Pith. sign in

REVIEW 3 major objections 4 minor 48 references

A multi-modal dataset and tracker pipeline can put temporary roadwork objects on the map to within one meter of ground truth GPS.

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

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

T0 review · grok-4.5

2026-07-11 20:00 UTC pith:W35FVWEE

load-bearing objection Useful public multi-modal roadwork dataset with object-level GPS; the geo-localization numbers rest on only 43 real objects and the pipeline is a thin AB3DMOT extension. the 3 major comments →

arxiv 2607.04330 v1 pith:W35FVWEE submitted 2026-07-05 cs.CV

Framework and Multi-modal Dataset for Roadwork Zone Detection and Geo-localization

Zhiran Yan , Yutong Xin , S Shyam Shenoi , Rui Song , Gordon Elger This is my paper
classification cs.CV
keywords roadwork zone detectionobject geo-localizationmulti-modal dataset3D multi-object trackingHD map updateautonomous drivingLiDAR camera fusion
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.

High-definition maps used by autonomous vehicles omit temporary roadwork zones that change the road network. The paper supplies a public multi-modal dataset (camera, LiDAR, GPS/IMU) with both real German scenes and CARLA simulations, annotated for semantic masks, 3D boxes, and true global coordinates of barriers and road beacons. On top of that data it builds a tracker-based pipeline that detects those objects in 3D, links repeated detections across frames into tracklets, fuses each tracklet into a single position, and transforms that position into latitude-longitude using the ego vehicle’s GPS/IMU. A prediction counts as correct when it lies within one meter of the measured ground-truth GPS. Under that strict criterion the best real-world configuration reaches precision 0.565, recall 0.898 and F1 0.597, with slightly higher numbers on simulation. The practical claim is that the missing semi-static layer of an HD map can be filled on the fly by ordinary vehicle sensors.

Core claim

The authors show that an extension of a standard 3-D multi-object tracker, fed by camera- or LiDAR-based detectors and ego GPS/IMU, can convert repeated local detections of roadwork objects into unique global coordinates that stay inside a one-meter Haversine ball of RTK ground truth on both real and simulated data.

What carries the argument

The RZDG pipeline: 3-D detections are tracked into tracklets, each tracklet is collapsed to a single local center (last-frame or weighted-average), and that center is transformed into WGS84 latitude-longitude by the ego vehicle’s GPS/IMU and heading.

Load-bearing premise

That the 43 real objects whose GPS was measured by parking an RTK receiver over their centers, drawn from only three consecutive scenes, are enough to prove the pipeline works in general.

What would settle it

Collect RTK ground truth for a larger, independent set of real roadwork objects (different cities, weather, denser clutter) and recompute the one-meter Haversine precision/recall; a clear drop below the reported F1 would falsify the claim of practical accuracy.

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

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

3 major / 4 minor

Summary. The paper introduces the RZDG multi-modal dataset (real + CARLA-simulated) with camera/LiDAR/GPS-IMU data and annotations for semantic segmentation, 3D bounding boxes, and object global positions of roadwork elements (barriers, road beacons). It also presents a tracker-based geo-localization pipeline that extends AB3DMOT: 3D detections are tracked into tracklets, fused (last-frame or weighted-average), and transformed from the camera frame to WGS84 via ego GPS/IMU using the spherical trigonometry of Eqs. (1)–(7). Under a 1 m Haversine true-positive rule the pipeline reports Precision/Recall/F1 of 0.565/0.898/0.597 (real) and 0.615/0.809/0.665 (sim) for the best detector–fusion combination; standard AP and mIoU numbers are also given for three detectors and three segmentors.

Significance. If the numbers hold under broader evaluation, the work supplies a missing public multi-modal benchmark for temporary roadwork zones—an acknowledged gap in HD-map maintenance for autonomous driving—and a simple, detector-agnostic pipeline that converts local detections into global coordinates. Explicit strengths include the promised open release of both real and simulated data plus code, multi-task annotations in KITTI format, and transparent reporting of train/val splits and three detector families. The contribution is therefore primarily infrastructural rather than algorithmic; the pipeline itself is a straightforward extension of AB3DMOT plus textbook coordinate conversion.

