pith. sign in

arxiv: 2606.21165 · v1 · pith:YHLQGVQInew · submitted 2026-06-19 · 💻 cs.RO · cs.AI

OmniV2X: A Generative Foundation Planner for Efficient End-to-End Cooperative Driving

Juntong Peng , Juanwu Lu , Yupeng Zhou , Can Cui , Yaobin Chen , Ziran Wang This is my paper

Pith reviewed 2026-06-26 14:22 UTC · model grok-4.3

classification 💻 cs.RO cs.AI
keywords generative foundation modelcooperative drivingV2Xend-to-end planningtrajectory generationmulti-modal observationscross-attentionfoundation model adaptation
0
0 comments X

The pith

OmniV2X is a generative foundation model for efficient end-to-end cooperative driving that adapts from single-agent pre-training using lightweight V2X tokens.

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

This paper presents OmniV2X as a generative foundation planner for vehicle-to-everything cooperative driving. The model interprets independent context sequences of multi-modal and multi-agent observations directly, avoiding the need to fuse inputs into a shared representation. It is trained end-to-end with a trajectory generation loss that allows the high-capacity planner to learn steering and input leverage through cross-attention. Pre-trained on single-agent datasets, it adapts to cooperative settings with standard-compliant lightweight tokens. On the DAIR-V2X-Seq dataset, it achieves state-of-the-art results while using less than 10 percent of the fine-tuning data and less than 1 percent of the communication bandwidth compared to existing methods.

Core claim

OmniV2X directly interprets independent context sequences comprising multi-modal and multi-agent observations. For training, an end-to-end supervised pipeline uses a downstream trajectory generation loss in which a high-capacity generative sequence planner implicitly learns to steer the model and leverage multi-modal inputs via cross-attention injection. As a foundation model pre-trained on large-scale single-agent planning datasets, it efficiently adapts to cooperative environments by integrating the conditioning context with lightweight, standard-compliant V2X tokens, achieving state-of-the-art performance on the DAIR-V2X-Seq dataset with less than 10% of the fine-tune V2X dataset and less

What carries the argument

High-capacity generative sequence planner that implicitly steers the model via cross-attention injection in an end-to-end supervised pipeline using downstream trajectory generation loss

If this is right

  • Reduces computational cost of dense 3D perception by processing independent sequences.
  • Mitigates vulnerability to data scarcity in cooperative scenarios through foundation model adaptation.
  • Improves compliance with standardized messaging using lightweight V2X tokens.
  • Achieves better performance than existing end-to-end cooperative driving baselines.
  • Requires less than 10% of the fine-tune V2X dataset and less than 1% of the communication bandwidth.

Where Pith is reading between the lines

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

  • The design could enable cooperative driving systems to scale with limited V2X deployment.
  • Pre-training on single-agent data might transfer to other multi-agent domains with similar token integration.
  • Lower bandwidth use could support larger numbers of agents in real-time cooperative planning.
  • End-to-end training might reduce the need for separate perception and planning modules in V2X systems.

Load-bearing premise

The high-capacity generative sequence planner can implicitly learn to steer the model and leverage multi-modal inputs via cross-attention injection to enable efficient adaptation from single-agent pre-training via lightweight V2X tokens.

What would settle it

Demonstrating that performance does not improve over baselines on DAIR-V2X-Seq when fine-tuned with less than 10% of the dataset and less than 1% bandwidth, or that removing the cross-attention injection prevents efficient adaptation.

Figures

Figures reproduced from arXiv: 2606.21165 by Can Cui, Juanwu Lu, Juntong Peng, Yaobin Chen, Yupeng Zhou, Ziran Wang.

Figure 1
Figure 1. Figure 1: Top: Comparison among different paradigm in end￾to-end cooperative driving. Bottom: OmniV2X successfully provides low latency, low communication bandwidth, and a reasonable number of parameters. across sources. Therefore, it requires a disruptive paradigm shift to enable the effective use of multi-modal V2X inputs. To address these limitations, we propose shifting the focus of cooperative driving from rigi… view at source ↗
Figure 2
Figure 2. Figure 2: The OmniV2X architecture. All modalities are independently encoded and normalized, then concatenated into a unified context sequence, which is consumed by a generative planner via computationally efficient cross-attention. The model is first pretrained on a large-scale single-agent dataset and adapted to a smaller-scale V2X dataset. planning engine, optimized for the long-context conditions from different … view at source ↗
Figure 5
Figure 5. Figure 5: Real-world demonstration of V2X cooperative driv [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: Robustness to ego localization error. Thor) without quantization. In Fig. 3b, we demonstrate the flexible inference-time computation uniquely enabled by flow matching. An acceleration to a maximum of 29 FPS can be achieved without sacrificing much planning performance. E. System Robustness & V2X-Augmented Field Testing While open-loop metrics evaluate the method using stan￾dardized evaluation protocols, co… view at source ↗
read the original abstract

