Pith. sign in

REVIEW 2 major objections 2 minor 41 references

Spherical voxel indexing derives a conservative bound that safely prunes redundant distance checks in farthest point sampling without altering the output.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-28 00:59 UTC pith:JQVKNPU7

load-bearing objection RadiusFPS claims exact FPS equivalence via spherical voxel pruning and a fused GPU kernel, delivering reported 2.5x speedups and lower memory on standard benchmarks, but the bound's conservativeness needs explicit checking. the 2 major comments →

arxiv 2606.06255 v1 pith:JQVKNPU7 submitted 2026-06-04 cs.RO cs.CVcs.DC

RadiusFPS: Efficient Farthest Point Sampling on CPUs and GPUs via Spherical Voxel Pruning

Ziyang Yu , Xiang Li , Qiong Chang , Jun Miyazaki This is my paper
classification cs.RO cs.CVcs.DC
keywords farthest point samplingpoint cloud downsamplingspherical voxelsGPU accelerationLiDAR processingrobotic perceptionreal-time sampling
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper sets out to show that standard farthest point sampling can be made substantially faster on both CPUs and GPUs by indexing the input cloud in spherical voxels and using the resulting geometric bound to skip distance calculations that cannot affect the next selected point. A coordinate-wise skip test removes further redundant updates. The GPU kernel fuses voxel selection, pruning, and updates into coalesced operations. On indoor and outdoor LiDAR benchmarks the resulting sampler delivers the claimed speedups while producing identical sample sets to classical FPS and comparable downstream segmentation accuracy.

Core claim

RadiusFPS indexes the point cloud with spherical voxels to obtain a conservative geometric bound that prunes redundant distance computations in each FPS iteration while preserving the exact update rule, initialization, and tie-breaking policy of standard FPS; a coordinate-wise point-skip test eliminates residual updates; the GPU variant fuses these steps into memory-coalesced kernels.

What carries the argument

Spherical voxel pruning: indexing the point cloud with spherical voxels to derive a conservative geometric bound that safely eliminates distance computations guaranteed not to contain the next farthest point.

Load-bearing premise

The spherical voxel bound is guaranteed to be conservative, so that no point that would have been chosen by the standard FPS rule is ever pruned.

What would settle it

Run both classical FPS and RadiusFPS on the same point cloud with identical initialization and tie-breaking; if any sampled point differs, the pruning bound is not conservative.

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

If this is right

  • RadiusFPS-G reaches up to 2.5 times the speed of GPU-based FPS on the evaluated indoor and outdoor datasets.
  • It matches or exceeds QuickFPS performance while using roughly half the GPU memory.
  • Segmentation accuracy remains comparable to standard FPS on S3DIS, ScanNet, and SemanticKITTI.
  • Pairing RadiusFPS-G with a learning-based sampler yields the fastest end-to-end inference among tested configurations.

Where Pith is reading between the lines

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

  • The same voxel-bound technique could be applied to other iterative farthest-point or diversity-sampling loops that rely on repeated distance queries.
  • Because the method preserves exact FPS output, it can serve as a drop-in replacement inside any pipeline that assumes the classical sampling distribution.
  • The coordinate-wise skip test may generalize to other coordinate-aligned data structures beyond spherical voxels.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 2 minor

Summary. The paper proposes RadiusFPS, a framework for accelerating classical Farthest Point Sampling (FPS) on point clouds via spherical voxel pruning that is claimed to preserve the exact FPS update rule and tie-breaking policy. It introduces CPU and GPU (RadiusFPS-G) implementations with fused kernels, reporting up to 2.5x speedup over GPU FPS, roughly half the memory of QuickFPS, and comparable segmentation accuracy on S3DIS, ScanNet, and SemanticKITTI when used in downstream pipelines, including with learning-based samplers.

