Pith. sign in

REVIEW 4 major objections 6 minor 56 references

RoofSeg: An edge-aware transformer-based network for end-to-end roof plane segmentation

T0 review · 4 major / 6 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 RoofSeg claims that roof planes in airborne LiDAR point clouds can be segmented end to end by a transformer using learnable plane queries, with edge-aware refinement and geometric losses, reporting roughly 3-4 point improvements over the pr

desk verdict RoofSeg is a serious end-to-end transformer for roof-plane segmentation with thorough ablations; the SOTA claim needs a missing baseline and a differentiability clarification, both fixable. read the letter →

arxiv 2508.19003 v1 pith:D4ZU43YJ submitted 2025-08-26 cs.CV cs.AI

Siyuan You , Guozheng Xu , Pengwei Zhou , Qiwen Jin , Jian Yao , Li Li This is my paper
classification cs.CVcs.AI
keywords roofplanesegmentationLiDARpointcloudstransformeredge-awaremaskmodule3Dbuildingreconstructioninstancecloud
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

RoofSeg argues that roof plane segmentation from airborne LiDAR can be done truly end to end: a transformer with learnable plane queries directly predicts plane instance masks, eliminating hand-tuned clustering and post-processing. To fix poor edge discrimination, it adds an Edge-Aware Mask Module that injects point-to-plane distances at predicted edge points, plus an adaptive-weighting mask loss and a plane geometric loss that reduce misclassified points and enforce planarity. On RoofNTNU, Roofpc3D, and Building3D, it reports mean coverage of 0.9589, 0.9601, and 0.9374, roughly 3-4 percentage points above the strongest prior approach. If correct, this would make roof plane extraction for 3D building reconstruction less sensitive to geometric clustering hyperparameters.

What carries the argument

The load-bearing mechanism is the set of learnable plane queries interacting with multi-scale point features through hierarchical cross-attention (the Query Refinement Decoders), followed by the Edge-Aware Mask Module (EAMM). EAMM augments the features of predicted edge points with the tangent distance from each edge point to a PCA-fitted plane of the predicted mask, then runs a self-attention layer over the merged edge and non-edge features to produce refined masks. Two loss terms carry the training signal: an adaptive-weighting BCE+Dice mask loss that suppresses misclassified points, and a plane geometric loss that minimizes distance from in-plane points to the fitted plane. The claim of '

What would settle it

Measure the gradient of the total loss L with respect to the pre-binarization plane-mask logits (the sigmoid(A) values before the 0.5 threshold in Eq. 3) and to the fused edge features E_fuse in EAMM. If RoofSeg trains end-to-end as claimed, these gradients are nonzero and the tangent-distance vector changes training; if the binarization and KNN outlier selection block the gradient, the edge-aware module cannot be learning from the geometric cue, and the ablation gains would have to come from another mechanism. A simpler discriminator: replace the tangent-distance vector in EAMM with random va

Watch

Extended reading notes

Core claim

The central claim is that roof planes in airborne LiDAR point clouds can be segmented in a single network forward pass by a query-based transformer, without iterative region growing, RANSAC, or clustering. A fixed set of learnable plane queries is refined through hierarchical cross-attention over multi-scale PointNet++ features; each query then predicts an initial plane mask. The Edge-Aware Mask Module (EAMM) refines these masks by extracting predicted edge points, fitting a PCA plane to the predicted interior points, and fusing the edge points' tangent distances to that plane back into the point features. An adaptive weighting scheme in the mask loss lowers the influence of points whose lab

Load-bearing premise

The hard threshold at 0.5 in Eq. (3), the binarized edge mask, and the KNN majority-rule outlier selection in Algorithm 1 are treated as trainable parts of an end-to-end network, but the paper never specifies how gradients pass through them; if they do not, the claim of truly end-to-end edge-aware training is not established.

Editorial extensions

If this is right

  • Roof plane segmentation for LoD2/LoD3 building reconstruction could become a single network forward pass, removing the hyperparameter tuning burden of clustering and region growing.
  • Edge accuracy should improve on complex roofs because point-to-plane distance gives the network an explicit geometric cue where features are usually least discriminative.
  • The plane geometric loss should make predicted segments lie closer to true planar surfaces, simplifying downstream model fitting and boundary extraction.
  • On all three benchmarks, RoofSeg outperforms both traditional pipelines and the strongest deep baseline, suggesting a new state of the art for this task.
  • The adaptive weighting and geometric loss, being largely insensitive to neighbor-count settings, point toward a stable training recipe that could transfer to other planar instance segmentation problems.

