Pith. sign in

REVIEW 4 major objections 5 minor 36 references

Inferring Driving Maps by Deep Learning-based Trail Map Extraction

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper argues that an offline mapping system can reconstruct directed lane centerlines from crowdsourced driving trails, and that this trail-based approach outperforms camera-based online mapping on both original and geographically…

desk verdict A solid trail-based offline mapping system with a real but overclaimed comparison to online mapping; the method works, yet the 'comprehensive global map' claim only covers trail-covered lanes. read the letter →

arxiv 2505.10258 v1 pith:WFRMQSLZ submitted 2025-05-15 cs.CV cs.RO

Michael Hubbertz , Pascal Colling , Qi Han , Tobias Meisen This is my paper · ORCID
classification cs.CVcs.RO
keywords trail-basedmappingHDmapgenerationofflinetransformerlanecenterlineextractioncrowdsourcedsensor-agnosticgeneralization
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

The paper argues that high-definition driving maps can be built offline from crowdsourced trail data—the routes actually driven by ego vehicles and observed traffic participants—rather than from onboard cameras or LiDAR at run time. It claims that when trails are aggregated into direction- and speed-tagged grid tiles and fed to a transformer decoder, the model reconstructs directed lane centerlines more accurately than state-of-the-art online mapping, and generalizes better to new cities and sensor setups. If true, this would let fleets of ordinary vehicles with localization and object tracking continuously update global maps without expensive HD map labeling or high-resolution sensors.

What carries the argument

The central object is the trail grid map $G \in \mathbb{R}^{w \times h \times (n+1)}$: aggregated vehicle trails are rasterized into tiles, binned by the yaw angle $\varphi$ of each trail segment into $n=6$ direction channels, plus one average-speed channel, then Gaussian-smoothed. This multi-channel grid replaces the camera and LiDAR input of a MapTRv2-style query-based transformer decoder, whose set-prediction loss with fixed-direction centerline permutations outputs directed lane centerlines as ordered point sequences. Tile-wise processing and random-tile augmentation let the model be trained on any dataset with trails and an HD centerline map.

What would settle it

Run the model on a held-out tile whose HD map contains several lanes that no vehicle ever traversed, and compare predictions against the full HD map including those lanes; if the model cannot predict them, the approach fails to produce a complete driving map and the reported AP depends on masking.

Watch

Extended reading notes

Core claim

On nuScenes, the proposed TrailTR model reaches 56.4 AP on the original validation split versus 53.3 for MapTRv2, and 44.6 AP versus 30.5 on the geographically disjoint near-extrapolation split, a relative improvement of 46.2 percent when the comparison is masked to trail-covered areas. Training and evaluating on nuPlan yields 70.3 AP. The same model transfers across sensor foundations: ground-truth tracks, camera-plus-LiDAR detections from FocalFormer3D, and Radar-based detections on a closed Las Vegas dataset, with only minimal degradation. The paper concludes that trail-based offline mapping is a viable alternative to online mapping, offering tile-wise updates and improved generalization.

Load-bearing premise

The method assumes every lane that matters has at least one recorded trail; centerlines never driven by the ego vehicle or other tracked traffic are removed from the ground truth, so the evaluation measures reconstruction only of trail-covered lanes.

Editorial extensions

If this is right

  • Map providers could replace per-region HD mapping campaigns with continuous fleet-data aggregation, updating only tiles where trails change.
  • Autonomous vehicles could operate in previously unmapped areas once a few passes by trail-equipped vehicles cover the road network.
  • The sensor-agnostic input means the same trained model can ingest detections from cameras, LiDAR, or Radar without retraining per sensor.
  • The reported gains on geographically disjoint splits suggest the trail representation captures road geometry that generalizes beyond the training cities.
  • Combining trail-based offline maps with online mapping could fill the occlusion and consistency gaps that plague purely online approaches.

