Pith. sign in

REVIEW 3 major objections 3 minor 54 references

TCFNet: Bidirectional face-bone transformation via a Transformer-based coarse-to-fine point movement network

T0 review · 3 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read TCFNet claims that a Transformer-based coarse-to-fine point movement network, refined by local geometric aggregation and guided by global features through a gated recurrent unit, can simulate face-bone shape changes after orthognathic surge

desk verdict A plausible engineering contribution in a narrow clinical niche, but the SOTA claim is currently unverifiable because the dataset, correspondence construction, and error bars are missing. read the letter →

arxiv 2508.14373 v1 pith:WFB2Y3OO submitted 2025-08-20 cs.CV

Runshi Zhang , Bimeng Jie , Yang He , Junchen Wang This is my paper
classification cs.CV
keywords face-bonetransformationorthognathicsurgerypointcloudTransformercoarse-to-finemovementnetworklocalinformationaggregationcomputer-aidedsurgicalsimulation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that the face-bone shape change involved in orthognathic (jaw-correcting) surgery can be predicted directly from point clouds by learning point movement paths, without the registration-based preprocessing that burdens earlier deep learning approaches. The proposed network, TCFNet, works in two reinforcing stages: a Transformer stage captures global context and moves points coarsely, then a local information aggregation network refines the movement by modeling edges, orientations, and relative positions, with a gated recurrent unit passing global features into the local stage. An auxiliary loss inspired by deformable medical image registration adds expert knowledge for reconstructing critical organs. If correct, surgical planning would get a faster, end-to-end way to predict postoperative facial shape from skeletal and facial surfaces, and the architecture would be a candidate for other dense point-cloud morphing tasks.

What carries the argument

The load-bearing design is the coarse-to-fine point movement with cross-stage global guidance. A Transformer-based network first computes a global receptive field and patch-level displacement; a Local Information Aggregation Network (LIA-Net) then refines point-level paths using geometric edge, orientation, and relative-position features, repairing the neighborhood precision that global attention loses. A gated recurrent unit carries the prior global features into the local displacement computation, and the auxiliary organ-reconstruction loss regularizes the deformation so critical anatomical structures are preserved.

What would settle it

Take a held-out set of real patients with pre- and post-operative CT or surface scans, run TCFNet without the authors' own registration preprocessing, and compare predicted versus actual postoperative facial surfaces point-by-point. If accuracy collapses when correspondences are built by an independent pipeline, the claimed end-to-end advantage is not general.

Watch

Extended reading notes

Core claim

The central claim is that point-level correspondence between skeletal and facial surfaces can be learned directly as movement paths, and that TCFNet learns them more faithfully than existing methods on the collected datasets. The first stage uses a Transformer-based network to produce coarse, patch-level displacement; the second stage, LIA-Net, compensates for the Transformer's limited neighborhood precision by encoding local geometric structure—edges, orientations, and relative position features—and generates refined point-level paths. The global features from the first stage are fed through a gated recurrent unit to guide the local displacement, and an auxiliary loss, borrowed from deforma

Load-bearing premise

The claim depends on the collected training pairs of facial and skeletal surfaces being anatomically meaningful correspondences; if the pairing or registration is biased, the network may simply be learning those pairing artifacts rather than true face-bone relationships.

Editorial extensions

If this is right

  • Orthognathic surgical planning could move from slow biomechanical simulation and registration-heavy pipelines to a single end-to-end prediction from skeletal and facial point clouds.
  • Methods that previously could not handle large-scale point sets or had limited receptive fields can benefit from combining global attention with local geometric refinement.
  • The auxiliary organ-reconstruction loss demonstrates a route for injecting expert anatomical knowledge directly into a point-movement network.
  • The two-stage, globally guided refinement scheme is a reusable recipe for dense point-cloud transformation tasks beyond face-bone prediction.

