Pith. sign in

REVIEW 3 major objections 7 minor 1 cited by

TurboReg: TurboClique for Robust and Efficient Point Cloud Registration

T0 review · 3 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Three mutually compatible correspondences are enough to estimate a rigid transformation as reliably as a maximal clique, turning exponential-time search into a linear-time GPU-parallel algorithm.

desk verdict A genuinely faster clique-based registration estimator with solid empirical results, but the paper's central stability justification is a hand-wavy limit that doesn't survive scrutiny. read the letter →

arxiv 2507.01439 v3 pith:UKVGCR6Q submitted 2025-07-02 cs.CV

Shaocheng Yan , Pengcheng Shi , Zhenjun Zhao , Kaixin Wang , Kuang Cao , Ji Wu , Jiayuan Li This is my paper · ORCID
classification cs.CV
keywords pointcloudregistrationrobustestimationcompatibilitygraphmaximalclique3-cliquepivot-guidedsearchsecond-orderspatialconsistencyGPUparallel
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

TurboReg's central claim is that a fixed 3-clique — three correspondences that are mutually compatible under a tight distance-difference bound — estimates a rigid transformation as stably as the maximal cliques used by leading graph-based registration methods, and that this swap turns an exponential-time search into a linear-time, GPU-parallel one. The reasoning is a variance trade-off: the least-squares pose estimate is stabilized either by feeding it many matches, as maximal cliques do, or by suppressing the effective noise through a stringent compatibility threshold $\tau$, which is the mode TurboClique relies on. To keep the three-match hypotheses rich in inliers, the Pivot-Guided Search ranks the highest second-order-compatibility edges as pivots and expands each into a third compatible match, producing only $K_1 K_2$ hypotheses per frame. If the claim holds, robust registration stops being a combinatorial bottleneck: on the 3DMatch dataset with FCGF features, TurboReg (1K) reports more than 200× the GPU throughput of the 3DMAC maximal-clique method while raising registration recall from 92.79% to 93.59%.

What carries the argument

The load-bearing objects are (1) TurboClique, a 3-clique in a compatibility graph whose edge exists when $|\,\|x_i-x_j\|-\|y_i-y_j\|\,| \le \tau$, with $\tau$ set near a quarter of the point-cloud resolution; (2) the Pivot-Guided Search (PGS), which orders edges of the second-order compatibility graph $\hat{G}_{ij} = G_{ij}\sum_k G_{ik}G_{jk}$, takes the top $K_1$ edges as pivots, and expands each pivot by its common neighbors into triplets scored by summed edge weight, keeping the top $K_2$ per pivot; and (3) the Ordered SC² Graph (O2Graph), a directed version that orients each edge from the lower to the higher node index and provably assigns every TurboClique to a unique pivot. The theoretical engine is the variance identity for the least-squares estimator, which frames clique size and residual noise as interchangeable sources of stability and licenses the swap from maximal to minimal cliques.

What would settle it

Run the discrepancy experiment from App. A.2 on datasets beyond 3DMatch+FPFH: for a sample of pairs, extract 10-cliques and all their $\binom{10}{3}$ 3-clique subsets at the recommended threshold $\tau \approx 0.25\times$ resolution, and compare the estimated rotations and translations; if on 3DLoMatch, KITTI, or with FCGF/Predator features the median disagreement grows to a substantial fraction of the success thresholds (15° rotation, 30 cm translation) — or if TurboReg's recall drops noticeably versus a 10-clique variant on the same pairs — then the compensating-stability assumption does not transfer and the reported speedup is specific to the 3DMatch+FPFH regime.

Watch

Extended reading notes

Core claim

The paper's central claim is that a TurboClique — a 3-clique in a compatibility graph built with a stringent (small) threshold $\tau$ (Definition 1) — carries enough spatial-consistency information to estimate a rigid transformation as stably as a maximal clique does. The supporting analysis treats the Kabsch solver as a least-squares estimator with variance $\mathrm{Var}(\hat{\beta}|X) = \sigma^2(X'X)^{-1}$, isolating two stability mechanisms: data-scaling stability from using many matches, and pairwise compatibility-induced stability from enforcing $|\,\|x_i-x_j\| - \|y_i-y_j\|\,| \le \tau$ on every pair. The paper argues that as $\tau$ shrinks, the constraint truncates the matching-noise distribution so that its effective variance $\sigma^2_{\text{eff}}$ tends to zero, and that the recommended threshold $\tau \approx 0.25\times$ point-cloud resolution makes a 3-clique estimate agree with a 10-clique estimate to within $0.1^\circ$ rotation and $0.5$ mm translation on 3DMatch+FPFH (App. A.2). The search side is carried by PGS, which ranks edges of the SC² graph by their second-order score, takes the top $K_1$ as pivots, and keeps the top $K_2$ third matches per pivot, giving $K_1 K_2$ hypotheses in $O(K_1 N)$ time; the directed O2Graph variant guarantees that each TurboClique is generated by exactly one pivot.

Load-bearing premise

The load-bearing premise is that tightening the compatibility threshold squeezes the matching-noise variance down fast enough that three compatible matches estimate a transformation as stably as a large clique would, a trade-off supported by a heuristic noise-truncation argument and a single 3-clique-versus-10-clique experiment on 3DMatch+FPFH rather than by a quantitative bound that holds across scenes, descriptors, and noise levels.