Reading between the lines

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

  • The reported AP likely overstates completeness on the full road network: because centerlines lacking trails are filtered from ground truth, the evaluation measures reconstruction only of lanes that were actually driven, and a map built this way is only as complete as the trail coverage.
  • A direct test of coverage bias would be to sweep trail density per tile and plot AP against the fraction of HD-map lanes that have trail support; a strong dependence would show the method is a trail densifier rather than a full map generator.
  • The methodology should extend naturally to lane boundaries, pedestrian crossings, and traffic signs if the input channels can encode corresponding trail or object evidence, as the paper hints in its conclusion.
  • The cross-dataset drop from 70.3 AP on nuPlan to 34.0 AP when trained on nuScenes suggests the model is sensitive to trail-density statistics; quantifying this mismatch could guide data collection requirements for new cities.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes TrailTR, an offline map-generation approach that converts aggregated vehicle trails (ego trajectories and tracked object trajectories) into multi-channel bird's-eye-view grid tiles and trains a MapTRv2-inspired transformer decoder to predict directed lane centerlines. The input representation encodes trail density, yaw-angle direction bins, and average speed, with Gaussian smoothing and random tile/rotation/subset augmentation. The method is evaluated on nuScenes and nuPlan, compared against a re-trained MapTRv2 baseline on original and geographically disjoint splits, and extended to detector-based trail extraction and a closed Las Vegas Radar dataset. The authors report that TrailTR outperforms MapTRv2 on both nuScenes splits and that the ablations support the chosen input design.

Significance. If the reported results are taken at face value, the paper makes a useful engineering contribution to crowdsourced offline HD-map generation: it adapts a state-of-the-art online-mapping architecture to trail-grid inputs, validates the design with systematic ablations (smoothing, direction binning, speed profile, augmentation), and evaluates on a geographically disjoint split, which is a real strength. The paper also demonstrates cross-dataset transfer between nuScenes and nuPlan and a modest drop when trails are produced by a learned detector, which supports the sensor-agnostic framing. However, the headline claims of superiority over online mapping and of constructing a 'comprehensive global map' are conditional on a trail-filtered evaluation protocol that removes untraversed lanes from the ground truth; this scoping issue is acknowledged in the manuscript itself but not resolved quantitatively.

major comments (4)
  1. [§3.1, §4.1, Fig. 4] The ground truth is filtered to centerlines that have associated trails, so every reported AP in Tables 1-5 measures performance only on the trail-covered subset of the road network. Since the model's only input is trails, it cannot infer untraversed lanes by construction, and the abstract/conclusion claim of constructing a 'comprehensive global map' is therefore not supported by the experiments. The manuscript explicitly acknowledges this in §4.1 and in the Fig. 4 caption. Please quantify trail coverage (e.g., fraction of centerline length or tile area with trails), report results on the full unfiltered ground truth where feasible, and either demonstrate a mechanism for recovering untraversed lanes or scale the claim back to trail-covered regions.
  2. [§4.2, Table 1] The comparison to MapTRv2 is not information-equivalent: MapTRv2 receives a short current sensor snippet at inference, while TrailTR receives a tile of trails accumulated over the whole recording period, and the MapTRv2 evaluation is additionally masked to trail-covered areas. The reported +46.2% advantage on the geographically disjoint split therefore reflects the difference in input/task setup as much as model quality. Please state this asymmetry explicitly, add an offline baseline that consumes comparable aggregated information, or reframe the headline as 'trail-based offline mapping outperforms a single-frame online baseline on trail-covered regions.'
  3. [§4.2, Table 1] The claim of 'superior generalization' is supported only on the nuScenes geographically disjoint split. The cross-dataset rows (34.0 AP from nuScenes-geo to nuPlan and 35.0 AP from nuPlan to nuScenes-geo) are far below the within-dataset scores and have no MapTRv2 comparison, so they demonstrate transferability of the input representation but not superiority over online mapping in unseen environments or sensor configurations. Please add corresponding MapTRv2 cross-dataset numbers or adjust the generalization claim to match the evidence.
  4. [§4.2, Table 2] The sensor-agnostic claim is supported on nuScenes by a single detector (FocalFormer3D, 43.7 AP vs 44.6 AP with ground-truth annotations), while the Radar Las Vegas row (30.0 AP) has no ground-truth-trail baseline and no comparison method on the same closed dataset. As reported, the Radar result cannot be interpreted as evidence of robustness across sensor configurations. Please provide a same-dataset baseline for the Radar row or report the comparison conditions and the expected performance level.
minor comments (5)
  1. [Fig. 4 caption] There is a typo in 'geograhical'; it should read 'geographical'.
  2. [Table 2] The row labeled 'Ground truth' is not a sensor foundation; rename it to something like 'Ground-truth annotations (no sensor noise)' and clarify which rows use sensor-based detections.
  3. [§4.1] The statement that the datasets 'do not provide trail data for every centerline element' should be accompanied by quantitative coverage statistics, since the coverage fraction determines the scope of every reported AP value.
  4. [§4] The paper says the experiments are designed for reproducibility but does not report key hyperparameters (tile dimensions, grid resolution, number of decoder queries, number of points m, training schedule, optimizer settings) or release code; please include these details.
  5. [Eq. (1)] The half-open interval notation for the last yaw bin is slightly unclear; consider writing the last interval explicitly as (π - 2π/n, π].