Reading between the lines

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

  • If the end-to-end claim holds, the same recipe of learnable queries plus explicit point-to-plane residuals could transfer to other planar primitive tasks, such as indoor wall and floor segmentation, where edge precision is the bottleneck.
  • The non-differentiable binarization and KNN selection inside EAMM could be replaced with differentiable soft assignments (e.g., straight-through estimators or weighted least-squares plane fitting), which would make the 'truly end-to-end' claim unambiguous and potentially improve the edge cue.
  • A testable prediction from the ablations is that EAMM's gain scales with the density of edge points: on roofs with few adjacent planes, the module should contribute less than the reported 4-5 points in mCov, while on highly fragmented roofs it should contribute more.
  • The reported insensitivity to the number of nearest neighbors in adaptive weighting suggests the mechanism acts as a broad regularizer rather than a finely tuned outlier filter, which may simplify transfer to datasets with different point densities.
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 / 6 minor

Summary. The paper presents RoofSeg, a query-based transformer for end-to-end roof plane instance segmentation from airborne LiDAR point clouds. It combines a PointNet++ encoder with attention-based feature propagation, multiple query refinement decoders, an Edge-Aware Mask Module (EAMM) that augments features at predicted edge points with point-to-plane distances, and a loss function with adaptive point weighting plus a plane geometric term. Experiments are reported on RoofNTNU, Roofpc3D, and Building3D; RoofSeg is claimed to outperform Region Growing, RANSAC, GoCoPP, PointGroup, Mask3D, and DeepRoofPlane on mCov/mWCov/mPrec/mRec. Ablations and efficiency analyses support the main design choices.

Significance. If the results are reproducible and the comparison is complete, this is a useful advance for LoD2/LoD3 building reconstruction. The paper's strengths include consistent gains across three datasets, a systematic ablation of EAMM and the loss components, and an efficiency analysis of decoder count and point resolution. The central claims, however, rest on a comparison set that omits a directly relevant transformer baseline, on an unclear treatment of non-differentiable operations inside the claimed end-to-end training loop, and on single-run point estimates without variance. These gaps do not by themselves invalidate the method, but they need to be resolved before the state-of-the-art claim can be accepted.

major comments (4)
  1. [Section 2.2 / Table 1] SPPSFormer (Zeng et al., 2025) is described in Section 2.2 as a superpoint-based transformer for roof plane instance segmentation, yet it does not appear in the quantitative comparison of Table 1. This is a near-contemporary, directly comparable transformer baseline for the exact task. Without it, the claim of "new state-of-the-art" is not established. Please add the comparison or explicitly justify why it is excluded.
  2. [Section 3.3 / Eq. (3) / Algorithm 1] The paper claims truly end-to-end training. However, Eq. (3) binarizes predicted masks at a 0.5 threshold; the edge mask is also binarized; EAMM extracts edge/interior points, fits a plane with PCA, and computes point-to-plane distances; and Algorithm 1 identifies outliers via a KNN majority rule. These are non-differentiable operations, and the manuscript does not state whether gradients are detached, softened, or approximated. Without this information, the end-to-end claim and the attribution of the reported gains to EAMM are not established. Please specify the gradient flow through these components or revise the claim.
  3. [Tables 1-9] All metrics are reported as single-run point estimates. Some differences are small (e.g., Table 5 shows mCov differences under 0.5 percentage points; Table 9 shows differences around 0.0005-0.001 between multi-scale and full-resolution features). The phrase "significantly outperforms" is therefore not supported by the presented evidence. Please report mean and standard deviation over at least three independent runs, or otherwise quantify variability.
  4. [Section 4.2] Roofpc3D was generated by Li et al. (2024) and the Building3D plane labels were manually annotated by Li et al. (2024); the evaluation metrics are also defined in that paper. Since the authors are from the same group, there is a risk of annotation or evaluation bias. Please state the relationship explicitly, describe the labeling protocol, and, if possible, include a fully external benchmark beyond RoofNTNU.
minor comments (6)
  1. [Eq. (9)] In the weighted Dice loss, the denominator appears as w_j(Σ a_j + Σ lgt_j) + ε with w_j outside the sums. This is dimensionally inconsistent; the intended form is likely 2Σ w_j a_j lgt_j / (Σ w_j a_j + Σ w_j lgt_j). Please correct.
  2. [Algorithm 1] Line 10 compares "Nmis" which is not defined; it should presumably be N_j^dif. Also the loop in line 3 says "for j = 0 to N" but the text indexing is 1-based elsewhere.
  3. [Abstract] Typo: "aiborne LiDAR" should be "airborne LiDAR".
  4. [Section 4.1] Typos: "sampling radio" should be "sampling ratio"; "output demision" should be "output dimension"; "dropout threshold" is likely "dropout rate".
  5. [Figure 10 caption] Caption reads "BAMM" in the third row description; this should be "EAMM".
  6. [Section 3.1] Typo: "The input point clouds can be donated" should be "denoted".

