Pith. sign in

REVIEW 3 major objections 6 minor 36 references

PriorFusion: Unified Integration of Priors for Robust Road Perception in Autonomous Driving

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Fusing semantic, geometric, and generative priors lifts road-element perception to 70.4% mAP and sharply boosts strict-threshold accuracy.

desk verdict Plausible plug-and-play prior integration for vectorized map perception, but the anchor-construction protocol is under-specified and the SOTA claim rests on an epoch mismatch. read the letter →

arxiv 2507.23309 v2 pith:F3G2BX5G submitted 2025-07-31 cs.CV

Xuewei Tang , Mengmeng Yang , Tuopu Wen , Peijin Jia , Le Cui , Mingshang Luo , Kehua Sheng , Bo Zhang
show 2 more authors
Diange Yang Kun Jiang
This is my paper
classification cs.CV
keywords autonomousdrivingroadelementperceptionvectorizedHDmapconstructionshapepriorsdiffusionmodelSVDtemplatespacebird'seyeviewnuScenes
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

PriorFusion claims that road perception for autonomous driving can be made markedly more accurate by systematically feeding the model three kinds of prior knowledge about road elements: semantic segmentation information, geometric shape templates, and a generative diffusion prior. The paper builds a low-dimensional 'shape template space' by applying SVD to road-element coordinates, clusters the dataset in that space to obtain typical anchor shapes, and uses those anchors plus a truncated diffusion loop in the vector decoder. On the nuScenes benchmark, the full system (PriorFusion-V2) reaches 70.4% mAP with a ResNet50 backbone, and under a stricter Chamfer-distance threshold of 0.2 the mAP jumps from 8.55% for the baseline to 15.81%. The stated point is that incorporating priors makes predicted road elements more regular, complete, and closely aligned with ground truth, especially in cluttered or occluded scenes. The paper also claims the modules are plug-and-play, adding small runtime and parameter overhead.

What carries the argument

The load-bearing object is the shape template space: a low-dimensional subspace spanned by the first $M$ left singular vectors of the road-element matrix $A$ (Eqs. 10-14), obtained by SVD, $A = U\Sigma V^\top$. Because each road element $r$ can be approximated as $\tilde{r} = U_M c$ with $c = U_M^\top r$, the network regresses $M$ template coefficients rather than raw coordinates. The orthogonal basis makes the map into this space isometric, so K-Means clustering of the coefficient matrix $C_A = U_M^\top A$ preserves Euclidean geometry while cutting dimensionality. The resulting prior anchors $P_A = U_M P_C$ serve as learned initial reference points for the transformer instance decoder, and the truncated diffusion loop (Equation 22) adds Gaussian noise to these anchors and denoises them through the decoder for $T=2$ iterations, with the decoder output feeding back as new reference points.

What would settle it

Recompute the SVD basis and the K-Means prior anchors using only the nuScenes training split, retrain PriorFusion-V2 under the same settings, and re-evaluate at τ=0.2; if the mAP gain from 8.55% to 15.81% largely disappears, the central claim of generalizable prior learning is not supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that road element perception improves substantially when semantic, geometric, and generative priors are integrated into the decoder pipeline rather than treated as post-processing. In concrete terms, the authors propose PriorFusion, which (1) extracts shape-aware weights from the semantic segmentation head and multiplies them into the detection queries; (2) builds a shape template space by SVD of the road element coordinate matrix, clusters that space to obtain prior anchors that replace random initial reference points; and (3) runs a truncated diffusion process that adds noise to the anchors and denoises them through the decoder over two steps. The authors report that on nuScenes with camera input and ResNet50, PriorFusion-V2 achieves 70.4% mAP, outperforming the replicated MapQR baseline (65.2%) and all compared prior methods, and that at the stricter Chamfer threshold of 0.2 the mAP rises from 8.55% to 15.81%, which they attribute to predictions whose shapes align more closely with ground truth.

Load-bearing premise

The prior anchors and shape template space are computed from ground-truth road element coordinates across the nuScenes dataset, and the paper does not state that only the training split is used; if validation or test annotations enter the SVD basis or the K-Means cluster centers, the reported accuracy gains would reflect evaluation-set information rather than learned generalization.

