Pith. sign in

REVIEW 2 major objections 5 minor 96 references

This paper claims that a CARLA-based generator can produce the largest V2X cooperative-perception dataset to date, with multi-task labels, and that adding fused axial attention to a max-pooling fusion baseline improves cooperative perceptio

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 · deepseek-v4-flash

2026-07-31 23:32 UTC pith:I4VUZRHX

load-bearing objection Dataset is a real contribution; overpass ground-truth fallback is an unquantified risk, and the model claim overreaches. the 2 major comments →

arxiv 2607.23910 v1 pith:I4VUZRHX submitted 2026-07-27 cs.CV cs.LGcs.MAcs.RO

SimBEV2X: A Large-Scale Dataset and Data Generation Tool for Multi-Task Vehicle-to-Everything Cooperative Perception

Goodarz Mehr , Sepideh Gohari , Montasir Abbas , Azim Eskandarian This is my paper
classification cs.CV cs.LGcs.MAcs.RO
keywords cooperative perceptionV2Xsynthetic datasetbird's-eye viewsemantic occupancydata generation toolautonomous drivingBEV segmentation
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's central claim is that cooperative self-driving perception has been held back by small, single-task datasets, and that a simulator-based generator can remove that bottleneck at scale. To show this, the authors built SimBEV2X, a CARLA-based tool that randomly generates diverse multi-agent driving scenes with up to eight connected vehicles and four roadside units, and used it to release a dataset roughly ten times larger than existing V2X perception datasets. The dataset pairs lidar, camera, radar, and GNSS/IMU data with 3D bounding boxes, HD map information, BEV segmentation maps, and semantic occupancy voxel grids for every agent. As a demonstration, they benchmark cooperative detection and BEV segmentation and present CoBEVFusion, which adds fused axial attention to an existing max-pooling fusion model and improves results, most clearly on small objects and per-cell segmentation.

Core claim

On its own terms, the paper establishes that a configurable synthetic pipeline can generate cooperative-perception data at a scale and annotation richness that real-world collection cannot currently match: 258 scenes, 102,200 frames, 588,520 point clouds, 3,241,560 images, and 27,337,066 bounding boxes across 11 road networks. The authors argue this is the largest V2X dataset to date and that its multi-task annotations—unique track IDs, HD maps, BEV segmentation, semantic occupancy—make it a broad testing ground for cooperative perception algorithms. The accompanying architecture claim is that replacing element-wise max-pooling fusion with fused axial attention (FAX), which applies local win

What carries the argument

The load-bearing object is the SimBEV2X annotation engine. For BEV ground truth, it exploits the one-way see-through property of CARLA's ground materials: cameras placed 1 km above and below each agent capture both the ground and objects occluded from the top, merged with OpenDRIVE-parsed road geometry and binary closing for the road class. For 3D semantic occupancy, a custom voxel sensor checks every voxel for overlap with object meshes and resolves conflicts by fixed class priority, with interior voxels filled afterwards by a six-direction scan. CoBEVFusion, the proposed baseline, is a cooperative fusion module that adds fused axial attention (FAX)—local window attention plus sparse global

Load-bearing premise

The load-bearing premise is the paper's own acknowledged compromise for overpass scenes (Section III-D-c): when roads at different elevations cross, BEV non-map ground truth falls back to bounding-box projection plus elevation filtering, and if that fallback is systematically biased, the BEV segmentation and occupancy benchmarks for those scenes are unreliable.

What would settle it

Check the overpass scenes: for frames where any two waypoints within 48 m differ in elevation by at least 6.4 m, compare the generated BEV segmentation ground truth against labels derived from semantic lidar or manual annotation; systematic disagreement under bridges or overpasses would mean those benchmark numbers are not trustworthy.

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

If this is right

  • The dataset gives the cooperative-perception community a common testbed roughly ten times larger than prior V2X sets, with enough frames to train deep models and enough scenes to evaluate generalization across maps, including a held-out Town13 subset.
  • Multi-task annotations—3D boxes with track IDs, HD map information, BEV segmentation, and semantic occupancy—allow a single benchmark to track detection, segmentation, and occupancy prediction under the same multi-agent conditions.
  • Replay and scene-expansion features in the tool mean researchers can generate custom V2X data or extend the released dataset with additional scenarios.
  • CoBEVFusion's gains over CoopDet3D are larger for BEV segmentation than for 3D detection, implying that fusion quality shows up where per-cell labels are required, while detection heads built on learned queries are robust to local feature noise.
  • The 160 m by 160 m BEV grid and 128 m by 128 m by 8 m occupancy grids support training-time augmentation beyond the usual 100 m range.

Where Pith is reading between the lines

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

  • If the overpass fallback for BEV ground truth—bounding-box projection plus elevation filtering—carries a systematic bias, then BEV segmentation and occupancy benchmarks in scenes with overpasses could be skewed; this is testable because the trigger condition is explicitly defined (two waypoints within 48 m with elevation gap at least 6.4 m).
  • The 'largest V2X dataset' claim depends on comparing against published dataset sizes; as the field grows, that record is inherently temporary, and the durable contribution is likely the generator and annotation machinery rather than the record itself.
  • Since the sensor configuration is patterned after a widely used real-world driving set, models trained on SimBEV2X may transfer to real-world data if the multi-agent alignment domain gap is addressed—a direction the authors name as future work.
  • The baseline results suggest camera-only cooperative perception remains weak on small objects; a testable extension is whether higher-resolution camera-to-BEV lifting with attention can close the lidar gap without adding lidar sensors.