Editorial extensions

If this is right

  • Maximal-clique enumeration — costing on the order of $O(d(N-d)3^{d/3})$ for graph degeneracy $d$ — can be replaced by an $O(N)$ fixed-size search, and with enough parallel units on a GPU the search approaches constant time.
  • Only $K_1 K_2 = 2000$ hypotheses are needed per point-cloud pair at the default settings, orders of magnitude fewer than the millions of RANSAC iterations, and this budget yields the highest registration recall among the compared estimators on 3DMatch, 3DLoMatch, and KITTI.
  • The speed and accuracy transfer across descriptor families: FPFH, FCGF, and Predator features all feed the same TurboClique pipeline without retraining or per-descriptor tuning beyond the threshold rule.
  • Runtime becomes predictable: unlike maximal-clique enumeration, whose cost depends on graph density and branching, PGS has a tightly clustered per-frame runtime, which the paper positions as suitable for real-time SLAM-style applications.
  • The ablations give a simple operating recipe: set $\tau$ near a quarter of the point-cloud resolution, $K_1 = 1000$, and $K_2 = 2$, and recall stays within a couple of points of its peak over a wide range of $\tau$ (above 82% for $\tau$ from 0.01 m to 0.08 m on 3DMatch+FPFH).

Reading between the lines

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

  • My inference: the stability trade-off is argued heuristically — a noise-truncation limit rather than a quantitative bound on transformation-error variance as a function of $\tau$ and clique size — so the compensation claim is a design principle awaiting a theorem; a closed-form variance bound would predict the optimal $\tau$ without per-dataset ablations.
  • My inference: the fixed-size-clique-plus-pivot recipe is generic and should transfer to any consensus problem with a pairwise geometric consistency graph, including loop-closure detection in SLAM, multi-view image matching, and wide-baseline stereo, where the same exponential graph-search bottleneck appears.
  • My inference: the paper's own supplementary analysis suggests the remaining bottleneck is hypothesis ranking, not search — the top-1 inlier-count rule misses correct transformations that sit inside the top-$K$ set, so a better re-ranker on top of PGS could raise recall further without slowing the search.
  • My inference: the practical ceiling of the method is reached when fewer than three true inliers survive feature matching — at extreme low overlap the 3-clique itself cannot exist, and the method would need a denser or better descriptor rather than a bigger search.
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 / 7 minor

Summary. The paper proposes TurboReg, a correspondence-based point cloud registration estimator. Instead of searching for maximal cliques in a compatibility graph, TurboReg defines a TurboClique as a 3-clique in a highly constrained compatibility graph, and finds these cliques with a pivot-guided search (PGS) that uses the K1 highest-weighted SC2 edges as pivots and keeps the top-K2 third nodes per pivot. An ordered variant of the SC2 graph (O2Graph) is introduced to avoid duplicate TurboClique detections, and each candidate triple is scored by Kabsch-based inlier count. Experiments on 3DMatch, 3DLoMatch, and KITTI with FPFH, FCGF, and Predator descriptors report state-of-the-art or competitive registration recall with large reported speedups, and code is released.

Significance. If the empirical results hold, TurboReg is a practically valuable contribution: the fixed-size hypothesis structure is GPU-friendly, the benchmark coverage is broad, and the paper includes a detailed ablation study plus released code. The paper also correctly identifies a real limitation of maximal-clique enumeration and demonstrates a credible alternative. However, the theoretical stability argument that underpins the TurboClique definition is not valid as stated, and the headline speedup is based on an asymmetric GPU-vs-CPU comparison. The significance is therefore conditional on revising those claims.

major comments (3)
  1. [App. A.1, Eq. (11)] The argument that the constraint |∥d_ij+e_ij∥−∥d_ij∥| ≤ τ forces e_ij→0 as τ→0 is geometrically false. The set {e_ij : ∥d_ij+e_ij∥=∥d_ij∥} is a sphere of codimension 1 and contains vectors of norm up to 2∥d_ij∥, so a single pairwise compatibility constraint does not restrict the component of e_ij perpendicular to d_ij. Consequently the claimed reduction of the effective noise variance to zero, σ_eff→0, is unsupported, and the central premise that a stringent τ lets a 3-clique match the stability of a maximal clique is not established by the provided derivation. The supporting experiment in App. A.2 compares 3-cliques with 10-cliques only on 3DMatch+FPFH and does not vary noise level, scene geometry, or descriptor, so it cannot carry that weight. Please replace the limit argument with a rigorous analysis, or explicitly reframe the stability claim as a heuristic validated only empirically.
  2. [Table 1 and Sec. 4.1.2] The abstract's headline claim that TurboReg (1K) is 208.22× faster than 3DMAC compares TurboReg's GPU time (64.55 FPS) against 3DMAC's CPU time (0.31 FPS). The same table reports TurboReg's CPU implementation at 2.73 FPS, i.e. an 8.8× CPU-to-CPU speedup. Since 3DMAC has no GPU implementation, the GPU-vs-CPU comparison should be clearly labeled as such, and the abstract and contributions should not present a GPU-to-CPU ratio as the primary speed comparison without this caveat.
  3. [Sec. 4.1.1 and Sec. 4.3] The hyperparameters τ, K1, and K2 are tuned on the same benchmarks on which the final registration recall is reported. Section 4.1.1 states that τ is 'adjusted based on empirical evaluation' and the ablation summary recommends fine-tuning to 1.2 cm for 3DMatch, while K1 is set to 1K/2K for indoor datasets and 0.25K/0.5K for KITTI based on the per-dataset ablations in Table 5. This selection procedure biases the reported state-of-the-art recall, since the test set is used for parameter choice. Please add a validation split or a sensitivity analysis that shows the method's performance across a fixed parameter range without per-dataset tuning, and report the corresponding recall.
minor comments (7)
  1. [Sec. 3.4] The heading 'Implemetation Details of PGS' contains a typo; it should read 'Implementation Details of PGS'.
  2. [App. C] In the proof of the Unique Assignment Property, the phrase 'denoeted as TC(πz)' should be 'denoted as TC(πz)'.
  3. [Eq. (4)] The definition of the pivot set P is ambiguous when several edges tie at the K1-th largest weight: the text says 'all such edges are included until |P| = K1', but including all tied edges can make |P| exceed K1. Please specify the tie-breaking rule precisely.
  4. [Table 5] The ablation table skips rows 20–22, yet the text refers to 'rows 10 vs. 23'. Please renumber the rows and adjust the references so that every cited row exists.
  5. [App. A.2, Fig. 5] Figure 5 would be easier to read if the horizontal axis labels and units were shown explicitly for each panel, and if the caption stated how many point-cloud pairs contributed to the distributions.
  6. [Abstract and Fig. 1] The notation 'TurboReg (1K)' is used in the abstract without defining K; please define K1 and K2 at first use. Also, Fig. 1 labels the proposed method as 'ICCV2025'; if this is not the accepted venue, this label should be removed.
  7. [Sec. 3.4] The claim that the GPU implementation 'effectively achieves near-constant time complexity, approximating O(1)' assumes R ≫ K1N and ignores data-transfer and synchronization costs. The measured times in Table 7 show that PGS consumes about 11.4–11.9 ms and model estimation scales with K1; please soften the O(1) claim to match the actual implementation behavior.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity: the stability argument in App. A.1 is an invalid (not circular) limit inference; self-citations are minor and non-load-bearing; benchmarks are external.