Significance. If the claimed exact equivalence holds, the work would provide a practical, low-memory acceleration for a core operator in robotic perception pipelines handling high-rate LiDAR data. The explicit comparisons to QuickFPS on public benchmarks, memory measurements, and end-to-end timing with FastPoint are concrete strengths that would make the result useful for latency-constrained systems.

major comments (2)
  1. [Abstract] Abstract (pruning derivation paragraph): the central claim that spherical voxel indexing produces a conservative geometric bound safely pruning distance computations without discarding the true farthest point (under standard FPS update and tie-breaking) rests on an unshown geometric derivation. No explicit proof, exhaustive check, or handling of floating-point/degenerate cases is provided, making equivalence to classical FPS unverifiable and rendering the reported accuracy parity non-guaranteed.
  2. [§4] §4 (evaluation on S3DIS/ScanNet/SemanticKITTI): the speedup, memory, and accuracy claims are presented without error bars, multiple random seeds, or statistical tests. Given that any deviation in selected points from the unverified pruning stages would affect downstream segmentation, the absence of variability analysis undermines the reliability of the 'matches or exceeds QuickFPS' and 'comparable accuracy' statements.
minor comments (2)
  1. [§3] The interaction between the spherical voxel pruning stage and the coordinate-wise point-skip test is described at a high level; a small example or pseudocode showing their combined effect on a single iteration would improve clarity without altering the central claim.
  2. Table captions and axis labels in the runtime/memory plots should explicitly state the number of points sampled (e.g., 1024 or 4096) and whether timings include voxel construction overhead.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and will revise the manuscript to incorporate the suggested improvements.

read point-by-point responses

  1. Referee: [Abstract] Abstract (pruning derivation paragraph): the central claim that spherical voxel indexing produces a conservative geometric bound safely pruning distance computations without discarding the true farthest point (under standard FPS update and tie-breaking) rests on an unshown geometric derivation. No explicit proof, exhaustive check, or handling of floating-point/degenerate cases is provided, making equivalence to classical FPS unverifiable and rendering the reported accuracy parity non-guaranteed.

    Authors: We agree that an explicit formal proof is needed for verifiability. Section 3 derives the conservative spherical voxel bound, but the revised manuscript will expand this with a complete geometric proof, including analysis of floating-point precision and degenerate cases (e.g., points on voxel boundaries or ties), to rigorously establish preservation of the exact FPS update rule and tie-breaking policy. revision: yes

  2. Referee: [§4] §4 (evaluation on S3DIS/ScanNet/SemanticKITTI): the speedup, memory, and accuracy claims are presented without error bars, multiple random seeds, or statistical tests. Given that any deviation in selected points from the unverified pruning stages would affect downstream segmentation, the absence of variability analysis undermines the reliability of the 'matches or exceeds QuickFPS' and 'comparable accuracy' statements.

    Authors: We acknowledge the lack of error bars, multi-seed analysis, and statistical tests in the current evaluation. The revised manuscript will report accuracy metrics averaged over multiple random seeds (minimum 5) with standard deviations for all datasets and include statistical tests to support the comparable accuracy claims. revision: yes

Circularity Check

0 steps flagged

No circularity: algorithmic performance claims rest on external benchmarks and geometric pruning argument

full rationale

The paper describes a pruning-based acceleration of classical FPS that claims to preserve the original update rule via conservative spherical-voxel bounds and coordinate-wise skips. No equations reduce a reported quantity to a fitted parameter or to the method's own output by construction. No self-citations are invoked as load-bearing uniqueness theorems. Speedup, memory, and accuracy figures are measured on public datasets (S3DIS, ScanNet, SemanticKITTI) rather than being derived from internal definitions. The equivalence claim is an algorithmic assertion whose verification status is a correctness question, not a circularity reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The method is purely algorithmic and relies on standard geometric properties of point clouds and spherical coordinates; no free parameters are introduced, no new entities are postulated, and the only background assumptions are the correctness of distance metrics and voxel partitioning.

axioms (2)
  • standard math Euclidean distance is the metric used for FPS selection
    Invoked in the description of the standard FPS update rule that RadiusFPS preserves.
  • domain assumption Spherical voxel partitioning induces a conservative bound on maximum possible distance
    Central to the pruning step; location is the paragraph describing derivation of the geometric bound.

pith-pipeline@v0.9.1-grok · 5840 in / 1438 out tokens · 24729 ms · 2026-06-28T00:59:13.916722+00:00 · methodology

