Pith. sign in

REVIEW 4 major objections 6 minor 109 references

GS-Occ3D: Scaling Vision-only Occupancy Reconstruction with Gaussian Splatting

T0 review · 4 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 GS-Occ3D claims a camera-only pipeline can reconstruct 3D occupancy from driving video, matching LiDAR-based labels on Waymo and beating them for zero-shot transfer to nuScenes.

desk verdict A real vision-only occupancy label curation pipeline with strong geometry results, but the headline claim of scalability is undercut by the deliberate exclusion of ego-static scenes that the authors admit the method cannot handle. read the letter →

arxiv 2507.19451 v3 pith:SN6AQMZX submitted 2025-07-25 cs.CV

Baijun Ye , Minghui Qin , Saining Zhang , Moonjun Gong , Shaoting Zhu , Zebang Shen , Luan Zhang , Lu Zhang
show 2 more authors
Hao Zhao Hang Zhao
This is my paper
classification cs.CV MSC 68T4568U05 PACS 07.05.Pj42.30.Sy
keywords occupancyreconstructionvision-onlyGaussiansplattingauto-labelingautonomousdrivingoctreeWaymozero-shotgeneralization
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

GS-Occ3D sets out to show that 3D occupancy labels for autonomous driving—normally produced from LiDAR sweeps—can be reconstructed entirely from camera images at scale. The paper builds an octree-based Gaussian surfel scene representation that separates static background, ground, and dynamic vehicles, and uses it to reconstruct the full Waymo dataset with no LiDAR input. On the Waymo Static-32 geometry benchmark it reports the lowest Chamfer distance (0.56) among the compared implicit and explicit methods, including LiDAR-supervised baselines. The reconstructed point clouds are converted into binary occupancy labels that train the CVT-Occ occupancy model to reach IoU comparable to LiDAR-based Occ3D labels on Occ3D-Waymo and higher IoU (33.4 versus 31.4) in zero-shot evaluation on Occ3D-nuScenes. If the claim holds, occupancy auto-labeling can move from expensive LiDAR fleets to crowdsourced camera data.

What carries the argument

The load-bearing object is the Octree-based Gaussian Surfel: a hierarchical octree of voxels, each holding up to m Gaussian surfels that locally approximate surface geometry, with level-of-detail that captures coarse structures like walls and roads alongside fine details like vegetation and object edges. Ground is handled by dedicated Gaussian surfels initialized by projecting camera poses onto the xy-plane with a fixed height offset, then regularized to stay planar; dynamic vehicles are initialized from a vision-based 3D tracker, with learnable pose corrections refining each box. The label-curation stage turns the reconstructed point cloud into training labels by slicing per-frame ranges, aggregating dynamic-object points in box coordinates, and ray-casting so only first-hit voxels are marked as observed, which separates free space from unobserved space.

What would settle it

Measure the IoU of CVT-Occ trained on GS-Occ3D labels versus Occ3D labels on the full Occ3D-Waymo validation set, including the ego-static scenes that the paper excludes; if the vision-only labels do worse than LiDAR labels when those scenes are kept, the claimed matching generalization is broken.

Watch

Extended reading notes

Core claim

The paper's central claim is that a vision-only pipeline can replace LiDAR-based annotation for downstream occupancy models, and that the bottleneck is not the sensor but the scene representation. GS-Occ3D optimizes an octree-based Gaussian Surfel field over each long driving sequence, with explicit ground surfels and separately tracked dynamic vehicles, producing dense point clouds that are converted into occupancy labels by frame-wise division, multi-frame aggregation, and ray-casting. Experiments on Waymo report state-of-the-art geometry reconstruction (CD 0.56) and, when the resulting labels train CVT-Occ, generalization on Occ3D-Waymo that is described as reasonable and overall comparable to LiDAR labels, with superior zero-shot IoU on nuScenes. The authors state this is the first vision-only reconstruction of the full Waymo dataset.

Load-bearing premise

The evaluation assumes that removing ego-static scenes is a fair way to compare labels, even though those are exactly the scenes where vision-only reconstruction is known to fail.