Editorial extensions

If this is right

  • With a ResNet50 backbone and camera inputs, PriorFusion-V2 achieves 70.4% mAP on nuScenes, surpassing the replicated MapQR baseline at 65.2%.
  • At the stricter Chamfer threshold of 0.2, mAP rises from 8.55% to 15.81%, a 7.26 percentage-point gain, indicating predictions that align more closely with ground-truth shape.
  • The shape-prior anchor module adds only 9M parameters and 0.25 FPS in the 24-epoch ablation, and the SPG query refinement adds 62M parameters with minimal speed loss, supporting the plug-and-play claim.
  • The truncated diffusion loop runs at 7.98 FPS with two decoder iterations, in contrast to the 0.60 FPS and 0.32 FPS reported for MapPrior and DiffMap, making the approach far more practical for onboard use.
  • Applying the modules to the MapQR decoder improves all three road element classes at τ=0.2, with dividers improving most (19.49% to 30.82%).

Reading between the lines

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

  • If the prior anchors are recomputed on the training split only, the same SVD-plus-clustering recipe could be a general plug-in for other vectorized-map decoders; the paper itself notes re-clustering is needed for new datasets, so the transferable part is the decoder integration, not the anchor values.
  • The τ=0.2 result is a shape-fidelity signal: downstream planning that needs precise boundary distances could benefit even where standard mAP gains are modest.
  • Attaching the three modules to a different strong decoder, such as MapTRv2 or HIMap, would test the plug-and-play claim more broadly than the MapQR-only demonstration.
  • Because the paper reports that nonlinear VAE shape embedding was not successfully integrated, the linear SVD template space may cap achievable shape fidelity; a nonlinear template space is a plausible next step.
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

3 major / 6 minor

Summary. PriorFusion proposes a unified framework for vectorized road-element perception in BEV space, combining three priors: a shape-prior-guided query refinement module that injects semantic segmentation features into detection queries, a data-driven shape template space built via SVD of road-element coordinates with K-means clustering to generate prior anchors, and a truncated diffusion process initialized from those anchors. The method is evaluated on nuScenes, reporting a state-of-the-art mAP of 70.4% (PriorFusion-V2) and a 7.26 percentage-point improvement over the MapQR baseline at the stricter Chamfer threshold of 0.2. Ablations are provided for each module, along with runtime and parameter overhead measurements.

Significance. If the reported results hold, PriorFusion offers a practical way to inject structured geometric priors into existing DETR-style map perception decoders with modest overhead, and the SVD-then-cluster procedure is a sensible extension of prior work on eigen-shape descriptors. The paper's strengths include module-level ablations, explicit runtime/parameter reporting, and the plug-and-play framing. However, the central empirical claims rest on two protocol conditions that are not currently documented: the prior anchors and shape template basis must be computed only from the training split, and the headline comparison must not conflate architectural gains with longer training. Both are verifiable and fixable, but they are load-bearing for the paper's main conclusions.