0 comments
read the original abstract

Point clouds are a primary sensory representation for robotic perception, underpinning LiDAR-based autonomous driving, simultaneous localization and mapping (SLAM), and navigation. Within these pipelines, Farthest Point Sampling (FPS) is the most well-known downsampling operator, as its uniform coverage preserves the geometric structure on which downstream perception relies. However, the large time complexity of classical FPS scales poorly with the million-point-per-second rates of modern 3D sensors, making it a dominant latency bottleneck that conflicts with the real-time and limited onboard compute budgets of robotic systems. Therefore, we propose RadiusFPS, an FPS acceleration framework based on spherical voxel pruning that preserves the standard FPS update rule under the same initialization and tie-breaking policy. By indexing the point cloud with spherical voxels, RadiusFPS derives a conservative geometric bound that prunes redundant distance computations in each iteration, complemented by a coordinate-wise point-skip test that removes residual updates. We further introduce RadiusFPS-G, a warp-level GPU implementation that fuses voxel selection, pruning, and distance update into memory-coalesced kernels, eliminating costly global-memory round-trips. On indoor (S3DIS, ScanNet) and outdoor LiDAR (SemanticKITTI) benchmarks, RadiusFPS-G attains up to 2.5x speedup over GPU-based FPS and matches or exceeds QuickFPS among the evaluated methods while using roughly half its GPU memory, with comparable segmentation accuracy. When coupled with the learning-based FastPoint sampler, the resulting pipeline achieves the fastest End-to-End inference among all evaluated configurations. These properties make high-quality FPS-style sampling practical for latency- and memory-constrained robotic vision.

Figures

Figures reproduced from arXiv: 2606.06255 by Jun Miyazaki, Qiong Chang, Xiang Li, Ziyang Yu.

Figure 1
Figure 1. Figure 1: Efficiency analysis of FPS on the S3DIS and ScanNet datasets. Top: point cloud size distribution and sam [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Sampling uniformity of FPS versus random sampling on the Stanford dragon point cloud. (a)-(b) sampling [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Percentage of ineffective distance updates at each sampling step of standard FPS on the Stanford dragon [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: An Overview of the RadiusFPS Algorithm. The algorithm mainly consists of three stages: (a) Point Cloud [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Standard Voxel Design (Left) and Spherical Voxel Design (Right). The main difference between the two [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Illustration of the standard voxel pruning versus our proposed spherical voxel pruning. The top row defines [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Architectural workflow of our proposed GPU-based RadiusFPS (RadiusFPS-G). Following data transfer to [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Overview of Fusion Kernel 1. The kernel fuses two sequential parallel reductions into a single operation. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Illustration of the two-stage warp-level reduction within a single CUDA block. To minimize synchronization [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Execution pipeline of Fusion Kernel 2. This kernel efficiently integrates radius-based voxel pruning with [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Comprehensive runtime and speedup analysis across S3DIS, ScanNet, and SemanticKITTI datasets. The [PITH_FULL_IMAGE:figures/full_fig_p021_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Visualization display of point cloud downsampling using our proposed RadiusFPS and RadiusFPS-G. We [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Close-up qualitative comparison of local geometric details. Zoomed-in views of the Stanford Dragon, [PITH_FULL_IMAGE:figures/full_fig_p023_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Performance scalability analysis across varying sample sizes. The line charts compare the Latency (ms), [PITH_FULL_IMAGE:figures/full_fig_p024_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Impact of varying voxel resolutions (v) on sampling speedup. The figures evaluate the performance scaling of (a) our GPU-accelerated RFPS-G against standard FPS and QuickFPS (QFPS), and (b) our CPU-based RFPS against FPS and FPS+NPDU, across three datasets. The results demonstrate that appropriately tuning the voxel resolution v maximizes the performance gains, achieving peak speedups of up to 52.4× on GP… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

41 extracted references · 1 canonical work pages

  1. [1]

    Comparative evaluation of lidar systems for transport infrastructure: case studies and perfor- mance analysis.European Journal of Remote Sensing, 57(1):2316304, 2024

    Rabia Rashdi, Iv ´an Garrido, Jes ´us Balado, Pablo Del R ´ıo-Barral, Juan Luis Rodr´ıguez-Somoza, and Joaqu´ın Mart´ınez-S´anchez. Comparative evaluation of lidar systems for transport infrastructure: case studies and perfor- mance analysis.European Journal of Remote Sensing, 57(1):2316304, 2024

    2024

  2. [2]

    Meta architecture for point cloud analysis

    Haojia Lin, Xiawu Zheng, Lijiang Li, Fei Chao, Shanshan Wang, Yan Wang, Yonghong Tian, and Rongrong Ji. Meta architecture for point cloud analysis. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17682–17691, 2023

    2023

  3. [3]