Editorial extensions

If this is right

  • If the method scales as claimed, occupancy label curation no longer requires LiDAR-equipped fleets; ordinary camera-equipped vehicles can supply auto-labeling data.
  • Downstream occupancy models trained on vision-only labels inherit wider coverage than LiDAR labels, including regions beyond LiDAR range and tall structures such as high-rise buildings.
  • Zero-shot generalization on Occ3D-nuScenes indicates that vision-only labels are less tied to a particular sensor configuration or scene distribution than LiDAR-based labels.
  • Reconstructing the full Waymo dataset provides a large camera-only label base for future occupancy and scene-completion research.
  • The richer 66-class semantics available from RGB segmentation compared with Occ3D's 16 classes could make vision-only labels useful beyond geometry, for objects like motorcycles, lane markings, and crosswalks.

Reading between the lines

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

  • If vision-only reconstruction proves robust to different camera rigs and weather, the same octree-surfel decomposition could be adapted to other offboard auto-labeling tasks, such as lane geometry extraction or object-level mesh generation.
  • The paper's admitted failure on ego-static scenes suggests a practical hybrid: keep vision-only reconstruction for moving-vehicle stretches and add a lightweight motion estimator or multi-view prior only for stationary segments.
  • A direct test of the scalability claim would be to reconstruct labels from a different camera rig, such as nuScenes images, and evaluate on Occ3D-Waymo, checking whether the zero-shot advantage is symmetric across sensor setups.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes GS-Occ3D, a vision-only framework for reconstructing 3D occupancy from multi-view driving sequences. The method initializes an octree-based Gaussian surfel representation from SfM point clouds, explicitly models ground planes and dynamic vehicles, and trains with photometric and geometric losses. After reconstructing the Waymo dataset, the authors curate binary occupancy labels via frame-wise division, multi-frame aggregation, and ray-casting, then use these labels to train the CVT-Occ downstream occupancy model. Experiments report state-of-the-art Chamfer Distance on the Waymo Static-32 geometry split (0.56 vs. 0.82 for the best explicit baseline), comparable downstream occupancy performance on a restricted Occ3D-Waymo validation set (44.7 vs. 57.4 IoU), superior zero-shot generalization on Occ3D-nuScenes (33.4 vs. 31.4 IoU), and favorable fitting results on the authors' own label sets.

Significance. If the central claim holds, the paper introduces a scalable vision-only alternative to LiDAR-based occupancy label curation, potentially enabling crowdsourced data to be used for auto-labeling at scale. The work is significant in scope: it reports reconstructing the full Waymo dataset, and the downstream experiment is a direct comparison of vision-only labels against an established LiDAR-based benchmark. The main strengths are the clear decomposition into static, ground, and dynamic components; the efficient octree-surfel representation (80 MB storage, 0.8 h training per scene); and the explicit downstream validation with a strong occupancy model. The geometry result is substantial, and the zero-shot nuScenes improvement, even if modest, is a concrete and falsifiable positive result. The main weakness is the benchmark restriction to non-ego-static scenes, which is acknowledged by the authors and must be resolved before the headline claims can be taken at face value.