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 / 5 minor

Summary. The paper introduces SimBEV2X, a CARLA-based configurable data-generation tool for V2X cooperative perception, and a dataset of 258 scenes, 102,200 frames, 588,520 lidar point clouds, 3.24M images, and 27.3M 3D bounding boxes, together with HD-map, BEV-segmentation, and 3D-semantic-occupancy annotations. It claims the dataset is the largest V2X perception dataset to date and an order of magnitude larger than prior ones. The authors benchmark the dataset with BEVFusion, CoopDet3D, and a proposed CoBEVFusion architecture that adds fused axial attention (FAX) to CoopDet3D's max-pooling fuser, reporting improvements in BEV segmentation and lidar/fused 3D detection. The dataset and tool are released.

Significance. If the dataset is sound, it is a valuable resource: it is substantially larger in annotation count than prior V2X datasets, covers multiple maps and dynamic weather, and provides multi-task annotations that are rare in this area. The open-source generation tool is a strength, as it enables reproducibility and user-customized extensions. The CoBEVFusion results are of secondary interest but provide useful baselines. The main risk is annotation correctness in the BEV ground truth, whose fallback mode for overpass scenes is unvalidated, and the over-stated uniform superiority of CoBEVFusion in the abstract.

major comments (2)
  1. [III-D-c (BEV ground truth)] The paper states that when two waypoints within 48.0 m have an elevation difference of at least 6.4 m, the BEV ground truth falls back to calculating non-map classes from 3D bounding boxes and filtering out map boundary points with an elevation difference above 4.8 m, and labels the result 'acceptable' without quantitative support. No fraction of the 102,200 frames or 258 scenes that trigger this fallback is reported, and no validation of the fallback against the primary camera-based method or any reference is given. Since BEV segmentation (Table VII) is a central advertised task of the dataset, a substantial prevalence of overpass frames would systematically degrade the road, road-line, sidewalk, crosswalk, and object-footprint labels in exactly those frames. Please report the fallback frequency per map/scene, validate the fallback labels on overpass segments, and quantify the sensitivi
  2. [Abstract and V-C (Table VIII)] The abstract and conclusion claim that CoBEVFusion achieves 'superior performance,' but Table VIII shows this is not true for the camera-only variant: CoBEVFusion-C obtains mAP 11.1 versus 24.2 for CoopDet3D-C and 25.8 for BEVFusion-C, and Table IX shows it is worse on nearly every class. The advantage of CoBEVFusion over CoopDet3D appears only in the lidar and fused variants, and no repeated-seed variance is reported, so the headline gains (e.g., 76.2 vs. 74.2 mAP for the fused models) could be within run-to-run noise. Please qualify the 'superior performance' claim, explain the camera-only regression, and add variance estimates or a statistical comparison.
minor comments (5)
  1. [Abstract and Table I] 'An order of magnitude larger' is not supported for all quantities: Table I shows 102,200 frames versus 39,000 for DAIR-V2X-C (2.6x), while the bounding-box count is >10x. Please specify which data modality/quantity supports the order-of-magnitude claim.
  2. [III and IV] The text refers to 'Appendix A.I' (custom CARLA changes) and 'Appendix A.II' (additional dataset statistics), but no appendices appear in the manuscript. Include them or remove the references.
  3. [Table VII] The column headers for Table VII are hard to parse: the grouping of IoU thresholds (@0.5/@0.7/@0.9), distance bands (<20 m/20-40 m/>40 m), and model/FPS/VRAM columns is unclear. Please reformat or add a caption explaining the column hierarchy.
  4. [V-B] The sentence 'We only use one FAX module (instead of the three used in [40])' appears to cite OpenCDA [40]; the three-FAX-module design is from CoBEVT [8]. Please correct the citation.
  5. [V-A, Eq. (4)] The notation TP is used both for the set of true-positive metrics and for individual metric values in the SDS equation. Please disambiguate, e.g., by writing mTP = (mATE, mAOE, mASE, mAVE).

Circularity Check

0 steps flagged

No circularity: the dataset is a newly generated artifact and the benchmark comparisons are empirical; the only self-citations are contextual and not load-bearing.

full rationale

The paper's central claims are (1) that SimBEV2X can generate a large multi-task V2X dataset and (2) that CoBEVFusion performs better than baseline cooperative-perception models on that dataset. There is no derivation chain in which a predicted quantity reduces by construction to a fitted input. The dataset is produced by direct simulator queries (CARLA API, OpenDRIVE parsing, camera projection, voxel overlap checks), not by inferring labels from the benchmarked models. The BEV ground-truth fallback for overpass scenes is explicitly labeled as less accurate but 'acceptable'; this is an annotation-quality and validation concern, not a circular reduction, because the fallback labels are generated independently of the models and no model output or fitted parameter is being renamed as a prediction. CoBEVFusion's reported superiority is an empirical result on a common benchmark: both CoBEVFusion and CoopDet3D are trained and evaluated on the same dataset, so the comparison is not forced by construction. The only self-citations (SimBEV [16], X-CAR [88]) are contextual: SimBEV2X is described as built on the authors' prior single-agent tool, but this self-citation does not carry the load of any derivation; the uniqueness/size claims are compared against externally published datasets in Table I. No self-citation chain is used to forbid alternatives or to justify the architecture's design choice. Consequently, the paper contains no identifiable circular step under the stated criteria.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

The paper introduces no free parameters in a derivation sense; its contributions are artifacts. The central claims rely on the fidelity of CARLA simulation, the custom ground-truth methods, and the validity-filter heuristic, all of which are taken as domain assumptions without external validation in the paper.