major comments (3)
  1. Section III-B/C and VI-B: real-world geo-localization metrics in Table VI rest exclusively on 43 RTK-measured objects collected across only three consecutive scenes. No confidence intervals, bootstrap estimates, leave-one-scene-out results, or uncertainty on the RTK ground truth itself are reported. With N=43 the headline F1 of 0.597 (and the abstract’s claim of “high accuracy”) is statistically under-powered and cannot be taken as a reliable estimate of pipeline performance.
  2. Abstract and Section VIII: the language “high accuracy” is inconsistent with the reported real-data Precision of 0.565 (Table VI, LF-MVXNet) and with the modest 3D detection APs in Table V (MVXNet ~33–39 % AP3D on real data). The discrepancy between low detection quality and high geo-localization recall needs explicit discussion or a more conservative claim.
  3. Section V-B and Table VI: the two tracklet-fusion strategies (LF vs WA) produce dramatically different rankings on real versus simulated data, yet no analysis of the failure modes (e.g., track fragmentation, ego-pose noise, or calibration residual) is supplied. Because the coordinate transform (Eqs. 1–7) is deterministic once a tracklet is chosen, the fusion choice is load-bearing; its sensitivity should be quantified.
minor comments (4)
  1. Throughout: inconsistent spacing and hyphenation (“A Vs”, “V oxelFusion”, “W A”, “H. thres.a”) should be cleaned for readability.
  2. Table I: the column “# Object’s global position 43/673” is useful but the caption should clarify that 43 is the only real-world geo-localization ground truth used for evaluation.
  3. Section III-A: the sensor-car figure caption mentions seven cameras and three radars while the text and dataset use only one camera + one LiDAR + GPS/IMU; a clarifying sentence would avoid confusion.
  4. Equations (1)–(7): the Earth radius is given as 6 372 800 m without citation or discussion of the approximation error relative to the 1 m TP threshold.

Circularity Check

0 steps flagged

No circularity: empirical multi-sensor pipeline and dataset benchmarked against independent RTK/simulator ground truth with fixed 1 m Haversine rule; no fitted parameters or self-definitional reductions.

full rationale

The paper presents a new multi-modal dataset (RZDG-real/sim) with independent annotations (semantic masks, 3D boxes, and 43 real RTK-measured object GPS positions obtained by placing a receiver over object centers) plus a tracker-based pipeline that extends AB3DMOT with standard 3D Kalman/Hungarian tracking, two simple tracklet-fusion heuristics (last-frame or weighted average), and textbook coordinate transforms (Eqs. 1-7 from an external navigation reference, plus Haversine distance). Performance numbers (Precision/Recall/F1 under a fixed 1 m TP threshold) are pure empirical evaluations of trained detectors + tracker on held-out scenes; no free parameter is fitted to the geo-localization F1 itself, no quantity is defined in terms of the reported metrics, and no uniqueness theorem or ansatz is imported via overlapping-author citation. The derivation chain is therefore self-contained engineering evaluation against external ground truth, not a circular reduction.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The central geo-localization claim rests on standard sensor models, the AB3DMOT tracking assumptions, a fixed 1 m acceptance radius, and a small set of RTK-measured ground-truth points. No new physical constants or free parameters are fitted to produce the F1 scores; the main modeling choices are conventional for the domain.

free parameters (2)
  • Haversine TP threshold = 1 meter
    Fixed at 1 m by author choice; directly defines true-positive counts that produce the reported Precision/Recall/F1.
  • Tracklet fusion strategy (LF vs WA) = LF preferred on real data
    Two discrete heuristics for collapsing a track into one global position; choice affects reported scores and is not derived from first principles.
axioms (4)
  • domain assumption Constant-velocity 3D Kalman filter + Hungarian association (AB3DMOT) correctly links detections of the same static roadwork object across frames.
    Invoked throughout Section V-B; objects are stationary, so the motion model is reasonable but still an assumption.
  • domain assumption RTK GPS placed over object centers yields ground-truth latitude/longitude accurate to a few centimeters, sufficient for a 1 m evaluation threshold.
    Section III-B; only 43 objects measured this way.
  • standard math Earth radius R = 6 372 800 m and the spherical Haversine formulas (Eqs. 8–10) are adequate for meter-scale distances.
    Used in the evaluation metric definition.
  • domain assumption CARLA sensor noise models and imported CAD assets are sufficiently realistic that sim results transfer to real performance claims.
    Section IV; sim F1 is higher than real, indicating a domain gap.

