Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

SimBEV: A Synthetic Multi-Task Multi-Sensor Driving Data Generation Tool and Dataset

T0 review · 4 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read SimBEV generates bird's-eye-view labels for every object class by merging a top-down camera with one placed under the road.

desk verdict Useful synthetic dataset tool with a clever BEV annotation trick; the 'accurate' claim is under-validated but the resource and code are worth engaging. read the letter →

arxiv 2502.01894 v2 pith:4GL2FSVU submitted 2025-02-04 cs.CV cs.LGcs.RO

Goodarz Mehr , Azim Eskandarian This is my paper · ORCID
classification cs.CVcs.LGcs.RO
keywords bird's-eye-viewsegmentationsyntheticdrivingdatasetautonomousBEVgroundtruthmulti-sensorfusiondomainrandomization3Dobjectdetectionsimulator
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

SimBEV is a synthetic data generation tool, built on an open urban driving simulator, whose central claim is that it can produce accurate multi-class bird's-eye-view (BEV) segmentation ground truth over a 144 m by 144 m area centered on the ego vehicle. The paper's key move is to pair a semantic-segmentation camera placed one kilometer above the vehicle with a second camera placed one kilometer below it, exploiting the fact that the simulator's ground materials are visible from only one direction, so the lower camera sees objects that the overhead view would miss. The tool also outputs 3D bounding boxes and supports a wide array of sensors, and the authors use it to release a 320-scene, 102,400-frame dataset spanning all of the simulator's maps. If the ground-truth procedure is sound, SimBEV fills a gap no existing public dataset covers: full BEV segmentation for both static and dynamic classes, which would let researchers train and benchmark BEV perception models without hand annotation.

What carries the argument

The load-bearing mechanism is the one-way visibility of ground materials in the simulator: roads and sidewalks render as solid from above but as see-through from below. A semantic-segmentation camera placed 1 km beneath the ego vehicle therefore collects labels for objects hidden from an overhead camera by vegetation, bridges, and other structures, while a second camera 1 km above the ego vehicle captures the unobstructed top-down view; the two are merged with a logical OR to form the non-road BEV labels. Road labels are computed from simulator waypoints placed a small distance apart and their lane widths, combined with the overhead camera and binary closing to fill gaps. When the ego vehicle is near roads with large elevation differences, such as under an overpass, the method falls back to 3D bounding boxes and waypoints, which the authors state is not as accurate but still acceptable.

What would settle it

Render a fixed scene with a known object layout, then compare SimBEV's merged BEV labels against a direct non-occluded top-down ground-truth render of the same layout; any object located under a bridge, tree canopy, or elevated road that is absent from the merged labels would falsify the accuracy claim. A complementary check is to inspect the simulator's ground assets across all maps for any mesh material that is not one-way visible.

Watch

Extended reading notes

Core claim

The paper's central claim is that SimBEV incorporates information from multiple independent sources to capture accurate BEV ground truth. Road labels come from densely spaced simulator waypoints and lane widths, refined by merging with an overhead camera and binary closing; non-road labels for car, truck, bus, motorcycle, bicycle, rider, and pedestrian come from logically OR-ing the overhead camera with an underground camera that sees through one-way-visible ground materials. This makes SimBEV, to the authors' knowledge, the first public single-vehicle driving dataset with full multi-class BEV segmentation support. The authors further claim that the resulting data are useful for benchmarking BEV perception models, and they report that fused camera-lidar models substantially outperform camera-only models on both BEV segmentation and 3D object detection.

Load-bearing premise

The non-road BEV labels are only trustworthy if the simulator's ground surfaces remain one-way see-through in every map, weather condition, and custom asset; if any ground element is two-sided or opaque, the underground camera will silently miss the objects hidden behind it.

Editorial extensions

If this is right

  • If the ground-truth procedure is sound, SimBEV becomes the first public single-vehicle driving dataset that fully supports multi-class BEV segmentation, giving researchers a benchmark for BEV perception without hand annotation.
  • The 144 m by 144 m BEV grid is larger than the typical 100 m by 100 m area, supporting data augmentations such as rotation, translation, and scaling during training.
  • Because scene parameters such as weather, traffic density, lights, jaywalking, and reckless driving are randomized, the dataset offers diverse and occasionally adversarial scenarios for testing perception and domain adaptation.
  • The benchmark results on two recent multi-sensor models show that lidar-only and fused camera-lidar variants far outperform camera-only variants, consistent with the importance of explicit geometric information for BEV localization.
  • The tool can regenerate ground truth for custom maps and sensor configurations, so researchers can create new annotated datasets without manual labeling.

Reading between the lines

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

  • Editorial inference: the one-way-material technique could transfer to any simulator with one-sided ground meshes, and its accuracy could be tested directly by comparing the merged BEV map against an independent voxel-based ground-truth render of the same scene.
  • Editorial inference: because the overpass fallback is admitted to be less accurate, the dataset's label quality is not uniform; benchmark users may want to identify and analyze those frames separately rather than averaging them into overall scores.
  • Editorial inference: the road labels inherit whatever errors exist in the simulator's waypoint and lane-width data, so claimed accuracy should be quantified per class and per map rather than taken as a single global property.
  • Editorial inference: with 20 Hz capture, dense 128-channel lidar, and a large number of pedestrians, the dataset may also serve as a testbed for velocity prediction and motion forecasting, not only static perception.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper introduces SimBEV, a CARLA-based synthetic data generation tool and an associated dataset for multi-task, multi-sensor autonomous driving perception, with emphasis on BEV segmentation and 3D object detection. The tool randomizes weather, traffic, and scene parameters, supports a wide array of sensors, and produces BEV ground truth for eight classes over a 144 m by 144 m ego-centered grid via a claimed novel procedure: overhead and underground semantic segmentation cameras exploiting CARLA's one-way visible ground materials, merged with waypoint-derived road labels, plus a fallback using 3D bounding boxes under elevated roads. The authors create a 320-scene, 102,400-frame dataset and benchmark BEVFusion and UniTR on BEV segmentation and 3D object detection.

