Qian, Y. Zeng, J. Zhang, and Y. He, ``Flattening-net: Deep regular 2d representation for 3d point cloud analysis,'' IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 8, pp. 9726--9742, 2023

    work page 2023

  30. [30]

    S. Ren, J. Hou, X. Chen, Y. He, and W. Wang, ``Geoudf: Surface reconstruction from 3d point clouds via geometry-guided distance representation,'' in Proceedings of IEEE International Conference on Computer Vision, 2023, pp. 14\,214--14\,224

    work page 2023

  31. [31]

    J. Ho, A. Jain, and P. Abbeel, ``Denoising diffusion probabilistic models,'' Proceedings of Advanced Neural Information Processing System, vol. 33, pp. 6840--6851, 2020

    work page 2020

  32. [32]

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, ``Pointnet: Deep learning on point sets for 3d classification and segmentation,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 652--660

    work page 2017

  33. [33]

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, ``Pointnet++: Deep hierarchical feature learning on point sets in a metric space,'' Proceedings of Advanced Neural Information Processing System, vol. 30, 2017

    work page 2017

  34. [34]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, ``Generative adversarial nets,'' in Proceedings of Advanced Neural Information Processing System, vol. 27, 2014, pp. 5543--5553

    work page 2014

  35. [35]

    Yifan, S

    W. Yifan, S. Wu, H. Huang, D. Cohen-Or, and O. Sorkine-Hornung, ``Patch-based progressive 3d point set upsampling,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2019, pp. 5958--5967

    work page 2019

  36. [36]

    G. Wei, H. Pan, S. Zhuang, Y. Zhou, and C. Li, ``ipunet: Iterative cross field guided point cloud upsampling,'' IEEE Transactions on Visualization and Computer Graphics, vol. 30, no. 9, pp. 6089--6103, 2024

    work page 2024

  37. [37]

    Y. Rong, H. Zhou, K. Xia, C. Mei, J. Wang, and T. Lu, ``Repkpu: Point cloud upsampling with kernel point representation and deformation,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2024, pp. 21\,050--21\,060

    work page 2024

  38. [38]

    B. Yang, P. Pfreundschuh, R. Siegwart, M. Hutter, P. Moghadam, and V. Patil, ``Tulip: Transformer for upsampling of lidar point clouds,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2024, pp. 15\,354--15\,364

    work page 2024

  39. [39]

    Dhariwal and A

    P. Dhariwal and A. Nichol, ``Diffusion models beat gans on image synthesis,'' Proceedings of Advanced Neural Information Processing System, vol. 34, pp. 8780--8794, 2021

    work page 2021

  40. [40]

    A. Q. Nichol and P. Dhariwal, ``Improved denoising diffusion probabilistic models,'' in Proceedings of International Conference on Machine Learning. 1em plus 0.5em minus 0.4em PMLR, 2021, pp. 8162--8171

    work page 2021

  41. [41]

    Peebles and S

    W. Peebles and S. Xie, ``Scalable diffusion models with transformers,'' in Proceedings of IEEE International Conference on Computer Vision, 2023, pp. 4195--4205

    work page 2023

  42. [42]

    Ramesh, M

    A. Ramesh, M. Pavlov, G. Goh, S. Gray, C. Voss, A. Radford, M. Chen, and I. Sutskever, ``Zero-shot text-to-image generation,'' in Proceedings of International Conference on Machine Learning. 1em plus 0.5em minus 0.4em Pmlr, 2021, pp. 8821--8831

    work page 2021

  43. [43]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, ``High-resolution image synthesis with latent diffusion models,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2022, pp. 10\,684--10\,695

    work page 2022

  44. [44]

    Y. Zhou, R. Zhang, C. Chen, C. Li, C. Tensmeyer, T. Yu, J. Gu, J. Xu, and T. Sun, ``Towards language-free training for text-to-image generation,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2022, pp. 17\,907--17\,917

    work page 2022

  45. [45]

    Y. Zhou, B. Liu, Y. Zhu, X. Yang, C. Chen, and J. Xu, ``Shifted diffusion for text-to-image generation,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2023, pp. 10\,157--10\,166

    work page 2023

  46. [46]

    Balaji, S

    Y. Balaji, S. Nah, X. Huang, A. Vahdat, J. Song, Q. Zhang, K. Kreis, M. Aittala, T. Aila, S. Laine et al., ``ediff-i: Text-to-image diffusion models with an ensemble of expert denoisers,'' arXiv preprint arXiv:2211.01324, 2022

    Pith/arXiv arXiv 2022

  47. [47]

    Ho and T

    J. Ho and T. Salimans, ``Classifier-free diffusion guidance,'' arXiv preprint arXiv:2207.12598, 2022

    Pith/arXiv arXiv 2022

  48. [48]

    Zhang, A

    L. Zhang, A. Rao, and M. Agrawala, ``Adding conditional control to text-to-image diffusion models,'' in Proceedings of IEEE International Conference on Computer Vision, 2023, pp. 3836--3847

    work page 2023

  49. [49]

