Pith. sign in

REVIEW 2 major objections 6 minor 1 cited by

VESPA: Towards un(Human)supervised Open-World Pointcloud Labeling for Autonomous Driving

T0 review · 2 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 VESPA claims state-of-the-art unsupervised 3D object discovery on nuScenes, reaching 52.95 class-agnostic AP and 48.12 NDS without human labels.

desk verdict Solid VLM-guided autolabeling pipeline with strong nuScenes results, but the SOTA claim rests on an unverified UNION reimplementation and undisclosed LLM priors. read the letter →

arxiv 2507.20397 v1 pith:G3YRV5RD submitted 2025-07-27 cs.CV

Levente Tempfli , Esteban Rivera , Markus Lienkamp This is my paper · ORCID
classification cs.CV
keywords 3Dobjectdetectionautolabelingvision-languagemodelsLiDARpointcloudspseudo-labelingopen-vocabularyautonomousdrivingself-supervisedlearning
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

VESPA's central claim is that a multimodal pipeline combining LiDAR point clouds with camera images processed by vision-language models can generate high-quality 3D bounding-box pseudolabels for autonomous driving scenes with no human-annotated ground truth. The paper reports that these pseudolabels yield a class-agnostic object-discovery AP of 52.95 and an NDS of 48.12 on the nuScenes validation set, and that training a CenterPoint detector on them reaches 46.54 mAP in a 3-class setting, outperforming prior un(human)-supervised methods. The practical stake is the annotation bottleneck: if VESPA transfers, new scenes, sensor configurations, and even novel object categories can be labeled from off-the-shelf VLMs and LiDAR alone.

What carries the argument

The load-bearing mechanism is a cascade: Grounding DINO proposes open-vocabulary 2D boxes (confidence at least 0.3), SAM converts them to segmentation masks, and LiDAR points are reprojected into each mask using the dataset's extrinsic and ego-pose calibration. The resulting object clusters are denoised by DBSCAN in the XY plane, with its distance threshold set to an average class width retrieved from a large language model rather than tuned on the dataset. Multi-camera proposals are merged, DINOv2 appearance embeddings drive temporal tracking, and ICP provides motion and orientation refinement; final boxes are inflated to LLM-provided class priors. The LLM-queried values (class widths, urban velocity bounds) are what let the pipeline avoid dataset-specific tuning, and the ablations show that the tracking and denoising modules account for the largest performance swings.

What would settle it

Run the pipeline on the nuScenes validation set while sweeping the DBSCAN distance threshold over a small range around the quoted class widths (e.g., 0.5 m to 3 m in 0.25 m steps). If the class-agnostic AP moves by more than a few points across that sweep, the denoising step is not the stable mechanism the paper claims; if the paper's threshold values are simply missing, reproducing the reported 52.95 AP requires them.

Watch

Extended reading notes

Core claim

The core discovery is that VLM-based open-vocabulary 2D detections and masks, when reprojected into the LiDAR point cloud and passed through geometric denoising, appearance-based tracking, and motion/box refinement, become reliable 3D object proposals without any ground-truth labels or HD maps. On nuScenes, VESPA outperforms the reimplemented UNION baseline by roughly 14.6 points in class-agnostic AP (52.95 versus 38.4) and reaches about 64% of fully supervised class-agnostic AP, a level the paper situates between training on 5% and 10% of the fully supervised data. The authors state that VESPA's classification capacity is inherited from the VLM, making the pseudolabels open-world to the extent the VLM is, and they frame the work as a step toward eliminating manual supervision while acknowledging that VLM distillation is itself a form of supervision.

Load-bearing premise

VESPA's results assume that DBSCAN clustering in the XY plane, with a distance threshold set to an average class width queried from a language model, reliably separates true object points from background points projected into foreground masks through occlusion and sensor parallax; the exact threshold values are never reported, so this assumption cannot be checked from the paper.