Significance. If the BEV ground-truth accuracy claim holds, SimBEV fills a real gap: it would be the first public driving dataset with full multi-class BEV segmentation support over a large ego-centered area, and the tool would let researchers generate custom datasets with a wide range of sensors and randomized scenarios. The paper is strong in its engineering contributions: the tool is open-source, the dataset is large-scale, the configuration and sensor details are extensively documented in the supplementary material, and the benchmarking includes several model variants and both IoU-based and distance-based 3D detection metrics. However, the load-bearing claim of accurate BEV ground truth is currently asserted rather than demonstrated; the paper needs quantitative validation of the label generation procedure before the dataset can be fully trusted as a benchmark.

major comments (4)
  1. [Section 3.4, 'BEV ground truth'] The central claim that SimBEV 'capture[s] accurate BEV ground truth' is not quantitatively supported. The paper provides no comparison of the generated labels against an independent reference (e.g., dense semantic lidar projected to BEV, or manually inspected frames), no per-class precision/recall or error metrics, and no ablation of the OR merge against each camera alone. Because this claim motivates the dataset's use for training and benchmarking, the authors should add a validation study over a stratified sample of frames spanning maps, weather conditions, and distances, and report per-class agreement.
  2. [Section 3.4 and Fig. 4] The dual-camera scheme rests on the assumption that CARLA's ground elements use one-way visible materials consistently across all maps, weather conditions, and the custom content library. This property is asserted but never tested. Supplementary Material A notes that some features 'may not work properly' with standard CARLA, which is exactly a setting in which material behavior could differ. The authors should either verify the property empirically across the relevant conditions or explicitly document the assumption and its failure modes as a limitation.
  3. [Section 3.4, fallback paragraph] The fallback used under overpasses replaces the camera-based BEV labels with 3D bounding boxes and is admitted to be 'not as accurate,' but the paper does not report how many frames use the fallback, which maps and scenes are affected, or the magnitude of the resulting label errors. Since the test set includes maps with non-zero elevation (Supp. B.1), the fallback could affect a non-negligible fraction of the benchmark; the authors should quantify the fallback's usage and validate its accuracy separately.
  4. [Section 5.2, Table 3] The statement that road IoU values 'are consistent with [33]' is not an external validation of ground-truth accuracy, because nuScenes road annotations are limited to static map elements and the class definitions and evaluation protocols differ across datasets. Absolute IoU comparability conflates model performance, label semantics, and ground-truth quality; the authors should remove this claim or replace it with a direct comparison of label geometry, not model scores.
minor comments (6)
  1. [Section 3.4] 'we do not discussed them here' should be 'we do not discuss them here.'
  2. [Figure 5 caption and Section 3.4] 'binray closing' should be 'binary closing.'
  3. [Table 10 header] 'BEV labes' should be 'BEV labels.'
  4. [Abstract and repository link] The paper states the dataset is 'open and available to the public' but gives only a code preview link; please provide a direct link or access procedure for the dataset itself.
  5. [Section 5.2, discussion of Table 3] The phrase 'compared to smaller ones ( motorcycle, bicycle, rider, and pedestrian)' contains an awkward extra space before 'motorcycle' and could be rewritten for readability.
  6. [Table 1] Table 1 lists the dataset year as 2024 while the arXiv version is dated 2025; please make the year consistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SimBEV's ground-truth generation is a constructive pipeline with no fitted parameters, no load-bearing self-citation chain, and no prediction that reduces to its inputs by definition.

full rationale

SimBEV is a dataset and tool paper; its central output is BEV labels produced by a deterministic pipeline: an overhead semantic segmentation camera 1 km above and an underground camera 1 km below the ego vehicle, merged via logical OR, plus CARLA waypoint-derived road labels with binary closing (Sec. 3.4, Figs. 4-5). No equation in the paper is fitted to a target, and no result is derived from a parameter estimated on the same quantity it claims to predict. The phrase 'accurate BEV ground truth' is asserted from the tool's design rather than independently validated, but asserted accuracy without external validation is a correctness or evaluation concern, not circularity: the stated grounds are CARLA's one-way-visible ground materials and camera geometry, not the BEV labels themselves. The only author self-citation identified is [11] (Eskandarian et al., IEEE TITS 2019), which supports a general introductory statement about autonomous driving and is not load-bearing for any derivation. The road-class IoU consistency with nuScenes and the BEVFusion/UniTR benchmarks are external anchors, and the models are evaluated against the dataset rather than used to define it. The overpass fallback is admitted to be 'not as accurate' (Sec. 3.4), which is a limitation and potential correctness risk, but it does not reduce the final ground truth to its own input by construction. Therefore no circular step meets the required quoted-equation or fitted-input reduction test.

