Pith. sign in

REVIEW 3 major objections 5 minor 53 references

Querying Autonomous Vehicle Point Clouds: Enhanced by 3D Object Counting with CounterNet

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read CounterNet lifts LiDAR object counting and point-cloud query accuracy

desk verdict Sensible system, permissive evaluation: the counting-vs-detection claim is plausible but the abstract overstates it. read the letter →

arxiv 2507.19209 v2 pith:X5UQRMKL submitted 2025-07-25 cs.CV cs.MM

Xiaoyu Zhang , Zhifeng Bao , Hai Dong , Ziwei Wang , Jiajun Liu This is my paper · ORCID
classification cs.CVcs.MM
keywords pointcloudquerying3Dobjectcountingheatmapregressionautonomousdrivingbird'seyeviewmodelselectionLiDAR
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

Autonomous-vehicle point-cloud analysis needs frame-level object counts, but the paper argues that off-the-shelf 3D detectors count poorly, which silently corrupts queries that retrieve, count, or aggregate frames by object category. The authors propose CounterNet, a network that predicts a heatmap of object centers in a bird's-eye view and derives counts from thresholded peaks, deliberately skipping box fitting and orientation estimation. Partitioning the feature map with overlapping regions plus a per-frame dynamic model-selection step extend the idea to dense scenes and varied object sizes. If the reported 5–20% counting-accuracy gains hold, RETRIEVAL, COUNT, and AGGREGATION queries over point-cloud stores become substantially more reliable across nuScenes, KITTI, and Waymo.

What carries the argument

The load-bearing mechanism is a center-detection heatmap plus two refinement stages. First, a bird's-eye-view feature map feeds a heatmap head whose peaks indicate object centers; thresholded local-maxima detection turns peaks into counts, and the model is trained with focal loss plus an L1 counter loss. Second, the feature map is split into partitions to tame dense scenes, each partition gets an Otsu-derived dynamic threshold, and overlapping partitions prevent objects near boundaries from being split; finally, a per-frame model-selection step estimates each model variant's center in feature space and uses a Chernoff bound to compute confidence, so the closest center, adjusted by that confidence, picks the best partition-and-overlap configuration for each frame.

What would settle it

Measure per-frame counting accuracy of CounterNet against ground-truth annotations on a contiguous held-out sequence of several thousand frames; if the claimed 5–20% improvement over detection baselines does not appear, or if the Chernoff-adjusted model selection picks configurations that give lower average count accuracy than a fixed best single model, the core claims are falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that object counting—not localization—is the weak link in point-cloud querying, and that a heatmap-based center detector trained with a counting-specific loss repairs it. CounterNet projects a point-cloud frame to a bird's-eye-view feature map, produces a K-channel heatmap of object centers, and counts objects via thresholded local maxima; it is supervised jointly by focal loss on the heatmap and an L1 count loss. The paper reports that this beats state-of-the-art detection models on counting accuracy by 5–20% across object categories on nuScenes, KITTI, and Waymo, and that the gain transfers to higher accuracy for all three query types, with the overlapped-partition variant typically the strongest.

Load-bearing premise

The model-selection step assumes point cloud frames are independent samples when it uses the Chernoff bound to estimate confidence in each model's center, yet consecutive LiDAR frames are temporally correlated, so a miscalibrated bound could select suboptimal models and cut into the reported query accuracy.

Editorial extensions

If this is right

  • On nuScenes, KITTI, and Waymo, counting accuracy improves by 5–20% per object category, with the overlapped-partition CounterNet usually the best.
  • RETRIEVAL, COUNT, and AGGREGATION queries all become more accurate, since each query type depends on correct per-frame counts.
  • Overlap between partitions specifically fixes the failures that plain partitioning causes for large, sparsely distributed objects.
  • Model selection with the Chernoff-based adjustment reallocates frames from the dominant model to complementary ones, improving the worst-category counting accuracy while preserving overall performance.