Reading between the lines

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

  • The same coarse-to-fine, global-guides-local design could transfer to other non-rigid point-cloud morphing tasks, such as soft-tissue prediction after other craniofacial procedures or generic shape interpolation.
  • Testing TCFNet on an independently built, openly described correspondence dataset would show whether the reported margin over prior methods comes from the architecture itself or from the way training pairs were constructed.
  • The organ-reconstruction auxiliary loss could be developed into a clinical safety constraint that flags predicted movements endangering critical anatomical structures.
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

3 major / 3 minor

Summary. The paper proposes TCFNet, a Transformer-based coarse-to-fine point movement network for dense face-bone point cloud transformation, combining a global Transformer stage, a local information aggregation network (LIA-Net), and GRU-guided local displacement. An auxiliary loss inspired by deformable registration is introduced to reconstruct critical organs. The method is claimed to achieve state-of-the-art metrics and visual quality on 'gathered datasets' for predicting facial shape changes after orthognathic surgery, with code available on GitHub.

Significance. If the empirical claims are verifiable, the work could advance computer-aided orthognathic surgical planning by providing an end-to-end, registration-free-at-inference method for dense face-bone shape prediction. The architecture is plausible and the code-release commitment is a positive reproducibility signal. However, the central claim is an empirical SOTA statement, and the submission as provided does not supply the three essential ingredients for assessing it: a named or described dataset, a description of the label/correspondence construction, or any uncertainty quantification on the reported metrics.

major comments (3)
  1. [Abstract ('gathered datasets')] The central empirical claim rests on 'gathered datasets' that are unnamed, undescribed, and unreleased. No patient count, imaging modality, inclusion criteria, data split, or preprocessing details are given. In particular, the ground-truth dense correspondences between source and target point clouds—the labels used for both training and evaluation—are not described. Since the method is trained against these correspondences, any bias or noise in their construction (e.g., from non-rigid registration, surface-proximity matching, or template warping) would be baked into the reported gains. This is load-bearing for the SOTA claim and must be addressed with a complete dataset and label-generation description, and ideally data release.
  2. [Quantitative result tables (fragments in full text)] The readable table fragments report point-wise metrics such as CD and HD, but no error bars, per-folding or per-seed statistics, significance tests, or standard deviations appear. Without these, the claimed improvements over prior SOTA methods cannot be distinguished from random variation, especially given that the differences in several rows are small. The authors should report mean ± std over repeated runs or cross-validation folds, and perform pairwise significance tests against each baseline.
  3. [Full text (as provided)] The body of the manuscript supplied to the referee is largely corrupted/illegible; most equations, table numbers, and architectural details are unreadable. For example, the specific definitions of the Transformer patch embeddings, the LIA-Net edge/orientation features, the GRU update equations, and the auxiliary loss are not recoverable from the provided text. A referee cannot verify that the method is correctly specified or that the experiments are internally consistent. The authors must resubmit a clean, readable manuscript with complete equations and tables.
minor comments (3)
  1. [Abstract] The statement 'end-to-end and without registration-based preprocessing' is ambiguous: it appears to describe inference-time behavior, but the training labels may have required substantial registration. Please clarify what preprocessing is used during training and evaluation, and what exactly is 'registration-free'.
  2. [Auxiliary loss (Section on multi-loss)] The auxiliary loss 'inspired by deformable medical image registration' is described only verbally. Please provide the explicit formula, the weight(s) used in the total loss, and an ablation showing its contribution independently of the other components.
  3. [Notation] Several terms such as 'previous global features', 'patch-level and point-level correspondences', and 'critical organs' are used without precise definitions. Please define them in the method section and indicate how 'critical organs' are represented in the point cloud setting.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: TCFNet's contribution is an empirical architecture and SOTA comparison, not a theoretical derivation, and no load-bearing step reduces to its inputs.

full rationale

The paper claims an end-to-end deep network for face-bone point cloud transformation and evaluates it against SOTA on gathered datasets. There is no derivation chain in which a predicted quantity is defined in terms of the outcome, and no fitted parameter is renamed as a prediction. The auxiliary organ-reconstruction loss uses expert anatomical labels for training, but evaluation metrics are computed on held-out data in the standard way, so using the same type of distance losses during training is not circular. No uniqueness theorem or ansatz is imported from prior work by the same authors; the only external anchor mentioned is the generic inspiration from deformable registration. The main weakness, unstated registration/preprocessing of the private correspondence labels, is a correctness/reproducibility concern, not a circularity in the derivation. Under the hard rule that circularity must be exhibited by quote and reduction, no such exhibit exists.