major comments (3)
  1. [Section 3.2.1, Algorithm 1, Section 4.1] The shape template basis U_M and prior anchor matrix PA are constructed from "all road elements in the dataset" (Eq. 10) and "Element matrix of the dataset A" (Algorithm 1), with no statement that only the nuScenes training split is used. Because these anchors replace the decoder's initial reference points and initialize the truncated diffusion process (Eq. 22), computing them from validation or test ground truth would leak positional and shape information from the evaluation set into inference. This is especially consequential for Table 2, where the +7.26 pp gain at threshold 0.2 could be inflated by anchors that coincide with evaluation-set ground-truth locations. The paper's own limitation statement in Section 5.1 that priors are dataset-specific makes the missing split specification material. Please clarify explicitly that A and Algorithm 1 use only the training split, and if that is not the case, rerun the affected experiments.
  2. [Table 1 and Section 4.2] The state-of-the-art comparison is confounded with training epochs: PriorFusion-V2 is trained for 110 epochs, while most baselines (MapTRv2, MapQR, InsightMapper, PivotNet, MapTR) are trained for 24 epochs. The 70.4% mAP may thus reflect longer training rather than the proposed priors. The controlled ablation in Table 5 (Model3 vs Model4, both 110 epochs) does show a benefit from the diffusion step, and PriorFusion-V1 at 24 epochs outperforms MapQR (66.3% vs 65.2%), but the headline claim should either use matched training budgets or explicitly state that the 110-epoch comparison is not apples-to-apples.
  3. [Section 3.2.1, Eq. (10)-(13)] The claim that the SVD-based templates exhibit rotation and translation invariance is not justified by the described construction. SVD of the raw coordinate matrix A in Eq. (10) is not invariant to translation or rotation unless the element coordinates are aligned or centered, and no such preprocessing is described. This matters because the prior anchors are used as reference points in a fixed ego-vehicle coordinate frame. Please describe the alignment/centering procedure, or temper the invariance claim accordingly.
minor comments (6)
  1. [Eq. (13)] The text contains a dangling reference "equation ??" instead of a proper citation to the Frobenius-norm expression; please fix the cross-reference.
  2. [Algorithm 1] The algorithm labels the inverse transformation as "Step 5" after Step 3, with no Step 4; renumber the steps consistently.
  3. [Section 3.4.7] "out model kept in line" should read "our model kept in line".
  4. [Table 6 and surrounding text] The text states that each additional decoder execution incurs an average of 46.89 ms, but the first diffusion step adds only 3.11 ms; please reconcile this discrepancy by explaining why the first decoder pass is much cheaper than subsequent ones.
  5. [Table 2] The training-epoch settings for Baseline, PriorFusion-V1, and PriorFusion-V2 are not stated in the table or its caption; specify them for interpretability.
  6. [Introduction and Section 3.4.7] Several instances of "plug-and-play" are malformed ("asplug-and-play", "aplug-and-play"); fix the spacing.

Circularity Check

1 steps flagged · score 6.0 of 10

Prior anchors and shape templates are fitted to 'the dataset' ground-truth matrix A with no documented train/val split, and the same nuScenes ground truth is then used to compute AP; if evaluation labels enter A, the reported +7.26 pp gain at tau=0.2 is forced by construction.

  1. fitted input called prediction [Section 3.2.1 Eq. (10)-(17), Algorithm 1, Section 4.1 Dataset/Evaluation Metrics, Table 2]
    "'All road elements in the dataset can be combined into the road element matrix A of the dataset' (Eq. 10); Algorithm 1 takes as input 'Element matrix of the dataset A' and selects the coefficient 'closest to the cluster centers as the coefficients of prior anchor'; the evaluation metric is 'calculated based on Chamfer distance..., which determines whether a predicted element matches the ground truth (GT).'"

    Prior anchors PA are cluster medoids selected from the dataset's ground-truth element matrix A (Eqs. 10-17, Algorithm 1); the shape basis UM is the SVD of the same matrix, and the paper never restricts A to the nuScenes training split in Section 3.2 or Section 4.1. The same nuScenes ground truth is the target of the AP metric (Eqs. 26-27, Table 2), and PA is injected as decoder reference points and as the starting points of truncated diffusion (Eq. 22). If validation/test elements enter A, inference is initialized at cluster medoids drawn from the evaluation labels, so low Chamfer-distance matches, especially the +7.26 pp mAP gain at tau=0.2, are forced by construction rather than by learned generalization.