pith-pipeline@v1.1.0-grok45 · 18113 in / 2904 out tokens · 23262 ms · 2026-07-11T20:00:42.272170+00:00 · methodology

0 comments
read the original abstract

Autonomous vehicles often rely on high-definition (HD) maps for navigation; however, these maps are not frequently updated and often lack semi-static information, such as temporary roadwork zones, which can significantly alter the road network. This limitation underscores the urgent need for an accurate global position of roadwork zones. However, the absence of publicly available datasets for evaluating roadwork zone detection and geo-localization models has hindered the development of reliable autonomous driving systems. To address this challenge, we propose the Roadwork Zone Detection and Geo-localization (RZDG) dataset, which includes both simulated and real-world data, providing multimodal sensor inputs along with comprehensive annotations. The dataset supports multiple perception tasks, including image semantic segmentation, 3D object detection, and object geo-localization. In addition, we introduce a tracker-based roadwork zone detection and geo-localization (RZDG) pipeline, an extension of AB3DMOT, for accurate object geo-localization in roadwork zones. We benchmark our approach on the RZDG dataset, demonstrating its effectiveness in detecting roadwork zones and transforming object positions from the local coordinate system to the global coordinate system. A prediction is considered a true positive (TP) if its estimated position falls within one meter of the ground truth. Our experimental results show that our approach achieves high accuracy on both real and simulated data. Specifically, we report: Precision: 0.565 (real) / 0.615 (simulated) Recall: 0.898 (real) / 0.809 (simulated) F1-score: 0.597 (real) / 0.665 (simulated).

Figures

Figures reproduced from arXiv: 2607.04330 by Gordon Elger, Rui Song, S Shyam Shenoi, Yutong Xin, Zhiran Yan.

Figure 1
Figure 1. Figure 1: Examples of diverse roadwork zones depicted in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Data Collection Platform. This is an Audi A6 [36] equipped with seven Basler cameras, one Ouster OS1-128 LiDAR, three ARS548 RADARs and one Certus GPS/IMU sensor. (Note that the dataset in this work contains only one camera, one LiDAR, and one GPS/IMU.) to ensure uniform system time across all sensors. Camera is hardware-triggered with LiDAR’s 20 Hz pulse signal and operate at a frequency of 20 Hz. This se… view at source ↗
Figure 3
Figure 3. Figure 3: Driving Route. Driving routes are depicted using the sensor car’s GPS data on OpenStreetMap, with the paths highlighted in blue to represent the car’s trajectory. license plates and heads of pedestrians on all RGB images, which is achieved by utilizing a fine-tuning YOLOv8 [40], an object detection algorithm. B. Data Annotation 3D Bounding Boxes Annotation. For the collected Li￾DAR data, we sample key fram… view at source ↗
Figure 4
Figure 4. Figure 4: Overview of the Roadwork Zone Detection and Geo-localization (RZDG) Pipeline. This comprehensive pipeline [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Required parameters for the coordinate transform [PITH_FULL_IMAGE:figures/full_fig_p005_5.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

48 extracted references · 4 linked inside Pith

  1. [1]

    ildm: An interoperable graph-based local dynamic map,

    M. Garc ´ıa, I. Urbieta, M. Nieto, J. Gonz ´alez de Mendibil, and O. Otaegui, “ildm: An interoperable graph-based local dynamic map,” Vehicles, vol. 4, no. 1, pp. 42–59, 2022

    2022

  2. [2]

    Intelligent transport systems (its); vehicular communica- tions; basic set of applications; definitions,

    T. ETSI, “Intelligent transport systems (its); vehicular communica- tions; basic set of applications; definitions,”Tech. Rep. ETSI TR 102 6382009, 2009

    2009

  3. [3]

    Characterization, detection, and segmentation of work-zone scenes from naturalistic driving data,

    V . Sundharam, A. Sarkar, A. Svetovidov, J. S. Hickman, and A. L. Abbott, “Characterization, detection, and segmentation of work-zone scenes from naturalistic driving data,”Transportation research record, vol. 2677, no. 3, pp. 490–504, 2023

    2023

  4. [4]

    Low-cost object detection models for traffic control devices through domain adaption of geographical regions,