Assumptions & free parameters 2 free parameters · 4 assumptions · 2 invented entities

For an empirical deep learning paper, the central claim rests on hand-chosen loss weights and architecture hyperparameters (free parameters), on the domain assumption that point cloud movement is a valid model of face-bone transformation with trustworthy ground-truth correspondences, and on the ad hoc assumption that the registration-inspired auxiliary loss helps. No external benchmarks or machine-checked guarantees support the invented network components.

free parameters (2)
  • Auxiliary loss weight(s) in the total training loss = not reported
    The total loss combines reconstruction and the auxiliary 'expert knowledge' loss; such weights are typically tuned by hand and directly control how strongly critical-organ reconstruction is enforced.
  • Architecture hyperparameters (Transformer layers, patch size, point sampling density, GRU dimensions) = not reported
    Hand-chosen design choices that determine the receptive field and noise properties the paper claims as advantages.
assumptions (4)
  • domain assumption Facial and skeletal anatomy can be represented as point clouds with learnable point-to-point movement paths
    Stated in the abstract as modeling 'face-bone shape transformations' as point cloud movement; no biomechanical guarantee is offered.
  • domain assumption Reliable ground-truth correspondence pairs exist in the gathered datasets, presumably built by registering CT-derived surfaces
    The supervised losses require paired skeletal and facial point clouds; the construction and accuracy of these pairs is not described in the abstract.
  • ad hoc to paper The auxiliary loss inspired by deformable medical image registration reconstructs critical organs better than plain point movement losses
    Introduced by the authors (abstract, final contribution); its effectiveness is asserted and tied to expert knowledge, not derived.
  • domain assumption Patch-level Transformer attention followed by local geometric aggregation covers both global structure and local detail
    Core design claim of the two-stage framework; relies on standard point cloud learning assumptions and is validated only by the paper's internal experiments.
invented entities (2)
  • LIA-Net (local information aggregation network)
    purpose: Recover neighborhood precision lost by the Transformer by modeling edges, orientations, and relative position features
    Validated only on the authors' gathered datasets; no external benchmark or theoretical characterization is provided.
  • GRU-guided local displacement mechanism
    purpose: Use previous global features to guide local point displacements in the second stage
    New gating design introduced for this task; its benefit is shown only through internal ablations that are not readable in the corrupted text.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TCFNet: Bidirectional face-bone transformation via a Transformer-based coarse-to-fine point movement network." pith.science (2026). https://pith.science/paper/WFB2Y3OO

@misc{pith2026250814373,
  author       = {Pith},
  title        = {Pith review of: TCFNet: Bidirectional face-bone transformation via a Transformer-based coarse-to-fine point movement network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WFB2Y3OO}},
  note         = {Machine review of arXiv:2508.14373}
}
read the original abstract