    Pointvector: A vector representation in point cloud analysis

    Xin Deng, WenYu Zhang, Qing Ding, and XinMing Zhang. Pointvector: A vector representation in point cloud analysis. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9455– 9465, 2023

    2023

  4. [4]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space.Advances in neural information processing systems, 30, 2017

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space.Advances in neural information processing systems, 30, 2017

    2017

  5. [5]

    Point transformer v3: Simpler, faster, stronger

    Xiaoyang Wu, Li Jiang, Peng-Shuai Wang, Zhijian Liu, Xihui Liu, Yu Qiao, Wanli Ouyang, Tong He, and Hengshuang Zhao. Point transformer v3: Simpler, faster, stronger. InCVPR, 2024

    2024

  6. [6]

    Towards large-scale 3d representation learning with multi-dataset point prompt training

    Xiaoyang Wu, Zhuotao Tian, Xin Wen, Bohao Peng, Xihui Liu, Kaicheng Yu, and Hengshuang Zhao. Towards large-scale 3d representation learning with multi-dataset point prompt training. InCVPR, 2024

    2024

  7. [7]

    Point transformer v2: Grouped vector attention and partition-based pooling

    Xiaoyang Wu, Yixing Lao, Li Jiang, Xihui Liu, and Hengshuang Zhao. Point transformer v2: Grouped vector attention and partition-based pooling. InNeurIPS, 2022

    2022

  8. [8]

    Pointcept: A codebase for point cloud perception research.https://github.com/ Pointcept/Pointcept, 2023

    Pointcept Contributors. Pointcept: A codebase for point cloud perception research.https://github.com/ Pointcept/Pointcept, 2023

    2023

  9. [9]

    Randla-net: Efficient semantic segmentation of large-scale point clouds.Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2020

    Qingyong Hu, Bo Yang, Linhai Xie, Stefano Rosa, Yulan Guo, Zhihua Wang, Niki Trigoni, and Andrew Markham. Randla-net: Efficient semantic segmentation of large-scale point clouds.Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2020

    2020

  10. [10]

    Pointrcnn: 3d object proposal generation and detection from point cloud

    Shaoshuai Shi, Xiaogang Wang, and Hongsheng Li. Pointrcnn: 3d object proposal generation and detection from point cloud. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 770–779, 2019

    2019

  11. [11]

    Not all points are equal: Learning highly efficient point-based detectors for 3d lidar point clouds

    Yifan Zhang, Qingyong Hu, Guoquan Xu, Yanxin Ma, Jianwei Wan, and Yulan Guo. Not all points are equal: Learning highly efficient point-based detectors for 3d lidar point clouds. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2022

    2022

  12. [12]

    3dssd: Point-based 3d single stage object detector

    Zetong Yang, Yanan Sun, Shu Liu, and Jiaya Jia. 3dssd: Point-based 3d single stage object detector. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11040–11048, 2020

    2020

  13. [13]

    STD: sparse-to-dense 3d object detector for point cloud

    Zetong Yang, Yanan Sun, Shu Liu, Xiaoyong Shen, and Jiaya Jia. STD: sparse-to-dense 3d object detector for point cloud. In2019 IEEE/CVF International Conference on Computer Vision, ICCV 2019, Seoul, Korea (South), October 27 - November 2, 2019, 2019

    2019

  14. [14]

    Point- mamba: A simple state space model for point cloud analysis

    Dingkang Liang, Xin Zhou, Wei Xu, Xingkui Zhu, Zhikang Zou, Xiaoqing Ye, Xiao Tan, and Xiang Bai. Point- mamba: A simple state space model for point cloud analysis. InAdvances in Neural Information Processing Systems, 2024

    2024

  15. [15]

    Point-bert: Pre-training 3d point cloud transformers with masked point modeling

    Xumin Yu, Lulu Tang, Yongming Rao, Tiejun Huang, Jie Zhou, and Jiwen Lu. Point-bert: Pre-training 3d point cloud transformers with masked point modeling. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022

    2022

  16. [16]

    Masked autoencoders for point cloud self-supervised learning

    Yatian Pang, Wenxiao Wang, Francis EH Tay, Wei Liu, Yonghong Tian, and Li Yuan. Masked autoencoders for point cloud self-supervised learning. InComputer Vision–ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part II, pages 604–621. Springer, 2022. 26 Arxiv preprint - June 5, 2026

    2022

  17. [17]

    3dctn: 3d convolution-transformer network for point cloud classification.IEEE Transactions on Intelligent Transportation Systems, pages 1–12, 2022