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical architecture paper with no derivation that reduces to its own inputs.

full rationale

RoofSeg does not claim a first-principles derivation whose output is equivalent to its input by construction. The plane masks are generated from query-point affinities (Eqs. 2-3), refined by the EAMM, and merged by Eq. 4; these are ordinary network operations, not fitted quantities renamed as predictions. The losses in Eqs. 7-14 are training objectives, and the ablations in Table 2 compare variants of the same architecture, so the reported gains are not forced by the loss definition itself. The main self-citation is to Li et al. (2024) for the Roofpc3D and Building3D preparations and for the metric definitions. That prior work is from the same group, which is a fairness and independence concern, but it is not circular: RoofNTNU is an external benchmark, the metrics are standard segmentation metrics, and the results are produced by an independent training/evaluation protocol rather than by algebraic identity. The omission of SPPSFormer (Zeng et al., 2025) from Table 1 is a baseline-completeness issue and weakens the SOTA claim, but it is not a circularity. Similarly, the non-differentiable thresholding in Eq. (3) and Algorithm 1 concerns the validity of the 'truly end-to-end' claim, not circular reasoning. Overall, no load-bearing step reduces to its own inputs by definition or by self-citation chain.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

RoofSeg introduces no new physical entities. Its free parameters are standard architectural hyperparameters (query count, decoder depth, neighbor count, thresholds), all tuned on validation performance. The main assumptions are domain-level: roofs are planar, the benchmark labels are trustworthy, and point-to-plane distance is a good edge cue. The ad hoc outlier and threshold rules are embedded in the loss design and are not independently justified.

free parameters (6)
  • Number of plane queries K = 32 (RoofNTNU, Roofpc3D), 64 (Building3D)
    Chosen by validation scan in Table 4; performance drops at 16 and 128 queries.
  • Number of query refinement decoders Ndec = 8
    Set from ablation in Tables 6 and 7; 4 decoders lose accuracy, 12 add compute without significant gain.
  • Nearest neighbors Nnbr for adaptive weighting loss = 30
    Table 5 shows insensitivity across 10 to 50; 30 was chosen empirically.
  • Binarization and semantic thresholds = 0.5
    Plane masks (Eq. 3), edge masks, and positive semantic masks all use a fixed 0.5 threshold; not learned.
  • Training point cloud resample size = 2048 points
    All training inputs are uniformly resampled to 2048 points; edge density and coverage depend on this choice.
  • Set Abstraction receptive radii = [0.05, 0.1], [0.1, 0.2], [0.2, 0.4], [0.4, 0.8]
    Chosen for the point density of the three benchmarks; affects multi-scale feature aggregation.
assumptions (5)
  • domain assumption Roofs can be decomposed into planar patches
    Core task definition used in Section 1 and throughout; LoD2/LoD3 reconstruction assumes planar roof structures.
  • domain assumption Ground-truth plane labels in RoofNTNU, Roofpc3D, and Building3D are correct and consistent
    All quantitative claims in Table 1 are computed against these labels; Building3D labels were manually annotated by the authors' group (Li et al. 2024).
  • domain assumption Point-to-plane tangent distance from an interior-point PCA fit is a sufficient edge-discriminative cue
    The EAMM design in Section 3.3 is built entirely on this geometric cue, with no comparison to alternative edge features.
  • ad hoc to paper A point is an outlier when more than half of its nearest neighbors disagree, and masks are positive when score >= 0.5
    Algorithm 1 and the semantic branch use these hand-set rules; they are not learned and are central to the adaptive weighting loss.
  • standard math PCA plane fitting and point-to-plane distances remain stable on small, noisy edge neighborhoods
    Used in EAMM and in the plane geometric loss, Eq. (11); no robustness analysis is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RoofSeg: An edge-aware transformer-based network for end-to-end roof plane segmentation." pith.science (2026). https://pith.science/paper/D4ZU43YJ