Reading between the lines

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

  • The count-first, localize-later design could transfer to other 3D sparse data, such as warehouse robots or indoor scanning, where aggregate statistics matter more than boxes.
  • If counting becomes reliable enough, query engines could build indexes from cheap counting models rather than running full detectors, an efficiency lever the paper leaves unexplored.
  • The independence assumption behind the Chernoff adjustment is testable: on a contiguous sequence, shuffling frame order should change selection accuracy if the confidence probabilities are miscalibrated, pointing to a temporal-aware bound as a direct extension.
  • The baseline uses a fixed heatmap threshold of 0.5 while the partitioned variant uses per-partition Otsu thresholds; per-category thresholds, rather than per-partition ones, might be an even sharper refinement.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes CounterNet, a heatmap-based object counting network for autonomous vehicle point clouds, together with three performance enhancements: feature map partitioning, overlapping partitions, and per-frame dynamic model selection. It formalizes three query types over point cloud sequences (RETRIEVAL, COUNT, AGGREGATION) and evaluates query accuracy on the nuScenes, KITTI, and Waymo datasets against several state-of-the-art 3D detection baselines. The abstract claims a 5% to 20% improvement in counting accuracy across object categories and more reliable query outcomes across all query types.

Significance. If the claims are substantiated, the paper addresses a real and practical problem: existing point cloud query pipelines assume that detection models provide reliable object counts, which is often false in 3D autonomous driving data. The formalization of three query types and the emphasis on counting as a first-class operation are useful contributions, and the experimental breadth (three datasets, multiple detectors, three query types) is a strength. The code appears to be available. However, the method is largely a composition of established techniques (heatmap center detection, focal loss, Otsu thresholding, a Chernoff heuristic), and the reported gains are uneven across categories. The significance is therefore moderate rather than high, subject to the evaluation concerns below.

major comments (3)
  1. [§6.1, Evaluation Metrics] The RETRIEVAL and COUNT query accuracies are computed with tolerances derived from the maximum number of objects in a scene multiplied by the error tolerance, rather than from the query threshold. When the scene maximum is large, this tolerance can exceed the query threshold itself: with M=30 cars and a 10% tolerance, a frame whose true count is 3 and predicted count is 0 would be scored correct for a "car >= 3" query even though the predicted count does not satisfy the condition. Similarly, the COUNT query error is applied to the number of returned frames, which does not verify that each frame satisfies the condition in Definition 2. The reported query-accuracy numbers therefore do not directly measure correctness under Definitions 1–3, and the central claim of "more reliable query outcomes" is not supported by the current protocol. The authors should re-evaluate with exact matching or with a tolerance proportional to the query threshold, and report per-category results under that stricter protocol.
  2. [§5.3, Eq. (4) and Algorithm 2] The Chernoff bound is applied to high-dimensional feature vectors p_i.f without defining a proper scalar concentration quantity. In Algorithm 2, the "sample variance" ξ^2 is written as the average of (p_i.f − ω̂)^2, which is undefined for vector-valued features. The bound also assumes independent samples, but LiDAR frames in a continuous sequence are temporally correlated; this premise is not discussed. Because the dynamic model selection contribution and its reported gains (Figure 11) depend on this confidence adjustment, the authors must either replace the bound with a valid concentration inequality (e.g., a vector-valued or empirically calibrated bound) or provide direct empirical evidence that the adjustment improves selection accuracy rather than acting as an uncontrolled heuristic.
  3. [Abstract and Table 13] The abstract claims that CounterNet "improves counting accuracy by 5% to 20% across object categories," but the per-category COUNT query accuracy in Table 13 shows several categories where the best CounterNet variant is below the best baseline (bus: CN_o_pt 0.860 vs. SN 0.995; truck: CN_o_pt 0.592 vs. SN 0.602). Additionally, the "Object Count Accuracy" metric used in Figure 2 is never defined in Section 6.1 or in the appendices, and the 5–20% range is not traced to any specific table or definition. The quantitative claim should be revised to match the reported data, and the metric should be explicitly defined before it is used to support the paper's main selling point.
minor comments (5)
  1. [Table 4 caption] The caption reads "ETRIEV AL-MULTIPLE (KITTI)"; this should be "RETRIEVAL-MULTIPLE (KITTI)".
  2. [Appendix B, Figure reference] The text in Appendix B refers to "Figure 10" for the KITTI subset experiment, but the experiment result is a table (Table 10), not a figure. The cross-reference appears to be a mistake.
  3. [§5.3, typo] The sentence "p_i.f demoted the feature vector" should read "p_i.f denotes the feature vector".
  4. [Appendix C.1, JSON data model] In the example JSON, the object "count" is stored as a string ("10"); it should be a numeric value for arithmetic query operations such as SUM and AGGREGATION.
  5. [Tables 8, 17–19] The column headings "dist" and "rate" in the model selection tables are not explained. Defining these abbreviations in the table caption or in the text would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: CounterNet's counting and query results are trained and evaluated on held-out data; no prediction reduces by construction to a fitted input or self-citation.