Editorial extensions

If this is right

  • A CenterPoint detector trained on VESPA pseudolabels reaches 46.54 mAP in the 3-class setting, more than double the 27.19 mAP of the same detector on UNION pseudolabels.
  • VESPA discovers small and rare classes (bicycles, motorcycles, construction vehicles) that prior LiDAR-based autolabelers miss entirely, suggesting the image semantics recover objects whose LiDAR signature alone is too sparse.
  • The pipeline runs on a different sensor setup (AnonymousScenes) with no retuning, although low-resolution surround cameras degrade performance, indicating image resolution is the current ceiling.
  • Swapping the VLM from Grounding DINO to OWLv2 changes results only modestly, so future improvements in VLMs should transfer directly into better pseudolabels without pipeline changes.
  • Removing the tracking module drops mAP by 21% and removing cluster denoising drops it by up to 19%, so the advertised results depend on the geometric and temporal correction steps, not on the 2D detections alone.

Reading between the lines

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

  • One reproducibility gap the paper leaves open is that the LLM-queried class widths that set the DBSCAN threshold are not reported; publishing them and testing a threshold sweep would show whether the denoising step's 19% contribution is stable or tuned to a lucky value.
  • Because classification is inherited from the VLM, a natural test of the 'open-world' claim is to query with categories absent from nuScenes classes (road debris, animals, trailers without tractors) and see whether the pipeline produces coherent 3D pseudolabels for them.
  • The resolution sensitivity suggests a concrete next experiment: running the same pipeline with higher-resolution surround cameras (or super-resolution preprocessing) should narrow the gap to fully supervised detection more than adding LiDAR channels did, a prediction the paper's data already hints at but does not test.
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

2 major / 6 minor

Summary. The paper introduces VESPA, a multimodal autolabeling pipeline that generates 3D bounding-box pseudolabels for autonomous driving scenes by combining LiDAR point clouds with open-vocabulary vision-language model (VLM) outputs. The pipeline consists of ground removal, Grounding DINO + SAM 2D proposal generation, LiDAR-to-image reprojection, DBSCAN-based cluster denoising with LLM-provided class-width priors, multi-camera merging, DINOv2-based appearance tracking with ICP motion estimation, and LLM-prior-based box refinement. The authors evaluate on nuScenes, reporting class-agnostic AP of 52.95 and NDS of 48.12 (improving on a reimplemented UNION baseline by 12.5 AP points), and a 3-class downstream CenterPoint mAP of 46.54. They also report ablations (Table 4) showing the tracking and cluster-denoising modules have the largest impact, and an additional evaluation on a proprietary dataset.

Significance. If correct, VESPA is a strong autolabeling result: it closes much of the gap to fully supervised detection, and the component ablations are informative. The pipeline is modular and uses publicly available foundation models. However, the state-of-the-art claim is contingent on the fidelity of the reimplemented UNION baseline and on the reproducibility of the LLM-derived parameters; both need to be addressed before the claim can be accepted.