@misc{pith2026250819003,
  author       = {Pith},
  title        = {Pith review of: RoofSeg: An edge-aware transformer-based network for end-to-end roof plane segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D4ZU43YJ}},
  note         = {Machine review of arXiv:2508.19003}
}
read the original abstract

Roof plane segmentation is one of the key procedures for reconstructing three-dimensional (3D) building models at levels of detail (LoD) 2 and 3 from airborne light detection and ranging (LiDAR) point clouds. The majority of current approaches for roof plane segmentation rely on the manually designed or learned features followed by some specifically designed geometric clustering strategies. Because the learned features are more powerful than the manually designed features, the deep learning-based approaches usually perform better than the traditional approaches. However, the current deep learning-based approaches have three unsolved problems. The first is that most of them are not truly end-to-end, the plane segmentation results may be not optimal. The second is that the point feature discriminability near the edges is relatively low, leading to inaccurate planar edges. The third is that the planar geometric characteristics are not sufficiently considered to constrain the network training. To solve these issues, a novel edge-aware transformer-based network, named RoofSeg, is developed for segmenting roof planes from LiDAR point clouds in a truly end-to-end manner. In the RoofSeg, we leverage a transformer encoder-decoder-based framework to hierarchically predict the plane instance masks with the use of a set of learnable plane queries. To further improve the segmentation accuracy of edge regions, we also design an Edge-Aware Mask Module (EAMM) that sufficiently incorporates planar geometric prior of edges to enhance its discriminability for plane instance mask refinement. In addition, we propose an adaptive weighting strategy in the mask loss to reduce the influence of misclassified points, and also propose a new plane geometric loss to constrain the network training.

Figures

Figures reproduced from arXiv: 2508.19003 by the authors.

Figure 1
Figure 1. The overview of the end-to-end edge-aware transformer-based network (RoofSeg). Notably, [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. The illustration of the Attention-based Feature Propagation (AFP) and Query Refinement [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. The illustration of the Edge-Aware Mask Module (EAMM). The green dots in the edge mask [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: The illustration of outliers. The red and blue dots represent the predicted in-plane points [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Visualization of some roof samples from RoofNTNU benchmark (the first line), Roofpc3D [PITH_FULL_IMAGE:figures/full_fig_p019_5.png]
Figure 6
Figure 6. Figure 6: Visual results of roof plane segmentation for all comparative approaches and RoofSeg on [PITH_FULL_IMAGE:figures/full_fig_p022_6.png]
Figure 7
Figure 7. Figure 7: Visual results of roof plane segmentation for all comparative approaches and RoofSeg on [PITH_FULL_IMAGE:figures/full_fig_p023_7.png]
Figure 8
Figure 8. Figure 8: Visual results of roof plane segmentation for all comparative approaches and RoofSeg on [PITH_FULL_IMAGE:figures/full_fig_p024_8.png]
Figure 9
Figure 9. Figure 9: Large scale segmentation results of RoofSeg on the real-world Building3D benchmark. In (a), [PITH_FULL_IMAGE:figures/full_fig_p025_9.png]
Figure 10
Figure 10. Figure 10: Segmentation results of two roof samples using the proposed network with different com [PITH_FULL_IMAGE:figures/full_fig_p027_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

56 extracted references · 55 canonical work pages

  1. [1]

    , author Veksler, O

    author Boykov, Y. , author Veksler, O. , year 2006 . title Graph cuts in vision and graphics: Theories and applications , in: booktitle Handbook of Mathematical Models in Computer Vision , pp. pages 79--96

    work page 2006

  2. [2]

    , author Karsli, F

    author Canaz Sevgen, S. , author Karsli, F. , year 2020 . title An improved RANSAC algorithm for extracting roof planes from airborne LiDAR data . journal The Photogrammetric Record volume 35 , pages 40--57

    work page 2020

  3. [3]