We present OmniV2X, a generative foundation model for vehicle-to-everything (V2X) cooperative driving. The model directly interprets independent context sequences comprising multi-modal and multi-agent observations. The new design mitigates the computational cost of dense 3D perception, the vulnerability to data scarcity in cooperative scenarios, and the poor compliance with standardized messaging in existing methods that fuse multi-modal inputs into a shared representation. For training, we present an end-to-end supervised pipeline using a downstream trajectory generation loss, in which a high-capacity generative sequence planner implicitly learns to steer the model and leverage multi-modal inputs via cross-attention injection. As a foundation model, we demonstrate that OmniV2X pre-trained on large-scale single-agent planning datasets can efficiently adapt to cooperative environments by integrating the conditioning context with lightweight, standard-compliant V2X tokens. Evaluated on the DAIR-V2X-Seq dataset, OmniV2X outperforms existing end-to-end cooperative driving baselines, achieving state-of-the-art performance with less than 10% of the fine-tune V2X dataset and less than 1% of the communication bandwidth. We conduct comprehensive evaluations to demonstrate its computational efficiency and robustness under real-world constraints.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript introduces OmniV2X, a generative foundation model for end-to-end V2X cooperative driving. It processes independent multi-modal multi-agent context sequences via cross-attention injection in a high-capacity generative sequence planner trained with a downstream trajectory generation loss. Pre-trained on large-scale single-agent planning data, the model adapts to cooperative scenarios using lightweight standard-compliant V2X tokens. The central claim is that it achieves state-of-the-art performance on the DAIR-V2X-Seq dataset while using less than 10% of the fine-tune V2X dataset and less than 1% of the communication bandwidth, with additional claims of computational efficiency and robustness.

Significance. If the experimental claims hold, the work could be significant for cooperative autonomous driving by demonstrating efficient adaptation of foundation models to multi-agent V2X settings, potentially reducing data and bandwidth requirements compared to dense fusion approaches.

major comments (2)
  1. [Abstract] Abstract: The assertion of outperforming 'existing end-to-end cooperative driving baselines' and achieving 'state-of-the-art performance' with quantified efficiency gains (<10% fine-tune data, <1% bandwidth) is presented without any supporting experimental details, baseline descriptions, metrics, dataset splits, error bars, or ablation results, rendering the central claim impossible to assess from the manuscript.
  2. [Abstract] Abstract: The description of the 'end-to-end supervised pipeline' and 'implicit' learning via cross-attention and trajectory loss lacks any equations, architecture diagrams, or training procedure details that would allow verification of how the high-capacity planner steers multi-modal inputs or enables parameter-efficient adaptation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the review and the opportunity to address these points. The abstract is intentionally concise, but the full manuscript provides the supporting details referenced in the comments. We respond to each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The assertion of outperforming 'existing end-to-end cooperative driving baselines' and achieving 'state-of-the-art performance' with quantified efficiency gains (<10% fine-tune data, <1% bandwidth) is presented without any supporting experimental details, baseline descriptions, metrics, dataset splits, error bars, or ablation results, rendering the central claim impossible to assess from the manuscript.

    Authors: The abstract summarizes key results; the full manuscript contains a complete Experiments section (Section 4) with all requested elements: comparisons against end-to-end cooperative baselines on DAIR-V2X-Seq, metrics (ADE, FDE, collision rate), dataset splits and fine-tuning protocol, error bars from repeated runs, and ablations on data volume and bandwidth usage. These directly support the quantified claims of <10% fine-tune data and <1% bandwidth. The manuscript therefore allows full assessment of the central claim. revision: no

  2. Referee: [Abstract] Abstract: The description of the 'end-to-end supervised pipeline' and 'implicit' learning via cross-attention and trajectory loss lacks any equations, architecture diagrams, or training procedure details that would allow verification of how the high-capacity planner steers multi-modal inputs or enables parameter-efficient adaptation.