axioms (4)
  • domain assumption CARLA's simulated sensor and traffic models are a valid approximation of real-world V2X perception conditions.
    The whole dataset's value as a benchmark rests on the sim-to-real transferability of CARLA data; the paper does not validate on real V2X data (Section IV-V).
  • domain assumption The BEV ground truth obtained through the one-way see-through material trick plus OpenDRIVE parsing is accurate except in overpass cases, where the fallback is 'still acceptable'.
    Section III-D-c asserts the fallback accuracy without quantitative validation; if wrong, BEV segmentation benchmarks are unreliable.
  • domain assumption The custom semantic occupancy sensor using Unreal Engine collision/overlap checks correctly labels voxels; post-processing fills interiors.
    Section III-D-d describes the method but provides no validation against a reference.
  • domain assumption The 'valid' object filter based on lidar/radar point counts and instance segmentation IDs accurately represents observability.
    Section III-E defines validity heuristically; detection difficulty labels depend on it.

pith-pipeline@v1.3.0-alltime-deepseek · 27108 in / 13230 out tokens · 129433 ms · 2026-07-31T23:32:15.428094+00:00 · methodology

0 comments
read the original abstract

Cooperative perception through vehicle-to-everything (V2X) communication can overcome the inherent physical limitations of individual autonomous vehicles, such as occlusions and limited sensor range. However, the development of robust V2X algorithms, particularly those relying on unified spatial representations like bird's-eye view (BEV) representation, is hampered by the lack of large-scale, multi-modal, multi-task datasets. Moreover, collecting and annotating a large set of synchronized, real-world multi-agent data is prohibitively expensive. This has resulted in a landscape where existing V2X datasets are notably limited in both size and scope. To overcome this, we introduce SimBEV2X, an advanced synthetic data generation tool built on the CARLA simulator. SimBEV2X automatically creates randomized driving scenarios to collect multi-modal sensor data alongside various types of ground truth including 3D bounding boxes with unique track IDs, HD map information, BEV segmentation maps, and semantic occupancy voxel grids from both vehicles and RSUs. We also present the SimBEV2X dataset, the largest V2X perception dataset to date. The dataset comprises 258 scenes, each involving up to 8 connected vehicles and up to 4 RSUs across a variety of road networks. The SimBEV2X dataset is an order of magnitude larger than existing V2X datasets and contains 102,200 frames, 588,520 lidar point clouds, more than 3 million images, over 27 million bounding boxes, and a comprehensive set of other annotations. Finally, we establish a strong baseline on the SimBEV2X dataset using CoopDet3D and propose CoBEVFusion, a novel architecture that combines CoopDet3D with fused axial attention (FAX) for context-aware multi-agent feature aggregation, resulting in superior performance. SimBEV2X, the SimBEV2X dataset, and CoBEVFusion are available at https://simbev2x.org and https://github.com/GoodarzMehr/SimBEV2X.

Figures

Figures reproduced from arXiv: 2607.23910 by Azim Eskandarian, Goodarz Mehr, Montasir Abbas, Sepideh Gohari.

Figure 1
Figure 1. Figure 1: Three annotation types provided by SimBEV2X: 3D object bounding [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of SimBEV2X’s modular design. The [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of SimBEV2X’s logic. Steps for Vehicle/RSU Manager, Sensor Manager, and Ground Truth Manager are repeated for each agent. Shaded [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Coordinate frames of the perception sensors attached to data collection [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Transition of weather from cloudy to rainy (top), day to night (middle), and clear to dense fog (bottom). [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Randomly generated traffic accidents and the associated warning sign [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Left: BEV road data calculated using the OpenDRIVE map; there [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Our custom CARLA sensor can produce semantic voxel grids of [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Number of 3D object bounding boxes, BEV ground truth labels, and 3D semantic occupancy labels per class in the SimBEV2X dataset. The hatching [PITH_FULL_IMAGE:figures/full_fig_p010_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Distribution of various parameters across the scenes and frames of the SimBEV2X dataset. While scene duration is chosen randomly from a uniform [PITH_FULL_IMAGE:figures/full_fig_p010_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Logarithmic heat map of BEV labels for each class aggregated across the SimBEV2X dataset. The biggest concentration of labels is along the path [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Transition matrices over the SimBEV2X dataset for select weather parameters. For (a) through (c), [PITH_FULL_IMAGE:figures/full_fig_p011_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: CoBEVFusion’s architecture builds on CoopDet3D’s by augmenting multi-agent BEV features using a fused axial attention (FAX) module that [PITH_FULL_IMAGE:figures/full_fig_p013_14.png] 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

96 extracted references · 10 linked inside Pith

  1. [1]

    Autonomous driving system: A comprehensive survey,

    J. Zhao, W. Zhao, B. Deng, Z. Wang, F. Zhang, W. Zheng, W. Cao, J. Nan, Y . Lian, and A. F. Burke, “Autonomous driving system: A comprehensive survey,”Expert Systems with Applications, p. 122836,

  2. [2]

    Multi-modal 3D object detection in autonomous driving: A survey and taxonomy,

    L. Wang, X. Zhang, Z. Song, J. Bi, G. Zhang, H. Wei, L. Tang, L. Yang, J. Li, C. Jiaet al., “Multi-modal 3D object detection in autonomous driving: A survey and taxonomy,”IEEE Transactions on Intelligent Vehicles, vol. 8, no. 7, pp. 3781–3798, 2023. 1

    2023

  3. [3]

    Research advances and challenges of autonomous and connected ground vehicles,

    A. Eskandarian, C. Wu, and C. Sun, “Research advances and challenges of autonomous and connected ground vehicles,”IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 2, pp. 683–711, 2019. 1

    2019

  4. [4]