full rationale

The SVD/K-means/diffusion pipeline is internally coherent and largely self-contained: the equations for PCA reconstruction (Eqs. 10-16), isometric clustering in the coefficient space (Eq. 17), and truncated diffusion (Eqs. 18-23) are standard and do not by themselves reduce to the evaluation metric. The self-citations in the related work (e.g., DiffMap, which shares authors) are used for speed comparisons and motivation, not as load-bearing proof, so they do not raise the score. The one material circularity is the provenance of the fitted prior: Algorithm 1 consumes 'the dataset' element matrix A, and Section 4.1 evaluates on 'the nuScenes dataset', with no sentence saying A is built only from the training split. Because the anchors are literally chosen as the dataset elements closest to K-means centers (Algorithm 1, Step 16), and because they initialize the decoder and diffusion process, any contribution from validation/test elements to A would make the reported mAP, especially the 7.26 pp jump at the strict tau=0.2 threshold, an artifact of fitting to the evaluation labels. The paper's own limitation that the priors are dataset-specific and need re-clustering across datasets confirms that this is not a fixed external prior. If the authors clarify that A is train-split-only and the anchors are held fixed before evaluation, the remaining derivation is independent and the score would drop to 1-2.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim relies on the low-rank assumption for road element shapes and on an evaluation protocol that uses anchors computed from the same dataset's ground truth. The SVD basis and cluster centers are data-fitted, not derived from first principles. No new physical entities are introduced.

free parameters (6)
  • Shape template space dimension M = 20
    Chosen by hand; controls the truncation error of the SVD reconstruction and the dimensionality of the clustering space (Section 3.2.1).
  • Number of prior anchors NP = equal to number of queries
    Set to the number of detection queries, not optimized (Section 3.2.2).
  • Diffusion step number T = 2
    Selected as a speed-accuracy trade-off; the ablation reports timing for T=1,2,3 but not accuracy for T=1 and T=3.
  • Loss weights lambda_cls, lambda_pts, lambda_dir, lambda_seg, lambda_shape = 2, 5, 0.005, 2, 0.001
    Set manually in Section 4.1; no sensitivity analysis is provided.
  • SVD basis U_M (shape templates) = data-derived left singular vectors
    Fitted to the ground-truth road element matrix A of the dataset (Eq. 10-11); this is a data-fitted quantity used as the core prior.
  • K-Means cluster centers (prior anchors) = data-derived cluster centers
    Computed by clustering the template-space coefficients of dataset ground-truth elements (Algorithm 1); used as decoder reference points.
assumptions (4)
  • domain assumption Road element shapes lie approximately in a low-dimensional linear subspace of dimension M=20.
    Section 3.2.1 truncates the SVD to M=20 components and claims the reconstruction error is tolerable because later PCA coefficients are small. This is an empirical assumption about the dataset, not a mathematical guarantee.
  • standard math Mapping to the shape template space is isometric because U_M is orthogonal.
    Used in Section 3.2.2 (Eq. 17 and surrounding text) to argue that K-Means in template space is equivalent to clustering in Euclidean space. This follows from orthogonality of the singular vectors.
  • standard math The Gaussian diffusion forward process in Eqs. 18-20 is a valid noise augmentation for road element coordinates.
    The equations are the standard DDPM forward process; the assumption is that adding Gaussian noise to keypoint coordinates is a meaningful perturbation for the diffusion training procedure.
  • domain assumption The reconstruction error from discarding N-M components is small enough not to harm perception accuracy.
    Section 3.2.2 states the later principal components have smaller coefficients, so the error is tolerable. This is a heuristic based on the data distribution, not a proven bound.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PriorFusion: Unified Integration of Priors for Robust Road Perception in Autonomous Driving." pith.science (2026). https://pith.science/paper/F3G2BX5G

@misc{pith2026250723309,
  author       = {Pith},
  title        = {Pith review of: PriorFusion: Unified Integration of Priors for Robust Road Perception in Autonomous Driving},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F3G2BX5G}},
  note         = {Machine review of arXiv:2507.23309}
}
read the original abstract

With the growing interest in autonomous driving, there is an increasing demand for accurate and reliable road perception technologies. In complex environments without high-definition map support, autonomous vehicles must independently interpret their surroundings to ensure safe and robust decision-making. However, these scenarios pose significant challenges due to the large number, complex geometries, and frequent occlusions of road elements. A key limitation of existing approaches lies in their insufficient exploitation of the structured priors inherently present in road elements, resulting in irregular, inaccurate predictions. To address this, we propose PriorFusion, a unified framework that effectively integrates semantic, geometric, and generative priors to enhance road element perception. We introduce an instance-aware attention mechanism guided by shape-prior features, then construct a data-driven shape template space that encodes low-dimensional representations of road elements, enabling clustering to generate anchor points as reference priors. We design a diffusion-based framework that leverages these prior anchors to generate accurate and complete predictions. Experiments on large-scale autonomous driving datasets demonstrate that our method significantly improves perception accuracy, particularly under challenging conditions. Visualization results further confirm that our approach produces more accurate, regular, and coherent predictions of road elements.