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

The central contribution is a tool and dataset, not a mathematical derivation, so the ledger lists design parameters still part of the BEV accuracy claim and the simulator assumptions on which the labels depend.

free parameters (3)
  • BEV grid resolution and cell size (l=360, d=0.4 m) = 360 x 360 grid, 0.4 m cells
    Hand-chosen defaults that determine the 144 m x 144 m coverage and annotation granularity; they affect the ground truth quality but are configurable.
  • Overhead/underground camera height (1 km) = 1 km
    Chosen as 'far enough to minimize perspective distortion'; it is a design parameter that affects accuracy.
  • Elevation difference threshold for fallback method (6.4 m) and distance threshold (48.0 m) = 6.4 m / 48.0 m
    Hand-picked constants that trigger the switch to the less accurate fallback BEV method; different thresholds change where accuracy degrades.
assumptions (3)
  • domain assumption Ground elements in CARLA use one-way visible materials consistently across maps and content
    The non-road BEV ground truth is obtained by a camera below the ground looking up; if the material property is inconsistent, labels would be wrong. Invoked in Sec. 3.4 and Fig. 4.
  • domain assumption CARLA-generated waypoints and lane widths provide accurate road geometry
    Road class is computed from waypoint distances and lane widths, patched with overhead camera; accuracy depends on CARLA's road model. See Sec. 3.4.
  • domain assumption CARLA's 3D bounding boxes for actors are ground-truth perfect
    All object labels and BEV object labels derive from simulator state; no validation against real sensor noise is provided. See Sec. 3.4.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SimBEV: A Synthetic Multi-Task Multi-Sensor Driving Data Generation Tool and Dataset." pith.science (2026). https://pith.science/paper/4GL2FSVU

@misc{pith2026250201894,
  author       = {Pith},
  title        = {Pith review of: SimBEV: A Synthetic Multi-Task Multi-Sensor Driving Data Generation Tool and Dataset},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4GL2FSVU}},
  note         = {Machine review of arXiv:2502.01894}
}
read the original abstract

Bird's-eye view (BEV) perception has garnered significant attention in autonomous driving in recent years, in part because BEV representation facilitates multi-modal sensor fusion. BEV representation enables a variety of perception tasks including BEV segmentation, a concise view of the environment useful for planning a vehicle's trajectory. However, this representation is not fully supported by existing datasets, and creation of new datasets for this purpose can be a time-consuming endeavor. To address this challenge, we introduce SimBEV. SimBEV is a randomized synthetic data generation tool that is extensively configurable and scalable, supports a wide array of sensors, incorporates information from multiple sources to capture accurate BEV ground truth, and enables a variety of perception tasks including BEV segmentation and 3D object detection. SimBEV is used to create the SimBEV dataset, a large collection of annotated perception data from diverse driving scenarios. SimBEV and the SimBEV dataset are open and available to the public.

Figures

Figures reproduced from arXiv: 2502.01894 by the authors.

Figure 1
Figure 1. A data sample generated by SimBEV. The left half depicts a 360-degree view of the ego vehicle’s surroundings using different [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. SimBEV’s logic flow when creating a new dataset. The arrow exiting green nodes at the top indicates the action taken when the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Ground elements (roads, sidewalks, etc.) in CARLA use [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (12 more)
Figure 5
Figure 5. Figure 5: Left: BEV road data calculated using CARLA-generated waypoints; there are clear gaps where lanes diverge. Middle: BEV road [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Front camera image samples from the SimBEV dataset. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: From left to right, the Bugatti Chiron, Ford Mustang, [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: In contrast to CARLA’s segmented light design ap [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 10
Figure 10. Figure 10: Coordinate frames of the perception sensors used in [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]
Figure 11
Figure 11. Figure 11: Distribution of weather across SimBEV dataset scenes. [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]
Figure 12
Figure 12. Figure 12: Distribution of the number of spawned vehicles (cars, [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 13
Figure 13. Figure 13: Breakdown of the number of valid 3D object bounding boxes per frame by class across the SimBEV dataset [PITH_FULL_IMAGE:figures/full_fig_p012_13.png]
Figure 14
Figure 14. Figure 14: Distribution of the distance of valid objects from the ego vehicle across the SimBEV dataset. [0, 0.5) [0.5, 1) [1, 1.5) [1.5, 2) [2, 2.5) [2.5, 3) [3, 3.5) [3.5, 4) [4, 4.5) [4.5, 5) [5, 5.5) [5.5, 6) [6, 6.5) [6.5, 7) [7, 7.5) [7.5, 8) 8+ 102 103 104 105 106 Number …
Figure 16
Figure 16. Figure 16: Distribution of the number of lidar (top) and radar (bot [PITH_FULL_IMAGE:figures/full_fig_p013_16.png]
Figure 15
Figure 15. Figure 15: Breakdown of the speed of valid objects across the SimBEV dataset. 100 101 102 103 104 105 Number of points Car Pedestrian Truck Bus Motorcycle Bicycle 0 20 40 60 80 100 120 100 101 102 103 104 Distance (m) Number of points Car Pedestrian Truck Bus Motorcycle Bicycle …
Figure 17
Figure 17. Figure 17: Logarithmic BEV heat maps of the SimBEV dataset for different classes. [PITH_FULL_IMAGE:figures/full_fig_p014_17.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

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

    cs.CV 2026-07 conditional novelty 6.0 of 10

    SimBEV2X delivers a large-scale synthetic V2X dataset and generator with multi-task annotations, plus an attention fusion model that improves fused/lidar performance.