full rationale

The derivation chain is self-contained. CounterNet is trained with heatmap focal loss plus an L1 count loss (Eqs. 1-2) against ground-truth object centers/counts; at test time, counts are obtained from thresholded local maxima on held-out frames, so no target quantity is fitted and then reported as a prediction. Partitioning, overlap, and Otsu thresholding are architectural/data-dependent transforms, not fitted parameters that reappear as query outputs. The model selection in Sec. 5.3 and Algorithm 2 computes model centers on training frames and selects nearest centers on validation frames; the Chernoff adjustment is a heuristic confidence multiplier, not a quantity that defines query accuracy, so its looseness is a modeling concern rather than a circular step. No load-bearing self-citation or imported uniqueness theorem is used; the only self-reference is the code repository [1], and the evaluations compare against external detection baselines on held-out query frames. The per-scene maximum-count tolerance in Sec. 6.1 can exceed a query threshold (e.g., 10% of the maximum car count), which may inflate absolute query-accuracy numbers, but every method is scored under the same tolerance and the CounterNet advantage is an empirical comparison, not a definitional reduction. The undefined "Object Count Accuracy" in Fig. 2 is a reporting gap, not a circular step. Overall, no prediction in the paper reduces by construction to its inputs.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The central method is a conventional deep network with hand-set hyperparameters (threshold, partition count, overlap ratio, epsilon, merge radius). The main domain assumptions are that BEV center heatmaps capture counts, that frames are independent for the confidence bound, and that the evaluation tolerance is meaningful. No new physical or conceptual entities beyond the proposed network architecture are introduced.

free parameters (5)
  • Fixed heatmap threshold t = 0.5
    Chosen as a mid-range value for peak detection in base CounterNet; hand-selected, not learned (Section 4).
  • Partition count pt = 4 (default); 2 and 9 studied
    Controls the size of feature map regions; tuned per dataset and object size (Section 5.1, Figure 7).
  • Overlap expansion ratio delta = 0.2 (default)
    Determines region expansion for overlapping partitions; set by experiments (Section 5.2, Figure 10).
  • Chernoff margin epsilon = 0.1 to 0.2 (nuScenes); 0.05 to 0.15 (KITTI)
    Tuning knob for the confidence adjustment in model selection; selected per dataset (Section 6.5, Appendix E.3).
  • Duplicate merge radius gamma = not specified
    Radius within which repeated center detections are merged; required by Algorithm 1 but its value is not reported.
assumptions (4)
  • domain assumption BEV-projected object centers are sufficient as counting supervision; box size and orientation provide no necessary information for counts.
    CounterNet trains only on center heatmaps; if adjacent objects collapse to one peak in dense scenes, counts will be underestimated. Section 4.
  • domain assumption Point cloud frames are independent, justifying the Chernoff-bound confidence estimate in model selection.
    Section 5.3 states the assumption without validation; LiDAR sequences are autocorrelated in time, so the bound may be miscalibrated.
  • domain assumption The ground-truth 3D annotations in nuScenes, KITTI, and Waymo are accurate enough to define heatmap targets.
    Occluded or truncated objects may have noisy centers, which would corrupt training and evaluation. Implicit in Section 6.1 data selection.
  • domain assumption The 10% query error tolerance (scaled by per-scene maximum object count) is a meaningful correctness criterion.
    Used for RETRIEVAL and COUNT metrics; lenient for categories with high max counts, potentially inflating reported accuracy. Section 6.1 Evaluation Metrics.
invented entities (1)
  • CounterNet heatmap counting head
    purpose: Counts objects by detecting peaks on a per-category BEV heatmap, replacing full 3D detection for query-accuracy purposes.
    Introduced in this paper; its effectiveness rests solely on the reported experiments, with no external replication, formal guarantee, or alternative-data confirmation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Querying Autonomous Vehicle Point Clouds: Enhanced by 3D Object Counting with CounterNet." pith.science (2026). https://pith.science/paper/X5UQRMKL