major comments (2)
  1. [4.2, Table 1] The UNION baseline is a reimplementation, not the original authors' released numbers, and no verification of this reimplementation is provided. Because the paper's central claim of state-of-the-art AP rests on the 12.5 AP point improvement over this reimplemented baseline, the comparison is not yet convincing. Please either report the official UNION results as originally published, or, if a reimplementation is necessary, demonstrate its fidelity by reproducing one or more reported UNION metrics using the original code/checkpoints on the same protocol. Without this, the claimed margin over the prior state of the art is not established.
  2. [3.4, 3.7, 3.8, 4.1] The pipeline's critical hyperparameters are not transparent. Section 3.4 states that the DBSCAN distance threshold is 'the average width of the object class' retrieved from an LLM; Section 3.7 uses an LLM-queried maximum velocity; Section 3.8 uses LLM-sourced box-size priors. Section 4.1 says all hyperparameters were 'set by visual tuning on three representative training scenes,' but neither the exact values, the prompts, nor the identities of the three scenes are given. This is load-bearing because the ablation in Table 4 shows that removing the DBSCAN denoising module (DN) reduces 3-class mAP from 46.76 to 27.86, nearly equal to the reimplemented UNION baseline (27.19 in Table 2). The full configuration of these parameters must be disclosed (in the paper or a supplement) to allow reproduction and to substantiate the claim that the LLM priors avoid dataset-specific tuning.
minor comments (6)
  1. [Table 1] In Table 1, the method label 'CMD3D' appears; the text and reference [9] use 'CM3D'. Please correct this typo.
  2. [4.3, Table 4] The text reports the effect of removing the tracking module as a '21% drop in mAP' and '22% drop in NDS.' The numbers in Table 4 (46.76 to 25.48 mAP; 43.47 to 21.84 NDS) correspond to drops of about 21 percentage points (and about 45% relative). Please clarify whether you mean percentage points or relative percentages throughout the ablation discussion.
  3. [4.1, Table 2] There is a discrepancy about the split used for pseudo-label quality metrics: Section 4.1 says pseudo labels are compared to ground truth on the validation set, but Table 2's caption says 'pseudo label quality on the training set.' Please align the text and caption.
  4. [Abstract, 5] The abstract and discussion claim 'open-world' and 'novel category' discovery, but all experiments use the predefined nuScenes classes as queries and evaluate only on those classes. Please provide a demonstration on a class not in the standard set, or soften the claim to 'open-vocabulary queries supported, evaluated on the nuScenes class set.'
  5. [3.2, 3.4, 3.7] For reproducibility, please specify which LLM was used to obtain the priors, the exact prompt templates, and the exact threshold values used for DBSCAN distance, tracking proximity, and box inflation. Also report the Grounding DINO prompt list.
  6. [Table 2, Table 4] The full VESPA configuration in Table 4 reports mAP of 46.76 and NDS of 43.47, while Table 2 reports 46.54 and 43.45 for the same setting. Please explain whether these are from different runs or a typo; if random variation, state the number of runs.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: VESPA's pseudolabels are evaluated against held-out nuScenes ground truth, and the load-bearing comparisons are external benchmarks rather than reconstructed fits.

full rationale

VESPA's derivation chain is self-contained against external benchmarks. All critical inputs are externally pretrained models (Grounding DINO, SAM, DINOv2), dataset-provided calibration, and LLM-queried class priors; no parameter is fitted to the nuScenes validation or test ground truth. Pseudolabel quality is measured by training CenterPoint on training-set pseudolabels and evaluating on held-out validation ground truth (Section 4.1), so the reported AP and NDS numbers are external evaluations rather than quantities rebuilt from the inputs. The cluster-denoising DBSCAN distance threshold is an LLM-provided average class width (Section 3.4), not a value fitted to target boxes, and the tracking velocity threshold is similarly LLM-sourced (Section 3.7); these are external priors, not self-referential fits. The only self-referential element is the footnote stating the work is based on an author's unpublished Master's thesis, which is not invoked as evidence for any claim. The SOTA comparison does depend on a reimplementation of UNION (Table 1 footnote: 'We reimplemented UNION based on the original code'), but that is a baseline-fidelity and correctness risk, not a circular reduction: the VESPA numbers are not derived from UNION's numbers, and no equation reduces the claimed 12.5-point AP improvement to an input of the VESPA pipeline. The unreported exact LLM-provided width values are an auditability limitation, but the values are still external inputs rather than quantities defined by the target labels. No step satisfies the quoted-reduction standard for circularity, so the score is 0.