Reference graph

Works this paper leans on

72 extracted references · 66 canonical work pages · cited by 1 Pith paper

  1. [33]

    BEVFusion: Multi- task multi-sensor fusion with unified bird’s-eye view repre- sentation

    Zhijian Liu, Haotian Tang, Alexander Amini, Xinyu Yang, Huizi Mao, Daniela Rus, and Song Han. BEVFusion: Multi- task multi-sensor fusion with unified bird’s-eye view repre- sentation. In Proceedings of the IEEE International Confer- ence on Robotics and Automation (ICRA), pages 2774–2781. IEEE, 2023. 1, 7, 6

    work page 2023

  2. [1]

    TransFusion: Robust lidar-camera fusion for 3D object detection with transform- ers

    Xuyang Bai, Zeyu Hu, Xinge Zhu, Qingqiu Huang, Yilun Chen, Hongbo Fu, and Chiew-Lan Tai. TransFusion: Robust lidar-camera fusion for 3D object detection with transform- ers. In CVPR, pages 1090–1099, 2022. 7

    work page 2022

  3. [2]

    Seeing through fog without seeing fog: Deep multimodal sensor fusion in unseen adverse weather

    Mario Bijelic, Tobias Gruber, Fahim Mannan, Florian Kraus, Werner Ritter, Klaus Dietmayer, and Felix Heide. Seeing through fog without seeing fog: Deep multimodal sensor fusion in unseen adverse weather. In CVPR, pages 11682– 11692, 2020. 1

    work page 2020

  4. [3]

    nuScenes: A mul- timodal dataset for autonomous driving

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuScenes: A mul- timodal dataset for autonomous driving. In CVPR, pages 11621–11631. IEEE, 2020. 1, 3, 4, 5, 6, 7, 8

    work page 2020

  5. [4]

    BEVFusion4D: Learn- ing lidar-camera fusion under bird’s-eye view via cross- modality guidance and temporal aggregation

    Hongxiang Cai, Zeyuan Zhang, Zhenyu Zhou, Ziyin Li, Wenbo Ding, and Jiuhua Zhao. BEVFusion4D: Learn- ing lidar-camera fusion under bird’s-eye view via cross- modality guidance and temporal aggregation. arXiv preprint arXiv:2303.17099, 2023. 1

    arXiv 2023

  6. [5]

    Argoverse: 3D tracking and forecasting with rich maps

    Ming-Fang Chang, John Lambert, Patsorn Sangkloy, Jag- jeet Singh, Slawomir Bak, Andrew Hartnett, De Wang, Peter Carr, Simon Lucey, Deva Ramanan, et al. Argoverse: 3D tracking and forecasting with rich maps. In CVPR, pages 8748–8757, 2019. 3, 6

    work page 2019

  7. [6]

    TransFuser: Imitation with transformer-based sensor fusion for autonomous driv- ing

    Kashyap Chitta, Aditya Prakash, Bernhard Jaeger, Zehao Yu, Katrin Renz, and Andreas Geiger. TransFuser: Imitation with transformer-based sensor fusion for autonomous driv- ing. IEEE TPAMI, 45(11):12878–12895, 2023. 1

    work page 2023

  8. [7]

    The CityScapes dataset for semantic urban scene understanding

    Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The CityScapes dataset for semantic urban scene understanding. In CVPR, pages 3213–3223, 2016. 3, 7 Class Model Modality mAP mATE mAOE mASE mA VE (%) ↑ (m) ↓ (rad) ↓ ↓ (m/s) ↓ Car BEVFusion-C C 23.3 0.824 0.896 0.217...

    work page 2016

Show all 72 references
  1. [8]

    A survey on multimodal large lan- guage models for autonomous driving

    Can Cui, Yunsheng Ma, Xu Cao, Wenqian Ye, Yang Zhou, Kaizhao Liang, Jintai Chen, Juanwu Lu, Zichong Yang, Kuei-Da Liao, et al. A survey on multimodal large lan- guage models for autonomous driving. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vi...

    2024

  2. [9]

    CARLA: An open urban driv- ing simulator

    Alexey Dosovitskiy, German Ros, Felipe Codevilla, Antonio Lopez, and Vladlen Koltun. CARLA: An open urban driv- ing simulator. In Proceedings of the Conference on Robot Learning, pages 1–16. PMLR, 2017. 2, 3, 1

    2017

  3. [10]

    ViT-BEVSeg: A hierarchi- cal transformer network for monocular bird’s-eye view seg- mentation

    Pramit Dutta, Ganesh Sistu, Senthil Yogamani, Edgar Galv´an, and John McDonald. ViT-BEVSeg: A hierarchi- cal transformer network for monocular bird’s-eye view seg- mentation. In Proceedings of the 2022 International Joint Conference on Neural Networks (IJCNN), pages 1–7. IEEE,

    2022

  4. [11]

    Re- search advances and challenges of autonomous and con- nected ground vehicles

    Azim Eskandarian, Chaoxian Wu, and Chuanyang Sun. Re- search advances and challenges of autonomous and con- nected ground vehicles. IEEE Transactions on Intelligent Transportation Systems, 22(2):683–711, 2019. 1

    2019

  5. [12]

    Vision meets robotics: The KITTI dataset

    Andreas Geiger, Philip Lenz, Christoph Stiller, and Raquel Urtasun. Vision meets robotics: The KITTI dataset. The International Journal of Robotics Research , 32(11):1231– 1237, 2013. 3, 7

    2013

  6. [13]