    D. Oh, K. Kang, S. Seo, J. Xiao, K. Jang, K. Kim, H. Park, and J. Won, “Low-cost object detection models for traffic control devices through domain adaption of geographical regions,”Remote Sensing, vol. 15, no. 10, p. 2584, 2023

    2023

  5. [5]

    Tracon: A novel dataset for real-time traffic cones detection using deep learning,

    I. Katsamenis, E. E. Karolou, A. Davradou, E. Protopapadakis, A. Doulamis, N. Doulamis, and D. Kalogeras, “Tracon: A novel dataset for real-time traffic cones detection using deep learning,” in Novel & Intelligent Digital Systems Conferences. Springer, 2022, pp. 382–391

    2022

  6. [6]

    Work zone detection for autonomous ve- hicles,

    W. Shi and R. R. Rajkumar, “Work zone detection for autonomous ve- hicles,” in2021 IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2021, pp. 1585–1591

    2021

  7. [7]

    Ab3dmot: A baseline for 3d multi-object tracking and new evaluation metrics,

    X. Weng, J. Wang, D. Held, and K. Kitani, “Ab3dmot: A baseline for 3d multi-object tracking and new evaluation metrics,”arXiv preprint arXiv:2008.08063, 2020

    Pith/arXiv arXiv 2008

  8. [8]

    Effects of highway work zone temporary countermeasures,

    F. Shahin, W. Elias, and T. Toledo, “Effects of highway work zone temporary countermeasures,”European Transport Research Review, vol. 15, no. 1, p. 20, 2023

    2023

  9. [9]

    Encoder-decoder with atrous separable convolution for semantic image segmentation,

    L.-C. Chen, Y . Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder-decoder with atrous separable convolution for semantic image segmentation,” inProceedings of the European conference on computer vision (ECCV), 2018, pp. 801–818

    2018

  10. [10]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 012–10 022

    2021

  11. [11]

    Segformer: Simple and efficient design for semantic segmentation with transformers,

    E. Xie, W. Wang, Z. Yu, A. Anandkumar, J. M. Alvarez, and P. Luo, “Segformer: Simple and efficient design for semantic segmentation with transformers,”Advances in neural information processing sys- tems, vol. 34, pp. 12 077–12 090, 2021

    2021

  12. [12]

    Gs3d: An efficient 3d object detection framework for autonomous driving,

    B. Li, W. Ouyang, L. Sheng, X. Zeng, and X. Wang, “Gs3d: An efficient 3d object detection framework for autonomous driving,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 1019–1028

    2019

  13. [13]

    Monocular 3d object detection with decoupled structured polygon estimation and height-guided depth estimation,

    Y . Cai, B. Li, Z. Jiao, H. Li, X. Zeng, and X. Wang, “Monocular 3d object detection with decoupled structured polygon estimation and height-guided depth estimation,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 34, no. 07, 2020, pp. 10 478– 10 485

    2020

  14. [14]

    Monocular 3d object detection for autonomous driving,

    X. Chen, K. Kundu, Z. Zhang, H. Ma, S. Fidler, and R. Urtasun, “Monocular 3d object detection for autonomous driving,” inPro- ceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 2147–2156

    2016

  15. [15]

    Ground-aware monocular 3d object detection for autonomous driving,

    Y . Liu, Y . Yixuan, and M. Liu, “Ground-aware monocular 3d object detection for autonomous driving,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 919–926, 2021

    2021

  16. [16]

    Smoke: Single-stage monocular 3d object detection via keypoint estimation,

    Z. Liu, Z. Wu, and R. T ´oth, “Smoke: Single-stage monocular 3d object detection via keypoint estimation,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, 2020, pp. 996–997

    2020

  17. [17]

    Objects are different: Flexible monocu- lar 3d object detection,

    Y . Zhang, J. Lu, and J. Zhou, “Objects are different: Flexible monocu- lar 3d object detection,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 3289–3298

    2021

  18. [18]

    Pointnet: Deep learning on point sets for 3d classification and segmentation,

    R. Q. Charles, H. Su, M. Kaichun, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 77–85

    2017

  19. [19]