major comments (4)
  1. [Section 4.3, Datasets; Table 2] The exclusion of ego-static scenes from both training and evaluation is the most load-bearing assumption in the paper. The text states that vision-only methods 'cannot be reliably handled' on these scenes and that they are excluded 'for fairness', leaving 637/798 training and 165/202 validation scenes. Since the Occ3D-baseline model is not given this advantage, the Table 2 comparison of 44.7 vs. 57.4 IoU on 'Occ3D-Val(Waymo)' is not a comparison on the full official benchmark, and the abstract's claim of effectiveness on Occ3D-Waymo is therefore overstated. Please report results on the full official Occ3D-Waymo validation split, or at least provide a per-split breakdown (ego-static vs. non-ego-static), so the reader can assess the impact of this selection. The current presentation makes it impossible to determine whether the vision-only labels are comparable on the full distribution or only on a favorable subset.
  2. [Section 4.2; Table 1] The Chamfer Distance metric is computed against LiDAR point clouds, and the paper itself notes that reconstruction beyond the LiDAR coverage is counted as error ('our reasonable reconstruction beyond the LiDAR coverage is also counted as regions with high CD'). This conflation of range with inaccuracy undermines the single-number 'SOTA geometry' claim in Table 1. Please report CD restricted to a common evaluation mask, e.g., voxels or points within the LiDAR field of view, or provide a range-limited precision/recall curve. Without such a restriction, the 0.56 CD value may reflect both genuine geometric accuracy and the advantage of not emitting points in regions where LiDAR has no ground truth.
  3. [Section 3.1; Equations (4)-(5)] The experimental section does not specify the values of the key hyperparameters needed to reproduce the reported results, including the number of Gaussian surfels per voxel m, the base voxel size epsilon, the loss weights lambda_geo, lambda_obj, lambda_road, lambda_sky, lambda_s, lambda_d, and lambda_n, the learning rates, and the number of optimization iterations. Since no code release is mentioned, these omitted values are load-bearing for the paper's quantitative claims. Please add a supplementary table with all training hyperparameters and the exact protocol used for the Waymo Static-32 and full-dataset reconstructions.
  4. [Section 4.5; Table 3] Table 3 states that all methods use 'our Ground Gaussians for fairness'. The ground Gaussian component is a core novel contribution of the proposed method, and giving it to PGSR, 2DGS, and GVKF changes the comparison from 'our method vs. the baselines as published' to 'our method vs. baselines augmented with one of our components'. This could be fair as a controlled study, but it also makes the SOTA claim sensitive to an unreported interaction: if the ground gaussians are more compatible with some representations than others, the ranking in Table 3 may not reflect the methods' intrinsic performance. Please provide an additional comparison where baselines use their own default initialization/ground handling, or justify in more detail why the ground gaussians constitute a neutral initialization for all methods.
minor comments (6)
  1. [Section 4.1] The 'Waymo Static-32' split is referenced to [88] but never defined in the paper; please state the number of scenes, the selection criterion, and the split composition so the geometry comparison is self-contained.
  2. [Table 1] The F2-NeRF Chamfer Distance of 886.77 is several orders of magnitude larger than all other values in the table; please verify whether this is a typo, a unit issue, or the result of a catastrophic failure, and clarify it in the text.
  3. [Section 4.3; Table 2] F1, Precision, and Recall are reported in Table 2 but are not defined in the metrics paragraph; please specify how these are computed for voxel occupancy (e.g., binary voxel-level scores over the visible region).
  4. [References and author list] Several reference titles contain incorrect spacing ('V oteflow', 'V oxformer'), and the author name 'Moonjun Goon' in the author list differs from 'Moonjun Gong' in reference [41]; please harmonize these.
  5. [Section 3.2, Multi-frame Aggregation] The description 'following a process similar to [71]' is vague for the coordinate transformation used to aggregate dynamic-object points; please specify the box coordinate system, the association of points to boxes, and how occluded or mis-associated points are filtered.
  6. [Tables 1 and 3] The 3Cam rows for PGSR, 2DGS, and GVKF in Table 3 appear to coincide with the values in Table 1, but Table 1 does not state which camera configuration was used; please clarify the relationship between the two tables.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the proposed vision-only labels are validated against independent LiDAR-based Occ3D benchmarks, and no fitted quantity is renamed as a prediction.

full rationale

The paper's derivation chain is: detector-free SfM and ground gaussians initialize an octree Gaussian-surfel reconstruction; the reconstructed point cloud is divided into frames, aggregated, and voxelized; CVT-Occ is trained on those voxel labels; and the trained model is evaluated on the external Occ3D-Waymo and Occ3D-nuScenes validation sets whose labels are LiDAR-derived. None of the equations (1)-(5) or the curation steps define the predicted occupancy in terms of Occ3D labels, and no parameter is fitted to the evaluation target. The rows in Table 2 that use Ours-Val/Ours-Train as evaluation labels are explicitly presented as the paper's own labels; the load-bearing comparisons are the rows against Occ3D-Val(Waymo) (44.7 vs 57.4 IoU) and Occ3D-Val(nuScenes) (33.4 vs 31.4 IoU), which use independent LiDAR-based ground truth. The self-citations (Occ3D [71], CVT-Occ [94]) are used as benchmark and baseline model, not as the justification for the reconstruction; the octree and dynamic-scene inspirations cite external groups ([62], [87]). The manuscript's Limitations point (3) and Section 4.3 admit that ego-static scenes are excluded because vision-only reconstruction cannot reliably handle them, and this is a real benchmark-fairness concern for the Waymo comparison (the full validation set is 202 scenes, of which 165 are used), but it is not circularity: the exclusion is a selection criterion, not a fitted parameter, and the nuScenes zero-shot test is independent of that selection. No circular step can be quoted, so the correct finding is no significant circularity.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central claim rests on several domain assumptions inherited from prior systems: a flat road assumption for ground initialization, a vehicle-bbox assumption for dynamic objects, SfM as a reliable skeleton, and LiDAR CD as ground truth. Loss weights and structure sizes are unspecified, so part of the recipe is hidden. No new physical entities are introduced.