full rationale

This paper's derivation chain is self-contained rather than circular. The central stability claim (Sec. 3.2, Definition 1) rests on the in-paper analysis of App. A, not on an imported fit or a self-citation chain: App. A.1 attempts to infer sigma_eff^2 -> 0 from Eq. (11), App. A.2 empirically compares 3-clique vs 10-clique estimates on 3DMatch+FPFH, and App. C proves the O2Graph Unique Assignment Property locally. The App. A.1 limit inference is geometrically invalid (the tau=0 constraint | ||d+e|| - ||d|| | = 0 defines a sphere, not e -> 0, so tangential noise components survive and sigma_eff^2 need not vanish); that is a correctness/validity gap in the heuristic motivation, not a circular reduction, because the conclusion is not contained in the premise by construction. Evaluation claims are checked against external benchmarks (3DMatch, 3DLoMatch, KITTI) and independent baselines (3DMAC, FastMAC, SC2-PCR, RANSAC, TEASER++, PointDSC, VBReg), and the speed advantage (PGS with O(K1*N) work and GPU parallelism) is a complexity argument independent of the fitted parameters. Self-citations exist ([27] QGORE, [47] k-clique listing, [49] QuadricsReg, [50] ML-SemReg) but are contextual or motivational: [47] motivates the O2Graph only where App. C supplies the proof. The mildest concern is that tau, K1, and K2 are tuned via ablation on the same benchmarks whose final RR is then reported (Tab. 5 row 4 gives the 93.59% that appears in Tab. 1, and Sec. 4.1.1 states tau is 'adjusted based on empirical evaluation'); this is benchmark-level hyperparameter selection rather than statistical forcing, since RR is stable over a wide tau range (82-84% over tau = 0.010-0.08 m on 3DMatch+FPFH) and the same settings generalize to KITTI and other descriptors. No prediction reduces to its inputs by construction, so the score reflects only the minor self-citations and the transparent same-benchmark tuning.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central empirical claim rests on a few hyperparameters (tau, K1, K2) and several domain assumptions about noise and the guidance value of SC2 scores. No physical entities are introduced.

free parameters (3)
  • tau (compatibility threshold) = 0.012 m for 3DMatch; 0.25x point cloud resolution elsewhere, fine-tuned
    Controls graph sparsity and the stability/recall trade-off; chosen by ablation (Tab. 5 rows 1-9).
  • K1 (pivot count) = 1000 or 2000 (indoor), 250 or 500 (KITTI)
    Number of top SC2 edges used as pivots; ablation shows convergence around K1=1000 (Tab. 5 rows 11-16).
  • K2 (TurboCliques per pivot) = 2
    Top-K2 triangles kept per pivot; ablation (Tab. 5 rows 17-19) picks 2.
assumptions (4)
  • domain assumption Match noise is independent, zero-mean, isotropic Gaussian
    App. A.1 assumes r_i ~ N(0, sigma^2 I) to derive the effective-variance argument; real feature matching noise is not isotropic Gaussian.
  • ad hoc to paper Spatial compatibility truncation reduces effective noise variance toward zero as tau tends to zero
    App. A.1 claims sigma^2_eff tends to zero when tau tends to zero; the argument treats the norm constraint as if it forced e_ij toward zero coordinate-wise, which is a heuristic, not a derivation.
  • domain assumption Kabsch estimation can be analyzed as linear least squares
    Sec. 3.2 uses Var(beta_hat|X)=sigma^2(X'X)^-1 to motivate stability; Kabsch is nonlinear in rotation, so this is an analogy.
  • domain assumption SC2 edge weight correlates with inlier likelihood and TurboClique density
    PGS uses SC2 scores to select pivots (Sec. 3.3); this property is borrowed from SC2-PCR [9] and assumed to hold under low overlap.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TurboReg: TurboClique for Robust and Efficient Point Cloud Registration." pith.science (2026). https://pith.science/paper/UKVGCR6Q