Computer-aided surgical simulation is a critical component of orthognathic surgical planning, where accurately simulating face-bone shape transformations is significant. The traditional biomechanical simulation methods are limited by their computational time consumption levels, labor-intensive data processing strategies and low accuracy. Recently, deep learning-based simulation methods have been proposed to view this problem as a point-to-point transformation between skeletal and facial point clouds. However, these approaches cannot process large-scale points, have limited receptive fields that lead to noisy points, and employ complex preprocessing and postprocessing operations based on registration. These shortcomings limit the performance and widespread applicability of such methods. Therefore, we propose a Transformer-based coarse-to-fine point movement network (TCFNet) to learn unique, complicated correspondences at the patch and point levels for dense face-bone point cloud transformations. This end-to-end framework adopts a Transformer-based network and a local information aggregation network (LIA-Net) in the first and second stages, respectively, which reinforce each other to generate precise point movement paths. LIA-Net can effectively compensate for the neighborhood precision loss of the Transformer-based network by modeling local geometric structures (edges, orientations and relative position features). The previous global features are employed to guide the local displacement using a gated recurrent unit. Inspired by deformable medical image registration, we propose an auxiliary loss that can utilize expert knowledge for reconstructing critical organs.Compared with the existing state-of-the-art (SOTA) methods on gathered datasets, TCFNet achieves outstanding evaluation metrics and visualization results. The code is available at https://github.com/Runshi-Zhang/TCFNet.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

54 extracted references · 38 canonical work pages

  1. [1]

    , author Diamanti, O

    author Achlioptas, P. , author Diamanti, O. , author Mitliagkas, I. , author Guibas, L. , year 2018 . title Learning representations and generative models for 3d point clouds , in: booktitle International conference on machine learning , organization PMLR . pp. pages 40--49

    work page 2018

  2. [3]

    , author Zhang, X

    author Bao, J. , author Zhang, X. , author Xiang, S. , author Liu, H. , author Cheng, M. , author Yang, Y. , author Huang, X. , author Xiang, W. , author Cui, W. , author Lai, H.C. , author Huang, S. , author Wang, Y. , author Qian, D. , author Yu, H. , year 2024 b. title Deep learning-based facial and skeletal transformations for surgical planning . jour...

    work page doi:10.1177/00220345241253186 2024

  3. [4]

    , author Seidita, F

    author Bianchi, A. , author Seidita, F. , author Badiali, G. , author Lusetti, L. , author Saporosi, C. , author Pironi, M. , author Marchetti, C. , author Crimi, S. , year 2023 . title Is beauty a matter of volume distribution? proposal of a new aesthetic three-dimensional guide in orthognathic surgery . journal Journal of Personalized Medicine volume 13...

    work page 2023

  4. [5]

    , author Kumaralingam, L

    author Chen, H. , author Kumaralingam, L. , author Zhang, S. , author Song, S. , author Zhang, F. , author Zhang, H. , author Pham, T.T. , author Punithakumar, K. , author Lou, E.H. , author Zhang, Y. , et al., year 2024 . title Neural implicit surface reconstruction of freehand 3d ultrasound volume with geometric constraints . journal Medical Image Analy...

    work page 2024

  5. [6]

    , author Zhu, X

    author Chen, W. , author Zhu, X. , author Chen, G. , author Yu, B. , year 2022 . title Efficient point cloud analysis using hilbert curve , in: booktitle European Conference on Computer Vision , organization Springer . pp. pages 730--747

    work page 2022

  6. [7]

    , author Pai, B.C

    author Denadai, R. , author Pai, B.C. , author Lo, L.J. , year 2020 . title Balancing the dental occlusion and facial aesthetic features in cleft orthognathic surgery: Patient-centered concept for computer-aided planning . journal Biomedical Journal volume 43 , pages 143--145 . https://www.sciencedirect.com/science/article/pii/S231941701930530X, :https://...

    work page doi:10.1016/j.bj.2019.12.008 2020

  7. [8]

    , author Aronovich, S

    author Elnagar, M.H. , author Aronovich, S. , author Kusnoto, B. , year 2020 . title Digital workflow for combined orthodontics and orthognathic surgery . journal Oral and Maxillofacial Surgery Clinics of North America volume 32 , pages 1--14 . https://www.sciencedirect.com/science/article/pii/S1042369919300688, :https://doi.org/10.1016/j.coms.2019.08.004...

    work page doi:10.1016/j.coms.2019.08.004 2020

  8. [9]

    , author Kim, D

    author Fang, X. , author Kim, D. , author Xu, X. , author Kuang, T. , author Lampen, N. , author Lee, J. , author Deng, H.H. , author Liebschner, M.A. , author Xia, J.J. , author Gateno, J. , author Yan, P. , year 2024 . title Correspondence attention for facial appearance simulation . journal Medical Image Analysis volume 93 , pages 103094 . https://www....

    arXiv 2024

Show all 54 references
  1. [10]

    , author Engelcke, M

    author Graham, B. , author Engelcke, M. , author Maaten, L.v.d. , year 2018 . title 3d semantic segmentation with submanifold sparse convolutional networks , in: booktitle 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. pages 9224--9232 . :10.1109/CVP...