free parameters (5)
  • Number of Gaussian surfels per voxel, m = not specified
    Controls point-cloud density and reconstruction capacity; no value given in Section 3.1.
  • Base voxel size epsilon = not specified
    Sets the coarsest octree resolution for Eq. (2); no value given.
  • Ground height offset for ground surfels = not specified
    Fixes the height of ground Gaussians relative to the nearest camera pose; no value given in Ground Reconstruction.
  • Loss weights lambda_geo, lambda_obj, lambda_road, lambda_sky, lambda_s, lambda_d, lambda_n = not specified
    Equations (4) and (5) define a weighted loss, but the weights are not reported.
  • Frame-wise perception range and target point count per sweep = not specified
    Frame-Wise Division says points are sampled to match a real LiDAR sweep, but the range and count are not given.
assumptions (6)
  • domain assumption Road is approximately parallel to the camera plane, and ground elevation can be recovered from the nearest camera pose with a fixed height offset.
    Used in Section 3.1, Ground Reconstruction, to initialize and regularize the road surface.
  • domain assumption Dynamic objects are vehicles with 3D bounding boxes from a vision-based tracker, and all sampled points inside a box belong to that object.
    Used in Section 3.1, Dynamic Reconstruction, to model moving vehicles and later aggregate their points in box coordinates.
  • domain assumption The SfM sparse point cloud is a reliable scene skeleton and coverage proxy for occupancy.
    Sparse point clouds from detector-free SfM initialize the octree and define the perception volume in Section 3.1.
  • domain assumption LiDAR point clouds serve as ground truth for geometric accuracy, and geometry outside LiDAR coverage can still be scored by CD.
    Section 4.2 acknowledges that reconstruction beyond LiDAR coverage is counted as high CD, yet CD is the headline geometry metric.
  • ad hoc to paper Excluding ego-static scenes from training and evaluation is a fair way to compare vision-only labels with LiDAR labels.
    Section 4.3 reduces the benchmark from 798/202 scenes to 637/165 scenes because vision-only methods fail on ego-static scenes; the fairness of this reduction is not tested.
  • domain assumption Binary, non-semantic occupancy labels are sufficient to train and evaluate the downstream occupancy model.
    The curation pipeline in Section 3.2 produces binary voxels and Table 2 evaluates IoU, F1, precision, and recall on geometry only.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GS-Occ3D: Scaling Vision-only Occupancy Reconstruction with Gaussian Splatting." pith.science (2026). https://pith.science/paper/SN6AQMZX

@misc{pith2026250719451,
  author       = {Pith},
  title        = {Pith review of: GS-Occ3D: Scaling Vision-only Occupancy Reconstruction with Gaussian Splatting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SN6AQMZX}},
  note         = {Machine review of arXiv:2507.19451}
}
read the original abstract

Occupancy is crucial for autonomous driving, providing essential geometric priors for perception and planning. However, existing methods predominantly rely on LiDAR-based occupancy annotations, which limits scalability and prevents leveraging vast amounts of potential crowdsourced data for auto-labeling. To address this, we propose GS-Occ3D, a scalable vision-only framework that directly reconstructs occupancy. Vision-only occupancy reconstruction poses significant challenges due to sparse viewpoints, dynamic scene elements, severe occlusions, and long-horizon motion. Existing vision-based methods primarily rely on mesh representation, which suffer from incomplete geometry and additional post-processing, limiting scalability. To overcome these issues, GS-Occ3D optimizes an explicit occupancy representation using an Octree-based Gaussian Surfel formulation, ensuring efficiency and scalability. Additionally, we decompose scenes into static background, ground, and dynamic objects, enabling tailored modeling strategies: (1) Ground is explicitly reconstructed as a dominant structural element, significantly improving large-area consistency; (2) Dynamic vehicles are separately modeled to better capture motion-related occupancy patterns. Extensive experiments on the Waymo dataset demonstrate that GS-Occ3D achieves state-of-the-art geometry reconstruction results. By curating vision-only binary occupancy labels from diverse urban scenes, we show their effectiveness for downstream occupancy models on Occ3D-Waymo and superior zero-shot generalization on Occ3D-nuScenes. It highlights the potential of large-scale vision-based occupancy reconstruction as a new paradigm for scalable auto-labeling. Project Page: https://gs-occ3d.github.io/

Figures

Figures reproduced from arXiv: 2507.19451 by the authors.