@misc{pith2026250701439,
  author       = {Pith},
  title        = {Pith review of: TurboReg: TurboClique for Robust and Efficient Point Cloud Registration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UKVGCR6Q}},
  note         = {Machine review of arXiv:2507.01439}
}
abstract

Robust estimation is essential in correspondence-based Point Cloud Registration (PCR). Existing methods using maximal clique search in compatibility graphs achieve high recall but suffer from exponential time complexity, limiting their use in time-sensitive applications. To address this challenge, we propose a fast and robust estimator, TurboReg, built upon a novel lightweight clique, TurboClique, and a highly parallelizable Pivot-Guided Search (PGS) algorithm. First, we define the TurboClique as a 3-clique within a highly-constrained compatibility graph. The lightweight nature of the 3-clique allows for efficient parallel searching, and the highly-constrained compatibility graph ensures robust spatial consistency for stable transformation estimation. Next, PGS selects matching pairs with high SC$^2$ scores as pivots, effectively guiding the search toward TurboCliques with higher inlier ratios. Moreover, the PGS algorithm has linear time complexity and is significantly more efficient than the maximal clique search with exponential time complexity. Extensive experiments show that TurboReg achieves state-of-the-art performance across multiple real-world datasets, with substantial speed improvements. For example, on the 3DMatch+FCGF dataset, TurboReg (1K) operates $208.22\times$ faster than 3DMAC while also achieving higher recall. Our code is accessible at \href{https://github.com/Laka-3DV/TurboReg}{\texttt{TurboReg}}.

Figures

Figures reproduced from arXiv: 2507.01439 by the authors.

Figure 1
Figure 1. Registration Recall and Speed Comparison on the 3DMatch+FCGF Dataset. Our method (⋆) achieves the highest recall and significantly outperforms competing methods in speed. cause it does not rely on initial transformation guesses [37, 39]. It typically consists of two main steps: (1) fea￾ture matching to establish putative 3D keypoint correspon￾dences [11, 39, 42], and (2) robust transformation estimation through inli… view at source ↗
Figure 2
Figure 2. Pipeline of TurboReg. TurboReg takes correspondences as input. First, these matches are used to construct an Ordered SC2 Graph (O2Graph, defined in Definition 2). Next, the PGS algorithm (Sec. 3.3) is applied to the O2Graph, producing TurboCliques (de￾fined in Definition 1). Finally, during the Model Estimation step (Sec. 3.5), a transformation is estimated for each TurboClique, and the highest-scoring transformatio… view at source ↗
Figure 3
Figure 3. Comparison of Different Types of Cliques. In the compatibility graph, each node represents a match, and the edges between nodes indicate spatial compatibility. Prior stud￾ies [38, 49, 61, 62] primarily estimate transformation using (a) maximal clique—a subset that maximizes mutually compatible matches to ensure transformation stability. However, the computa￾tional complexity of maximal clique search grows exponentia… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Time comparison (ms) between various robust estimators, [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Distribution of discrepancies between transformations [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Demonstration of matches when τ = 0. confirms this inference by demonstrating that transformation discrepancies between 3-cliques and 10-cliques decrease as τ diminishes. Specifically, we assess transformation dis￾crepancies between estimates derived from 3-cliques and…
Figure 7
Figure 7. Figure 7: Undirected graph for demonstration. distinct vertices is adjacent: \forall \boldsymbol {u}, \boldsymbol {v} \in \mathcal {C}, (\boldsymbol {u}, \boldsymbol {v}) \in E(\mathcal {G}), (15) where E(G) represents the edge set of G. For example, the vertices {1, 3, 5} are f…
Figure 8
Figure 8. Figure 8: Insufficient Consensus Correspondences. Red indicates lower IN values, while green denotes higher IN values. N: 4213 IN: 46 IN: 85 RE: 7.21° TE: 1.55 m N: 6017 IN: 148 IN: 175 RE: 11.65° TE: 0.46 m N: 3398 IN: 94 IN: 112 RE: 11.47° TE: 0.12 m Ours Ground Truth [PITH_F…
Figure 9
Figure 9. Figure 9: Larger Consensus Set with Small Errors. Red indicates lower IN values, while green denotes higher IN values. and Model Estimation modules exhibit a positive correlation between K1 and runtime, since an increase in K1 leads to a higher number of TurboCliques. In GPU imp…
Figure 10
Figure 10. Figure 10: Larger Consensus Set with Large Errors. Red indicates lower IN values, while green denotes higher IN values. correct. This contradicts the maximum consensus set assumption. We categorize this scenario into two sub￾categories: (a) Small Errors: The estimated rigid tran…
Figure 11
Figure 11. Figure 11: Qualitative Comparison on 3DMatch. Red and green represent failed and successful registrations, respectively. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Qualitative Comparison on 3DLoMatch. Red and green represent failed and successful registrations, respectively. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_12.png]
Figure 13
Figure 13. Figure 13: Qualitative comparison on KITTI. Red and green represent failed and successful registrations, respectively. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_13.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. TopoLiDM: Topology-Aware LiDAR Diffusion Models for Interpretable and Realistic LiDAR Point Cloud Generation

    cs.CV 2025-07 conditional novelty 6.0 of 10

    A graph-based latent diffusion model with persistent homology regularization improves LiDAR scene generation fidelity on KITTI-360.