    A survey on multimodal large language models for autonomous driving,

    C. Cui, Y . Ma, X. Cao, W. Ye, Y . Zhou, K. Liang, J. Chen, J. Lu, Z. Yang, K.-D. Liaoet al., “A survey on multimodal large language models for autonomous driving,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 958–

    2024

  5. [5]

    Seeing through fog without seeing fog: Deep multi- modal sensor fusion in unseen adverse weather,

    M. Bijelic, T. Gruber, F. Mannan, F. Kraus, W. Ritter, K. Dietmayer, and F. Heide, “Seeing through fog without seeing fog: Deep multi- modal sensor fusion in unseen adverse weather,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 11 682–11 692. 1

    2020

  6. [6]

    State estimation and motion prediction of vehicles and vulnerable road users for cooper- ative autonomous driving: A survey,

    P. Ghorai, A. Eskandarian, Y .-K. Kim, and G. Mehr, “State estimation and motion prediction of vehicles and vulnerable road users for cooper- ative autonomous driving: A survey,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 10, pp. 16 983–17 002, 2022. 1

    2022

  7. [7]

    Vision-centric BEV perception: A survey,

    Y . Ma, T. Wang, X. Bai, H. Yang, Y . Hou, Y . Wang, Y . Qiao, R. Yang, and X. Zhu, “Vision-centric BEV perception: A survey,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 1

    2024

  8. [8]

    CoBEVT: Coop- erative bird’s-eye view semantic segmentation with sparse transformers,

    R. Xu, Z. Tu, H. Xiang, W. Shao, B. Zhou, and J. Ma, “CoBEVT: Coop- erative bird’s-eye view semantic segmentation with sparse transformers,” inConference on Robot Learning. PMLR, 2023, pp. 989–1000. 1, 2, 3, 12

    2023

  9. [9]

    MACP: Efficient model adaptation for cooperative perception,

    Y . Ma, J. Lu, C. Cui, S. Zhao, X. Cao, W. Ye, and Z. Wang, “MACP: Efficient model adaptation for cooperative perception,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 3373–3382. 1, 3

    2024

  10. [10]

    Collaborative semantic occupancy prediction with hybrid feature fusion in connected automated vehicles,

    R. Song, C. Liang, H. Cao, Z. Yan, W. Zimmer, M. Gross, A. Festag, and A. Knoll, “Collaborative semantic occupancy prediction with hybrid feature fusion in connected automated vehicles,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 17 996–18 006. 1

    2024

  11. [11]

    MDNet: Multimodal cooperative perception via spatial alignment of modal decision-making,

    J. He, X. Deng, J. Gui, T. Zhang, and X. He, “MDNet: Multimodal cooperative perception via spatial alignment of modal decision-making,” IEEE Internet of Things Journal, vol. 12, no. 11, pp. 16 142–16 154,

  12. [12]

    Challenges in de- signing datasets and validation for autonomous driving,

    M. Uric ´ar, D. Hurych, P. Krizek, and S. Yogamani, “Challenges in de- signing datasets and validation for autonomous driving,”arXiv preprint arXiv:1901.09270, 2019. 2

    Pith/arXiv arXiv 1901

  13. [13]

    A survey on autonomous driving datasets: Statistics, annotation quality, and a future outlook,

    M. Liu, E. Yurtsever, J. Fossaert, X. Zhou, W. Zimmer, Y . Cui, B. L. Zagar, and A. C. Knoll, “A survey on autonomous driving datasets: Statistics, annotation quality, and a future outlook,”IEEE Transactions on Intelligent Vehicles, 2024. 2

    2024

  14. [14]

    nuScenes: A multi- modal dataset for autonomous driving,

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Krishnan, Y . Pan, G. Baldan, and O. Beijbom, “nuScenes: A multi- modal dataset for autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2020, pp. 11 621–11 631. 2, 3, 5, 9, 11, 12, 13

    2020

  15. [15]

    CARLA: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V . Koltun, “CARLA: An open urban driving simulator,” inProceedings of the Conference on Robot Learning. PMLR, 2017, pp. 1–16. 2, 3

    2017

  16. [16]

    SimBEV: A synthetic multi-task multi-sensor driving data generation tool and dataset,

    G. Mehr and A. Eskandarian, “SimBEV: A synthetic multi-task multi-sensor driving data generation tool and dataset,”arXiv preprint arXiv:2502.01894, 2025. 2

    Pith/arXiv arXiv 2025

  17. [17]

    OPV2V: An open benchmark dataset and fusion pipeline for perception with vehicle-to- vehicle communication,

    R. Xu, H. Xiang, X. Xia, X. Han, J. Li, and J. Ma, “OPV2V: An open benchmark dataset and fusion pipeline for perception with vehicle-to- vehicle communication,” inProceedings of the International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 2583–2589. 2, 3

    2022

  18. [18]

    V2X- Sim: Multi-agent collaborative perception dataset and benchmark for autonomous driving,

    Y . Li, D. Ma, Z. An, Z. Wang, Y . Zhong, S. Chen, and C. Feng, “V2X- Sim: Multi-agent collaborative perception dataset and benchmark for autonomous driving,”IEEE Robotics and Automation Letters, vol. 7, no. 4, pp. 10 914–10 921, 2022. 2, 3

    2022

  19. [19]

    DAIR-V2X: A large-scale dataset for vehicle- infrastructure cooperative 3D object detection,

    H. Yu, Y . Luo, M. Shu, Y . Huo, Z. Yang, Y . Shi, Z. Guo, H. Li, X. Hu, J. Yuanet al., “DAIR-V2X: A large-scale dataset for vehicle- infrastructure cooperative 3D object detection,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 21 361–21 370. 2, 3

    2022

  20. [20]