Figure 1
Figure 1. Overview of occupancy reconstruction pipelines. a) Existing methods predominantly rely on LiDAR-based occu￾pancy annotations, requiring costly specialized surveying vehicles, which significantly limits scalability. b) In contrast, GS-Occ3D in￾troduces a scalable, vision-only occupancy reconstruction frame￾work that effectively harnesses abundant crowdsourced data from consumer-grade vehicles for auto-labeling. Our a… view at source ↗
Figure 2
Figure 2. Overview of the GS-Occ3D. Left: Panoramic street views captured along long trajectories are used to generate a sparse point cloud and ground surfels as initialization. We adopt Octree-based Gaussian Surfel representation that integrates ground, background, and dynamic objects to achieve scalable vision-only geometry reconstruction. Here, we present an uphill scene with colors indicating height. Middle: Given the vis… view at source ↗
Figure 3
Figure 3. Visualization of Geometry Reconstruction on Waymo. The color represents CD with respect to LiDAR, ranging from blue (lower CD) to red (higher CD). We exhibit improved reconstruction fidelity in weakly-textured regions compared to other methods, while maintaining structural completeness comparable to LiDAR point cloud, even in the absence of geometric priors. Ours LiDAR [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Details of Geometry Reconstruction. We present the detailed geometry of the red-boxed area in [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Generalization Results on the Occ3D-Waymo Vali￾dation Set. We evaluate the SOTA occupancy model CVT-Occ [94] trained with our labels and Occ3D, achieving reasonable and overall comparable results. RGB Ours Occ3D Motorcycle Lane marking [PITH_FULL_IMAGE:figures/full_fi…
Figure 7
Figure 7. Figure 7: More Visualization. Up: Richer semantic labels. Down: superior generalization on Occ3D-nuScenes. Pink indi￾cates binary prediction, others show errors. (e.g., Waymo’s forward-facing views versus LiDAR’s 360- degree coverage), we achieve slightly lower performance on th…
Figure 8
Figure 8. Figure 8: Ablation Results of Ground Gaussians. We show the effectiveness of our ground gaussians. Colors indicate height, ranging from blue to red. Views Method Geometry Rendering CD-Pcd ↓ CD-Mesh ↓ PSNR ↑ 5Cam PGSR [9] 3.63 4.41 19.18 2DGS [27] 1.25 2.14 23.42 GVKF [64] 0.82 1…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

109 extracted references · 46 canonical work pages

  1. [1]

    Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P

    Jonathan T. Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P. Srinivasan. Mip-nerf: A multiscale representation for anti-aliasing neu- ral radiance fields. In ICCV, 2021. 2

    2021

  2. [2]

    Se- mantickitti: A dataset for semantic scene understanding of lidar sequences

    Jens Behley, Martin Garbade, Andres Milioto, Jan Quen- zel, Sven Behnke, Cyrill Stachniss, and Jurgen Gall. Se- mantickitti: A dataset for semantic scene understanding of lidar sequences. In Proceedings of the IEEE/CVF inter- national conference on computer vision, pages 9297–9307,

  3. [3]

    Langocc: Self-supervised open vocabulary occupancy estimation via volume rendering

    Simon Boeder, Fabian Gigengack, and Benjamin Risse. Langocc: Self-supervised open vocabulary occupancy estimation via volume rendering. arXiv preprint arXiv:2407.17310, 2024. 3

    arXiv 2024

  4. [4]

    Gaussianflowocc: Sparse and weakly supervised occu- pancy estimation using gaussian splatting and temporal flow