Reference graph

Works this paper leans on

71 extracted references · 70 canonical work pages · cited by 1 Pith paper

  1. [1]

    4-points congruent sets for robust pairwise surface registration

    Dror Aiger, Niloy J Mitra, and Daniel Cohen-Or. 4-points congruent sets for robust pairwise surface registration. In ACM SIGGRAPH 2008 papers, pages 1–10. 2008. 2

    work page 2008

  2. [2]

    Parallelizing maximal clique enumeration on gpus

    Mohammad Almasri, Yen-Hsiang Chang, Izzat El Hajj, Rakesh Nagi, Jinjun Xiong, and Wen-mei Hwu. Parallelizing maximal clique enumeration on gpus. In 2023 32nd Interna- tional Conference on Parallel Architectures and Compilation Techniques (PACT), pages 162–175. IEEE, 2023. 2, 3

    work page 2023

  3. [3]

    Spinnet: Learning a general surface descriptor for 3d point cloud registration

    Sheng Ao, Qingyong Hu, Bo Yang, Andrew Markham, and Yulan Guo. Spinnet: Learning a general surface descriptor for 3d point cloud registration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 11753–11762, 2021. 2

    work page 2021

  4. [4]

    A survey of augmented reality

    Ronald T Azuma. A survey of augmented reality. Presence: teleoperators & virtual environments, 6(4):355–385, 1997. 1

    1997

  5. [5]

    Pointdsc: Robust point cloud registration using deep spatial consistency

    Xuyang Bai, Zixin Luo, Lei Zhou, Hongkai Chen, Lei Li, Zeyu Hu, Hongbo Fu, and Chiew-Lan Tai. Pointdsc: Robust point cloud registration using deep spatial consistency. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 15859–15869, 2021. 1, 3, 6, 7

    work page 2021

  6. [6]

    Simultaneous localiza- tion and mapping (slam): Part ii

    Tim Bailey and Hugh Durrant-Whyte. Simultaneous localiza- tion and mapping (slam): Part ii. IEEE robotics & automation magazine, 13(3):108–117, 2006. 1

    work page 2006

  7. [7]

    Graph-cut ransac

    Daniel Barath and Jiˇr´ı Matas. Graph-cut ransac. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 6733–6741, 2018. 6

    work page 2018

  8. [8]

    3d free-form object recognition in range images using local surface patches.Pattern Recognition Letters, 28(10):1252–1262, 2007

    Hui Chen and Bir Bhanu. 3d free-form object recognition in range images using local surface patches.Pattern Recognition Letters, 28(10):1252–1262, 2007. 2

    work page 2007

Show all 71 references
  1. [9]

    Sc2-pcr: A second order spatial compatibility for efficient and robust point cloud registration

    Zhi Chen, Kun Sun, Fan Yang, and Wenbing Tao. Sc2-pcr: A second order spatial compatibility for efficient and robust point cloud registration. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR), pages 13221–13231, 2022. 2, 3, 4, 5, 6

    2022

  2. [10]

    Sc2-pcr: A second order spatial compatibility for efficient and robust point cloud registration

    Zhi Chen, Kun Sun, Fan Yang, and Wenbing Tao. Sc2-pcr: A second order spatial compatibility for efficient and robust point cloud registration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 13221–13231, 2022. 6, 7

    2022

  3. [11]

    Fully convolutional geometric features

    Christopher Choy, Jaesik Park, and Vladlen Koltun. Fully convolutional geometric features. In Proceedings of the IEEE International Conference on Computer Vision, pages 8958– 8966, 2019. 1, 3, 5, 6

    2019

  4. [12]

    Deep global registration

    Christopher Choy, Wei Dong, and Vladlen Koltun. Deep global registration. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 2514–2523, 2020. 2, 6, 7

    2020

  5. [13]

    Ppfnet: Global context aware local features for robust 3d point match- ing

    Haowen Deng, Tolga Birdal, and Slobodan Ilic. Ppfnet: Global context aware local features for robust 3d point match- ing. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 195–205, 2018. 2

    2018

  6. [14]

    Accelerat- ing maximal clique enumeration via graph reduction

    Weiguo Zheng Deng, Wen and Hong Cheng. Accelerat- ing maximal clique enumeration via graph reduction. arXiv preprint arXiv:2311.00279, 2013. 2, 3

    2013 arXiv

  7. [15]

    Simultaneous local- ization and mapping: part i

    Hugh Durrant-Whyte and Tim Bailey. Simultaneous local- ization and mapping: part i. IEEE robotics & automation magazine, 13(2):99–110, 2006. 1

    2006

  8. [16]

    The maximum clique enumeration problem: algorithms, applications, and implementations

    John D Eblen, Charles A Phillips, Gary L Rogers, and Michael A Langston. The maximum clique enumeration problem: algorithms, applications, and implementations. In BMC bioinformatics, pages 1–11. Springer, 2012. 2

    2012

  9. [17]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography

    Martin A Fischler and Robert C Bolles. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography. Communications of the ACM, 24(6):381–395, 1981. 1, 2, 6, 7

    1981

  10. [18]

    Recognizing objects in range data us- ing regional point descriptors

    Andrea Frome, Daniel Huber, Ravi Kolluri, Thomas B¨ulow, and Jitendra Malik. Recognizing objects in range data us- ing regional point descriptors. In European conference on computer vision, pages 224–237. Springer, 2004. 2

    2004

  11. [19]

    Are we ready for autonomous driving? the kitti vision benchmark suite

    Andreas Geiger, Philip Lenz, and Raquel Urtasun. Are we ready for autonomous driving? the kitti vision benchmark suite. In Conference on Computer Vision and Pattern Recog- nition (CVPR), 2012. 6

    2012

  12. [20]

    Rops: A local feature descriptor for 3d rigid objects based on rotational projection statistics

    Yulan Guo, Ferdous A Sohel, Mohammed Bennamoun, Jian- wei Wan, and Min Lu. Rops: A local feature descriptor for 3d rigid objects based on rotational projection statistics. In 2013 1st International Conference on Communications, Sig- nal Processing, and their Applications (ICCS...