@misc{pith2026250719209,
  author       = {Pith},
  title        = {Pith review of: Querying Autonomous Vehicle Point Clouds: Enhanced by 3D Object Counting with CounterNet},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X5UQRMKL}},
  note         = {Machine review of arXiv:2507.19209}
}
read the original abstract

Autonomous vehicles generate massive volumes of point cloud data, yet only a subset is relevant for specific tasks such as collision detection, traffic analysis, or congestion monitoring. Effectively querying this data is essential to enable targeted analytics. In this work, we formalize point cloud querying by defining three core query types: RETRIEVAL, COUNT, and AGGREGATION, each aligned with distinct analytical scenarios. All these queries rely heavily on accurate object counts to produce meaningful results, making precise object counting a critical component of query execution. Prior work has focused on indexing techniques for 2D video data, assuming detection models provide accurate counting information. However, when applied to 3D point cloud data, state-of-the-art detection models often fail to generate reliable object counts, leading to substantial errors in query results. To address this limitation, we propose CounterNet, a heatmap-based network designed for accurate object counting in large-scale point cloud data. Rather than focusing on accurate object localization, CounterNet detects object presence by finding object centers to improve counting accuracy. We further enhance its performance with a feature map partitioning strategy using overlapping regions, enabling better handling of both small and large objects in complex traffic scenes. To adapt to varying frame characteristics, we introduce a per-frame dynamic model selection strategy that selects the most effective configuration for each input. Evaluations on three real-world autonomous vehicle datasets show that CounterNet improves counting accuracy by 5% to 20% across object categories, resulting in more reliable query outcomes across all supported query types.

Figures

Figures reproduced from arXiv: 2507.19209 by the authors.

Figure 1
Figure 1. Example of different query types car truckbus trailer constpedmotor bicycle cone barrier 0.5 0.6 0.7 0.8 0.9 1.0 Object Count Accuracy VoxelNeXt Transfusion CounterNet [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Counting ACC of baselines v.s. CounterNet (ours) [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. CounterNet Architecture. Given a point cloud frame, the features extracted from the backbone (e.g., VoxelNet) are [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Feature Map Partition object type. To train the heatmap, we utilise the ground truth posi￾tions of objects, annotating object centers with higher values and surrounding areas with lower values [26, 46, 49]. Object Counting via Threshold-based Peak Detection. We count o…
Figure 6
Figure 6. Figure 6: Process of Model Selection This raises a key question: how do we handle cases where an object’s center falls near the boundary of a partition? An intuitive approach to prevent objects from being split across partitions is to use feature-based clustering methods such as…
Figure 8
Figure 8. Figure 8: Different partitions(KITTI) vehicle ped cyclist 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 Object Count Acc CN CNpt = 2 CN o pt = 2 CNpt = 4 CN o pt = 4 CNpt = 9 CN o pt = 9 [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 11
Figure 11. Figure 11: Model selection(nuScenes) car truckbus trailer constpedmotor bicycle cone barrier 0.0 0.2 0.4 0.6 0.8 1.0 Object Count Acc k=0.5 k=0.52 k=0.54 Dynamic [PITH_FULL_IMAGE:figures/full_fig_p008_11.png]
Figure 13
Figure 13. Figure 13: Model selection with 9 models (nuScenes) car truckbus trailer constpedmotor bicycle cone barrier 0.6 0.7 0.8 0.9 1.0 Object Count Acc CNo pt w/o.Adjust =0.1 =0.15 =0.2 [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]
Figure 16
Figure 16. Figure 16: The relationship between 𝜖 and probability dominant models from the 9-model set (i.e., 𝐶𝑁𝑜=0 𝑝𝑡=4 , 𝐶𝑁𝑜=0.1 𝑝𝑡=4 , and 𝐶𝑁𝑜=0.2 𝑝𝑡=4 ). According to [PITH_FULL_IMAGE:figures/full_fig_p013_16.png]
Figure 17
Figure 17. Figure 17: Result of Dynamic Thresholding From the results in [PITH_FULL_IMAGE:figures/full_fig_p014_17.png]
Figure 19
Figure 19. Figure 19: Model selection (KITTI) H Case studies We evaluate the performance of the actual query across different scenarios and levels of selectivity on the nuScenes dataset. The experimental results assess accuracy, precision, and recall. The key observations are as follows: (…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 44 canonical work pages