    , author Cao, J

    author Cao, Y. , author Cao, J. , author Xing, L. , author Zhu, B. , author Bao, Q. , author Li, T. , author Wang, Q. , year 2025 . title An enhanced- RANSAC -fusion and K d-tree-filtering pipeline method for multi-scale 3D substation components extraction . journal IEEE Sensors Journal volume 25 , pages 30897--30905

    work page 2025

  4. [4]

    , author Wang, Y

    author Cao, Y. , author Wang, Y. , author Xue, Y. , author Zhang, H. , author Lao, Y. , year 2022 . title FEC : F ast E uclidean clustering for point cloud segmentation . journal Drones volume 6 , pages 325

    work page 2022

  5. [5]

    , author Xie, F

    author Chen, F. , author Xie, F. , author Sun, L. , author Gu, Y. , author Zhang, Z. , author Chen, J. , author Zhang, J. , author Yi, M. , year 2024 . title Point cloud segmentation algorithm based on improved E uclidean clustering . journal IEEE Access volume 12 , pages 152959--152971

    work page 2024

  6. [6]

    , author Mao, J

    author Chen, X. , author Mao, J. , author Zhao, B. , author Wu, C. , author Qin, M. , year 2025 . title Facet-segmentation of point cloud based on multiscale hypervoxel region growing . journal Journal of the Indian Society of Remote Sensing , pages 1--22

    work page 2025

  7. [7]

    , author Misra, I

    author Cheng, B. , author Misra, I. , author Schwing, A.G. , author Kirillov, A. , author Girdhar, R. , year 2022 . title Masked-attention mask transformer for universal image segmentation , in: booktitle IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 1290--1299

    work page 2022

  8. [8]

    , author Schwing, A

    author Cheng, B. , author Schwing, A. , author Kirillov, A. , year 2021 . title Per-pixel classification is not all you need for semantic segmentation , in: booktitle Advances in Neural Information Processing Systems (NeurIPS) , pp. pages 17864--17875

    work page 2021

Show all 56 references
  1. [9]

    , author Gwak, J

    author Choy, C. , author Gwak, J. , author Savarese, S. , year 2019 . title 4D spatio-temporal convnets: Minkowski convolutional neural networks , in: booktitle IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 3075--3084

    2019

  2. [10]

    , author Sankaran, B

    author Czerniawski, T. , author Sankaran, B. , author Nahangi, M. , author Haas, C. , author Leite, F. , year 2018 . title 6D DBSCAN -based segmentation of building point clouds for planar object classification . journal Automation in Construction volume 88 , pages 44--58

    2018

  3. [11]

    , author Nagi, R

    author Date, K. , author Nagi, R. , year 2016 . title GPU -accelerated Hungarian algorithms for the linear assignment problem . journal Parallel Computing volume 57 , pages 52--72

    2016

  4. [12]

    , author Zhuang, C

    author Fang, Z. , author Zhuang, C. , author Lu, Z. , author Wang, Y. , author Liu, L. , author Xiao, J. , year 2025 . title BGPSeg : B oundary-guided primitive instance segmentation of point clouds . journal IEEE Transactions on Image Processing (TIP) volume 34 , pages 1454--1468

    2025

  5. [13]

    , author Cao, Z

    author Gu, Y. , author Cao, Z. , author Dong, L. , year 2017 . title A hierarchical energy minimization method for building roof segmentation from airborne LiDAR data . journal Multimedia Tools and Applications (MTA) volume 76 , pages 4197--4210

    2017

  6. [14]

    , author Zhang, C

    author He, Y. , author Zhang, C. , author Fraser, C.S. , year 2013 . title A line-based spectral clustering method for efficient planar structure extraction from LiDAR data . journal ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences volume 2 ,...