    GitNet: Geometric prior- based transformation for bird’s-eye view segmentation

    Shi Gong, Xiaoqing Ye, Xiao Tan, Jingdong Wang, Errui Ding, Yu Zhou, and Xiang Bai. GitNet: Geometric prior- based transformation for bird’s-eye view segmentation. In ECCV, pages 396–411. Springer, 2022. 1

    2022

  7. [14]

    Lift-Attend-Splat: Bird’s-eye view camera-lidar fusion using transformers

    James Gunn, Zygmunt Lenyk, Anuj Sharma, Andrea Donati, Alexandru Buburuzan, John Redford, and Romain Mueller. Lift-Attend-Splat: Bird’s-eye view camera-lidar fusion using transformers. In CVPR, pages 4526–4536, 2024. 1

    2024

  8. [15]

    Is it safe to drive? An overview of factors, metrics, and datasets for driveability assessment in autonomous driving

    Junyao Guo, Unmesh Kurup, and Mohak Shah. Is it safe to drive? An overview of factors, metrics, and datasets for driveability assessment in autonomous driving. IEEE Trans- actions on Intelligent Transportation Systems , 21(8):3135– 3151, 2019. 1

    2019

  9. [16]

    One thousand and one hours: Self-driving motion prediction dataset

    John Houston, Guido Zuidhof, Luca Bergamini, Yawei Ye, Long Chen, Ashesh Jain, Sammy Omari, Vladimir Iglovikov, and Peter Ondruska. One thousand and one hours: Self-driving motion prediction dataset. In Conference on Robot Learning, pages 409–418. PMLR, 2021. 1, 3

    2021

  10. [17]

    BEVDet: High-performance multi-camera 3D object de- tection in bird’s-eye view

    Junjie Huang, Guan Huang, Zheng Zhu, Yun Ye, and Dalong Du. BEVDet: High-performance multi-camera 3D object de- tection in bird’s-eye view. arXiv preprint arXiv:2112.11790,

    arXiv

  11. [18]

    The ApolloScape dataset for autonomous driving

    Xinyu Huang, Xinjing Cheng, Qichuan Geng, Binbin Cao, Dingfu Zhou, Peng Wang, Yuanqing Lin, and Ruigang Yang. The ApolloScape dataset for autonomous driving. In CVPRW, pages 954–960, 2018. 3

    2018

  12. [19]

    Multi- modal sensor fusion-based deep neural network for end-to- end autonomous driving with scene understanding

    Zhiyu Huang, Chen Lv, Yang Xing, and Jingda Wu. Multi- modal sensor fusion-based deep neural network for end-to- end autonomous driving with scene understanding. IEEE Sensors Journal, 21(10):11781–11790, 2020. 1

    2020

  13. [20]

    Fuller: Unified multi-modality multi-task 3D perception via multi-level gradient calibration

    Zhijian Huang, Sihao Lin, Guiyu Liu, Mukun Luo, Chao- qiang Ye, Hang Xu, Xiaojun Chang, and Xiaodan Liang. Fuller: Unified multi-modality multi-task 3D perception via multi-level gradient calibration. In ICCV, pages 3502–3511,

  14. [21]

    MSMDFusion: Fusing lidar and camera at multiple scales with multi-depth seeds for 3D ob- ject detection

    Yang Jiao, Zequn Jie, Shaoxiang Chen, Jingjing Chen, Lin Ma, and Yu-Gang Jiang. MSMDFusion: Fusing lidar and camera at multiple scales with multi-depth seeds for 3D ob- ject detection. In CVPR, pages 21643–21652, 2023. 1

    2023

  15. [22]

    Ford Mustang, 2023

    Kentik khudosovtsev. Ford Mustang, 2023. Accessed: 2024- 11-16. 1

    2023

  16. [23]

    Time3D: End-to-end joint monocu- lar 3D object detection and tracking for autonomous driving

    Peixuan Li and Jieyu Jin. Time3D: End-to-end joint monocu- lar 3D object detection and tracking for autonomous driving. In CVPR, pages 3885–3894, 2022. 1

    2022

  17. [24]

    V2X-Sim: Multi-agent col- laborative perception dataset and benchmark for autonomous driving

    Yiming Li, Dekun Ma, Ziyan An, Zixun Wang, Yiqi Zhong, Siheng Chen, and Chen Feng. V2X-Sim: Multi-agent col- laborative perception dataset and benchmark for autonomous driving. IEEE Robotics and Automation Letters, 7(4):10914– 10921, 2022. 3

    2022

  18. [25]

    Fast-BEV: A fast and strong bird’s-eye view perception baseline

    Yangguang Li, Bin Huang, Zeren Chen, Yufeng Cui, Feng Liang, Mingzhu Shen, Fenggang Liu, Enze Xie, Lu Sheng, Wanli Ouyang, et al. Fast-BEV: A fast and strong bird’s-eye view perception baseline. IEEE TPAMI, 2024. 1

    2024

  19. [26]

    BEVNeXt: Reviving dense BEV frameworks for 3D object detection

    Zhenxin Li, Shiyi Lan, Jose M Alvarez, and Zuxuan Wu. BEVNeXt: Reviving dense BEV frameworks for 3D object detection. In CVPR, pages 20113–20123, 2024

    2024

  20. [27]

    UniMODE: Unified monocular 3D object detection

    Zhuoling Li, Xiaogang Xu, SerNam Lim, and Hengshuang Zhao. UniMODE: Unified monocular 3D object detection. In CVPR, pages 16561–16570, 2024

    2024

  21. [28]