Assumptions & free parameters 6 free parameters · 3 assumptions · 0 invented entities

The central contribution is an engineering pipeline, not a mathematical derivation. It depends on the reliability of external pretrained models, provided sensor calibration, and undisclosed LLM-derived parameters. No new physical entities are introduced.

free parameters (6)
  • Grounding DINO confidence threshold = 0.3
    Detections with confidence below 0.3 are discarded (Section 3.2). This threshold was chosen during visual tuning and affects recall and precision.
  • Mask NMS IoA threshold = 0.5
    Proposals with intersection-over-area greater than 0.5 are suppressed (Section 3.2). A fixed design choice.
  • DBSCAN distance threshold (LLM-queried average class width) = Not reported
    Used as the clustering epsilon in Section 3.4 to separate object points from background. Queried from an unspecified LLM to avoid dataset-specific tuning; exact values not given.
  • Tracking spatial proximity threshold (LLM-queried max velocity) = Not reported
    Used in Section 3.7 to limit candidate matches across frames. Queried from an unspecified LLM; exact value not given.
  • Box inflation class priors (LLM-queried) = Not reported
    Used in Section 3.8 to inflate too-small boxes to class means. Sourced from an unspecified LLM; values not given.
  • Ground removal thresholds = 30 cm plane distance; >40 m range; <4 m height
    Used in Section 3.1 to filter ground and distant points. Chosen based on sensor characteristics, not fitted to labels.
assumptions (3)
  • domain assumption Accurate sensor calibration (extrinsics, ego-pose) is provided by the dataset
    LiDAR-to-camera reprojection in Section 3.3 assumes the calibration is correct; errors would propagate to all mask associations and downstream boxes.
  • domain assumption Pre-trained vision-language and segmentation models generalize to the driving scene distribution
    Grounding DINO and SAM are used zero-shot; their detection and segmentation quality bounds pipeline performance (Section 3.2, also acknowledged in Discussion).
  • domain assumption LLM-queried priors are appropriate for the evaluation environment
    DBSCAN epsilon, tracking proximity threshold, and box inflation dimensions rely on these priors (Sections 3.4, 3.7, 3.8); the source LLM and values are not disclosed, so their validity is unverified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of VESPA: Towards un(Human)supervised Open-World Pointcloud Labeling for Autonomous Driving." pith.science (2026). https://pith.science/paper/G3YRV5RD

@misc{pith2026250720397,
  author       = {Pith},
  title        = {Pith review of: VESPA: Towards un(Human)supervised Open-World Pointcloud Labeling for Autonomous Driving},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G3YRV5RD}},
  note         = {Machine review of arXiv:2507.20397}
}
read the original abstract

Data collection for autonomous driving is rapidly accelerating, but manual annotation, especially for 3D labels, remains a major bottleneck due to its high cost and labor intensity. Autolabeling has emerged as a scalable alternative, allowing the generation of labels for point clouds with minimal human intervention. While LiDAR-based autolabeling methods leverage geometric information, they struggle with inherent limitations of lidar data, such as sparsity, occlusions, and incomplete object observations. Furthermore, these methods typically operate in a class-agnostic manner, offering limited semantic granularity. To address these challenges, we introduce VESPA, a multimodal autolabeling pipeline that fuses the geometric precision of LiDAR with the semantic richness of camera images. Our approach leverages vision-language models (VLMs) to enable open-vocabulary object labeling and to refine detection quality directly in the point cloud domain. VESPA supports the discovery of novel categories and produces high-quality 3D pseudolabels without requiring ground-truth annotations or HD maps. On Nuscenes dataset, VESPA achieves an AP of 52.95% for object discovery and up to 46.54% for multiclass object detection, demonstrating strong performance in scalable 3D scene understanding. Code will be available upon acceptance.

Figures

Figures reproduced from arXiv: 2507.20397 by the authors.

Figure 1
Figure 1. VESPA pipeline overview. Stages: (1) ground removal; (2) zero-shot segmentation; (3) reprojection; (4) cluster denoising; (5) [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Qualitative comparison between UNION (left) and [PITH_FULL_IMAGE:figures/full_fig_p015_10.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. ReSAGE-PAR: Representational Similarity Assessment for Generative Expansion in Pedestrian Attribute Recognition

    cs.CV 2026-06 unverdicted novelty 6.0 of 10

    ReSAGE-PAR adapts diffusion models with LoRA, scores generated images via vision-language prompts, and applies Bayesian classification to produce pseudo-labels, yielding up to 8.7% gains when used to expand PAR datasets.