    Dynamic graph cnn for learning on point clouds,

    Y . Wang, Y . Sun, Z. Liu, S. E. Sarma, M. M. Bronstein, and J. M. Solomon, “Dynamic graph cnn for learning on point clouds,” ACM Trans. Graph., vol. 38, no. 5, Oct. 2019. [Online]. Available: https://doi.org/10.1145/3326362

    doi:10.1145/3326362 2019

  20. [20]

    Kpconv: Flexible and deformable convolution for point clouds,

    H. Thomas, C. R. Qi, J.-E. Deschaud, B. Marcotegui, F. Goulette, and L. J. Guibas, “Kpconv: Flexible and deformable convolution for point clouds,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October 2019

    2019

  21. [21]

    V oxelnet: End-to-end learning for point cloud based 3d object detection,

    Y . Zhou and O. Tuzel, “V oxelnet: End-to-end learning for point cloud based 3d object detection,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018

    2018

  22. [22]

    Transformer-based global pointpillars 3d object detection method,

    L. Zhang, H. Meng, Y . Yan, and X. Xu, “Transformer-based global pointpillars 3d object detection method,”Electronics, vol. 12, no. 14, p. 3092, 2023. [Online]. Available: https://doi.org/10.3390/ electronics12143092

    2023

  23. [23]

    Pv- rcnn: Point-voxel feature set abstraction for 3d object detection,

    S. Shi, C. Guo, L. Jiang, Z. Wang, J. Shi, X. Wang, and H. Li, “Pv- rcnn: Point-voxel feature set abstraction for 3d object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2020

    2020

  24. [24]

    Pointpillars: Fast encoders for object detection from point clouds,

    A. H. Lang, S. V ora, H. Caesar, L. Zhou, J. Yang, and O. Beijbom, “Pointpillars: Fast encoders for object detection from point clouds,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2019

    2019

  25. [25]

    Fusion of 3d lidar and camera data for object detection in autonomous vehicle applications,

    X. Zhao, P. Sun, Z. Xu, H. Min, and H. Yu, “Fusion of 3d lidar and camera data for object detection in autonomous vehicle applications,” IEEE Sensors Journal, vol. 20, no. 9, pp. 4901–4913, 2020

    2020

  26. [26]

    Mvx-net: Multimodal voxelnet for 3d object detection,

    V . A. Sindagi, Y . Zhou, and O. Tuzel, “Mvx-net: Multimodal voxelnet for 3d object detection,” in2019 International Conference on Robotics and Automation (ICRA). IEEE, 2019, pp. 7276–7282

    2019

  27. [27]

    3d dual-fusion: Dual-domain dual-query camera-lidar fusion for 3d object detection,

    Y . Kim, K. Park, M. Kim, D. Kum, and J. W. Choi, “3d dual-fusion: Dual-domain dual-query camera-lidar fusion for 3d object detection,” arXiv preprint arXiv:2211.13529, 2022

    Pith/arXiv arXiv 2022

  28. [28]

    Image and object geo-localization,

    D. Wilson, X. Zhang, W. Sultani, and S. Wshah, “Image and object geo-localization,”International Journal of Computer Vision, vol. 132, no. 4, pp. 1350–1392, 2024

    2024

  29. [29]

    Road deformation detection,

    K. A. Govada, H. P. Jonnalagadda, P. Kapavarapu, S. Alavala, and K. S. Vani, “Road deformation detection,” in2020 7th International Conference on Smart Structures and Systems (ICSSS). IEEE, 2020, pp. 1–5

    2020

  30. [30]

    Object tracking and geo-localization from street images,

    D. Wilson, T. Alshaabi, C. Van Oort, X. Zhang, J. Nelson, and S. Wshah, “Object tracking and geo-localization from street images,” Remote Sensing, vol. 14, no. 11, p. 2575, 2022

    2022

  31. [31]

    End-to-end learning improves static object geo-localization from video,

    M. Chaabane, L. Gueguen, A. Trabelsi, R. Beveridge, and S. O’Hara, “End-to-end learning improves static object geo-localization from video,” inProceedings of the IEEE/CVF Winter Conference on Ap- plications of Computer Vision, 2021, pp. 2063–2072

    2021

  32. [32]

    Automatic discovery and geotagging of objects from street view imagery,

    V . A. Krylov, E. Kenny, and R. Dahyot, “Automatic discovery and geotagging of objects from street view imagery,”Remote Sensing, vol. 10, no. 5, p. 661, 2018