    Y. Tang, Q. Zhang, Y. Liu, and J. Hou, ``Human as points: Explicit point-based 3d human reconstruction from single-view rgb images,'' IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 7, pp. 5884--5900, 2025

    work page 2025

  50. [50]

    Karras, M

    T. Karras, M. Aittala, T. Aila, and S. Laine, ``Elucidating the design space of diffusion-based generative models,'' Proceedings of Advanced Neural Information Processing System, vol. 35, pp. 26\,565--26\,577, 2022

    work page 2022

  51. [51]

    C. Lu, Y. Zhou, F. Bao, J. Chen, C. Li, and J. Zhu, ``Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps,'' Proceedings of Advanced Neural Information Processing System, vol. 35, pp. 5775--5787, 2022

    work page 2022

  52. [52]

    1--22, 2025

    ------, ``Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models,'' Machine Intelligence Research, pp. 1--22, 2025

    work page 2025

  53. [53]

    J. Song, C. Meng, and S. Ermon, ``Denoising diffusion implicit models,'' arXiv preprint arXiv:2010.02502, 2020

    Pith/arXiv arXiv 2010

  54. [54]

    C. Meng, R. Rombach, R. Gao, D. Kingma, S. Ermon, J. Ho, and T. Salimans, ``On distillation of guided diffusion models,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2023, pp. 14\,297--14\,306

    work page 2023

  55. [55]

    Mersch, T

    B. Mersch, T. Guadagnino, X. Chen, I. Vizzo, J. Behley, and C. Stachniss, ``Building volumetric beliefs for dynamic environments exploiting map-based moving object segmentation,'' IEEE Robotics and Automation Letters, 2023

    work page 2023

  56. [56]

    J. Lee, W. Im, S. Lee, and S.-E. Yoon, ``Diffusion probabilistic models for scene-scale 3d categorical data,'' arXiv preprint arXiv:2301.00527, 2023

    Pith/arXiv arXiv 2023

  57. [57]

    Nakashima and R

    K. Nakashima and R. Kurazume, ``Lidar data synthesis with denoising diffusion probabilistic models,'' in IEEE International Conference on Robotics and Automation. 1em plus 0.5em minus 0.4em IEEE, 2024, pp. 14\,724--14\,731

    work page 2024

  58. [58]

    Zyrianov, X

    V. Zyrianov, X. Zhu, and S. Wang, ``Learning to generate realistic lidar point clouds,'' in Proceedings of European Conference on Computer Vision. 1em plus 0.5em minus 0.4em Springer, 2022, pp. 17--35

    work page 2022

  59. [59]

    Roldao, R

    L. Roldao, R. de Charette, and A. Verroust-Blondet, ``Lmscnet: Lightweight multiscale 3d semantic completion,'' in Proceedings of International Conference on 3D Vision. 1em plus 0.5em minus 0.4em IEEE, 2020, pp. 111--119

    work page 2020

  60. [60]

    Fix, Discriminatory analysis: nonparametric discrimination, consistency properties

    E. Fix, Discriminatory analysis: nonparametric discrimination, consistency properties. 1em plus 0.5em minus 0.4em USAF school of Aviation Medicine, 1985, vol. 1

    work page 1985

  61. [61]

    Behley, M

    J. Behley, M. Garbade, A. Milioto, J. Quenzel, S. Behnke, C. Stachniss, and J. Gall, `` SemanticKITTI: A Dataset for Semantic Scene Understanding of LiDAR Sequences ,'' in Proceedings of IEEE International Conference on Computer Vision, 2019

    work page 2019

  62. [62]

    Geiger, P

    A. Geiger, P. Lenz, and R. Urtasun, ``Are we ready for autonomous driving? the kitti vision benchmark suite,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. 1em plus 0.5em minus 0.4em IEEE, 2012, pp. 3354--3361

    work page 2012

  63. [63]

    Y. Liao, J. Xie, and A. Geiger, ``Kitti-360: A novel dataset and benchmarks for urban scene understanding in 2d and 3d,'' IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 3, pp. 3292--3310, 2022

    work page 2022

  64. [64]

    Menze and A

    M. Menze and A. Geiger, ``Object scene flow for autonomous vehicles,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2015

    work page 2015

  65. [65]

    D. P. Kingma and J. Ba, ``Adam: A method for stochastic optimization,'' arXiv preprint arXiv:1412.6980, 2014

    Pith/arXiv arXiv 2014

  66. [66]

    R. Cui, S. Qiu, S. Anwar, J. Liu, C. Xing, J. Zhang, and N. Barnes, ``P2c: Self-supervised point cloud completion from single partial clouds,'' in Proceedings of IEEE International Conference on Computer Vision, 2023, pp. 14\,351--14\,360

    work page 2023

  67. [67]

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, ``3d shapenet: A large-scale repository of 3d shapes,'' in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 1--9

    work page 2016

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.