    V2V4Real: A real-world large-scale dataset for vehicle-to-vehicle cooperative perception,

    R. Xu, X. Xia, J. Li, H. Li, S. Zhang, Z. Tu, Z. Meng, H. Xiang, X. Dong, R. Songet al., “V2V4Real: A real-world large-scale dataset for vehicle-to-vehicle cooperative perception,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 13 712–13 722. 2, 3

    2023

  21. [21]

    V2X-Seq: A large-scale sequential dataset for vehicle-infrastructure cooperative perception and forecasting,

    H. Yu, W. Yang, H. Ruan, Z. Yang, Y . Tang, X. Gao, X. Hao, Y . Shi, Y . Pan, N. Sunet al., “V2X-Seq: A large-scale sequential dataset for vehicle-infrastructure cooperative perception and forecasting,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 5486–5495. 2, 3

    2023

  22. [22]

    TUMTraf V2X cooperative perception dataset,

    W. Zimmer, G. A. Wardana, S. Sritharan, X. Zhou, R. Song, and A. C. Knoll, “TUMTraf V2X cooperative perception dataset,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, 2024, pp. 22 668–22 677. 2, 3, 10, 12

    2024

  23. [23]

    V2X-Real: A large-scale dataset for vehicle- to-everything cooperative perception,

    H. Xiang, Z. Zheng, X. Xia, R. Xu, L. Gao, Z. Zhou, X. Han, X. Ji, M. Li, Z. Menget al., “V2X-Real: A large-scale dataset for vehicle- to-everything cooperative perception,” inProceedings of the European Conference on Computer Vision. Springer, 2024, pp. 455–470. 2, 3, 10

    2024

  24. [24]

    A synthetic benchmark for collaborative 3D semantic occupancy pre- diction in V2X autonomous driving,

    H. Wu, P. Lin, E. Javanmardi, N. Bao, B. Qian, H. Si, and M. Tsukada, “A synthetic benchmark for collaborative 3D semantic occupancy pre- diction in V2X autonomous driving,”arXiv preprint arXiv:2506.17004,

    arXiv

  25. [25]

    SHIFT: A synthetic driving dataset for contin- uous multi-task domain adaptation,

    T. Sun, M. Segu, J. Postels, Y . Wang, L. Van Gool, B. Schiele, F. Tombari, and F. Yu, “SHIFT: A synthetic driving dataset for contin- uous multi-task domain adaptation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 21 371–21 382. 2, 3, 6

    2022

  26. [26]

    Vision meets robotics: The KITTI dataset,

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The KITTI dataset,”The International Journal of Robotics Research, vol. 32, no. 11, pp. 1231–1237, 2013. 2

    2013

  27. [27]

    The CityScapes dataset for semantic urban scene understanding,

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benen- son, U. Franke, S. Roth, and B. Schiele, “The CityScapes dataset for semantic urban scene understanding,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2016, pp. 3213–3223. 2

    2016

  28. [28]

    The Mapillary vistas dataset for semantic understanding of street scenes,

    G. Neuhold, T. Ollmann, S. Rota Bulo, and P. Kontschieder, “The Mapillary vistas dataset for semantic understanding of street scenes,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2017, pp. 4990–4999. 2

    2017

  29. [29]

    A*3D dataset: Towards autonomous driving in challenging environments,

    Q.-H. Pham, P. Sevestre, R. S. Pahwa, H. Zhan, C. H. Pang, Y . Chen, A. Mustafa, V . Chandrasekhar, and J. Lin, “A*3D dataset: Towards autonomous driving in challenging environments,” inProceedings of the IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 2267–2273. 2

    2020

  30. [30]

    BDD100K: A diverse driving dataset for heterogeneous multitask learning,

    F. Yu, H. Chen, X. Wang, W. Xian, Y . Chen, F. Liu, V . Madhavan, and T. Darrell, “BDD100K: A diverse driving dataset for heterogeneous multitask learning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 2636–2645. 3

    2020

  31. [31]

    Scalability in perception for autonomous driving: Waymo open dataset,

    P. Sun, H. Kretzschmar, X. Dotiwalla, A. Chouard, V . Patnaik, P. Tsui, J. Guo, Y . Zhou, Y . Chai, B. Caineet al., “Scalability in perception for autonomous driving: Waymo open dataset,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 2446–2454. 3

    2020

  32. [32]

    The ApolloScape dataset for autonomous driving,

    X. Huang, X. Cheng, Q. Geng, B. Cao, D. Zhou, P. Wang, Y . Lin, and R. Yang, “The ApolloScape dataset for autonomous driving,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshop, 2018, pp. 954–960. 3

    2018

  33. [33]

    Argoverse 2: Next IEEE TRANSACTIONS ON IEEE TRANSACTIONS 16 generation datasets for self-driving perception and forecasting,

    B. Wilson, W. Qi, T. Agarwal, J. Lambert, J. Singh, S. Khandelwal, B. Pan, R. Kumar, A. Hartnett, J. K. Ponteset al., “Argoverse 2: Next IEEE TRANSACTIONS ON IEEE TRANSACTIONS 16 generation datasets for self-driving perception and forecasting,”arXiv preprint arXiv:2301.00493, 2023. 3

    Pith/arXiv arXiv 2023

  34. [34]

    OmniHD-Scenes: A next-generation multimodal dataset for autonomous driving,

    L. Zheng, L. Yang, Q. Lin, W. Ai, M. Liu, S. Lu, J. Liu, H. Ren, J. Mo, X. Baiet al., “OmniHD-Scenes: A next-generation multimodal dataset for autonomous driving,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026. 3

    2026

  35. [35]