    BEVFusion: A simple and robust lidar-camera fusion framework

    Tingting Liang, Hongwei Xie, Kaicheng Yu, Zhongyu Xia, Zhiwei Lin, Yongtao Wang, Tao Tang, Bing Wang, and Zhi Tang. BEVFusion: A simple and robust lidar-camera fusion framework. NeurIPS, 35:10421–10434, 2022. 7

    2022

  22. [29]

    RCBEVDet: Radar-camera fusion in bird’s-eye view for 3D object detection

    Zhiwei Lin, Zhe Liu, Zhongyu Xia, Xinhao Wang, Yongtao Wang, Shengxiang Qi, Yang Dong, Nan Dong, Le Zhang, and Ce Zhu. RCBEVDet: Radar-camera fusion in bird’s-eye view for 3D object detection. In CVPR, pages 14928–14937,

  23. [30]

    H-V2X: A large scale highway dataset for BEV perception

    Chang Liu, Mingxu Zhu, and Cong Ma. H-V2X: A large scale highway dataset for BEV perception. In ECCV, pages 139–157. Springer, 2025. 2, 3

    2025

  24. [31]

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

    Mingyu Liu, Ekim Yurtsever, Jonathan Fossaert, Xingcheng Zhou, Walter Zimmer, Yuning Cui, Bare Luka Zagar, and Alois C Knoll. A survey on autonomous driving datasets: Statistics, annotation quality, and a future outlook. IEEE Transactions on Intelligent Vehicles, 2024. 1, 2

    2024

  25. [32]

    PETRv2: A unified frame- work for 3D perception from multi-camera images

    Yingfei Liu, Junjie Yan, Fan Jia, Shuailin Li, Aqi Gao, Tian- cai Wang, and Xiangyu Zhang. PETRv2: A unified frame- work for 3D perception from multi-camera images. InICCV, pages 3262–3272, 2023. 1

    2023

  26. [34]

    SEED: A simple and effective 3D DETR in point clouds

    Zhe Liu, Jinghua Hou, Xiaoqing Ye, Tong Wang, Jingdong Wang, and Xiang Bai. SEED: A simple and effective 3D DETR in point clouds. arXiv preprint arXiv:2407.10749 , 2024

    2024 arXiv

  27. [35]

    DETR4D: Direct multi-view 3D object detection with sparse attention

    Zhipeng Luo, Changqing Zhou, Gongjie Zhang, and Shijian Lu. DETR4D: Direct multi-view 3D object detection with sparse attention. arXiv preprint arXiv:2212.07849, 2022. 1

    2022 arXiv

  28. [36]

    Vision-centric BEV perception: A survey

    Yuexin Ma, Tai Wang, Xuyang Bai, Huitong Yang, Yuenan Hou, Yaming Wang, Yu Qiao, Ruigang Yang, and Xinge Zhu. Vision-centric BEV perception: A survey. IEEE TPAMI, 2024. 1

    2024

  29. [37]

    BEV- guided multi-modality fusion for driving perception

    Yunze Man, Liang-Yan Gui, and Yu-Xiong Wang. BEV- guided multi-modality fusion for driving perception. In CVPR, pages 21960–21969, 2023. 1

    2023

  30. [38]

    BEVDetNet: Bird’s-eye view lidar point cloud based real-time 3D object detection for au- tonomous driving

    Sambit Mohapatra, Senthil Yogamani, Heinrich Gotzig, Ste- fan Milz, and Patrick Mader. BEVDetNet: Bird’s-eye view lidar point cloud based real-time 3D object detection for au- tonomous driving. In Proceedings of the 2021 IEEE Interna- tional Intelligent Transportation Systems ...

    2021

  31. [39]

    Towards compact autonomous driving perception with balanced learning and multi-sensor fusion

    Oskar Natan and Jun Miura. Towards compact autonomous driving perception with balanced learning and multi-sensor fusion. IEEE Transactions on Intelligent Transportation Sys- tems, 23(9):16249–16266, 2022. 1

    2022

  32. [40]

    The Mapillary vistas dataset for seman- tic understanding of street scenes

    Gerhard Neuhold, Tobias Ollmann, Samuel Rota Bulo, and Peter Kontschieder. The Mapillary vistas dataset for seman- tic understanding of street scenes. In ICCV, pages 4990– 4999, 2017. 3

    2017

  33. [41]

    BEVSegFormer: Bird’s-eye view se- mantic segmentation from arbitrary camera rigs

    Lang Peng, Zhirong Chen, Zhangjie Fu, Pengpeng Liang, and Erkang Cheng. BEVSegFormer: Bird’s-eye view se- mantic segmentation from arbitrary camera rigs. InProceed- ings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 5935–5943, 2023. 1

    2023

  34. [42]

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

    Quang-Hieu Pham, Pierre Sevestre, Ramanpreet Singh Pahwa, Huijing Zhan, Chun Ho Pang, Yuda Chen, Armin Mustafa, Vijay Chandrasekhar, and Jie Lin. A*3D dataset: Towards autonomous driving in challenging environments. In Proceedings of the IEEE International Conference on Roboti...