    Authors: The abstract gives a high-level summary. The manuscript provides the requested details in the main body: architecture diagram (Figure 2), cross-attention equations (Eq. 4), trajectory generation loss (Eq. 5), end-to-end training procedure (Section 3.2), and pre-training plus lightweight V2X token adaptation (Section 3.3). These explain how the generative planner processes multi-modal inputs and enables efficient adaptation. This organization follows standard practice for abstracts. revision: no

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper describes an empirical ML architecture and training pipeline for a generative planner, with claims resting on dataset evaluation rather than any mathematical derivation chain. No equations, first-principles results, fitted parameters renamed as predictions, or load-bearing self-citations appear in the provided text. The central claims concern model performance on DAIR-V2X-Seq and adaptation efficiency, which are externally falsifiable via standard benchmarks and do not reduce to self-definition or input renaming by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities; assessment is limited to the high-level description given.

pith-pipeline@v0.9.1-grok · 5761 in / 1253 out tokens · 16948 ms · 2026-06-26T14:22:17.483634+00:00 · methodology

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

33 extracted references · 6 linked inside Pith

  1. [1]

    End- to-End Autonomous Driving Through V2X Cooperation,

    H. Yu, W. Yang, J. Zhong, Z. Yang, S. Fan, P. Luo, and Z. Nie, “End- to-End Autonomous Driving Through V2X Cooperation,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 9, pp. 9598–9606, Apr. 2025

    2025

  2. [2]

    UniMM- V2X: MoE-Enhanced Multi-Level Fusion for End-to-End Cooperative Autonomous Driving,

    Z. Song, C. Xia, C. Wang, H. Yu, S. Zhou, and Z. Niu, “UniMM- V2X: MoE-Enhanced Multi-Level Fusion for End-to-End Cooperative Autonomous Driving,” Nov. 2025, arXiv:2511.09013 [cs]

    arXiv 2025

  3. [3]

    nuscenes: A multimodal 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 multimodal dataset for autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2020, pp. 11 621–11 631

    2020

  4. [4]

    NA VSIM: Data-Driven Non-Reactive Autonomous Vehicle Simula- tion and Benchmarking,

    D. Dauner, M. Hallgarten, T. Li, X. Weng, Z. Huang, Z. Yang, H. Li, I. Gilitschenski, B. Ivanovic, M. Pavone, A. Geiger, and K. Chitta, “NA VSIM: Data-Driven Non-Reactive Autonomous Vehicle Simula- tion and Benchmarking,”Advances in Neural Information Processing Systems, vol. 37, pp. 28 706–28 719, Dec. 2024

    2024

  5. [5]

    V2x-seq: A large- scale sequential dataset for vehicle-infrastructure cooperative percep- tion and forecasting,

    H. Yu, W. Yang, H. Ruan, Z. Yang, Y . Tang, X. Gao, X. Hao, Y . Shi, Y . Pan, N. Sun, J. Song, J. Yuan, P. Luo, and Z. Nie, “V2x-seq: A large- scale sequential dataset for vehicle-infrastructure cooperative percep- tion and forecasting,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 5486–5495

    2023

  6. [6]

    Planning-oriented autonomous driving,

    Y . Hu, J. Yang, L. Chen, K. Li, C. Sima, X. Zhu, S. Chai, S. Du, T. Lin, W. Wang, L. Lu, X. Jia, Q. Liu, J. Dai, Y . Qiao, and H. Li, “Planning-oriented autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 17 853–17 862

    2023

  7. [7]

    Vad: Vectorized scene representation for efficient autonomous driving,

    B. Jiang, S. Chen, Q. Xu, B. Liao, J. Chen, H. Zhou, Q. Zhang, W. Liu, C. Huang, and X. Wang, “Vad: Vectorized scene representation for efficient autonomous driving,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October 2023, pp. 8340–8350

    2023

  8. [8]

    Hydra-MDP: End-to- end Multimodal Planning with Multi-target Hydra-Distillation,

    Z. Li, K. Li, S. Wang, S. Lan, Z. Yu, Y . Ji, Z. Li, Z. Zhu, J. Kautz, Z. Wu, Y .-G. Jiang, and J. M. Alvarez, “Hydra-MDP: End-to- end Multimodal Planning with Multi-target Hydra-Distillation,” Aug. 2024, arXiv:2406.06978 [cs]

    Pith/arXiv arXiv 2024

  9. [9]

    SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation,

    W. Sun, X. Lin, Y . Shi, C. Zhang, H. Wu, and S. Zheng, “SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation,” in2025 IEEE International Conference on Robotics and Automation (ICRA), May 2025, pp. 8795–8801

    2025

  10. [10]

    GenAD: Genera- tive End-to-End Autonomous Driving,

    W. Zheng, R. Song, X. Guo, C. Zhang, and L. Chen, “GenAD: Genera- tive End-to-End Autonomous Driving,” Apr. 2024, arXiv:2402.11502

    arXiv 2024

  11. [11]