    2013

  13. [21]

    A novel local surface feature for 3d object recognition under clutter and occlusion

    Yulan Guo, Ferdous Sohel, Mohammed Bennamoun, Jianwei Wan, and Min Lu. A novel local surface feature for 3d object recognition under clutter and occlusion. Information Sciences, 293:196–213, 2015. 2

    2015

  14. [22]

    Predator: Registration of 3d point clouds with low overlap

    Shengyu Huang, Zan Gojcic, Mikhail Usvyatsov, Andreas Wieser, and Konrad Schindler. Predator: Registration of 3d point clouds with low overlap. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4267–4276, 2021. 2, 5, 6

    2021

  15. [23]

    Efficient and robust point cloud registration via heuristics-guided parameter search

    Tianyu Huang, Haoang Li, Liangzu Peng, Yinlong Liu, and Yun-Hui Liu. Efficient and robust point cloud registration via heuristics-guided parameter search. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 2

    2024

  16. [24]

    Scalable 3d registration via truncated entry-wise absolute residuals

    Tianyu Huang, Liangzu Peng, Ren´e Vidal, and Yun-Hui Liu. Scalable 3d registration via truncated entry-wise absolute residuals. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 27477– 27487, 2024. 2

    2024

  17. [25]

    Robust outlier rejection for 3d registration with variational bayes

    Haobo Jiang, Zheng Dang, Zhen Wei, Jin Xie, Jian Yang, and Mathieu Salzmann. Robust outlier rejection for 3d registration with variational bayes. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1148–1157, 2023. 1, 3, 5, 6

    2023

  18. [26]

    Wolfgang Kabsch. A solution for the best rotation to relate two sets of vectors.Acta Crystallographica Section A: Crystal Physics, Diffraction, Theoretical and General Crystallogra- phy, 32(5):922–923, 1976. 3, 5

    1976

  19. [27]

    Qgore: Quadratic-time guaranteed outlier removal for point 9 cloud registration

    Jiayuan Li, Pengcheng Shi, Qingwu Hu, and Yongjun Zhang. Qgore: Quadratic-time guaranteed outlier removal for point 9 cloud registration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(9):11136–11151, 2023. 1, 2

    2023

  20. [28]

    Sarnet: Semantic aug- mented registration of large-scale urban point clouds

    Chao Liu, Jianwei Guo, Dong-Ming Yan, Zhirong Liang, Xiaopeng Zhang, and Zhanglin Cheng. Sarnet: Semantic aug- mented registration of large-scale urban point clouds. arXiv preprint arXiv:2206.13117, 2022. 2

    2022 arXiv

  21. [29]

    Translo: A window-based masked point transformer framework for large-scale lidar odometry

    Jiuming Liu, Guangming Wang, Chaokang Jiang, Zhe Liu, and Hesheng Wang. Translo: A window-based masked point transformer framework for large-scale lidar odometry. In Proceedings of the AAAI Conference on Artificial Intelligence, pages 1683–1691, 2023. 1

    2023

  22. [30]

    Regformer: An efficient projection-aware transformer network for large-scale point cloud registration

    Jiuming Liu, Guangming Wang, Zhe Liu, Chaokang Jiang, Marc Pollefeys, and Hesheng Wang. Regformer: An efficient projection-aware transformer network for large-scale point cloud registration. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 8451–8460,

  23. [31]

    Difflow3d: Toward robust uncertainty-aware scene flow estimation with iterative diffusion-based refinement

    Jiuming Liu, Guangming Wang, Weicai Ye, Chaokang Jiang, Jinru Han, Zhe Liu, Guofeng Zhang, Dalong Du, and Hesh- eng Wang. Difflow3d: Toward robust uncertainty-aware scene flow estimation with iterative diffusion-based refinement. In Proceedings of the IEEE/CVF Conference on Co...

    2024

  24. [32]

    Mamba4d: Ef- ficient 4d point cloud video understanding with disentangled spatial-temporal state space models

    Jiuming Liu, Jinru Han, Lihao Liu, Angelica I Aviles-Rivero, Chaokang Jiang, Zhe Liu, and Hesheng Wang. Mamba4d: Ef- ficient 4d point cloud video understanding with disentangled spatial-temporal state space models. In Proceedings of the Computer Vision and Pattern Recognition ...

    2025

  25. [33]

    Dvlo: Deep visual- lidar odometry with local-to-global feature fusion and bi- directional structure alignment

    Jiuming Liu, Dong Zhuo, Zhiheng Feng, Siting Zhu, Chen- sheng Peng, Zhe Liu, and Hesheng Wang. Dvlo: Deep visual- lidar odometry with local-to-global feature fusion and bi- directional structure alignment. In European Conference on Computer Vision, pages 475–493. Springer, 2025. 1

    2025

  26. [34]

    Clipper: A graph-theoretic framework for robust data association

    Parker C Lusk, Kaveh Fathian, and Jonathan P How. Clipper: A graph-theoretic framework for robust data association. In 2021 IEEE International Conference on Robotics and Au- tomation (ICRA), pages 13828–13834. IEEE, 2021. 2

    2021

  27. [35]

    3dregnet: A deep neural network for 3d point registration

    G Dias Pais, Srikumar Ramalingam, Venu Madhav Govindu, Jacinto C Nascimento, Rama Chellappa, and Pedro Miraldo. 3dregnet: A deep neural network for 3d point registration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7193–7203, 2020. 2

    2020

  28. [36]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high-performance deep learning library. Ad- vances in neural information processing systems, 3...