Circularity Check

1 steps flagged · score 2.0 of 10

Trail-coverage filtering makes the evaluation target a subset defined by the input; the rest of the derivation is self-contained and externally benchmarked.

  1. self definitional [Section 3.1 (Input and ground truth representation) and Section 4.2 (Comparison to online HD mapping approaches), with confirming Figure 4 caption]
    "To generate the training data, we extract centerlines from the HD map that correspond to regions with nonzero trail density while explicitly filtering out centerlines that lack associated trails. ... For training and evaluation of our trail-based approach, we delete ground truth data for areas with no trails. ... The samples from the nuScenes dataset lack sufficient amounts of trails to depict the entirety of lanes in the map tile, so the ground truth was refined as described in Sec. 3.1."

    The ground-truth centerline set used for both training and evaluation is defined as the subset of HD centerlines that overlap with the trail input. Because the input is a trail-density grid, lanes without trails have zero input signal and are removed from the target. The reported AP values (56.4 and 44.6) therefore measure reconstruction of trail-covered lanes only, and the abstract/conclusion claim of constructing a 'comprehensive global map' is, by construction, a claim about the trail-covered subset rather than the full road network. The paper discloses this filtering and even masks MapTRv2 on the same trail areas, so the comparison is fair, but the headline generalization claim is not independently supported for untraversed regions.

full rationale

The core derivation chain is not circular: the input trail grids are derived from recorded trajectories and ground-truth centerlines come from external HD maps; training and validation use geographically disjoint splits; and the MapTRv2 baseline is retrained and evaluated under the same masking. The main circularity is a scoping one: the target set is filtered by trail presence, which is itself derived from the input modality. This makes 'global map construction' equivalent to 'construction of the trail-covered subset' and prevents the evaluation from measuring inference of untraversed lanes. This is a disclosed, explicit filtering rather than a hidden statistical fit, so it does not invalidate the comparison, but it does lower the circularity score to 2 rather than 0. No other circular step was found: the model architecture borrows from MapTRv2 as an external baseline, the sensor-agnostic claim is tested with FocalFormer3D detections and a closed Radar-based dataset, and the ablation studies vary input channels and augmentation without renaming fitted parameters as predictions.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No new physical entities are proposed. The central claim rests on standard supervised learning with external HD maps as ground truth; the main assumption burden is that trail coverage is sufficient to define the full mapping task and that the chosen hyperparameters generalize. The free parameters are model and input hyperparameters selected on validation data, not quantities derived from first principles.