Reference graph

Works this paper leans on

36 extracted references · 31 canonical work pages · cited by 1 Pith paper

  1. [1]

    Qwen-vl: A strong large vision-language model for practicable applications

    Jinze Bai, Shuai Tang, Fan Long, Jianwei Fang, Yang Li, Bo Zhang, Wenyuan Wang, Junyi Yang, Shujie Ding, Xin Ma, et al. Qwen-vl: A strong large vision-language model for practicable applications. arXiv preprint arXiv:2308.12966 ,

    arXiv

  2. [2]

    Liso: Lidar-only self-supervised 3d object detec- tion

    Stefan Andreas Baur, Frank Moosmann, and Andreas Geiger. Liso: Lidar-only self-supervised 3d object detec- tion. In European Conference on Computer Vision , pages 253–270. Springer, 2024. 1, 3, 5, 6

    work page 2024

  3. [3]

    Method for registration of 3-d shapes

    Paul J Besl and Neil D McKay. Method for registration of 3-d shapes. In Sensor fusion IV: control paradigms and data structures, pages 586–606. Spie, 1992. 4

    work page 1992

  4. [4]

    nuscenes: A multi- modal dataset for autonomous driving

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Gi- ancarlo Baldan, and Oscar Beijbom. nuscenes: A multi- modal dataset for autonomous driving. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11621–11631, 2020. 4, 5

    work page 2020

  5. [5]

    A density-based algorithm for discovering clusters in large spatial databases with noise

    Martin Ester, Hans-Peter Kriegel, J ¨org Sander, Xiaowei Xu, et al. A density-based algorithm for discovering clusters in large spatial databases with noise. In kdd, pages 226–231,

  6. [6]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Communications of the ACM, 24(6):381–395, 1981

    Martin A Fischler and Robert C Bolles. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Communications of the ACM, 24(6):381–395, 1981. 3

    work page 1981

  7. [7]

    Vision-language guidance for lidar-based unsupervised 3d object detection

    Christian Fruhwirth-Reisinger, Wei Lin, Du ˇsan Mali´c, Horst Bischof, and Horst Possegger. Vision-language guidance for lidar-based unsupervised 3d object detection. arXiv preprint arXiv:2408.03790, 2024. 3

    arXiv 2024

  8. [8]

    Are we ready for autonomous driving? the kitti vision benchmark suite

    Andreas Geiger, Philipp Lenz, and Raquel Urtasun. Are we ready for autonomous driving? the kitti vision benchmark suite. In 2012 IEEE conference on computer vision and pat- tern recognition, pages 3354–3361. IEEE, 2012. 1

    work page 2012

Show all 36 references
  1. [9]

    Shelf-supervised cross-modal pre-training for 3d ob- ject detection

    Mehar Khurana, Neehar Peri, James Hays, and Deva Ra- manan. Shelf-supervised cross-modal pre-training for 3d ob- ject detection. arXiv preprint arXiv:2406.10115, 2024. 3, 4, 5, 6, 7

    2024 arXiv

  2. [10]

    Segment any- thing

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment any- thing. In Proceedings of the IEEE/CVF international confer- ence on computer vision, pages 4015–4026, 2023. 2, 3

    2023

  3. [11]

    Sen- sors and sensor fusion in autonomous vehicles

    Jelena Koci ´c, Nenad Jovi ˇci´c, and Vujo Drndarevi ´c. Sen- sors and sensor fusion in autonomous vehicles. In 2018 26th Telecommunications Forum (TELFOR) , pages 420–

    2018

  4. [12]

    Union: Unsupervised 3d object detection using object appearance- based pseudo-classes

    Ted Lentsch, Holger Caesar, and Dariu Gavrila. Union: Unsupervised 3d object detection using object appearance- based pseudo-classes. Advances in Neural Information Pro- cessing Systems, 37:22028–22046, 2024. 1, 3, 4, 5, 6, 11

    2024

  5. [13]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. In European Conference on Computer Vision , pages 38–55. Springe...