    2013

  7. [15]

    , author Stoter, J

    author Huang, J. , author Stoter, J. , author Peters, R. , author Nan, L. , year 2022 . title City3D : L arge-scale building reconstruction from airborne LiDAR point clouds . journal Remote Sensing volume 14 , pages 2254

    2022

  8. [16]

    , author Zhang, Y

    author Huang, J. , author Zhang, Y. , author Sun, M. , year 2021 . title PrimitiveNet : P rimitive instance segmentation with local primitive embedding under adversarial metric , in: booktitle IEEE/CVF International Conference on Computer Vision (ICCV) , pp. pages 15343--15353

    2021

  9. [17]

    , author Zhao, H

    author Jiang, L. , author Zhao, H. , author Shi, S. , author Liu, S. , author Fu, C.W. , author Jia, J. , year 2020 . title PoinGroup : Dual -set point grouping for 3D instance segmentation , in: booktitle IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) ,...

    2020

  10. [18]

    , author Lattanzi, D

    author Khaloo, A. , author Lattanzi, D. , year 2017 . title Robust normal estimation and region growing segmentation of infrastructure 3D point cloud models . journal Advanced Engineering Informatics (AEI) volume 34 , pages 1--16

    2017

  11. [19]

    , author Yuan, Y

    author Lai, X. , author Yuan, Y. , author Chu, R. , author Chen, Y. , author Hu, H. , author Jia, J. , year 2023 . title Mask-attention-free transformer for 3 D instance segmentation , in: booktitle IEEE/CVF International Conference on Computer Vision (ICCV) , pp. pages 3693--3703

    2023

  12. [20]

    , author Li, Q

    author Li, L. , author Li, Q. , author Xu, G. , author Zhou, P. , author Tu, J. , author Li, J. , author Li, M. , author Yao, J. , year 2024 . title A boundary-aware point clustering approach in Euclidean and embedding spaces for roof plane segmentation . journal ISPRS Journal...

    2024

  13. [21]

    , author Yang, F

    author Li, L. , author Yang, F. , author Zhu, H. , author Li, D. , author Li, Y. , author Tang, L. , year 2017 . title An improved RANSAC for 3D point cloud plane segmentation based on normal distribution transformation cells . journal Remote Sensing volume 9 , pages 433

    2017

  14. [22]

    , author Yao, J

    author Li, L. , author Yao, J. , author Tu, J. , author Liu, X. , author Li, Y. , author Guo, L. , year 2020 . title Roof plane segmentation from airborne LiDAR data using hierarchical clustering and boundary relabeling . journal Remote Sensing volume 12 , pages 1363

    2020

  15. [23]

    , author Liu, S

    author Li, Y. , author Liu, S. , author Yang, X. , author Guo, J. , author Guo, J. , author Guo, Y. , year 2023 . title Surface and edge detection for primitive fitting of point clouds , in: booktitle ACM SIGGRAPH 2023 conference proceedings , pp. pages 1--10

    2023

  16. [24]

    , author Ma, H

    author Liu, K. , author Ma, H. , author Zhang, L. , author Liang, X. , author Chen, D. , author Liu, Y. , year 2023 . title Roof segmentation from airborne LiDAR using octree-based hybrid region growing and boundary neighborhood verification voting . journal IEEE Journal of Se...

    2023

  17. [25]

    , author Obukhov, A

    author Liu, Y. , author Obukhov, A. , author Wegner, J.D. , author Schindler, K. , year 2024 . title Point2Building : R econstructing buildings from airborne LiDAR point clouds . journal ISPRS Journal of Photogrammetry and Remote Sensing volume 215 , pages 351--368

    2024

  18. [26]

    , author Wang, J

    author Liu, Z. , author Wang, J. , author Liu, W. , year 2005 . title Building extraction from high resolution imagery based on multi-scale object oriented classification and probabilistic Hough Transform , in: booktitle International Geoscience and Remote Sensing Symposium (I...

    2005

  19. [27]

    , author Hutter, F

    author Loshchilov, I. , author Hutter, F. , et al., year 2017 . title Fixing weight decay regularization in A dam . journal arXiv preprint arXiv:1711.05101 volume 5 , pages 5

    2017 arXiv

  20. [28]

    , author Li, S

    author Miao, Y. , author Li, S. , author Wang, L. , author Li, H. , author Qiu, R. , author Zhang, M. , year 2023 . title A single plant segmentation method of maize point cloud based on E uclidean clustering and K -means clustering . journal Computers and Electronics in Agric...