free parameters (5)
  • Gaussian smoothing sigma = 2
    Selected by validation AP on the nuScenes geographic split (Table 3); sigma=2 gives the best AP (44.6).
  • Number of yaw direction bins n = 6
    Selected by validation AP (Table 4); n=6 gives the best AP (44.6) among {4, 6, 12, 36}.
  • Number of decoder queries (max predicted centerlines per tile)
    Tunable hyperparameter correlated with the number of centerlines; exact value not reported in the paper.
  • Fixed number of points m per centerline
    Ground truth and predictions use a fixed number of ordered points m; m is not reported.
  • Tile dimensions and grid resolution
    Real-world tile size w_glob x h_glob and resolution r determine input grid size; values are not reported, making exact replication impossible.
assumptions (4)
  • domain assumption nuScenes and nuPlan HD centerline maps are accurate and globally aligned with the trail coordinate system.
    Ground truth for training and evaluation is taken from these maps without independent verification of their accuracy.
  • domain assumption Vehicle trajectories approximate drivable lane centerlines.
    The method treats aggregated ego and object trails as the signal from which lane centerlines are inferred; this holds only if vehicles drive on lanes and tracking is accurate.
  • ad hoc to paper Centerlines without trails can be excluded from training and evaluation without materially changing the measured performance.
    The paper removes such centerlines from ground truth (Sec. 3.1) and masks Chamfer distances for no-trail areas (Sec. 4.2), restricting the problem to trail-covered regions.
  • domain assumption The AP metric with Chamfer thresholds at 0.5, 1.0, 1.5 meters is a valid measure of map quality.
    This is the standard online mapping metric, but it does not capture topological errors such as missing or disconnected lanes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Inferring Driving Maps by Deep Learning-based Trail Map Extraction." pith.science (2026). https://pith.science/paper/WFRMQSLZ

@misc{pith2026250510258,
  author       = {Pith},
  title        = {Pith review of: Inferring Driving Maps by Deep Learning-based Trail Map Extraction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WFRMQSLZ}},
  note         = {Machine review of arXiv:2505.10258}
}
read the original abstract

High-definition (HD) maps offer extensive and accurate environmental information about the driving scene, making them a crucial and essential element for planning within autonomous driving systems. To avoid extensive efforts from manual labeling, methods for automating the map creation have emerged. Recent trends have moved from offline mapping to online mapping, ensuring availability and actuality of the utilized maps. While the performance has increased in recent years, online mapping still faces challenges regarding temporal consistency, sensor occlusion, runtime, and generalization. We propose a novel offline mapping approach that integrates trails - informal routes used by drivers - into the map creation process. Our method aggregates trail data from the ego vehicle and other traffic participants to construct a comprehensive global map using transformer-based deep learning models. Unlike traditional offline mapping, our approach enables continuous updates while remaining sensor-agnostic, facilitating efficient data transfer. Our method demonstrates superior performance compared to state-of-the-art online mapping approaches, achieving improved generalization to previously unseen environments and sensor configurations. We validate our approach on two benchmark datasets, highlighting its robustness and applicability in autonomous driving systems.

Figures

Figures reproduced from arXiv: 2505.10258 by the authors.

Figure 1
Figure 1. Visualization of the ground truth lane centerlines and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall architecture of the trail-based mapping approach. Starting with the trail aggregation (a), object and ego trails are [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Visualization of an example sample from the nuScenes dataset. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visualization of ground truth (red), model predictions (blue) and its input grid map on an aerial image. The samples are from [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 20 canonical work pages

  1. [1]

    High-Definition Map Generation Technologies for Au- tonomous Driving, 2022

    Zhibin Bao, Sabir Hossain, Haoxiang Lang, and Xianke Lin. High-Definition Map Generation Technologies for Au- tonomous Driving, 2022. arXiv:2206.05400 [cs]. 1, 2

    arXiv 2022

  2. [2]

    Inferring Road Maps from Global Positioning System Traces: Survey and Com- parative Evaluation

    James Biagioni and Jakob Eriksson. Inferring Road Maps from Global Positioning System Traces: Survey and Com- parative Evaluation. Transportation Research Record: Jour- nal of the Transportation Research Board , 2291(1):61–71,

    work page

  3. [3]

    Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom

    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 multimodal dataset for autonomous driving, 2020. arXiv:1903.11027 [cs]. 6

    arXiv 2020

  4. [4]

    NuPlan: A closed-loop ML- based planning benchmark for autonomous vehicles, 2022

    Holger Caesar, Juraj Kabzan, Kok Seang Tan, Whye Kit Fong, Eric Wolff, Alex Lang, Luke Fletcher, Oscar Bei- jbom, and Sammy Omari. NuPlan: A closed-loop ML- based planning benchmark for autonomous vehicles, 2022. arXiv:2106.11810 [cs]. 6

    arXiv 2022

  5. [5]