    Simon Boeder, Fabian Gigengack, and Benjamin Risse. Gaussianflowocc: Sparse and weakly supervised occu- pancy estimation using gaussian splatting and temporal flow. arXiv preprint arXiv:2502.17288, 2025. 3

    arXiv 2025

  5. [5]

    nuscenes: A mul- timodal dataset for autonomous driving

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuscenes: A mul- timodal dataset for autonomous driving. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11621–11631, 2020. 3

    2020

  6. [6]

    Monoscene: Monocular 3d semantic scene completion

    Anh-Quan Cao and Raoul De Charette. Monoscene: Monocular 3d semantic scene completion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 3991–4001, 2022. 3

    2022

  7. [7]

    Gaussrender: Learning 3d occupancy with gaussian rendering, 2025

    Loïck Chambon, Eloi Zablocki, Alexandre Boulch, Mick- aël Chen, and Matthieu Cord. Gaussrender: Learning 3d occupancy with gaussian rendering, 2025. 3

    2025

  8. [8]

    Argoverse: 3d tracking and forecasting with rich maps

    Ming-Fang Chang, John Lambert, Patsorn Sangkloy, Jag- jeet Singh, Slawomir Bak, Andrew Hartnett, De Wang, Pe- ter Carr, Simon Lucey, Deva Ramanan, et al. Argoverse: 3d tracking and forecasting with rich maps. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8748–8757, 2019. 3

    2019

Show all 109 references
  1. [9]

    Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction

    Danpeng Chen, Hai Li, Weicai Ye, Yifan Wang, Weijian Xie, Shangjin Zhai, Nan Wang, Haomin Liu, Hujun Bao, and Guofeng Zhang. Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction. arXiv preprint arXiv:2406.06521, 2024. 2, 5, 6, 8

    2024 arXiv

  2. [10]

    Dogaussian: Distributed- oriented gaussian splatting for large-scale 3d reconstruction via gaussian consensus

    Yu Chen and Gim Hee Lee. Dogaussian: Distributed- oriented gaussian splatting for large-scale 3d reconstruction via gaussian consensus. arXiv preprint arXiv:2405.13943,

    arXiv

  3. [11]

    Omnire: Omni urban scene reconstruction

    Ziyu Chen, Jiawei Yang, Jiahui Huang, Riccardo de Lutio, Janick Martinez Esturo, Boris Ivanovic, Or Litany, Zan Go- jcic, Sanja Fidler, Marco Pavone, Li Song, and Yue Wang. Omnire: Omni urban scene reconstruction. arXiv preprint arXiv:2408.16760, 2024. 2

    2024 arXiv

  4. [12]

    Trackocc: Camera-based 4d panop- tic occupancy tracking

    Zhuoguang Chen, Kenan Li, Xiuyu Yang, Tao Jiang, Yim- ing Li, and Hang Zhao. Trackocc: Camera-based 4d panop- tic occupancy tracking. arXiv preprint arXiv:2503.08471,

    arXiv

  5. [13]

    Long3r: Long sequence streaming 3d re- construction, 2025

    Zhuoguang Chen, Minghui Qin, Tianyuan Yuan, Zhe Liu, and Hang Zhao. Long3r: Long sequence streaming 3d re- construction, 2025. 8

    2025

  6. [14]

    Mask2former for video instance segmentation

    Bowen Cheng, Anwesa Choudhuri, Ishan Misra, Alexan- der Kirillov, Rohit Girdhar, and Alexander G Schwing. Mask2former for video instance segmentation. arXiv preprint arXiv:2112.10764, 2021. 3, 8

    2021 arXiv

  7. [15]

    Streetsurfgs: Scalable ur- ban street surface reconstruction with planar-based gaus- sian splatting

    Xiao Cui, Weicai Ye, Yifan Wang, Guofeng Zhang, Wen- gang Zhou, and Houqiang Li. Streetsurfgs: Scalable ur- ban street surface reconstruction with planar-based gaus- sian splatting. arXiv preprint arXiv:2410.04354, 2024. 2

    2024 arXiv

  8. [16]

    High-quality surface re- construction using gaussian surfels

    Pinxuan Dai, Jiamin Xu, Wenxiang Xie, Xinguo Liu, Huamin Wang, and Weiwei Xu. High-quality surface re- construction using gaussian surfels. In ACM SIGGRAPH 2024 Conference Papers, pages 1–11, 2024. 2

    2024

  9. [17]

    Scaling rectified flow transformers for high-resolution image synthesis, 2024

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, Dustin Podell, Tim Dockhorn, Zion English, Kyle Lacey, Alex Goodwin, Yan- nik Marek, and Robin Rombach. Scaling rectified flow tran...