    2018

  2. [11]

    , author Cai, J.X

    author Guo, M.H. , author Cai, J.X. , author Liu, Z.N. , author Mu, T.J. , author Martin, R.R. , author Hu, S.M. , year 2021 . title Pct: Point cloud transformer . journal COMPUTATIONAL VISUAL MEDIA volume 7 , pages 187--199 . :10.1007/s41095-021-0229-5

    2021 doi

  3. [12]

    , author Uneri, A

    author Han, R. , author Uneri, A. , author Vijayan, R. , author Wu, P. , author Vagdargi, P. , author Sheth, N. , author Vogt, S. , author Kleinszig, G. , author Osgood, G. , author Siewerdsen, J. , year 2021 . title Fracture reduction planning and guidance in orthopaedic trau...

    2021

  4. [13]

    , author Ding, Z

    author Huang, T. , author Ding, Z. , author Zhang, J. , author Tai, Y. , author Zhang, Z. , author Chen, M. , author Wang, C. , author Liu, Y. , year 2023 . title Learning to measure the point cloud reconstruction loss in a representation space , in: booktitle Proceedings of t...

    2023

  5. [14]

    , author Yang, X

    author Huang, T. , author Yang, X. , author Zhang, J. , author Cui, J. , author Zou, H. , author Chen, J. , author Zhao, X. , author Liu, Y. , year 2022 . title Learning to train a point cloud reconstruction network without matching , in: booktitle European Conference on Compu...

    2022

  6. [15]

    , author Zhao, J

    author Jia, B. , author Zhao, J. , author Xin, S. , author Duan, F. , author Pan, Z. , author Wu, Z. , author Li, J. , author Zhou, M. , year 2021 . title Craniofacial reconstruction based on heat flow geodesic grid regression (hf-ggr) model . journal Computers and Graphics vo...

    2021

  7. [16]

    , author Bolitho, M

    author Kazhdan, M. , author Bolitho, M. , author Hoppe, H. , year 2006 . title Poisson surface reconstruction , in: booktitle Proceedings of the fourth Eurographics symposium on Geometry processing

    2006

  8. [17]

    , author Kuang, T

    author Kim, D. , author Kuang, T. , author Rodrigues, Y.L. , author Gateno, J. , author Shen, S.G.F. , author Wang, X. , author Deng, H. , author Yuan, P. , author Alfi, D.M. , author Liebschner, M.A.K. , author Xia, J.J. , year 2019 . title A new approach of predicting facial...

    2019

  9. [18]

    , author Kuang, T

    author Kim, D. , author Kuang, T. , author Rodrigues, Y.L. , author Gateno, J. , author Shen, S.G.F. , author Wang, X. , author Stein, K. , author Deng, H.H. , author Liebschner, M.A.K. , author Xia, J.J. , year 2021 . title A novel incremental simulation of facial changes fol...

    2021

  10. [19]

    , author Kim, D

    author Lampen, N. , author Kim, D. , author Xu, X. , author Fang, X. , author Lee, J. , author Kuang, T. , author Deng, H.H. , author Liebschner, M.A.K. , author Xia, J.J. , author Gateno, J. , author Yan, P. , year 2023 . title Spatiotemporal incremental mechanics modeling of...