    End-to-End Object Detection with Transformers, 2020

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nico- las Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to-End Object Detection with Transformers, 2020. arXiv:2005.12872 [cs]. 2

    arXiv 2020

  6. [6]

    MapTracker: Tracking with Strided Memory Fusion for Consistent Vector HD Mapping, 2024

    Jiacheng Chen, Yuefan Wu, Jiaqi Tan, Hang Ma, and Ya- sutaka Furukawa. MapTracker: Tracking with Strided Memory Fusion for Consistent Vector HD Mapping, 2024. arXiv:2403.15951 [cs]. 3

    arXiv 2024

  7. [7]

    Focal- Former3D : Focusing on Hard Instance for 3D Object Detec- tion, 2023

    Yilun Chen, Zhiding Yu, Yukang Chen, Shiyi Lan, Ani- mashree Anandkumar, Jiaya Jia, and Jose Alvarez. Focal- Former3D : Focusing on Hard Instance for 3D Object Detec- tion, 2023. arXiv:2308.04556 [cs]. 8

    arXiv 2023

  8. [8]

    HD Lane Map Generation Based on Trail Map Aggregation

    Pascal Colling, Dennis Muller, and Matthias Rottmann. HD Lane Map Generation Based on Trail Map Aggregation. In 2022 IEEE Intelligent Vehicles Symposium (IV), pages 600– 606, Aachen, Germany, 2022. IEEE. 3

    work page 2022

Show all 36 references
  1. [9]

    Parting with Misconceptions about Learning-based Vehicle Motion Planning, 2023

    Daniel Dauner, Marcel Hallgarten, Andreas Geiger, and Kashyap Chitta. Parting with Misconceptions about Learning-based Vehicle Motion Planning, 2023. arXiv:2306.07962 [cs]. 2

    2023 arXiv

  2. [10]

    FlowMap: Path Generation for Automated Vehicles in Open Space Using Traffic Flow

    Wenchao Ding, Jieru Zhao, Yubin Chu, Haihui Huang, Tong Qin, Chunjing Xu, Yuxiang Guan, and Zhongxue Gan. FlowMap: Path Generation for Automated Vehicles in Open Space Using Traffic Flow. In 2023 IEEE International Con- ference on Robotics and Automation (ICRA) , pages 1616– 1...

    2023

  3. [11]

    Christopher Doer, Michael Henzler, Heiner Messner, and Gert F. Trommer. HD Map Generation from Vehicle Fleet Data for Highly Automated Driving on Highways. In 2020 IEEE Intelligent Vehicles Symposium (IV), pages 2014–2020, Las Vegas, NV , USA, 2020. IEEE. 3

    2020

  4. [12]

    A Low-Cost Solution for Automatic Lane-Level Map Generation Using Conventional In-Car Sensors

    Chunzhao Guo, Kiyosumi Kidono, Junichi Meguro, Yoshiko Kojima, Masaru Ogawa, and Takashi Naito. A Low-Cost Solution for Automatic Lane-Level Map Generation Using Conventional In-Car Sensors. IEEE Transactions on Intelli- gent Transportation Systems, 17(8):2355–2366, 2016. 3

    2016

  5. [13]

    Machine Learning Assisted High-Definition Map Creation

    Jialin Jiao. Machine Learning Assisted High-Definition Map Creation. In 2018 IEEE 42nd Annual Computer Software and Applications Conference (COMPSAC), pages 367–373, Tokyo, Japan, 2018. IEEE. 1

    2018

  6. [14]

    HD Map Update for Autonomous Driving With Crowdsourced Data

    Kitae Kim, Soohyun Cho, and Woojin Chung. HD Map Update for Autonomous Driving With Crowdsourced Data. IEEE Robotics and Automation Letters , 6(2):1895–1901,

    1901

  7. [15]

    HDMapNet: An Online HD Map Construction and Evaluation Frame- work, 2022

    Qi Li, Yue Wang, Yilun Wang, and Hang Zhao. HDMapNet: An Online HD Map Construction and Evaluation Frame- work, 2022. arXiv:2107.06307 [cs]. 2

    2022 arXiv

  8. [16]