Figures

Figures reproduced from arXiv: 2507.23309 by the authors.

Figure 1
Figure 1. (a) shows the ground truth. (b) shows the problem of general methods [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The pipeline of our vectorized road element detection model. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The Segmentation-Vector instance consistency weight encoding module. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Visualization of the first 5 eigenvectors for all road elements, pedes [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Compare the visualization of prior reference points obtained through random initialization and clustering in template space. Randomly initialized points [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Compare the decoder of baseline model and our method. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: The overall pipeline of our vectorized road elements perception method. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Visualization results. The visualized scene images are di [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: Additional Visualization results. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

36 extracted references · 25 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    , author Oh, J

    author Bae, I. , author Oh, J. , author Jeon, H.G. , year 2023 . title Eigentrajectory: Low-rank descriptors for multi-modal trajectory forecasting , in: booktitle Proceedings of the IEEE/CVF International Conference on Computer Vision , pp. pages 10017--10029

    work page 2023

  3. [3]

    , author Bankiti, V

    author Caesar, H. , author Bankiti, V. , author Lang, A.H. , author Vora, S. , author Liong, V.E. , author Xu, Q. , author Krishnan, A. , author Pan, Y. , author Baldan, G. , author Beijbom, O. , year 2020 . title nuscenes: A multimodal dataset for autonomous driving , in: booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern rec...

    2020

  4. [4]

    , author Massa, F

    author Carion, N. , author Massa, F. , author Synnaeve, G. , author Usunier, N. , author Kirillov, A. , author Zagoruyko, S. , year 2020 . title End-to-end object detection with transformers , in: booktitle Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part I 16 , organization Springer . pp. pages 213--229

    work page 2020

  5. [5]

    PivotNet: Vectorized Pivot Learning for End-to-end HD Map Construction

    author Ding, W. , author Qiao, L. , author Qiu, X. , author Zhang, C. , year 2023 . title Pivotnet: Vectorized pivot learning for end-to-end hd map construction . journal arXiv preprint arXiv:2308.16477

    work page Pith review arXiv 2023

  6. [6]

    , author Guo, S

    author Feng, Z. , author Guo, S. , author Tan, X. , author Xu, K. , author Wang, M. , author Ma, L. , year 2022 . title Rethinking efficient lane detection via curve modeling , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. pages 17062--17070

    work page 2022

  7. [7]

    , author Shen, L

    author Hu, J. , author Shen, L. , author Sun, G. , year 2018 . title Squeeze-and-excitation networks , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 7132--7141

    work page 2018

  8. [8]

    , author Huang, G

    author Huang, J. , author Huang, G. , year 2022 . title Bevpoolv2: A cutting-edge implementation of bevdet toward deployment . journal arXiv preprint arXiv:2211.17111

    arXiv 2022

Show all 36 references
  1. [9]

    , author Wen, T

    author Jia, P. , author Wen, T. , author Luo, Z. , author Yang, M. , author Jiang, K. , author Liu, Z. , author Tang, X. , author Lei, Z. , author Cui, L. , author Zhang, B. , et al., year 2024 . title Diffmap: Enhancing map segmentation with map prior using diffusion model . ...

    2024

  2. [10]

    , author Park, W

    author Jin, D. , author Park, W. , author Jeong, S.G. , author Kwon, H. , author Kim, C.S. , year 2022 . title Eigenlanes: Data-driven lane descriptors for structurally diverse lanes , in: booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern recognit...