    2023

  11. [20]

    , author Yu, R

    author Lan, S. , author Yu, R. , author Yu, G. , author Davis, L.S. , year 2019 . title Modeling local geometric structure of 3d point clouds using geo-cnn , in: booktitle Proceedings of the IEEE/cvf conference on computer vision and pattern recognition , pp. pages 998--1008

    2019

  12. [21]

    , author Lin, Y

    author Liu, Z. , author Lin, Y. , author Cao, Y. , author Hu, H. , author Wei, Y. , author Zhang, Z. , author Lin, S. , author Guo, B. , year 2021 . title Swin transformer: Hierarchical vision transformer using shifted windows , in: booktitle Proceedings of the IEEE/CVF intern...

    2021

  13. [22]

    , author Yang, X

    author Liu, Z. , author Yang, X. , author Tang, H. , author Yang, S. , author Han, S. , year 2023 . title Flatformer: Flattened window attention for efficient point cloud transformer , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognit...

    2023

  14. [23]

    , author Du, D

    author Luo, Z. , author Du, D. , author Zhu, H. , author Yu, Y. , author Fu, H. , author Han, X. , year 2024 . title Sketchmetaface: A learning-based sketching interface for high-fidelity 3d character face modeling . journal IEEE Transactions on Visualization and Computer Grap...

    2024

  15. [24]

    , author Lian, C

    author Ma, L. , author Lian, C. , author Kim, D. , author Xiao, D. , author Wei, D. , author Liu, Q. , author Kuang, T. , author Ghanbari, M. , author Li, G. , author Gateno, J. , author Shen, S.G. , author Wang, L. , author Shen, D. , author Xia, J.J. , author Yap, P.T. , yea...

    2023

  16. [25]

    , author Xiao, D

    author Ma, L. , author Xiao, D. , author Kim, D. , author Lian, C. , author Kuang, T. , author Liu, Q. , author Deng, H. , author Yang, E. , author Liebschner, M.A.K. , author Gateno, J. , author Xia, J.J. , author Yap, P.T. , year 2023 b. title Simulation of postoperative fac...

    2023

  17. [26]

    , author Zimmermann, M

    author Millesi, G.A. , author Zimmermann, M. , author Eltz, M. , year 2023 . title Surgery first and surgery early treatment approach in orthognathic surgery . journal Oral and Maxillofacial Surgery Clinics volume 35 , pages 71--82

    2023

  18. [27]

    , author Schutyser, F

    author Mollemans, W. , author Schutyser, F. , author Nadjmi, N. , author Maes, F. , author Suetens, P. , year 2007 . title Predicting soft tissue deformations for a maxillofacial surgery planning system: From computational strategies to a complete clinical validation . journal...

    2007 doi

  19. [28]

    , author L \"u thi, M

    author Paysan, P. , author L \"u thi, M. , author Albrecht, T. , author Lerch, A. , author Amberg, B. , author Santini, F. , author Vetter, T. , year 2009 . title Face reconstruction from skull shapes and physical attributes , in: booktitle Pattern Recognition: 31st DAGM Sympo...

    2009

  20. [29]

    , author Wu, X

    author Peng, B. , author Wu, X. , author Jiang, L. , author Chen, Y. , author Zhao, H. , author Tian, Z. , author Jia, J. , year 2024 . title Oa-cnns: Omni-adaptive sparse cnns for 3d semantic segmentation , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Visi...

    2024

  21. [30]

    , author Su, H

    author Qi, C.R. , author Su, H. , author Mo, K. , author Guibas, L.J. , year 2017 a. title Pointnet: Deep learning on point sets for 3d classification and segmentation , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 652--660

    2017

  22. [31]

    , author Yi, L

    author Qi, C.R. , author Yi, L. , author Su, H. , author Guibas, L.J. , year 2017 b. title Pointnet plus plus : Deep hierarchical feature learning on point sets in a metric space , in: editor Guyon, I. , editor Luxburg, U. , editor Bengio, S. , editor Wallach, H. , editor Ferg...