    DTCLMapper: Dual Temporal Consistent Learning for Vectorized HD Map Con- struction, 2024

    Siyu Li, Jiacheng Lin, Hao Shi, Jiaming Zhang, Song Wang, You Yao, Zhiyong Li, and Kailun Yang. DTCLMapper: Dual Temporal Consistent Learning for Vectorized HD Map Con- struction, 2024. arXiv:2405.05518 [cs, eess]. 3

    2024 arXiv

  9. [17]

    MapTR: Structured Modeling and Learning for Online Vectorized HD Map Construction, 2023

    Bencheng Liao, Shaoyu Chen, Xinggang Wang, Tianheng Cheng, Qian Zhang, Wenyu Liu, and Chang Huang. MapTR: Structured Modeling and Learning for Online Vectorized HD Map Construction, 2023. arXiv:2208.14437 [cs]. 2, 3

    2023 arXiv

  10. [18]

    MapTRv2: An End-to-End Framework for Online Vector- ized HD Map Construction, 2023

    Bencheng Liao, Shaoyu Chen, Yunchi Zhang, Bo Jiang, Qian Zhang, Wenyu Liu, Chang Huang, and Xinggang Wang. MapTRv2: An End-to-End Framework for Online Vector- ized HD Map Construction, 2023. arXiv:2308.05736 [cs]. 2, 3, 5, 7

    2023 arXiv

  11. [19]

    Lane Graph as Path: Continuity-preserving Path- wise Modeling for Online Lane Graph Construction, 2024

    Bencheng Liao, Shaoyu Chen, Bo Jiang, Tianheng Cheng, Qian Zhang, Wenyu Liu, Chang Huang, and Xinggang Wang. Lane Graph as Path: Continuity-preserving Path- wise Modeling for Online Lane Graph Construction, 2024. arXiv:2303.08815 [cs]. 5

    2024 arXiv

  12. [20]

    Localization is All You Evaluate: Data Leak- age in Online Mapping Datasets and How to Fix it

    Adam Lilja, Junsheng Fu, Erik Stenborg, and Lars Ham- marstrand. Localization is All You Evaluate: Data Leak- age in Online Mapping Datasets and How to Fix it. In 2024 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 22150–22159, Seattle, W A, USA...

    2024

  13. [21]

    MGMap: Mask-Guided Learn- ing for Online Vectorized HD Map Construction

    Xiaolu Liu, Song Wang, Wentong Li, Ruizi Yang, Junbo Chen, and Jianke Zhu. MGMap: Mask-Guided Learn- ing for Online Vectorized HD Map Construction. In 2024 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 14812–14821, Seattle, W A, USA, 2024. IEEE. 3

    2024

  14. [22]

    VectorMapNet: End-to-end Vectorized HD Map Learning, 2023

    Yicheng Liu, Tianyuan Yuan, Yue Wang, Yilun Wang, and Hang Zhao. VectorMapNet: End-to-end Vectorized HD Map Learning, 2023. arXiv:2206.08920 [cs]. 2, 3

    2023 arXiv

  15. [23]

    Leveraging Enhanced Queries of Point Sets for Vectorized Map Construction, 2024

    Zihao Liu, Xiaoyu Zhang, Guangwei Liu, Ji Zhao, and Ningyi Xu. Leveraging Enhanced Queries of Point Sets for Vectorized Map Construction, 2024. arXiv:2402.17430 [cs]. 3

    2024 arXiv

  16. [24]

    Massow, B

    K. Massow, B. Kwella, N. Pfeifer, F. H ¨ausler, J. Pontow, I. Radusch, J. Hipp, F. D ¨olitzscher, and M. Haueis. Deriving 9 HD maps for highly automated driving from vehicular probe data. In 2016 IEEE 19th International Conference on Intelli- gent Transportation Systems (ITSC)...