    2019

  29. [37]

    Learning general and dis- tinctive 3d local deep descriptors for point cloud registration

    Fabio Poiesi and Davide Boscaini. Learning general and dis- tinctive 3d local deep descriptors for point cloud registration. In IEEE Trans. on Pattern Analysis and Machine Intelligence, (early access) 2022. 1, 3

    2022

  30. [38]

    G3reg: Pyramid graph-based global reg- istration using gaussian ellipsoid model

    Zhijian Qiao, Zehuan Yu, Binqian Jiang, Huan Yin, and Shaojie Shen. G3reg: Pyramid graph-based global reg- istration using gaussian ellipsoid model. arXiv preprint arXiv:2308.11573, 2023. 3, 4

    2023 arXiv

  31. [39]

    Geometric transformer for fast and robust point cloud registration

    Zheng Qin, Hao Yu, Changjian Wang, Yulan Guo, Yuxing Peng, and Kai Xu. Geometric transformer for fast and robust point cloud registration. arXiv preprint arXiv:2202.06688,

    arXiv

  32. [40]

    Compatibility-guided sampling consensus for 3-d point cloud registration

    Siwen Quan and Jiaqi Yang. Compatibility-guided sampling consensus for 3-d point cloud registration. IEEE Transactions on Geoscience and Remote Sensing, 58(10):7380–7392, 2020. 6, 7

    2020

  33. [41]

    Aligning point cloud views using persistent feature histograms

    Radu Bogdan Rusu, Nico Blodow, Zoltan Csaba Marton, and Michael Beetz. Aligning point cloud views using persistent feature histograms. In 2008 IEEE/RSJ international con- ference on intelligent robots and systems, pages 3384–3391. IEEE, 2008. 2

    2008

  34. [42]

    Fast point feature histograms (fpfh) for 3d registration

    Radu Bogdan Rusu, Nico Blodow, and Michael Beetz. Fast point feature histograms (fpfh) for 3d registration. In 2009 IEEE international conference on robotics and automation, pages 3212–3217. IEEE, 2009. 1, 2, 3, 5, 6

    2009

  35. [43]

    Trivoc: Efficient voting-based con- sensus maximization for robust point cloud registration with extreme outlier ratios

    Lei Sun and Lu Deng. Trivoc: Efficient voting-based con- sensus maximization for robust point cloud registration with extreme outlier ratios. IEEE Robotics and Automation Letters, 7(2):4654–4661, 2022. 2

    2022

  36. [44]

    Acne: Attentive context normalization for robust permutation-equivariant learning

    Weiwei Sun, Wei Jiang, Eduard Trulls, Andrea Tagliasacchi, and Kwang Moo Yi. Acne: Attentive context normalization for robust permutation-equivariant learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11286–11295, 2020. 2

    2020

  37. [45]

    Unique shape context for 3d data description

    Federico Tombari, Samuele Salti, and Luigi Di Stefano. Unique shape context for 3d data description. In Proceedings of the ACM workshop on 3D object retrieval, pages 57–62,

  38. [46]

    You only hypothesize once: Point cloud registration with rotation-equivariant descriptors

    Haiping Wang, Yuan Liu, Zhen Dong, and Wenping Wang. You only hypothesize once: Point cloud registration with rotation-equivariant descriptors. In Proceedings of the 30th ACM International Conference on Multimedia, pages 1630– 1641, 2022. 2

    2022

  39. [47]

    Efficient k- clique listing: An edge-oriented branching strategy

    Kaixin Wang, Kaiqiang Yu, and Cheng Long. Efficient k- clique listing: An edge-oriented branching strategy. Proceed- ings of the ACM on Management of Data, 2(1):1–26, 2024. 4

    2024

  40. [48]

    Generalized differentiable ransac

    Tong Wei, Yash Patel, Alexander Shekhovtsov, Jiri Matas, and Daniel Barath. Generalized differentiable ransac. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 17649–17660, 2023. 2

    2023

  41. [49]

    Quadricsreg: Large-scale point cloud registration using quadric primitives

    Ji Wu, Huai Yu, Shu Han, Xi-Meng Cai, Ming-Feng Wang, Wen Yang, and Gui-Song Xia. Quadricsreg: Large-scale point cloud registration using quadric primitives. arXiv preprint arXiv:2412.02998, 2024. 3, 4

    2024 arXiv

  42. [50]

    Ml-semreg: Boosting point cloud registration with multi-level semantic consistency

    Shaocheng Yan, Pengcheng Shi, and Jiayuan Li. Ml-semreg: Boosting point cloud registration with multi-level semantic consistency. In European Conference on Computer Vision, pages 19–37. Springer, 2025. 2

    2025

  43. [51]

    Teaser: Fast and certifiable point cloud registration

    Heng Yang, Jingnan Shi, and Luca Carlone. Teaser: Fast and certifiable point cloud registration. IEEE Transactions on Robotics, 37(2):314–333, 2020. 1, 2, 3, 6, 7

    2020

  44. [52]

    Mutual voting for ranking 3d correspondences

    Jiaqi Yang, Xiyu Zhang, Shichao Fan, Chunlin Ren, and Yan- ning Zhang. Mutual voting for ranking 3d correspondences. IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 2023. 2 10

    2023

  45. [53]