    2017

  23. [33]

    , author Borghi, A

    author Ruggiero, F. , author Borghi, A. , author Bevini, M. , author Badiali, G. , author Lunari, O. , author Dunaway, D. , author Marchetti, C. , year 2023 . title Soft tissue prediction in orthognathic surgery: Improving accuracy by means of anatomical details . journal PLOS...

    2023 doi

  24. [34]

    , author Chen, Z

    author Wang, L. , author Chen, Z. , author Yu, T. , author Ma, C. , author Li, L. , author Liu, Y. , year 2022 . title Faceverse: a fine-grained and detail-controllable 3d face morphable model from a hybrid dataset , in: booktitle 2022 IEEE/CVF Conference on Computer Vision an...

    2022

  25. [35]

    , year 2023

    author Wang, P.S. , year 2023 . title Octformer: Octree-based transformers for 3d point clouds . journal ACM Transactions on Graphics (TOG) volume 42 , pages 1--11

    2023

  26. [36]

    , author Sun, Y

    author Wang, Y. , author Sun, Y. , author Liu, Z. , author Sarma, S.E. , author Bronstein, M.M. , author Solomon, J.M. , year 2019 . title Dynamic graph cnn for learning on point clouds . journal ACM Transactions on Graphics (tog) volume 38 , pages 1--12

    2019

  27. [37]

    , author Xiang, P

    author Wen, X. , author Xiang, P. , author Han, Z. , author Cao, Y.P. , author Wan, P. , author Zheng, W. , author Liu, Y.S. , year 2021 . title Pmp-net: Point cloud completion by learning multi-step point moving paths , in: booktitle 2021 IEEE/CVF Conference on Computer Visio...

    2021

  28. [38]

    , author Xiang, P

    author Wen, X. , author Xiang, P. , author Han, Z. , author Cao, Y.P. , author Wan, P. , author Zheng, W. , author Liu, Y.S. , year 2023 . title Pmp-net++: Point cloud completion by transformer-enhanced multi-step point moving paths . journal IEEE Transactions on Pattern Analy...

    2023

  29. [39]

    , author Pan, L

    author Wu, T. , author Pan, L. , author Zhang, J. , author Wang, T. , author Liu, Z. , author Lin, D. , year 2021 . title Density-aware chamfer distance as a comprehensive metric for point cloud completion , in: editor Ranzato, M. , editor Beygelzimer, A. , editor Dauphin, Y. ...

    2021

  30. [40]

    , author Qi, Z

    author Wu, W. , author Qi, Z. , author Fuxin, L. , year 2019 . title Pointconv: Deep convolutional networks on 3d point clouds , in: booktitle 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 9613--9622 . :10.1109/CVPR.2019.00985

    2019

  31. [41]

    , author Jiang, L

    author Wu, X. , author Jiang, L. , author Wang, P.S. , author Liu, Z. , author Liu, X. , author Qiao, Y. , author Ouyang, W. , author He, T. , author Zhao, H. , year 2024 . title Point transformer v3: Simpler faster stronger , in: booktitle Proceedings of the IEEE/CVF Conferen...

    2024

  32. [42]

    , author Lao, Y

    author Wu, X. , author Lao, Y. , author Jiang, L. , author Liu, X. , author Zhao, H. , year 2022 . title Point transformer v2: Grouped vector attention and partition-based pooling . journal Advances in Neural Information Processing Systems volume 35 , pages 33330--33342

    2022

  33. [43]

    , author Lian, C

    author Xiao, D. , author Lian, C. , author Wang, L. , author Deng, H. , author Lin, H.Y. , author Thung, K.H. , author Zhu, J. , author Yuan, P. , author Perez, L. , author Gateno, J. , author Shen, S.G. , author Yap, P.T. , author Xia, J.J. , author Shen, D. , year 2021 . tit...

    2021

  34. [44]

    , author Wang, L

    author Xiao, D. , author Wang, L. , author Deng, H. , author Thung, K.H. , author Zhu, J. , author Yuan, P. , author Rodrigues, Y.L. , author Perez, L. , author Crecelius, C.E. , author Gateno, J. , author Kuang, T. , author Shen, S.G.F. , author Kim, D. , author Alfi, D.M. , ...