    2016

  17. [25]

    Anytime Lane-Level Intersection Estimation Based on Trajectories of Other Traffic Participants, 2019

    Annika Meyer, Jonas Walter, Martin Lauer, and Christoph Stiller. Anytime Lane-Level Intersection Estimation Based on Trajectories of Other Traffic Participants, 2019. arXiv:1906.02495 [cs]. 3

    2019 arXiv

  18. [26]

    Fast Lane-Level Intersection Estimation using Markov Chain Monte Carlo Sampling and B-Spline Refinement, 2020

    Annika Meyer, Jonas Walter, and Martin Lauer. Fast Lane-Level Intersection Estimation using Markov Chain Monte Carlo Sampling and B-Spline Refinement, 2020. arXiv:2007.06904 [cs]. 3

    2020 arXiv

  19. [27]

    LMT-Net: Lane Model Transformer Network for Auto- mated HD Mapping from Sparse Vehicle Observations,

    Michael Mink, Thomas Monninger, and Steffen Staab. LMT-Net: Lane Model Transformer Network for Auto- mated HD Mapping from Sparse Vehicle Observations,

  20. [28]

    How to Keep HD Maps for Automated Driv- ing Up To Date

    David Pannen, Martin Liebner, Wolfgang Hempel, and Wol- fram Burgard. How to Keep HD Maps for Automated Driv- ing Up To Date. In 2020 IEEE International Conference on Robotics and Automation (ICRA) , pages 2288–2294, Paris, France, 2020. IEEE. 2, 3

    2020

  21. [29]

    PrevPredMap: Explor- ing Temporal Modeling with Previous Predictions for Online Vectorized HD Map Construction, 2024

    Nan Peng, Xun Zhou, Mingming Wang, Xiaojun Yang, Songming Chen, and Guisong Chen. PrevPredMap: Explor- ing Temporal Modeling with Previous Predictions for Online Vectorized HD Map Construction, 2024. arXiv:2407.17378 [cs]. 3

    2024 arXiv

  22. [30]

    Traffic Flow-Based Crowdsourced Map- ping in Complex Urban Scenario

    Tong Qin, Haihui Huang, Ziqiang Wang, Tongqing Chen, and Wenchao Ding. Traffic Flow-Based Crowdsourced Map- ping in Complex Urban Scenario. IEEE Robotics and Au- tomation Letters, 8(8):5077–5083, 2023. 3

    2023

  23. [31]

    Towards a Functional System Architecture for Automated Vehicles, 2017

    Simon Ulbrich, Andreas Reschka, Jens Rieken, Susanne Ernst, Gerrit Bagschik, Frank Dierkes, Marcus Nolte, and Markus Maurer. Towards a Functional System Architecture for Automated Vehicles, 2017. arXiv:1703.08557 [cs]. 2

    2017 arXiv

  24. [32]

    Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecasting, 2023

    Benjamin Wilson, William Qi, Tanmay Agarwal, John Lam- bert, Jagjeet Singh, Siddhesh Khandelwal, Bowen Pan, Rat- nesh Kumar, Andrew Hartnett, Jhony Kaesemodel Pontes, Deva Ramanan, Peter Carr, and James Hays. Argoverse 2: Next Generation Datasets for Self-Driving Perception an...

    2023 arXiv

  25. [33]

    Neural Map Prior for Autonomous Driving

    Xuan Xiong, Yicheng Liu, Tianyuan Yuan, Yue Wang, Yilun Wang, and Hang Zhao. Neural Map Prior for Autonomous Driving. In 2023 IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR), pages 17535–17544, Vancouver, BC, Canada, 2023. IEEE. 3

    2023

  26. [34]

    StreamMapNet: Streaming Mapping Net- work for Vectorized Online HD Map Construction

    Tianyuan Yuan, Yicheng Liu, Yue Wang, Yilun Wang, and Hang Zhao. StreamMapNet: Streaming Mapping Net- work for Vectorized Online HD Map Construction. In 2024 IEEE/CVF Winter Conference on Applications of Com- puter Vision (WACV), pages 7341–7350, Waikoloa, HI, USA,

    2024

  27. [35]

    Transformer-based models and hardware acceleration analysis in autonomous driving: A survey, 2023

    Juan Zhong, Zheng Liu, and Xi Chen. Transformer-based models and hardware acceleration analysis in autonomous driving: A survey, 2023. arXiv:2304.10891 [cs]. 2 10

    2023 arXiv

  28. [2024]

    arXiv:2409.12409 [cs]. 3

    arXiv

Pith tools

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