  1. [1]

    Authors. 2025. Code. https://github.com/PetalZh/PCQuery

    work page 2025

  2. [2]

    Xuyang Bai, Zeyu Hu, Xinge Zhu, Qingqiu Huang, Yilun Chen, Hongbo Fu, and Chiew-Lan Tai. 2022. Transfusion: Robust lidar-camera fusion for 3d object detection with transformers. In CVPR. 1090–1099

    work page 2022

  3. [3]

    Jaeho Bang, Gaurav Tarlok Kakkar, Pramod Chunduri, Subrata Mitra, and Joy Arulraj. 2023. Seiden: Revisiting Query Processing in Video Database Systems. VLDB 16, 9 (2023), 2289–2301

    work page 2023

  4. [4]

    Behley, M

    J. Behley, M. Garbade, A. Milioto, J. Quenzel, S. Behnke, C. Stachniss, and J. Gall

  5. [5]

    Lang, Sourabh Vora, Venice Erin Li- ong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom

    Holger Caesar, Varun Bankiti, Alex H. Lang, Sourabh Vora, Venice Erin Li- ong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom

    work page

  6. [6]

    Jiashen Cao, Karan Sarkar, Ramyad Hadidi, Joy Arulraj, and Hyesoon Kim. 2022. Figo: Fine-grained query optimization in video analytics. In SIGMOD. 559–572

    work page 2022

  7. [7]

    arXiv preprint arXiv:1903.11027 (2019)

    nuScenes: A multimodal dataset for autonomous driving. arXiv preprint arXiv:1903.11027 (2019)

    arXiv 2019

  8. [8]

    Xiaozhi Chen, Huimin Ma, Ji Wan, Bo Li, and Tian Xia. 2017. Multi-view 3d object detection network for autonomous driving. In CVPR. 1907–1915

    work page 2017

Show all 53 references
  1. [9]

    Prithvijit Chattopadhyay, Ramakrishna Vedantam, Ramprasaath R Selvaraju, Dhruv Batra, and Devi Parikh. 2017. Counting everyday objects in everyday scenes. In CVPR. 1135–1144

    2017

  2. [10]

    Hisham Cholakkal, Guolei Sun, Salman Khan, Fahad Shahbaz Khan, Ling Shao, and Luc Van Gool. 2020. Towards partial supervision for generic object counting in natural scenes. IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 3 (2020), 1604–1622

    2020

  3. [11]

    Yukang Chen, Jianhui Liu, Xiangyu Zhang, Xiaojuan Qi, and Jiaya Jia. 2023. Voxelnext: Fully sparse voxelnet for 3d object detection and tracking. In CVPR. 21674–21683

    2023

  4. [12]

    Yuxing Han, Ziniu Wu, Peizhi Wu, Rong Zhu, Jingyi Yang, Liang Wei Tan, Kai Zeng, Gao Cong, Yanzhao Qin, Andreas Pfadler, Zhengping Qian, Jingren Zhou, Jiangneng Li, and Bin Cui. 2021. Cardinality Estimation in DBMS: A Comprehen- sive Benchmark Evaluation. Proc. VLDB Endow. 15,...

    2021

  5. [13]

    Vance Faber et al. 1994. Clustering and the continuous k-means algorithm. Los Alamos Science 22, 138144.21 (1994), 67

    1994

  6. [14]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In CVPR. 770–778

    2016

  7. [15]

    Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. 2017. Mask r-cnn. In ICCV. 2961–2969

    2017

  8. [16]

    Porter Jenkins, Kyle Armstrong, Stephen Nelson, Siddhesh Gotad, J Stockton Jenkins, Wade Wilkey, and Tanner Watts. 2023. Countnet3d: A 3d computer vision approach to infer counts of occluded objects. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer V...

    2023

  9. [17]

    Martin Hellman and Josef Raviv. 1970. Probability of error, equivocation, and the Chernoff bound. IEEE Transactions on Information Theory 16, 4 (1970), 368–372

    1970

  10. [18]

    Glenn Jocher, Alex Stoken, Jirka Borovec, NanoCode012, ChristopherSTAN, Liu Changyu, Laughing, tkianai, Adam Hogan, lorenzomammana, yxNONG, AlexWang1900, Laurentiu Diaconu, Marc, wanghaoyang0106, ml5ah, Doug, Fran- cisco Ingham, Frederik, Guilhen, Hatovix, Jake Poznanski, Jiac...