    2022

  3. [11]

    , author Wang, Y

    author Li, Q. , author Wang, Y. , author Wang, Y. , author Zhao, H. , year 2022 . title Hdmapnet: An online hd map construction and evaluation framework , in: booktitle 2022 International Conference on Robotics and Automation (ICRA) , organization IEEE . pp. pages 4628--4634

    2022

  4. [12]

    , author Chen, S

    author Liao, B. , author Chen, S. , author Wang, X. , author Cheng, T. , author Zhang, Q. , author Liu, W. , author Huang, C. , year 2022 . title Maptr: Structured modeling and learning for online vectorized hd map construction . journal arXiv preprint arXiv:2208.14437

    2022 arXiv

  5. [13]

    , author Chen, S

    author Liao, B. , author Chen, S. , author Yin, H. , author Jiang, B. , author Wang, C. , author Yan, S. , author Zhang, X. , author Li, X. , author Zhang, Y. , author Zhang, Q. , author Wang, X. , year 2024 . title Diffusiondrive: Truncated diffusion model for end-to-end auto...

    2024 arXiv

  6. [14]

    , author Chen, S

    author Liao, B. , author Chen, S. , author Zhang, Y. , author Jiang, B. , author Zhang, Q. , author Liu, W. , author Huang, C. , author Wang, X. , year 2023 . title Maptrv2: An end-to-end framework for online vectorized hd map construction . journal arXiv preprint arXiv:2308.05736

    2023 arXiv

  7. [15]

    , author Wang, Y

    author Liu, Y. , author Wang, Y. , author Wang, Y. , author Zhao, H. , year 2022 a. title Vectormapnet: End-to-end vectorized hd map learning . journal arXiv preprint arXiv:2206.08920

    2022 arXiv

  8. [16]

    , author Yan, J

    author Liu, Y. , author Yan, J. , author Jia, F. , author Li, S. , author Gao, A. , author Wang, T. , author Zhang, X. , author Sun, J. , year 2022 b. title Petrv2: A unified framework for 3d perception from multi-camera images . journal arXiv preprint arXiv:2206.01256

    2022 arXiv

  9. [17]

    , author Yuan, T

    author Liu, Y. , author Yuan, T. , author Wang, Y. , author Wang, Y. , author Zhao, H. , year 2023 . title Vectormapnet: End-to-end vectorized hd map learning , in: booktitle International Conference on Machine Learning , organization PMLR . pp. pages 22352--22369

    2023

  10. [18]

    , author Zhang, X

    author Liu, Z. , author Zhang, X. , author Liu, G. , author Zhao, J. , author Xu, N. , year 2024 . title Leveraging enhanced queries of point sets for vectorized map construction , in: booktitle European Conference on Computer Vision

    2024

  11. [19]

    , author De Brabandere, B

    author Neven, D. , author De Brabandere, B. , author Georgoulis, S. , author Proesmans, M. , author Van Gool, L. , year 2018 . title Towards end-to-end lane detection: an instance segmentation approach , in: booktitle 2018 IEEE intelligent vehicles symposium (IV) , organizatio...

    2018

  12. [20]

    , author Jin, D

    author Park, W. , author Jin, D. , author Kim, C.S. , year 2022 . title Eigencontours: Novel contour descriptors based on low-rank approximation , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. pages 2667--2675

    2022

  13. [21]

    , author Chen, Z

    author Peng, L. , author Chen, Z. , author Fu, Z. , author Liang, P. , author Cheng, E. , year 2023 . title Bevsegformer: Bird's eye view semantic segmentation from arbitrary camera rigs , in: booktitle Proceedings of the IEEE/CVF Winter Conference on Applications of Computer ...

    2023

  14. [22]

    , author Hu, X

    author Seif, H.G. , author Hu, X. , year 2016 . title Autonomous driving in the icity—hd maps as a key challenge of the automotive industry . journal Engineering volume 2 , pages 159--162

    2016

  15. [23]

    , author Wang, Y

    author Sun, P. , author Wang, Y. , author He, P. , author Pei, X. , author Yang, M. , author Jiang, K. , author Yang, D. , year 2022 . title Gcd-l: A novel method for geometric change detection in hd maps using low-cost sensors . journal Automotive Innovation volume 5 , pages 324--332

    2022

  16. [24]

    , author Jiang, K

    author Tang, X. , author Jiang, K. , author Yang, M. , author Liu, Z. , author Jia, P. , author Wijaya, B. , author Wen, T. , author Cui, L. , author Yang, D. , year 2023 . title High-definition maps construction based on visual sensor: A comprehensive survey . journal IEEE Tr...