    2024

  10. [18]

    Instantsplat: Un- bounded sparse-view pose-free gaussian splatting in 40 sec- onds

    Zhiwen Fan, Wenyan Cong, Kairun Wen, Kevin Wang, Jian Zhang, Xinghao Ding, Danfei Xu, Boris Ivanovic, Marco Pavone, Georgios Pavlakos, et al. Instantsplat: Un- bounded sparse-view pose-free gaussian splatting in 40 sec- onds. arXiv preprint arXiv:2403.20309, 2(3):4, 2024. 8

    2024 arXiv

  11. [19]

    Rogs: Large scale road surface reconstruction with meshgrid gaussian

    Zhiheng Feng, Wenhua Wu, Tianchen Deng, and Hesh- eng Wang. Rogs: Large scale road surface reconstruction with meshgrid gaussian. arXiv preprint arXiv:2405.14342,

    arXiv

  12. [20]

    Cc-3dt: Panoramic 3d object tracking via cross-camera fusion

    Tobias Fischer, Yung-Hsu Yang, Suryansh Kumar, Min Sun, and Fisher Yu. Cc-3dt: Panoramic 3d object tracking via cross-camera fusion. arXiv preprint arXiv:2212.01247,

    arXiv

  13. [21]

    Sugar: Surface- aligned gaussian splatting for efficient 3d mesh recon- struction and high-quality mesh rendering

    Antoine Guédon and Vincent Lepetit. Sugar: Surface- aligned gaussian splatting for efficient 3d mesh recon- struction and high-quality mesh rendering. arXiv preprint arXiv:2311.12775, 2023. 2

    2023 arXiv

  14. [22]

    Streetsurf: Extending multi-view im- plicit surface reconstruction to street views

    Jianfei Guo, Nianchen Deng, Xinyang Li, Yeqi Bai, Bo- tian Shi, Chiyu Wang, Chenjing Ding, Dongliang Wang, and Yikang Li. Streetsurf: Extending multi-view im- plicit surface reconstruction to street views. arXiv preprint arXiv:2306.04988, 2023. 2, 5, 6

    2023 arXiv

  15. [23]

    Dist-4d: Disentangled spatiotemporal diffusion with metric depth for 4d driving scene generation

    Jiazhe Guo, Yikang Ding, Xiwu Chen, Shuo Chen, Bohan Li, Yingshuang Zou, Xiaoyang Lyu, Feiyang Tan, Xiaojuan Qi, Zhiheng Li, et al. Dist-4d: Disentangled spatiotemporal diffusion with metric depth for 4d driving scene generation. arXiv preprint arXiv:2503.15208, 2025. 2

    2025 arXiv

  16. [24]

    Detector-free struc- ture from motion, 2023

    Xingyi He, Jiaming Sun, Yifan Wang, Sida Peng, Qixing Huang, Hujun Bao, and Xiaowei Zhou. Detector-free struc- ture from motion, 2023. 3

    2023

  17. [25]

    Splatad: Real-time li- dar and camera rendering with 3d gaussian splatting for au- tonomous driving, 2024

    Georg Hess, Carl Lindström, Maryam Fatemi, Christoffer Petersson, and Lennart Svensson. Splatad: Real-time li- dar and camera rendering with 3d gaussian splatting for au- tonomous driving, 2024. 2

    2024

  18. [26]

    Denoising dif- fusion probabilistic models, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models, 2020. 2

    2020

  19. [27]

    2d gaussian splatting for geometrically ac- curate radiance fields

    Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically ac- curate radiance fields. arXiv preprint arXiv:2403.17888 ,

    arXiv

  20. [28]

    Tri-perspective view for vision- based 3d semantic occupancy prediction

    Yuanhui Huang, Wenzhao Zheng, Yunpeng Zhang, Jie Zhou, and Jiwen Lu. Tri-perspective view for vision- based 3d semantic occupancy prediction. In Proceedings of the IEEE/CVF conference on computer vision and pat- tern recognition, pages 9223–9232, 2023. 3