    2020 doi

  11. [19]

    Ruixiang Jiang, Lingbo Liu, and Changwen Chen. 2023. Clip-count: Towards text- guided zero-shot object counting. In Proceedings of the 31st ACM International Conference on Multimedia. 4535–4545

    2023

  12. [20]

    Daniel Kang, John Guibas, Peter D Bailis, Tatsunori Hashimoto, and Matei Za- haria. 2022. Tasti: Semantic indexes for machine learning-based queries over unstructured data. In SIGMOD. 1934–1947

    2022

  13. [21]

    Daniel Kang, Peter Bailis, and Matei Zaharia. 2019. BlazeIt: Optimizing Declara- tive Aggregation and Limit Queries for Neural Network-Based Video Analytics. VLDB 13, 4 (2019)

    2019

  14. [22]

    Fiodar Kazhamiaka, Matei Zaharia, and Peter Bailis. 2021. Challenges and Op- portunities for Autonomous Vehicle Query Systems.. In CIDR

    2021

  15. [23]

    Seunggu Kang, WonJun Moon, Euiyeon Kim, and Jae-Pil Heo. 2024. Vlcounter: Text-aware visual representation for zero-shot object counting. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 38. 2714–2722

    2024

  16. [24]

    Alex H Lang, Sourabh Vora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom. 2019. Pointpillars: Fast encoders for object detection from point clouds. In CVPR. 12697–12705

    2019

  17. [25]

    Xin Lai, Yukang Chen, Fanbin Lu, Jianhui Liu, and Jiaya Jia. 2023. Spherical transformer for lidar-based 3d recognition. In CVPR. 17545–17555

    2023

  18. [26]

    Hei Law and Jia Deng. 2018. Cornernet: Detecting objects as paired keypoints. In ECCV. 734–750

    2018

  19. [27]

    Issam H Laradji, Negar Rostamzadeh, Pedro O Pinheiro, David Vazquez, and Mark Schmidt. 2018. Where are the blobs: Counting by localization with point supervision. In ECCV. 547–562

    2018

  20. [28]

    Jiangneng Li, Haitao Yuan, Gao Cong, Han Mao Kiah, and Shuhao Zhang. 2025. MAST: Towards Efficient Analytical Query Processing on Point Cloud Data. Proceedings of the ACM on Management of Data 3, 1 (2025), 1–27

    2025

  21. [29]

    Asad Lesani, Ehsan Nateghinia, and Luis F Miranda-Moreno. 2020. Development and evaluation of a real-time pedestrian counting system for high-volume condi- tions based on 2D LiDAR. Transportation research part C: emerging technologies 114 (2020), 20–35

    2020

  22. [30]

    Shaoshan Liu, Liangkai Liu, Jie Tang, Bo Yu, Yifan Wang, and Weisong Shi. 2019. Edge computing for autonomous driving: Opportunities and challenges. Proc. IEEE 107, 8 (2019), 1697–1716

    2019

  23. [31]

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. 2017. Focal loss for dense object detection. In ICCV. 2980–2988

    2017

  24. [32]

    Guo Ming. 2023. Exploration of the intelligent control system of autonomous vehicles based on edge computing. PLoS One 18, 2 (2023), e0281294

    2023

  25. [33]

    Zhijian Liu, Haotian Tang, Shengyu Zhao, Kevin Shao, and Song Han. 2021. Pvnas: 3d neural architecture search with point-voxel convolution. IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 11 (2021), 8552–8568

    2021

  26. [34]

    Nobuyuki Otsu et al. 1975. A threshold selection method from gray-level his- tograms. Automatica 11, 285-296 (1975), 23–27

    1975

  27. [35]

    Anjana K Nellithimaru and George A Kantor. 2019. Rols: Robust object-level slam for grape counting. In CVPR workshops. 0–0

    2019

  28. [36]

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. 2017. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. NeurIPS 30 (2017)

    2017

  29. [37]