  35. [43]

    Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly unprojecting to 3D

    Jonah Philion and Sanja Fidler. Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly unprojecting to 3D. In ECCV, pages 194–210. Springer, 2020. 7

    2020

  36. [44]

    Deep multi-task learning for joint localization, perception, and prediction

    John Phillips, Julieta Martinez, Ioan Andrei B ˆarsan, Sergio Casas, Abbas Sadat, and Raquel Urtasun. Deep multi-task learning for joint localization, perception, and prediction. In CVPR, pages 4679–4689, 2021. 1

    2021

  37. [45]

    Playing for data: Ground truth from computer games

    Stephan R Richter, Vibhav Vineet, Stefan Roth, and Vladlen Koltun. Playing for data: Ground truth from computer games. In ECCV, pages 102–118. Springer, 2016. 3

    2016

  38. [46]

    The SYNTHIA dataset: A large collection of synthetic images for semantic segmen- tation of urban scenes

    German Ros, Laura Sellart, Joanna Materzynska, David Vazquez, and Antonio M Lopez. The SYNTHIA dataset: A large collection of synthetic images for semantic segmen- tation of urban scenes. In CVPR, pages 3234–3243, 2016. 3

    2016

  39. [47]

    Safety-enhanced autonomous driving using inter- pretable sensor fusion transformer

    Hao Shao, Letian Wang, Ruobing Chen, Hongsheng Li, and Yu Liu. Safety-enhanced autonomous driving using inter- pretable sensor fusion transformer. In Conference on Robot Learning, pages 726–737. PMLR, 2023. 1

    2023

  40. [48]

    ViPER: Augmenting auto- matic information extraction with visual perceptions

    Kai Simon and Georg Lausen. ViPER: Augmenting auto- matic information extraction with visual perceptions. InPro- ceedings of the 14th ACM International Conference on Infor- mation and Knowledge Management, pages 381–388, 2005. 3

    2005

  41. [49]

    Synthetic datasets for autonomous driv- ing: A survey

    Zhihang Song, Zimin He, Xingyu Li, Qiming Ma, Ruibo Ming, Zhiqi Mao, Huaxin Pei, Lihui Peng, Jianming Hu, Danya Yao, et al. Synthetic datasets for autonomous driv- ing: A survey. IEEE Transactions on Intelligent Vehicles ,

  42. [50]

    Scalability in perception for autonomous driving: Waymo open dataset

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, et al. Scalability in perception for autonomous driving: Waymo open dataset. In CVPR, pages 2446–2454, 2020. 3, 6

    2020

  43. [51]

    SHIFT: A synthetic driving dataset for continuous multi-task domain adaptation

    Tao Sun, Mattia Segu, Janis Postels, Yuxuan Wang, Luc Van Gool, Bernt Schiele, Federico Tombari, and Fisher Yu. SHIFT: A synthetic driving dataset for continuous multi-task domain adaptation. In CVPR, pages 21371–21382, 2022. 3, 8

    2022

  44. [52]

    Challenges in designing datasets and validation for autonomous driving

    Michal Uric ´ar, David Hurych, Pavel Krizek, and Senthil Yo- gamani. Challenges in designing datasets and validation for autonomous driving. arXiv preprint arXiv:1901.09270,

    1901 arXiv

  45. [53]

    UniTR: A unified and efficient multi-modal transformer for bird’s-eye view representation

    Haiyang Wang, Hao Tang, Shaoshuai Shi, Aoxue Li, Zhen- guo Li, Bernt Schiele, and Liwei Wang. UniTR: A unified and efficient multi-modal transformer for bird’s-eye view representation. In ICCV, pages 6792–6802, 2023. 1, 7, 8, 6

    2023

  46. [54]

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

    Li Wang, Xinyu Zhang, Ziying Song, Jiangfeng Bi, Guoxin Zhang, Haiyue Wei, Liyao Tang, Lei Yang, Jun Li, Caiyan Jia, et al. Multi-modal 3D object detection in autonomous driving: A survey and taxonomy. IEEE Transactions on In- telligent Vehicles, 8(7):3781–3798, 2023. 1

    2023

  47. [55]

    Multi-sensor fusion technology for 3D object detection in autonomous driving: A review

    Xuan Wang, Kaiqiang Li, and Abdellah Chehri. Multi-sensor fusion technology for 3D object detection in autonomous driving: A review. IEEE Transactions on Intelligent Trans- portation Systems, 2023. 1

    2023

  48. [56]

    DETR3D: 3D object detection from multi-view images via 3D-to-2D queries

    Yue Wang, Vitor Campagnolo Guizilini, Tianyuan Zhang, Yilun Wang, Hang Zhao, and Justin Solomon. DETR3D: 3D object detection from multi-view images via 3D-to-2D queries. In Conference on Robot Learning, pages 180–191. PMLR, 2022. 1

    2022

  49. [57]

    Bi-LRFusion: Bi-directional lidar-radar fusion for 3D dy- namic object detection

    Yingjie Wang, Jiajun Deng, Yao Li, Jinshui Hu, Cong Liu, Yu Zhang, Jianmin Ji, Wanli Ouyang, and Yanyong Zhang. Bi-LRFusion: Bi-directional lidar-radar fusion for 3D dy- namic object detection. InCVPR, pages 13394–13403, 2023. 1

    2023

  50. [58]

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

    Xinshuo Weng, Yunze Man, Jinhyung Park, Ye Yuan, Matthew O’Toole, and Kris M Kitani. All-in-One Drive: A comprehensive perception dataset with high-density long- range point clouds. OpenReview submission, 2023. 3, 6

    2023

  51. [59]

    Argoverse 2: Next generation datasets for self-driving perception and forecasting

    Benjamin Wilson, William Qi, Tanmay Agarwal, John Lambert, Jagjeet Singh, Siddhesh Khandelwal, Bowen Pan, Ratnesh Kumar, Andrew Hartnett, Jhony Kaesemodel Pontes, et al. Argoverse 2: Next generation datasets for self-driving perception and forecasting. arXiv preprint arXiv:230...