    The SYNTHIA dataset: A large collection of synthetic images for semantic segmentation of urban scenes,

    G. Ros, L. Sellart, J. Materzynska, D. Vazquez, and A. M. Lopez, “The SYNTHIA dataset: A large collection of synthetic images for semantic segmentation of urban scenes,” inProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2016, pp. 3234–3243. 3

    2016

  36. [36]

    Playing for data: Ground truth from computer games,

    S. R. Richter, V . Vineet, S. Roth, and V . Koltun, “Playing for data: Ground truth from computer games,” inProcessings of the European Conference on Computer Vision. Springer, 2016, pp. 102–118. 3

    2016

  37. [37]

    Playing for benchmarks,

    S. R. Richter, Z. Hayder, and V . Koltun, “Playing for benchmarks,” in Proceedings of the IEEE International Conference on Computer Vision, 2017, pp. 2213–2222. 3

    2017

  38. [38]

    All- in-One Drive: A comprehensive perception dataset with high-density long-range point clouds,

    X. Weng, Y . Man, J. Park, Y . Yuan, M. O’Toole, and K. M. Kitani, “All- in-One Drive: A comprehensive perception dataset with high-density long-range point clouds,”OpenReview submission, 2023. 3

    2023

  39. [39]

    Collaborative perception datasets for autonomous driving: A review,

    N. Wang, D. Shang, Y . Gong, X. Hu, Z. Song, L. Yang, Y . Huang, X. Wang, and J. Lu, “Collaborative perception datasets for autonomous driving: A review,”IEEE Sensors Journal, 2025. 3

    2025

  40. [40]

    OpenCDA: An open cooperative driving automation framework integrated with co-simulation,

    R. Xu, Y . Guo, X. Han, X. Xia, H. Xiang, and J. Ma, “OpenCDA: An open cooperative driving automation framework integrated with co-simulation,” inProceedings of the IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2021, pp. 1155–

    2021

  41. [41]

    Multivehicle cooperative driving using cooperative perception: Design and experimental validation,

    S.-W. Kim, B. Qin, Z. J. Chong, X. Shen, W. Liu, M. H. Ang, E. Fraz- zoli, and D. Rus, “Multivehicle cooperative driving using cooperative perception: Design and experimental validation,”IEEE Transactions on Intelligent Transportation Systems, vol. 16, no. 2, pp. 663–680, 2014. 3

    2014

  42. [42]

    Cooperative perception in autonomous ground vehicles using a mobile-robot testbed,

    S. Sridhar and A. Eskandarian, “Cooperative perception in autonomous ground vehicles using a mobile-robot testbed,”IET Intelligent Transport Systems, vol. 13, no. 10, pp. 1545–1556, 2019. 3

    2019

  43. [43]

    Toward lightweight, privacy-preserving cooperative object classification for connected au- tonomous vehicles,

    J. Xiong, R. Bi, Y . Tian, X. Liu, and D. Wu, “Toward lightweight, privacy-preserving cooperative object classification for connected au- tonomous vehicles,”IEEE Internet of Things Journal, vol. 9, no. 4, pp. 2787–2801, 2021. 3

    2021

  44. [44]

    Cooper: Cooperative percep- tion for connected autonomous vehicles based on 3D point clouds,

    Q. Chen, S. Tang, Q. Yang, and S. Fu, “Cooper: Cooperative percep- tion for connected autonomous vehicles based on 3D point clouds,” inProceedings of the IEEE International Conference on Distributed Computing Systems (ICDCS). IEEE, 2019, pp. 514–524. 3

    2019

  45. [45]

    Incentivizing point cloud-based accurate cooperative perception for connected vehi- cles,

    Y . Zhang, Z. Fan, J. Hou, N. Chen, F. Lyu, and P. Yang, “Incentivizing point cloud-based accurate cooperative perception for connected vehi- cles,”IEEE Transactions on Vehicular Technology, 2024. 3

    2024

  46. [46]

    F-Cooper: Feature based cooperative perception for autonomous vehicle edge computing system using 3D point clouds,

    Q. Chen, X. Ma, S. Tang, J. Guo, Q. Yang, and S. Fu, “F-Cooper: Feature based cooperative perception for autonomous vehicle edge computing system using 3D point clouds,” inProceedings of the 4th ACM/IEEE Symposium on Edge Computing. IEEE, 2019, pp. 88–100. 3

    2019

  47. [47]

    Machine-learning-enabled cooperative perception for connected autonomous vehicles: Challenges and opportunities,

    Q. Yang, S. Fu, H. Wang, and H. Fang, “Machine-learning-enabled cooperative perception for connected autonomous vehicles: Challenges and opportunities,”IEEE Network, vol. 35, no. 3, pp. 96–101, 2021. 3

    2021

  48. [48]

    Learning distilled collaboration graph for multi-agent perception,

    Y . Li, S. Ren, P. Wu, S. Chen, C. Feng, and W. Zhang, “Learning distilled collaboration graph for multi-agent perception,”Advances in Neural Information Processing Systems, vol. 34, pp. 29 541–29 552, 2021. 3

    2021

  49. [49]

    Collaboration helps camera overtake lidar in 3D detection,

    Y . Hu, Y . Lu, R. Xu, W. Xie, S. Chen, and Y . Wang, “Collaboration helps camera overtake lidar in 3D detection,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2023, pp. 9243–9252. 3

    2023

  50. [50]

    FedBEVT: Federated learning bird’s eye view perception transformer in road traffic systems,

    R. Song, R. Xu, A. Festag, J. Ma, and A. Knoll, “FedBEVT: Federated learning bird’s eye view perception transformer in road traffic systems,” IEEE Transactions on Intelligent Vehicles, pp. 1–12, 2023. 3