    2023

  21. [29]

    , author Girdhar, R

    author Misra, I. , author Girdhar, R. , author Joulin, A. , year 2021 . title An end-to-end transformer model for 3 D object detection , in: booktitle IEEE/CVF International Conference on Computer Vision (ICCV) , pp. pages 2906--2917

    2021

  22. [30]

    , author Bodum, L

    author Overby, J. , author Bodum, L. , author Kjems, E. , author Iisoe, P. , year 2004 . title Automatic 3D building reconstruction from airborne laser scanning and cadastral data using Hough Transform . journal International Archives of Photogrammetry, Remote Sensing and Spat...

    2004

  23. [31]

    , author Mattes, C

    author Poux, F. , author Mattes, C. , author Selman, Z. , author Kobbelt, L. , year 2022 . title Automatic region-growing system for the segmentation of large point clouds . journal Automation in Construction volume 138 , pages 104250

    2022

  24. [32]

    , author Yi, L

    author Qi, C.R. , author Yi, L. , author Su, H. , author Guibas, L.J. , year 2017 . title PointNet++ : Deep hierarchical feature learning on point sets in a metric space , in: booktitle Advances in Neural Information Processing Systems (NeurIPS) , pp. pages 5099--5108

    2017

  25. [33]

    , author Jung, H

    author Roh, W. , author Jung, H. , author Nam, G. , author Yeom, J. , author Park, H. , author Yoon, S.H. , author Kim, S. , year 2024 . title Edge-aware 3 D instance segmentation network with intelligent semantic prior , in: booktitle IEEE/CVF Conference on Computer Vision an...

    2024

  26. [34]

    , author Wahl, R

    author Schnabel, R. , author Wahl, R. , author Klein, R. , year 2007 . title Efficient RANSAC for point-cloud shape detection . journal Computer Graphics Forum volume 26 , pages 214--226

    2007

  27. [35]

    , author Engelmann, F

    author Schult, J. , author Engelmann, F. , author Hermans, A. , author Litany, O. , author Tang, S. , author Leibe, B. , year 2023 . title Mask3D : M ask transformer for 3 D semantic instance segmentation , in: booktitle IEEE International Conference on Robotics and Automation...

    2023

  28. [36]

    , author Qing, C

    author Sun, J. , author Qing, C. , author Tan, J. , author Xu, X. , year 2023 . title Superpoint transformer for 3 D scene instance segmentation , in: booktitle AAAI Conference on Artificial Intelligence , pp. pages 2393--2401

    2023

  29. [37]

    , author Guo, B

    author Sun, X. , author Guo, B. , author Li, C. , author Sun, N. , author Wang, Y. , author Yao, Y. , year 2024 . title Semantic segmentation and roof reconstruction of urban buildings based on LiDAR point clouds . journal ISPRS International Journal of Geo-Information volume ...

    2024

  30. [38]

    , author Landes, T

    author Tarsha-Kurdi, F. , author Landes, T. , author Grussenmeyer, P. , year 2007 . title Hough-transform and extended RANSAC algorithms for automatic detection of 3D building roof planes from LiDAR data , in: booktitle ISPRS Workshop on Laser Scanning , pp. pages 407--412

    2007

  31. [39]

    , author Shazeer, N

    author Vaswani, A. , author Shazeer, N. , author Parmar, N. , author Uszkoreit, J. , author Jones, L. , author Gomez, A.N. , author Kaiser, . , author Polosukhin, I. , year 2017 . title Attention is all you need , in: booktitle Advances in Neural Information Processing Systems...

    2017

  32. [40]

    , author Truong-Hong, L

    author Vo, A.V. , author Truong-Hong, L. , author Laefer, D.F. , author Bertolotto, M. , year 2015 . title Octree-based region growing for point cloud segmentation . journal ISPRS Journal of Photogrammetry and Remote Sensing volume 104 , pages 88--100

    2015

  33. [41]

    , author Ji, M

    author Wang, C. , author Ji, M. , author Wang, J. , author Wen, W. , author Li, T. , author Sun, Y. , year 2019 . title An improved DBSCAN method for LiDAR data segmentation with automatic E ps estimation . journal Sensors volume 19 , pages 172

    2019

  34. [42]

    , author Huang, S

    author Wang, R. , author Huang, S. , author Yang, H. , year 2023 a. title Building3D : A urban-scale dataset and benchmarks for learning roof structures from point clouds , in: booktitle IEEE/CVF International Conference on Computer Vision (ICCV) , pp. pages 20076--20086