    2023 arXiv

  52. [60]

    M2BEV: Multi-camera joint 3D detection and segmentation with unified bird’s-eye view representation

    Enze Xie, Zhiding Yu, Daquan Zhou, Jonah Philion, Anima Anandkumar, Sanja Fidler, Ping Luo, and Jose M Alvarez. M2BEV: Multi-camera joint 3D detection and segmentation with unified bird’s-eye view representation. arXiv preprint arXiv:2204.05088, 2022. 1

    2022 arXiv

  53. [61]

    CAPE: Camera view position embedding for multi-view 3D object detection

    Kaixin Xiong, Shi Gong, Xiaoqing Ye, Xiao Tan, Ji Wan, Errui Ding, Jingdong Wang, and Xiang Bai. CAPE: Camera view position embedding for multi-view 3D object detection. In CVPR, pages 21570–21579, 2023

    2023

  54. [62]

    LXL: Lidar excluded lean 3D object detection with 4D imaging radar and camera fusion

    Weiyi Xiong, Jianan Liu, Tao Huang, Qing-Long Han, Yux- uan Xia, and Bing Zhu. LXL: Lidar excluded lean 3D object detection with 4D imaging radar and camera fusion. IEEE Transactions on Intelligent Vehicles, 2023. 1

    2023

  55. [63]

    PointFu- sion: Deep sensor fusion for 3D bounding box estimation

    Danfei Xu, Dragomir Anguelov, and Ashesh Jain. PointFu- sion: Deep sensor fusion for 3D bounding box estimation. In CVPR, pages 244–253. IEEE, 2018. 1

    2018

  56. [64]

    OPV2V: An open benchmark dataset and fu- sion pipeline for perception with vehicle-to-vehicle commu- nication

    Runsheng Xu, Hao Xiang, Xin Xia, Xu Han, Jinlong Li, and Jiaqi Ma. OPV2V: An open benchmark dataset and fu- sion pipeline for perception with vehicle-to-vehicle commu- nication. In Proceedings of the International Conference on Robotics and Automation (ICRA) , pages 2583–2589. IEEE,

  57. [65]

    CoBEVT: Cooperative bird’s-eye view semantic segmentation with sparse transformers

    Runsheng Xu, Zhengzhong Tu, Hao Xiang, Wei Shao, Bolei Zhou, and Jiaqi Ma. CoBEVT: Cooperative bird’s-eye view semantic segmentation with sparse transformers. In Confer- ence on Robot Learning, pages 989–1000. PMLR, 2023. 1

    2023

  58. [66]

    BEVFormer v2: Adapting modern image backbones to bird’s-eye view recognition via perspective su- pervision

    Chenyu Yang, Yuntao Chen, Hao Tian, Chenxin Tao, Xizhou Zhu, Zhaoxiang Zhang, Gao Huang, Hongyang Li, Yu Qiao, Lewei Lu, et al. BEVFormer v2: Adapting modern image backbones to bird’s-eye view recognition via perspective su- pervision. In CVPR, pages 17830–17839, 2023. 1

    2023

  59. [67]

    BDD100K: A diverse driving dataset for heterogeneous multitask learning

    Fisher Yu, Haofeng Chen, Xin Wang, Wenqi Xian, Yingying Chen, Fangchen Liu, Vashisht Madhavan, and Trevor Dar- rell. BDD100K: A diverse driving dataset for heterogeneous multitask learning. In CVPR, pages 2636–2645, 2020. 3

    2020

  60. [68]

    Uni3D: A unified baseline for multi-dataset 3D object detection

    Bo Zhang, Jiakang Yuan, Botian Shi, Tao Chen, Yikang Li, and Yu Qiao. Uni3D: A unified baseline for multi-dataset 3D object detection. In CVPR, pages 9253–9262, 2023. 1

    2023

  61. [69]

    Autonomous driving system: A comprehensive survey

    Jingyuan Zhao, Wenyi Zhao, Bo Deng, Zhenghong Wang, Feng Zhang, Wenxiang Zheng, Wanke Cao, Jinrui Nan, Yubo Lian, and Andrew F Burke. Autonomous driving system: A comprehensive survey. Expert Systems with Applications , page 122836, 2023. 1

    2023

  62. [70]

    Improving bird’s-eye view semantic segmentation by task decomposition

    Tianhao Zhao, Yongcan Chen, Yu Wu, Tianyang Liu, Bo Du, Peilun Xiao, Shi Qiu, Hongda Yang, Guozhen Li, Yi Yang, et al. Improving bird’s-eye view semantic segmentation by task decomposition. In CVPR, pages 15512–15521, 2024. 1

    2024

  63. [71]

    MaskBEV: Towards a unified framework for BEV detection and map segmentation

    Xiao Zhao, Xukun Zhang, Dingkang Yang, Mingyang Sun, Mingcheng Li, Shunli Wang, and Lihua Zhang. MaskBEV: Towards a unified framework for BEV detection and map segmentation. In ACM MM, pages 2652–2661, 2024. 1

    2024

  64. [72]

    Perception-aware multi-sensor fusion for 3D lidar semantic segmentation

    Zhuangwei Zhuang, Rong Li, Kui Jia, Qicheng Wang, Yuan- qing Li, and Mingkui Tan. Perception-aware multi-sensor fusion for 3D lidar semantic segmentation. In ICCV, pages 16280–16290. IEEE, 2021. 1

    2021

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.