    Diffusiondrive: Truncated diffu- sion model for end-to-end autonomous driving,

    B. Liao, S. Chen, H. Yin, B. Jiang, C. Wang, S. Yan, X. Zhang, X. Li, Y . Zhang, Q. Zhang, and X. Wang, “Diffusiondrive: Truncated diffu- sion model for end-to-end autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2025, pp. 12 037–12 047

    2025

  12. [12]

    DriveLM: Driving with Graph Visual Question Answering,

    C. Sima, K. Renz, K. Chitta, L. Chen, H. Zhang, C. Xie, J. Beißwenger, P. Luo, A. Geiger, and H. Li, “DriveLM: Driving with Graph Visual Question Answering,” Jul. 2024, arXiv:2312.14150

    arXiv 2024

  13. [13]

    DriveCoT: Integrating Chain-of-Thought Reasoning with End-to-End Driving,

    T. Wang, E. Xie, R. Chu, Z. Li, and P. Luo, “DriveCoT: Integrating Chain-of-Thought Reasoning with End-to-End Driving,” Mar. 2024, arXiv:2403.16996

    arXiv 2024

  14. [14]

    EMMA: End-to-End Multimodal Model for Autonomous Driving,

    J.-J. Hwang, R. Xu, H. Lin, W.-C. Hung, J. Ji, K. Choi, D. Huang, T. He, P. Covington, B. Sapp, J. Guo, D. Anguelov, and M. Tan, “EMMA: End-to-End Multimodal Model for Autonomous Driving,” Oct. 2024, arXiv:2410.23262 [cs] version: 1

    Pith/arXiv arXiv 2024

  15. [15]

    Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail,

    NVIDIA, Y . Wang, W. Luo, J. Bai, Y . Cao, T. Che, K. Chen, Y . Chen, J. Diamond, Y . Ding, W. Ding, L. Feng, G. Heinrich, J. Huang, P. Karkus, B. Li, P. Li, T.-Y . Lin, D. Liu, M.-Y . Liu, L. Liu, Z. Liu, J. Lu, Y . Mao, P. Molchanov, L. Pavao, Z. Peng, M. Ranzinger, E. Schmerling, S. Shen, Y . Shi, S. Tariq, R. Tian, T. Wekel, X. Weng, T. Xiao, E. Yang,...

    Pith/arXiv arXiv 2026

  16. [16]

    ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving,

    Y . Li, K. Xiong, X. Guo, F. Li, S. Yan, G. Xu, L. Zhou, L. Chen, H. Sun, B. Wang, K. Ma, G. Chen, H. Ye, W. Liu, and X. Wang, “ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving,” Sep. 2025, arXiv:2506.08052 [cs]

    Pith/arXiv arXiv 2025

  17. [17]

    ViLaD: A Large Vision Language Diffusion Framework for End-to-End Autonomous Driving,

    C. Cui, Y . Zhou, J. Peng, S.-Y . Park, Z. Yang, P. Sankaranarayanan, J. Zhang, R. Zhang, and Z. Wang, “ViLaD: A Large Vision Language Diffusion Framework for End-to-End Autonomous Driving,” Aug. 2025, arXiv:2508.12603 [cs]

    arXiv 2025

  18. [18]

    Accelerating Structured Chain- of-Thought in Autonomous Vehicles,

    Y . Gu, Y . Wang, Y . Chen, Y . You, W. Luo, Y . Wang, W. Ding, B. Li, H. Yang, B. Ivanovic, and M. Pavone, “Accelerating Structured Chain- of-Thought in Autonomous Vehicles,” Feb. 2026, arXiv:2602.02864 [cs]

    arXiv 2026

  19. [19]

    V2VNet: Vehicle-to-Vehicle Communication for Joint Perception and Prediction,

    T.-H. Wang, S. Manivasagam, M. Liang, B. Yang, W. Zeng, and R. Urtasun, “V2VNet: Vehicle-to-Vehicle Communication for Joint Perception and Prediction,” inComputer Vision – ECCV 2020, A. Vedaldi, H. Bischof, T. Brox, and J.-M. Frahm, Eds. Cham: Springer International Publishing, 2020, vol. 12347, pp. 605–621, series Title: Lecture Notes in Computer Science

    2020

  20. [20]

    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 (CVPR), June 2022, pp. 17 252–17 262

    2022

  21. [21]

    CoBEVT: Cooperative Bird’s Eye View Semantic Segmentation with Sparse Transformers,

    R. Xu, Z. Tu, H. Xiang, W. Shao, B. Zhou, and J. Ma, “CoBEVT: Cooperative Bird’s Eye View Semantic Segmentation with Sparse Transformers,” inProceedings of The 6th Conference on Robot Learn- ing. PMLR, Mar. 2023, pp. 989–1000