    2024

  6. [14]

    Reward fine- tuning for faster and more accurate unsupervised object dis- covery

    Katie Luo, Zhenzhen Liu, Xiangyu Chen, Yurong You, Sagie Benaim, Cheng Perng Phoo, Mark Campbell, Wen Sun, Bharath Hariharan, and Kilian Q Weinberger. Reward fine- tuning for faster and more accurate unsupervised object dis- covery. Advances in Neural Information Processing Sy...

    2023

  7. [15]

    hdbscan: Hierarchical density based clustering

    Leland McInnes, John Healy, Steve Astels, et al. hdbscan: Hierarchical density based clustering. J. Open Source Softw., 2(11):205, 2017. 2

    2017

  8. [16]

    Scaling open-vocabulary object detection

    Matthias Minderer, Alexey Gritsenko, and Neil Houlsby. Scaling open-vocabulary object detection. Advances in Neu- ral Information Processing Systems, 36:72983–73007, 2023. 7

    2023

  9. [17]

    Gpt-4 technical report, 2024

    OpenAI, Josh Achiam Marvin Zhang, Shengjia Zhao, Tian- hao Zheng, Juntang Zhuang, William Zhuk, and Barret Zoph. Gpt-4 technical report, 2024. 2

    2024

  10. [18]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193, 2023. 4

    2023 arXiv

  11. [19]

    Perception, planning, control, and coordi- nation for autonomous vehicles

    Scott Drew Pendleton, Hans Andersen, Xinxin Du, Xiaotong Shen, Malika Meghjani, You Hong Eng, Daniela Rus, and Marcelo H Ang. Perception, planning, control, and coordi- nation for autonomous vehicles. Machines, 5(1):6, 2017. 1

    2017

  12. [20]

    Towards long-tailed 3d detection

    Karthik Peri, Kuang Li, Jianing Tang, Xinyu Chen, Xuan Li, Xiaokang Wu, Xinge Yu, and Bo Yang. Towards long-tailed 3d detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Work- shops, pages 5556–5566, 2023. 2

    2023

  13. [21]

    SeMoLi: What Moves To- gether Belongs Together

    Jenny Seidenschwarz, Aljosa Osep, Franceso Ferroni, Simon Lucey, and Laura Leal-Taixe. SeMoLi: What Moves To- gether Belongs Together . In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 14685–14694, Los Alamitos, CA, USA, 2024. IEEE Com- pute...

    2024

  14. [22]

    Openpcdet: An open- source toolbox for 3d object detection from point clouds

    OpenPCDet Development Team. Openpcdet: An open- source toolbox for 3d object detection from point clouds. https://github.com/open-mmlab/OpenPCDet ,

  15. [23]

    Ms3d++: Ensemble of experts for multi-source unsupervised domain adaptation in 3d ob- 9 ject detection

    Darren Tsai, Julie Stephany Berrio, Mao Shan, Eduardo Nebot, and Stewart Worrall. Ms3d++: Ensemble of experts for multi-source unsupervised domain adaptation in 3d ob- 9 ject detection. IEEE Transactions on Intelligent Vehicles , pages 1–16, 2024. 1

    2024

  16. [24]

    Dsvt: Dynamic sparse voxel transformer with rotated sets

    Haiyang Wang, Chen Shi, Shaoshuai Shi, Meng Lei, Sen Wang, Di He, Bernt Schiele, and Liwei Wang. Dsvt: Dynamic sparse voxel transformer with rotated sets. In 2023 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 13520–13529, 2023. 1

    2023

  17. [25]