    2023

  51. [51]

    HM-ViT: Hetero-modal vehicle-to-vehicle cooperative perception with vision transformer,

    H. Xiang, R. Xu, and J. Ma, “HM-ViT: Hetero-modal vehicle-to-vehicle cooperative perception with vision transformer,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 284–295. 3

    2023

  52. [52]

    V2X-ViT: Vehicle-to-everything cooperative perception with vision transformer,

    R. Xu, H. Xiang, Z. Tu, X. Xia, M.-H. Yang, and J. Ma, “V2X-ViT: Vehicle-to-everything cooperative perception with vision transformer,” inProceedings of the 17th European Conference on Computer Vision (ECCV). Springer, 2022, pp. 107–124. 3

    2022

  53. [53]

    Complementarity-enhanced and redundancy-minimized collaboration network for multi-agent per- ception,

    G. Luo, H. Zhang, Q. Yuan, and J. Li, “Complementarity-enhanced and redundancy-minimized collaboration network for multi-agent per- ception,” inProceedings of the 30th ACM International Conference on Multimedia. ACM, 2022, pp. 3578–3586. 3

    2022

  54. [54]

    Coopernaut: End-to- end driving with cooperative perception for networked vehicles,

    J. Cui, H. Qiu, D. Chen, P. Stone, and Y . Zhu, “Coopernaut: End-to- end driving with cooperative perception for networked vehicles,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2022, pp. 17 252–17 262. 3

    2022

  55. [55]

    Federated vehicular transformers and their federations: Privacy-preserving com- puting and cooperation for autonomous driving,

    Y . Tian, J. Wang, Y . Wang, C. Zhao, F. Yao, and X. Wang, “Federated vehicular transformers and their federations: Privacy-preserving com- puting and cooperation for autonomous driving,”IEEE Transactions on Intelligent Vehicles, vol. 7, no. 3, pp. 456–465, 2022. 3

    2022

  56. [56]

    V2VNet: Vehicle-to-vehicle communication for joint perception and prediction,

    T.-H. Wang, S. Manivasagam, M. Liang, B. Yang, W. Zeng, and R. Ur- tasun, “V2VNet: Vehicle-to-vehicle communication for joint perception and prediction,” inProceedings of the 16th European Conference on Computer Vision (ECCV). Springer, 2020, pp. 605–621. 3

    2020

  57. [57]

    Cooperative perception with learning-based V2V communications,

    C. Liu, Y . Chen, J. Chen, R. Payton, M. Riley, and S.-H. Yang, “Cooperative perception with learning-based V2V communications,” IEEE Wireless Communications Letters, vol. 12, no. 11, pp. 1831–1835,

  58. [58]

    V2XP-ASG: Generating adversarial scenes for vehicle-to-everything perception,

    H. Xiang, R. Xu, X. Xia, Z. Zheng, B. Zhou, and J. Ma, “V2XP-ASG: Generating adversarial scenes for vehicle-to-everything perception,” in Proceedings of the IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 3584–3591. 3

    2023

  59. [59]

    SiCP: Simultaneous individual and cooperative perception for 3D object detection in connected and automated vehicles,

    D. Qu, Q. Chen, T. Bai, A. Qin, H. Lu, H. Fan, S. Fu, and Q. Yang, “SiCP: Simultaneous individual and cooperative perception for 3D object detection in connected and automated vehicles,” inProceedings of the 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 8905–8912. 3

    2024

  60. [60]

    V2V co- operative perception with adaptive communication loss for autonomous driving,

    J. Shi, J. Zhao, L. Zhuo, X. Wang, X. Zhan, and H. Liu, “V2V co- operative perception with adaptive communication loss for autonomous driving,”IEEE Transactions on Intelligent Transportation Systems, 2025. 3

    2025

  61. [61]

    PnPDA+: A meta feature-guided domain adapter for collaborative perception,

    L. Xin, G. Zhou, Z. Yu, D. Wang, T. Luo, X. Fu, and J. Li, “PnPDA+: A meta feature-guided domain adapter for collaborative perception,”World Electric Vehicle Journal, vol. 16, no. 7, p. 343, 2025. 3

    2025

  62. [62]

    Adaptive fusion of lidar features for 3D object detection in autonomous driving,

    M. Wang, D. Li, J. R. Casas, and J. Ruiz-Hidalgo, “Adaptive fusion of lidar features for 3D object detection in autonomous driving,”Sensors, vol. 25, no. 13, p. 3865, 2025. 3

    2025

  63. [63]

    Transformer-based latency compensation for cooperative perception,

    Y .-C. Cheng, H.-J. Lin, and C.-S. Shih, “Transformer-based latency compensation for cooperative perception,” inProceedings of the 2025 IEEE Vehicular Networking Conference (VNC). IEEE, 2025, pp. 1–8. 3

    2025

  64. [64]

    RoCooper: Robust cooperative perception under vehicle-to-vehicle communication impairments,

    T. Tang, C. Zhang, G. Chenet al., “RoCooper: Robust cooperative perception under vehicle-to-vehicle communication impairments,” in Proceedings of the 2025 IEEE Conference on Computer Communica- tions. IEEE, 2025, pp. 1–10. 3

    2025

  65. [65]

    Coop-WD: Cooperative perception with weighting and denoising for robust V2V communication,

    C. Liu, J. Chen, Y . Chen, Y . He, Z. Wei, H. Sun, H. Lu, and Q. Hao, “Coop-WD: Cooperative perception with weighting and denoising for robust V2V communication,”arXiv preprint arXiv:2505.03528, 2025. 3