    2023

  22. [22]

    V2X-VLM: End-to-End V2X Cooperative Au- tonomous Driving Through Large Vision-Language Models,

    J. You, H. Shi, Z. Jiang, Z. Huang, R. Gan, K. Wu, X. Cheng, X. Li, and B. Ran, “V2X-VLM: End-to-End V2X Cooperative Au- tonomous Driving Through Large Vision-Language Models,” Jun. 2025, arXiv:2408.09251 [cs]

    arXiv 2025

  23. [23]

    LangCoop: Collaborative Driving with Language,

    X. Gao, Y . Wu, R. Wang, C. Liu, Y . Zhou, and Z. Tu, “LangCoop: Collaborative Driving with Language,” Apr. 2025, arXiv:2504.13406 [cs]

    arXiv 2025

  24. [24]

    CoLMDriver: LLM- based Negotiation Benefits Cooperative Autonomous Driving,

    C. Liu, G. Liu, Z. Wang, J. Yang, and S. Chen, “CoLMDriver: LLM- based Negotiation Benefits Cooperative Autonomous Driving,” Mar. 2025, arXiv:2503.08683 [cs]

    arXiv 2025

  25. [25]

    M3CAD: Towards Generic Cooperative Autonomous Driving Bench- mark,

    M. Zhu, Y . Zhu, Y . Zhu, Q. Chen, D. Qu, S. Fu, and Q. Yang, “M3CAD: Towards Generic Cooperative Autonomous Driving Bench- mark,” May 2025, arXiv:2505.06746 [cs]

    arXiv 2025

  26. [26]

    From Features to Reference Points: Lightweight and Adaptive Fusion for Cooperative Autonomous Driving,

    Y . Zhu, M. Zhu, Q. Chen, D. Qu, I. Luo, S. Fu, and Q. Yang, “From Features to Reference Points: Lightweight and Adaptive Fusion for Cooperative Autonomous Driving,” Jan. 2026, arXiv:2511.18757 [cs]

    arXiv 2026

  27. [27]

    V2X Sensor-Sharing for Cooperative and Auto- mated Driving,

    SAE International, “V2X Sensor-Sharing for Cooperative and Auto- mated Driving,” SAE International, Warrendale, PA, USA, Standard J3224 202208, Aug. 2022

    2022

  28. [28]

    V2X Communications Message Set Dictionary,

    ——, “V2X Communications Message Set Dictionary,” SAE Interna- tional, Warrendale, PA, USA, Standard J2735 202409, Sep. 2024

    2024

  29. [29]

    RoFormer: Enhanced transformer with Rotary Position Embedding,

    J. Su, M. Ahmed, Y . Lu, S. Pan, W. Bo, and Y . Liu, “RoFormer: Enhanced transformer with Rotary Position Embedding,”Neurocom- puting, vol. 568, p. 127063, Feb. 2024

    2024

  30. [30]

    NuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles,

    H. Caesar, J. Kabzan, K. S. Tan, W. K. Fong, E. Wolff, A. Lang, L. Fletcher, O. Beijbom, and S. Omari, “NuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles,” Feb. 2022, arXiv:2106.11810 [cs]

    Pith/arXiv arXiv 2022

  31. [31]

    Sim ´eoni, H

    O. Sim ´eoni, H. V . V o, M. Seitzer, F. Baldassarre, M. Oquab, C. Jose, V . Khalidov, M. Szafraniec, S. Yi, M. Ramamonjisoa, F. Massa, D. Haziza, L. Wehrstedt, J. Wang, T. Darcet, T. Moutakanni, L. Sen- tana, C. Roberts, A. Vedaldi, J. Tolan, J. Brandt, C. Couprie, J. Mairal, H. J ´egou, P. Labatut, and P. Bojanowski, “DINOv3,” Aug. 2025, arXiv:2508.10104 [cs]

    Pith/arXiv arXiv 2025

  32. [32]

    Center-based 3D Object De- tection and Tracking,

    T. Yin, X. Zhou, and P. Krahenbuhl, “Center-based 3D Object De- tection and Tracking,” in2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, TN, USA: IEEE, Jun. 2021, pp. 11 779–11 788

    2021

  33. [33]

    BEVFormer: Learning Bird’s-Eye-View Representation From LiDAR-Camera via Spatiotemporal Transformers,

    Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Q. Yu, and J. Dai, “BEVFormer: Learning Bird’s-Eye-View Representation From LiDAR-Camera via Spatiotemporal Transformers,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 3, pp. 2020– 2036, Mar. 2025

    2020