    Learning to find good correspondences

    Kwang Moo Yi, Eduard Trulls, Yuki Ono, Vincent Lepetit, Mathieu Salzmann, and Pascal Fua. Learning to find good correspondences. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2666–2674,

  46. [54]

    Segregator: Global point cloud reg- istration with semantic and geometric cues

    Pengyu Yin, Shenghai Yuan, Haozhi Cao, Xingyu Ji, Shuyang Zhang, and Lihua Xie. Segregator: Global point cloud reg- istration with semantic and geometric cues. arXiv preprint arXiv:2301.07425, 2023. 2

    2023 arXiv

  47. [55]

    Cofinet: Reliable coarse-to-fine correspondences for ro- bust pointcloud registration

    Hao Yu, Fu Li, Mahdi Saleh, Benjamin Busam, and Slobodan Ilic. Cofinet: Reliable coarse-to-fine correspondences for ro- bust pointcloud registration. Advances in Neural Information Processing Systems, 34, 2021. 2

    2021

  48. [56]

    Rotation-invariant transformer for point cloud matching

    Hao Yu, Zheng Qin, Ji Hou, Mahdi Saleh, Dongsheng Li, Benjamin Busam, and Slobodan Ilic. Rotation-invariant transformer for point cloud matching. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5384–5393, 2023

    2023

  49. [57]

    Inlier confidence calibration for point cloud registration

    Yongzhe Yuan, Yue Wu, Xiaolong Fan, Maoguo Gong, Qiguang Miao, and Wenping Ma. Inlier confidence calibration for point cloud registration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 5312–5321, 2024. 2

    2024

  50. [58]

    Surface feature detection and description with ap- plications to mesh matching

    Andrei Zaharescu, Edmond Boyer, Kiran Varanasi, and Radu Horaud. Surface feature detection and description with ap- plications to mesh matching. In 2009 IEEE Conference on Computer Vision and Pattern Recognition, pages 373–380. IEEE, 2009. 2

    2009

  51. [59]

    3dmatch: Learning local geometric descriptors from rgb-d reconstructions

    Andy Zeng, Shuran Song, Matthias Nießner, Matthew Fisher, Jianxiong Xiao, and Thomas Funkhouser. 3dmatch: Learning local geometric descriptors from rgb-d reconstructions. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1802–1811, 2017. 2, 5

    2017

  52. [60]

    Learning two-view correspondences and geometry using order-aware network

    Jiahui Zhang, Dawei Sun, Zixin Luo, Anbang Yao, Lei Zhou, Tianwei Shen, Yurong Chen, Long Quan, and Hongen Liao. Learning two-view correspondences and geometry using order-aware network. arXiv preprint arXiv:1908.04964, 2019. 2

    1908 arXiv

  53. [61]

    3d registration with maximal cliques

    Xiyu Zhang, Jiaqi Yang, Shikun Zhang, and Yanning Zhang. 3d registration with maximal cliques. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17745–17754, 2023. 1, 2, 3, 4, 5, 6, 7, 15

    2023

  54. [62]

    Fast- mac: Stochastic spectral sampling of correspondence graph

    Yifei Zhang, Hao Zhao, Hongyang Li, and Siheng Chen. Fast- mac: Stochastic spectral sampling of correspondence graph. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17857–17867, 2024. 2, 3, 4, 6, 7

    2024

  55. [63]

    Nm-net: Mining reliable neighbors for robust feature corre- spondences

    Chen Zhao, Zhiguo Cao, Chi Li, Xin Li, and Jiaqi Yang. Nm-net: Mining reliable neighbors for robust feature corre- spondences. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 215–224,

  56. [65]

    Set τ ∈ {1 cm, 10 cm, 50 cm, 100 cm} to construct com- patibility graph G

  57. [66]

    Extract all 10-cliques fromG and compute multiple trans- formations T(10) = (R(10), t(10))

  58. [67]

    For each 10-clique, estimate transformations {T(3) k }K k=1 = {(R(3) k , t(3) k )} from all K = 10 3 3-clique subsets

  59. [68]

    Calculate rotation and translation errors: ∆Rk = arccos tr(R(10)⊤R(3) k ) − 1 2 ! , (12) ∆tk = ∥t(10) − t(3) k ∥2; (13)

  60. [69]

    Visualize error distributions across τ values in Fig. 5. Results show negligible discrepancies at τ = 1 cm (< 0.1◦ rotation, < 0.5 mm translation), with errors ris- ing proportionally to τ. This confirms that 3-cliques achieve accuracy comparable to larger cliques under tight ...

    1930

  61. [70]

    This occurs in scenarios with extremely low overlap or strong symmetry, as illustrated in Fig

    Insufficient Consensus Correspondences : TurboReg fails to identify a sufficiently large set of consensus corre- spondences. This occurs in scenarios with extremely low overlap or strong symmetry, as illustrated in Fig. 8

  62. [71]

    Larger Consensus Set with Large Errors

    Larger Consensus Set but Incorrect Transformation: The algorithm identifies a larger IN than that of the ground truth transformation, yet the result remains in- 16 N: 5131 IN: 25 IN: 82 RE: 179.68° TE: 3.29 m N: 3631 IN: 11 IN: 52 RE: 38.08° TE: 2.42 m N: 5388 IN: 27 IN: 80 RE...

  63. [2019]

    Specifically: (1) App

    2 11 TurboReg: TurboClique for Robust and Efficient Point Cloud Registration Supplementary Material In this supplementary material, we first provide additional analyses to elaborate on concepts introduced in the main paper. Specifically: (1) App. A details why TurboClique empl...

Pith tools

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