    2019

  35. [45]

    , author Zhang, Y

    author Yang, Q. , author Zhang, Y. , author Chen, S. , author Xu, Y. , author Sun, J. , author Ma, Z. , year 2023 a. title Mped: Quantifying point cloud distortion based on multiscale potential energy discrepancy . journal IEEE Transactions on Pattern Analysis and Machine Inte...

    2023

  36. [46]

    , author Guo, Y.X

    author Yang, Y.Q. , author Guo, Y.X. , author Xiong, J.Y. , author Liu, Y. , author Pan, H. , author Wang, P.S. , author Tong, X. , author Guo, B. , year 2023 b. title Swin3d: A pretrained transformer backbone for 3d indoor scene understanding . journal arXiv preprint arXiv:2304.06906

    2023 arXiv

  37. [47]

    , author Huang, H

    author Yin, K. , author Huang, H. , author Cohen-Or, D. , author Zhang, H. , year 2018 . title P2p-net: bidirectional point displacement net for shape transform . journal ACM Trans. Graph. volume 37 . https://doi.org/10.1145/3197517.3201288, :10.1145/3197517.3201288

    2018

  38. [48]

    , author Rao, Y

    author Yu, X. , author Rao, Y. , author Wang, Z. , author Lu, J. , author Zhou, J. , year 2023 . title Adapointr: Diverse point cloud completion with adaptive geometry-aware transformers . journal IEEE Transactions on Pattern Analysis and Machine Intelligence volume 45 , pages...

    2023

  39. [49]

    , author Ettinger, R.E

    author Zammit, D. , author Ettinger, R.E. , author Sanati-Mehrizy, P. , author Susarla, S.M. , year 2023 . title Current trends in orthognathic surgery . journal Medicina volume 59 . https://www.mdpi.com/1648-9144/59/12/2100, :10.3390/medicina59122100

    2023 doi

  40. [50]

    , author Liu, N

    author Zhang, D. , author Liu, N. , author Wu, Z. , author Wang, X. , year 2024 a. title 3d craniofacial similarity calculation and craniofacial relationships analysis based on spectral analysis method . journal Multimedia Tools and Applications volume 83 , pages 14063--14084

    2024

  41. [51]

    , author Jie, B

    author Zhang, R. , author Jie, B. , author He, Y. , author Zhu, L. , author Xie, Z. , author Liu, Z. , author Mo, H. , author Wang, J. , year 2024 b. title Craniomaxillofacial bone segmentation and landmark detection using semantic segmentation networks and an unbiased heatmap...

    2024

  42. [52]

    , author Mo, H

    author Zhang, R. , author Mo, H. , author Hu, W. , author Jie, B. , author Xu, L. , author He, Y. , author Ke, J. , author Wang, J. , year 2024 c. title Super-resolution landmark detection networks for medical images . journal Computers in Biology and Medicine volume 182 , pag...

    2024

  43. [53]

    , author Mo, H

    author Zhang, R. , author Mo, H. , author Wang, J. , author Jie, B. , author He, Y. , author Jin, N. , author Zhu, L. , year 2024 d. title Utsrmorph: A unified transformer and superresolution network for unsupervised medical image registration . journal IEEE Transactions on Me...

    2024

  44. [54]

    , author Jiang, L

    author Zhao, H. , author Jiang, L. , author Fu, C.W. , author Jia, J. , year 2019 . title Pointweb: Enhancing local neighborhood features for point cloud processing , in: booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pp. pages 5565--5573

    2019

  45. [55]

    , author Jiang, L

    author Zhao, H. , author Jiang, L. , author Jia, J. , author Torr, P. , author Koltun, V. , year 2021 . title Point transformer , in: booktitle 2021 IEEE/CVF International Conference on Computer Vision (ICCV) , pp. pages 16239--16248 . :10.1109/ICCV48922.2021.01595

    2021

  46. [56]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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