    Dening Lu, Qian Xie, Kyle Gao, Linlin Xu, and Jonathan Li. 3dctn: 3d convolution-transformer network for point cloud classification.IEEE Transactions on Intelligent Transportation Systems, pages 1–12, 2022

    2022

  18. [18]

    Point-gn: A non-parametric network using gaussian positional en- coding for point cloud classification

    Marzieh Mohammadi and Amir Salarpour. Point-gn: A non-parametric network using gaussian positional en- coding for point cloud classification. In2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 3487–3496, 2025

    2025

  19. [19]

    3d semantic parsing of large-scale indoor spaces

    Iro Armeni, Ozan Sener, Amir R Zamir, Helen Jiang, Ioannis Brilakis, Martin Fischer, and Silvio Savarese. 3d semantic parsing of large-scale indoor spaces. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 1534–1543, 2016

    2016

  20. [20]

    Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias Nießner

    Angela Dai, Angel X. Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias Nießner. Scan- net: Richly-annotated 3d reconstructions of indoor scenes. InProc. Computer Vision and Pattern Recognition (CVPR), IEEE, 2017

    2017

  21. [21]

    An efficient accelerator for point-based and voxel-based point cloud neural networks

    Xinhao Yang, Tianyu Fu, Guohao Dai, Shulin Zeng, Kai Zhong, Ke Hong, and Yu Wang. An efficient accelerator for point-based and voxel-based point cloud neural networks. In2023 60th ACM/IEEE Design Automation Conference (DAC), pages 1–6. IEEE, 2023

    2023

  22. [22]

    A point transformer accelerator with fine-grained pipelines and distribution-aware dynamic fps

    Yaoxiu Lian, Xinhao Yang, Ke Hong, Yu Wang, Guohao Dai, and Ningyi Xu. A point transformer accelerator with fine-grained pipelines and distribution-aware dynamic fps. In2023 IEEE/ACM International Conference on Computer Aided Design (ICCAD), pages 1–9. IEEE, 2023

    2023

  23. [23]

    Meng Han, Liang Wang, Limin Xiao, Hao Zhang, Chenhao Zhang, Xiangrong Xu, and Jianfeng Zhu. Quickfps: Architecture and algorithm co-design for farthest point sampling in large-scale point clouds.IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 42(11):4011–4024, 2023

    2023

  24. [24]

    Lee, and Hongil Yoon

    Donghyun Lee, Dawoon Jeong, Jae W. Lee, and Hongil Yoon. Fastpoint: Accelerating 3d point cloud model inference via sample point distance prediction. InIEEE/CVF International Conference on Computer Vision (ICCV 25), 2025

    2025

  25. [25]

    4d spatio-temporal convnets: Minkowski convolutional neural networks

    Christopher Choy, JunYoung Gwak, and Silvio Savarese. 4d spatio-temporal convnets: Minkowski convolutional neural networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3075–3084, 2019

    2019

  26. [26]

    Frustum pointnets for 3d object detection from rgb-d data

    Charles R Qi, Wei Liu, Chenxia Wu, Hao Su, and Leonidas J Guibas. Frustum pointnets for 3d object detection from rgb-d data. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 918–927, 2018

    2018

  27. [27]

    Semantic segmentation for real point cloud scenes via bilateral aug- mentation and adaptive fusion

    Shi Qiu, Saeed Anwar, and Nick Barnes. Semantic segmentation for real point cloud scenes via bilateral aug- mentation and adaptive fusion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1757–1767, 2021

    2021

  28. [28]

    A volumetric method for building complex models from range images

    Brian Curless and Marc Levoy. A volumetric method for building complex models from range images. In Proceedings of the 23rd annual conference on Computer graphics and interactive techniques, pages 303–312, 1996

    1996

  29. [29]

    Pointacc: Efficient point cloud accelerator

    Yujun Lin, Zhekai Zhang, Haotian Tang, Hanrui Wang, and Song Han. Pointacc: Efficient point cloud accelerator. InMICRO-54: 54th Annual IEEE/ACM International Symposium on Microarchitecture, pages 449–461, 2021

    2021

  30. [30]

    A point transformer acceler- ator with distribution-aware heuristic distance calculation.IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2024