    Commonsense prototype for outdoor unsuper- vised 3d object detection

    Hai Wu, Shijia Zhao, Xun Huang, Chenglu Wen, Xin Li, and Cheng Wang. Commonsense prototype for outdoor unsuper- vised 3d object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 14968–14977, 2024. 3

    2024

  18. [26]

    Center- based 3d object detection and tracking

    Tianwei Yin, Xingyi Zhou, and Philipp Krahenbuhl. Center- based 3d object detection and tracking. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11784–11793, 2021. 5, 7

    2021

  19. [27]

    Centernet3d: An anchor-free approach for scalable 3d object detection

    Zheng Yin, Jianping Shi, and Xiaojuan Sun. Centernet3d: An anchor-free approach for scalable 3d object detection. arXiv preprint arXiv:2104.09503, 2021. 1

    2021 arXiv

  20. [28]

    Learning to detect mobile objects from lidar scans without labels

    Yurong You, Katie Luo, Cheng Perng Phoo, Wei-Lun Chao, Wen Sun, Bharath Hariharan, Mark Campbell, and Kilian Q Weinberger. Learning to detect mobile objects from lidar scans without labels. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition,...

    2022

  21. [29]

    To- wards unsupervised object detection from lidar point clouds

    Lunjun Zhang, Anqi Joyce Yang, Yuwen Xiong, Sergio Casas, Bin Yang, Mengye Ren, and Raquel Urtasun. To- wards unsupervised object detection from lidar point clouds. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9317–9328, 2023. 3, 5, 6

    2023

  22. [30]

    To- wards Unsupervised Object Detection from LiDAR Point Clouds

    Lunjun Zhang, Anqi Joyce Yang, Yuwen Xiong, Sergio Casas, Bin Yang, Mengye Ren, and Raquel Urtasun. To- wards Unsupervised Object Detection from LiDAR Point Clouds . In 2023 IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR) , pages 9317–9328, Los Alamitos...

    2023

  23. [31]

    Harnessing uncertainty-aware bounding boxes for unsuper- vised 3d object detection

    Ruiyang Zhang, Hu Zhang, Hang Yu, and Zhedong Zheng. Harnessing uncertainty-aware bounding boxes for unsuper- vised 3d object detection. arXiv preprint arXiv:2408.00619,

    arXiv

  24. [32]

    Ef- ficient l-shape fitting for vehicle detection using laser scan- ners

    Xiao Zhang, Wenda Xu, Chiyu Dong, and John M Dolan. Ef- ficient l-shape fitting for vehicle detection using laser scan- ners. In 2017 IEEE Intelligent Vehicles Symposium (IV) , pages 54–59. IEEE, 2017. 4

    2017

  25. [33]

    A survey of deep learning-driven 3d object detection: Sensor modalities, technical architectures, and applications

    Xiang Zhang, Hai Wang, and Haoran Dong. A survey of deep learning-driven 3d object detection: Sensor modalities, technical architectures, and applications. Sensors, 25(12),

  26. [34]

    Center- point: Keypoint head for 3d object detection

    Xin Zhou, Daquan Wang, and Philipp Kr ¨ahenb¨uhl. Center- point: Keypoint head for 3d object detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12693–12702, 2020. 1

    2020

  27. [35]

    Class-balanced grouping and sampling for point cloud 3d object detection

    Benjin Zhu, Zhengkai Jiang, Xiangxin Zhou, Zeming Li, and Gang Yu. Class-balanced grouping and sampling for point cloud 3d object detection. arXiv preprint arXiv:1908.09492,

    1908 arXiv

  28. [36]

    We examine a set of randomly selected frames and analyze the 3D bounding boxes projected to the bird’s eye view

    Qualitative examples In this section, we qualitatively compare the pseudo-labels generated by UNION [12], the current state-of-the-art method, with those from our proposed approach, VESPA. We examine a set of randomly selected frames and analyze the 3D bounding boxes projected...

Pith tools

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