    2023

  17. [25]

    , author Pentland, A

    author Turk, M. , author Pentland, A. , year 1991 . title Eigenfaces for recognition . journal Journal of cognitive neuroscience volume 3 , pages 71--86

    1991

  18. [26]

    , author Lu, X

    author Wang, R. , author Lu, X. , author Liu, X. , author Zou, X. , author Cao, T. , author Li, Y. , year 2024 . title Priormapnet: Enhancing online vectorized hd map construction with priors . journal arXiv preprint arXiv:2408.08802

    2024 arXiv

  19. [27]

    , author Jiang, K

    author Wen, T. , author Jiang, K. , author Miao, J. , author Wijaya, B. , author Jia, P. , author Yang, M. , author Yang, D. , year 2022 . title Roadside hd map object reconstruction using monocular camera . journal IEEE Robotics and Automation Letters volume 7 , pages 7722--7729

    2022

  20. [28]

    , author Jiang, K

    author Wijaya, B. , author Jiang, K. , author Yang, M. , author Wen, T. , author Tang, X. , author Yang, D. , year 2022 . title Crowdsourced road semantics mapping based on pixel-wise confidence level . journal Automotive Innovation volume 5 , pages 43--56

    2022

  21. [29]

    , author Wong, K.K

    author Xu, Z. , author Wong, K.K. , author Zhao, H. , year 2023 . title Insightmapper: A closer look at inner-instance information for vectorized high-definition mapping . journal arXiv preprint arXiv:2308.08543

    2023 arXiv

  22. [30]

    , author Chen, Y

    author Yang, C. , author Chen, Y. , author Tian, H. , author Tao, C. , author Zhu, X. , author Zhang, Z. , author Huang, G. , author Li, H. , author Qiao, Y. , author Lu, L. , et al., year 2023 . title Bevformer v2: Adapting modern image backbones to bird's-eye-view recognitio...

    2023

  23. [31]

    , author Jiang, K

    author Yang, D. , author Jiang, K. , author Zhao, D. , author Yu, C. , author Cao, Z. , author Xie, S. , author Xiao, Z. , author Jiao, X. , author Wang, S. , author Zhang, K. , year 2018 . title Intelligent and connected vehicles: Current status and future perspectives . jour...

    2018

  24. [32]

    , author Liu, Y

    author Yuan, T. , author Liu, Y. , author Wang, Y. , author Wang, Y. , author Zhao, H. , year 2023 . title Streammapnet: Streaming mapping network for vectorized online hd map construction . journal arXiv preprint arXiv:2308.12570

    2023 arXiv

  25. [33]

    , author Zhang, Y

    author Zhang, Z. , author Zhang, Y. , author Ding, X. , author Jin, F. , author Yue, X. , year 2023 . title Online vectorized hd map construction using geometry . journal arXiv preprint arXiv:2312.03341

    2023 arXiv

  26. [34]

    , author Zhang, H

    author Zhou, Y. , author Zhang, H. , author Yu, J. , author Yang, Y. , author Jung, S. , author Park, S.I. , author Yoo, B. , year 2024 . title Himap: Hybrid representation learning for end-to-end vectorized hd map construction , in: booktitle Proceedings of the IEEE/CVF Confe...

    2024

  27. [35]

    , author Zyrianov, V

    author Zhu, X. , author Zyrianov, V. , author Liu, Z. , author Wang, S. , year 2023 . title Mapprior: Bird's-eye view map layout estimation with generative models , in: booktitle Proceedings of the IEEE/CVF International Conference on Computer Vision , pp. pages 8228--8239

    2023

  28. [36]

    2015, , 579, A101

    Aladro, R., Martín, S., Riquelme, D., et al. 2015, , 579, A101

    2015

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.