    2023

  35. [43]

    , author Qin, F

    author Wang, S. , author Qin, F. , author Tong, Y. , author Shang, X. , author Zhang, Z. , year 2023 b. title Probabilistic boundary-guided point cloud primitive segmentation network . journal IEEE Transactions on Instrumentation and Measurement volume 72 , pages 1--13

    2023

  36. [44]

    , author Zhang, Y

    author Wang, W. , author Zhang, Y. , author Ge, G. , author Jiang, Q. , author Wang, Y. , author Hu, L. , year 2023 c. title Indoor point cloud segmentation using a modified region growing algorithm and accurate normal estimation . journal IEEE Access volume 11 , pages 42510--42520

    2023

  37. [45]

    , author Ji, S

    author Wang, X. , author Ji, S. , year 2021 . title Roof plane segmentation from LiDAR point cloud data using region expansion based L0 gradient minimization and graph cut . journal IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (J-STARS) volu...

    2021

  38. [46]

    , author Zhang, X

    author Wu, H. , author Zhang, X. , author Shi, W. , author Song, S. , author Cardenas-Tristan, A. , author Li, K. , year 2019 . title An accurate and robust region-growing algorithm for plane segmentation of TLS point clouds using a multiscale tensor voting method . journal IE...

    2019

  39. [47]

    , author Jiang, W

    author Xu, B. , author Jiang, W. , author Shan, J. , author Zhang, J. , author Li, L. , year 2015 . title Investigation on the weighted RANSAC approaches for building roof plane segmentation from LiDAR point clouds . journal Remote Sensing volume 8 , pages 5

    2015

  40. [48]

    , author You, S

    author Xu, G. , author You, S. , author Liu, K. , author Li, L. , author Yao, J. , year 2025 . title A coarse-to-fine boundary relabeling approach for roof plane segmentation . journal IEEE Geoscience and Remote Sensing Letters volume 22 , pages 1--5

    2025

  41. [49]

    , author Li, J

    author Xu, J. , author Li, J. , year 2023 . title Optimal RANSAC method for segmentation of complex building roof planes . journal Geomatics and Information Science of Wuhan University volume 48 , pages 1531--1537

    2023

  42. [50]

    , author Shan, J

    author Yan, J. , author Shan, J. , author Jiang, W. , year 2014 . title A global optimization approach to roof segmentation from airborne LiDAR point clouds . journal ISPRS Journal of Photogrammetry and Remote Sensing volume 94 , pages 183--193

    2014

  43. [51]

    , author Yang, Z

    author Yan, S. , author Yang, Z. , author Ma, C. , author Huang, H. , author Vouga, E. , author Huang, Q. , year 2021 . title HpNet : Deep primitive segmentation using hybrid representations , in: booktitle IEEE/CVF International Conference on Computer Vision (ICCV) , pp. page...

    2021

  44. [52]

    , author Lafarge, F

    author Yu, M. , author Lafarge, F. , year 2022 . title Finding good configurations of planar primitives in unorganized point clouds , in: booktitle IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 6367--6376

    2022

  45. [53]

    , author Qi, X

    author Zeng, C. , author Qi, X. , author Chen, C. , author Sun, K. , author Zhang, W. , author Liu, Y. , author Meng, Y. , author Yang, B. , year 2025 . title SPPSFormer : H igh-quality superpoint-based transformer for roof plane instance segmentation from point clouds . journ...

    2025 arXiv

  46. [54]

    , author Fan, H

    author Zhang, C. , author Fan, H. , year 2022 . title An improved multi-task pointwise network for segmentation of building roofs in airborne laser scanning point clouds . journal The Photogrammetric Record volume 37 , pages 260--284

    2022

  47. [55]

    , author Li, Z

    author Zhang, L. , author Li, Z. , author Li, A. , author Liu, F. , year 2018 . title Large-scale urban point cloud labeling and reconstruction . journal ISPRS Journal of Photogrammetry and Remote Sensing volume 138 , pages 86--100

    2018

  48. [56]

    , author Persello, C

    author Zhao, W. , author Persello, C. , author Stein, A. , year 2022 . title Extracting planar roof structures from very high resolution images using graph neural networks . journal ISPRS Journal of Photogrammetry and Remote Sensing volume 187 , pages 34--45

    2022

Pith tools

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