    Yaoxiu Lian, Xinhao Yang, Ke Hong, Yu Wang, Ningyi Xu, and Guohao Dai. A point transformer acceler- ator with distribution-aware heuristic distance calculation.IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2024

    2024

  31. [31]

    Accelerating point cloud sampling by parallel structure deconstruction.IEEE Transactions on Parallel and Distributed Systems, 37(1):60–75, 2026

    Hengzhe Chi, Jihe Wang, Jinzhe Zhang, and Danghui Wang. Accelerating point cloud sampling by parallel structure deconstruction.IEEE Transactions on Parallel and Distributed Systems, 37(1):60–75, 2026

    2026

  32. [32]

    3D is here: Point Cloud Library (PCL)

    Radu Bogdan Rusu and Steve Cousins. 3D is here: Point Cloud Library (PCL). InIEEE International Conference on Robotics and Automation (ICRA), Shanghai, China, May 9-13 2011

    2011

  33. [33]

    Grid-gcn for fast and scalable point cloud learning

    Qiangeng Xu, Xudong Sun, Cho-Ying Wu, Panqu Wang, and Ulrich Neumann. Grid-gcn for fast and scalable point cloud learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5661–5670, 2020

    2020

  34. [34]

    Kpconv: Flexible and deformable convolution for point clouds

    Hugues Thomas, Charles R Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, Franc ¸ois Goulette, and Leonidas J Guibas. Kpconv: Flexible and deformable convolution for point clouds. InProceedings of the IEEE/CVF international conference on computer vision, pages 6411–6420, 2019

    2019

  35. [35]

    Pgformer: a point cloud segmentation network for urban scenes combining grouped transformer and kpconv.IEEE Transactions on Geoscience and Remote Sensing, 2025

    Jiakang Xia, Yanming Chen, Guoli Li, Yueqian Shen, Xincan Zou, Dong Chen, and Yufu Zang. Pgformer: a point cloud segmentation network for urban scenes combining grouped transformer and kpconv.IEEE Transactions on Geoscience and Remote Sensing, 2025. 27 Arxiv preprint - June 5, 2026

    2025

  36. [36]

    Edgepc: Ef- ficient deep learning analytics for point clouds on edge devices

    Ziyu Ying, Sandeepa Bhuyan, Yan Kang, Yingtian Zhang, Mahmut T Kandemir, and Chita R Das. Edgepc: Ef- ficient deep learning analytics for point clouds on edge devices. InProceedings of the 50th Annual International Symposium on Computer Architecture, pages 1–14, 2023

    2023

  37. [37]

    An adjustable farthest point sampling method for approximately-sorted point cloud data

    Jingtao Li, Jian Zhou, Yan Xiong, Xing Chen, and Chaitali Chakrabarti. An adjustable farthest point sampling method for approximately-sorted point cloud data. In2022 IEEE workshop on signal processing systems (SiPS), pages 1–6. IEEE, 2022

    2022

  38. [38]

    Van-icp: Gpu-accelerated approximate nearest neighbor search for icp regis- tration via voxel dilation

    Weimin Wang and Qiong Chang. Van-icp: Gpu-accelerated approximate nearest neighbor search for icp regis- tration via voxel dilation. InICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5. IEEE, 2023

    2023

  39. [39]

    Accelerating nearest neighbor search in 3d point cloud regis- tration on gpus.ACM Transactions on Architecture and Code Optimization, 22(1):1–24, 2025

    Qiong Chang, Weimin Wang, and Jun Miyazaki. Accelerating nearest neighbor search in 3d point cloud regis- tration on gpus.ACM Transactions on Architecture and Code Optimization, 22(1):1–24, 2025

    2025

  40. [40]

    Behley, M

    J. Behley, M. Garbade, A. Milioto, J. Quenzel, S. Behnke, C. Stachniss, and J. Gall. SemanticKITTI: A Dataset for Semantic Scene Understanding of LiDAR Sequences. InProc. of the IEEE/CVF International Conf. on Computer Vision (ICCV), 2019

    2019

  41. [41]

    Pointnext: Revisiting pointnet++ with improved training and scaling strategies.arXiv:2206.04670, 2022

    Guocheng Qian, Yuchen Li, Houwen Peng, Jinjie Mai, Hasan Hammoud, Mohamed Elhoseiny, and Bernard Ghanem. Pointnext: Revisiting pointnet++ with improved training and scaling strategies.arXiv:2206.04670, 2022. 28

    work page arXiv 2022