    2018

  33. [33]

    Geolocation estimation of target vehicles using image processing and geometric computation,

    E. Namazi, R. Mester, C. Lu, and J. Li, “Geolocation estimation of target vehicles using image processing and geometric computation,” Neurocomputing, vol. 499, pp. 35–46, 2022

    2022

  34. [34]

    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

    2013

  35. [35]

    Robot operating system 2: Design, architecture, and uses in the wild,

    S. Macenski, T. Foote, B. Gerkey, C. Lalancette, and W. Woodall, “Robot operating system 2: Design, architecture, and uses in the wild,” Science robotics, vol. 7, no. 66, p. eabm6074, 2022

    2022

  36. [36]

    Audi a6 avant

    Bibliocad, “Audi a6 avant.” [Online]. Available: https://www. bibliocad.com/de/library/audi-a6-avant- 12984/

  37. [37]

    Planet dump retrieved from https://planet.osm.org ,

    OpenStreetMap contributors, “Planet dump retrieved from https://planet.osm.org ,” https://www.openstreetmap.org, 2017

    2017

  38. [38]

    A flexible new technique for camera calibration,

    Z. Zhang, “A flexible new technique for camera calibration,”IEEE Transactions on pattern analysis and machine intelligence, vol. 22, no. 11, pp. 1330–1334, 2000

    2000

  39. [39]

    Calib-anything: Zero-training lidar-camera extrinsic calibration method using segment anything,

    Z. Luo, G. Yan, and Y . Li, “Calib-anything: Zero-training lidar-camera extrinsic calibration method using segment anything,”arXiv preprint arXiv:2306.02656, 2023

    Pith/arXiv arXiv 2023

  40. [40]

    Yolov8: A novel object detection algorithm with enhanced performance and robustness,

    R. Varghese and S. M., “Yolov8: A novel object detection algorithm with enhanced performance and robustness,” in2024 International Conference on Advances in Data Engineering and Intelligent Com- puting Systems (ADICS), 2024, pp. 1–6

    2024

  41. [41]

    Sustech points: A portable 3d point cloud interactive annotation platform system,

    E. Li, S. Wang, C. Li, D. Li, X. Wu, and Q. Hao, “Sustech points: A portable 3d point cloud interactive annotation platform system,” in 2020 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2020, pp. 1108–1115

    2020

  42. [42]

    Labelme: Online image annotation and applications,

    A. Torralba, B. C. Russell, and J. Yuen, “Labelme: Online image annotation and applications,”Proceedings of the IEEE, vol. 98, no. 8, pp. 1467–1484, 2010

    2010

  43. [43]

    Carla: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V . Koltun, “Carla: An open urban driving simulator,” inConference on robot learning. PMLR, 2017, pp. 1–16

    2017

  44. [44]

    H. D. Sch ¨onborn and W. Schulte,RSA Handbuch: Richtlinien f ¨ur die verkehrsrechtliche Sicherung von Arbeitsstellen an Straßen RSA 21, Ausgabe 2021: Handbuch und Kommentar, 2022

    2021

  45. [45]

    Distance, latitude, longitude, and navigation,

    C. Heidenreich, “Distance, latitude, longitude, and navigation,”Carto- graphica: The International Journal for Geographic Information and Geovisualization, vol. 13, no. 2, pp. 42–76, 1976

    1976

  46. [46]

    MMDetection3D: OpenMMLab next-generation platform for general 3D object detection,

    M. Contributors, “MMDetection3D: OpenMMLab next-generation platform for general 3D object detection,” https://github.com/ open-mmlab/mmdetection3d, 2020

    2020

  47. [47]

    Simultaneous multi- view instance detection with learned geometric soft-constraints,

    A. S. Nassar, S. Lef `evre, and J. D. Wegner, “Simultaneous multi- view instance detection with learned geometric soft-constraints,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 6559–6568

    2019

  48. [48]

    Mmdetection: Open mmlab detection toolbox and benchmark,

    K. Chen, J. Wang, J. Pang, Y . Cao, Y . Xiong, X. Li, S. Sun, W. Feng, Z. Liu, J. Xuet al., “Mmdetection: Open mmlab detection toolbox and benchmark,”arXiv preprint arXiv:1906.07155, 2019

    Pith/arXiv arXiv 1906