    Pith/arXiv arXiv 2025

  66. [66]

    Het- erogeneous multiscale cooperative perception for connected autonomous vehicles via V2X interaction,

    Y . Zha, W. Shangguan, J. Chen, L. Chai, W. Qiu, and A. M. L´opez, “Het- erogeneous multiscale cooperative perception for connected autonomous vehicles via V2X interaction,”IEEE Internet of Things Journal, 2025. 3

    2025

  67. [67]

    TraF-Align: Trajectory-aware fea- ture alignment for asynchronous multi-agent perception,

    Z. Song, L. Yang, F. Wen, and J. Li, “TraF-Align: Trajectory-aware fea- ture alignment for asynchronous multi-agent perception,” inProceedings of the Computer Vision and Pattern Recognition Conference (CVPR), 2025, pp. 12 048–12 057. 3

    2025

  68. [68]

    V2X-DG: Domain generalization for vehicle-to-everything cooperative perception,

    B. Li, Z. Xu, J. Li, X. Liu, J. Fang, X. Li, and H. Yu, “V2X-DG: Domain generalization for vehicle-to-everything cooperative perception,”arXiv preprint arXiv:2503.15435, 2025. 3

    Pith/arXiv arXiv 2025

  69. [69]

    SparseAlign: A fully sparse framework for cooperative object detection,

    Y . Yuan, Y . Xia, D. Cremers, and M. Sester, “SparseAlign: A fully sparse framework for cooperative object detection,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 22 296–22 305. 3

    2025

  70. [70]

    V2X-ReaLO: An open online framework and dataset for cooperative perception in reality,

    H. Xiang, Z. Zheng, X. Xia, S. Z. Zhao, L. Gao, Z. Zhou, T. Cai, Y . Zhang, and J. Ma, “V2X-ReaLO: An open online framework and dataset for cooperative perception in reality,”arXiv preprint arXiv:2503.10034, 2025. 3

    Pith/arXiv arXiv 2025

  71. [71]

    CoopDETR: A unified cooperative perception framework for 3D detection via object query,

    Z. Wang, S. Xu, X. Zhuang, T. Xu, Y . Wang, J. Liu, Y . Chen, and Y .-Q. Zhang, “CoopDETR: A unified cooperative perception framework for 3D detection via object query,”arXiv preprint arXiv:2502.19313, 2025. 3

    Pith/arXiv arXiv 2025

  72. [72]

    Bandwidth-efficient communication mod- elling for autonomous vehicle collaborative perception,

    D. Jin, Y . Zeng, and Y . Gong, “Bandwidth-efficient communication mod- elling for autonomous vehicle collaborative perception,” inProceedings of the 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). IEEE, 2025, pp. 6146–6155. 3

    2025

  73. [73]

    LCV2I: Communication-efficient and high-performance collaborative perception framework with low- resolution lidar,

    X. Feng, H. Sun, and H. Zheng, “LCV2I: Communication-efficient and high-performance collaborative perception framework with low- resolution lidar,”arXiv preprint arXiv:2502.17039, 2025. 3

    Pith/arXiv arXiv 2025

  74. [74]

    mmCooper: A multi-agent multi-stage communication-efficient and IEEE TRANSACTIONS ON IEEE TRANSACTIONS 17 collaboration-robust cooperative perception framework,

    B. Liu, J. Teng, H. Xue, E. Wang, C. Zhu, P. Wang, and L. Wu, “mmCooper: A multi-agent multi-stage communication-efficient and IEEE TRANSACTIONS ON IEEE TRANSACTIONS 17 collaboration-robust cooperative perception framework,”arXiv preprint arXiv:2501.12263, 2025. 3

    arXiv 2025

  75. [75]

    A lightweight two-stage multi-vehicle feature fusion method guided by global feature,

    Y . Wang, B. Zhao, S. Chen, D. Tian, C. Hu, and X. Zhang, “A lightweight two-stage multi-vehicle feature fusion method guided by global feature,”IEEE Sensors Journal, 2025. 3

    2025

  76. [76]

    AgentAlign: Misalignment-adapted multi-agent perception for resilient inter-agent sensor correlations,

    Z. Meng, Y . Zhang, Z. Zheng, Z. Zhao, and J. Ma, “AgentAlign: Misalignment-adapted multi-agent perception for resilient inter-agent sensor correlations,”arXiv preprint arXiv:2412.06142, 2024. 3

    Pith/arXiv arXiv 2024

  77. [77]

    Bridging the domain gap for multi-agent perception,

    R. Xu, J. Li, X. Dong, H. Yu, and J. Ma, “Bridging the domain gap for multi-agent perception,” inProceedings of the IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 6035–6042. 3

    2023

  78. [78]

    Soft-NMS – improving object detection with one line of code,

    N. Bodla, B. Singh, R. Chellappa, and L. S. Davis, “Soft-NMS – improving object detection with one line of code,” inProceedings of the IEEE International Conference on Computer Vision. IEEE, 2017, pp. 5561–5569. 3

    2017

  79. [79]

    Model-agnostic multi-agent perception framework,

    R. Xu, W. Chen, H. Xiang, X. Xia, L. Liu, and J. Ma, “Model-agnostic multi-agent perception framework,” inProceedings of the 2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 1471–1478. 3

    2023

  80. [80]

    AICP: Augmented informative cooperative perception,

    P. Zhou, P. Kortoc ¸i, Y .-P. Yau, B. Finley, X. Wang, T. Braud, L.- H. Lee, S. Tarkoma, J. Kangasharju, and P. Hui, “AICP: Augmented informative cooperative perception,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 11, pp. 22 505–22 518, 2022. 3

    2022

Showing first 80 references.