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. 2017. Pointnet: Deep learning on point sets for 3d classification and segmentation. In CVPR. 652–660

    2017

  30. [38]

    Tobias Stahl, Silvia L Pintea, and Jan C Van Gemert. 2018. Divide and count: Generic object counting by image divisions.IEEE Transactions on Image Processing 28, 2 (2018), 1035–1044

    2018

  31. [39]

    Shaoshuai Shi, Chaoxu Guo, Li Jiang, Zhe Wang, Jianping Shi, Xiaogang Wang, and Hongsheng Li. 2020. Pv-rcnn: Point-voxel feature set abstraction for 3d object detection. In CVPR. 10529–10538

    2020

  32. [40]

    Pei Sun, Mingxing Tan, Weiyue Wang, Chenxi Liu, Fei Xia, Zhaoqi Leng, and Dragomir Anguelov. 2022. SWFormer: Sparse Window Transformer for 3D Object Detection in Point Clouds. In ECCV, Shai Avidan, Gabriel J. Brostow, Moustapha Cissé, Giovanni Maria Farinella, and Tal Hassner ...

    2022

  33. [41]

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, Vijay Vasudevan, Wei Han, Jiquan Ngiam, Hang Zhao, Aleksei Timofeev, Scott Ettinger, Maxim Krivokon, Amy Gao, Aditya Joshi, Yu Zhang,...

    2020

  34. [42]

    OpenPCDet Development Team. 2020. OpenPCDet: An Open-source Toolbox for 3D Object Detection from Point Clouds. https://github.com/open-mmlab/ OpenPCDet

    2020

  35. [43]

    Sihai Tang, Bruce Chen, Harold Iwen, Jason Hirsch, Song Fu, Qing Yang, Paparao Palacharla, Nannan Wang, Xi Wang, and Weisong Shi. 2021. Vecframe: A vehicu- lar edge computing framework for connected autonomous vehicles. In 2021 IEEE International Conference on Edge Computing (...

    2021

  36. [44]

    Yan Yan, Yuxing Mao, and Bo Li. 2018. Second: Sparsely embedded convolutional detection. Sensors 18, 10 (2018), 3337

    2018

  37. [45]

    Jingyi Xu, Hieu Le, Vu Nguyen, Viresh Ranjan, and Dimitris Samaras. 2023. Zero-shot object counting. In CVPR. 15548–15557

    2023

  38. [46]

    Tianwei Yin, Xingyi Zhou, and Philipp Krahenbuhl. 2021. Center-based 3d object detection and tracking. In CVPR. 11784–11793

    2021

  39. [47]

    Zetong Yang, Yanan Sun, Shu Liu, and Jiaya Jia. 2020. 3dssd: Point-based 3d single stage object detector. In CVPR. 11040–11048

    2020

  40. [48]

    Hui Zhou, Xinge Zhu, Xiao Song, Yuexin Ma, Zhe Wang, Hongsheng Li, and Dahua Lin. 2020. Cylinder3d: An effective 3d framework for driving-scene lidar semantic segmentation. arXiv preprint arXiv:2008.01550 (2020)

    2020 arXiv

  41. [49]

    Gang Zhang, Junnan Chen, Guohuan Gao, Jianmin Li, Si Liu, and Xiaolin Hu

  42. [50]

    f r a m e _ i d

    Yin Zhou and Oncel Tuzel. 2018. Voxelnet: End-to-end learning for point cloud based 3d object detection. In CVPR. 4490–4499. MM ’25, October 27–31, 2025, Dublin, Ireland. Xiaoyu Zhang, Zhifeng Bao, Hai Dong, Ziwei Wang, and Jiajun Liu A Table of Notations P set of point cloud ...

    2018

  43. [52]

    Xingyi Zhou, Vladlen Koltun, and Philipp Krähenbühl. 2020. Tracking Objects as Points. In ECCV, Andrea Vedaldi, Horst Bischof, Thomas Brox, and Jan-Michael Frahm (Eds.), Vol. 12349. Springer, 474–490

    2020

  44. [2019]

    SemanticKITTI: A Dataset for Semantic Scene Understanding of LiDAR Sequences. In ICCV

  45. [2024]

    In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, W A, USA, June 16-22, 2024

    SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, W A, USA, June 16-22, 2024. IEEE, 14477–14486

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.