    2023

  21. [29]

    Selfocc: Self-supervised vision-based 3d oc- cupancy prediction

    Yuanhui Huang, Wenzhao Zheng, Borui Zhang, Jie Zhou, and Jiwen Lu. Selfocc: Self-supervised vision-based 3d oc- cupancy prediction. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 19946–19956, 2024. 3

    2024

  22. [30]

    Gaussianformer: Scene as gaussians for vision-based 3d semantic occupancy prediction

    Yuanhui Huang, Wenzhao Zheng, Yunpeng Zhang, Jie Zhou, and Jiwen Lu. Gaussianformer: Scene as gaussians for vision-based 3d semantic occupancy prediction. In Eu- ropean Conference on Computer Vision , pages 376–393. Springer, 2024. 1, 3

    2024

  23. [31]

    Gausstr: Foundation model-aligned gaussian transformer for self-supervised 3d spatial understanding

    Haoyi Jiang, Liu Liu, Tianheng Cheng, Xinjie Wang, Tian- wei Lin, Zhizhong Su, Wenyu Liu, and Xinggang Wang. Gausstr: Foundation model-aligned gaussian transformer for self-supervised 3d spatial understanding. arXiv preprint arXiv:2412.13193, 2024. 3

    2024 arXiv

  24. [32]

    Horizon-gs: Unified 3d gaussian splatting for large-scale aerial-to-ground scenes, 2024

    Lihan Jiang, Kerui Ren, Mulin Yu, Linning Xu, Junt- ing Dong, Tao Lu, Feng Zhao, Dahua Lin, and Bo Dai. Horizon-gs: Unified 3d gaussian splatting for large-scale aerial-to-ground scenes, 2024. 2

    2024

  25. [33]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph., 42(4):139–1,

  26. [34]

    A hierarchical 3d gaussian representation for real- time rendering of very large datasets

    Bernhard Kerbl, Andreas Meuleman, Georgios Kopanas, Michael Wimmer, Alexandre Lanvin, and George Dret- takis. A hierarchical 3d gaussian representation for real- time rendering of very large datasets. ACM Transactions on Graphics (TOG), 43(4):1–15, 2024. 2

    2024

  27. [35]

    Grounding image matching in 3d with mast3r

    Vincent Leroy, Yohann Cabon, and Jérôme Revaud. Grounding image matching in 3d with mast3r. InEuropean Conference on Computer Vision , pages 71–91. Springer,

  28. [36]

    Uniscene: Unified occupancy-centric driving scene generation

    Bohan Li, Jiazhe Guo, Hongsi Liu, Yingshuang Zou, Yikang Ding, Xiwu Chen, Hu Zhu, Feiyang Tan, Chi Zhang, Tiancai Wang, et al. Uniscene: Unified occupancy-centric driving scene generation. arXiv preprint arXiv:2412.05435, 2024. 2

    2024 arXiv

  29. [37]

    Occmamba: Semantic occu- pancy prediction with state space models

    Heng Li, Yuenan Hou, Xiaohan Xing, Yuexin Ma, Xiao Sun, and Yanyong Zhang. Occmamba: Semantic occu- pancy prediction with state space models. InProceedings of the Computer Vision and Pattern Recognition Conference , pages 11949–11959, 2025. 3

    2025

  30. [38]

    Mtgs: Multi- traversal gaussian splatting, 2025

    Tianyu Li, Yihang Qiu, Zhenhua Wu, Carl Lindström, Peng Su, Matthias Nießner, and Hongyang Li. Mtgs: Multi- traversal gaussian splatting, 2025. 2

    2025

  31. [39]

    Bevstereo: Enhancing depth estimation in multi-view 3d object detection with temporal stereo

    Yinhao Li, Han Bao, Zheng Ge, Jinrong Yang, Jianjian Sun, and Zeming Li. Bevstereo: Enhancing depth estimation in multi-view 3d object detection with temporal stereo. InPro- ceedings of the AAAI Conference on Artificial Intelligence, pages 1486–1494, 2023. 3

    2023

  32. [40]

    V oxformer: Sparse voxel transformer for camera-based 3d semantic scene completion

    Yiming Li, Zhiding Yu, Christopher Choy, Chaowei Xiao, Jose M Alvarez, Sanja Fidler, Chen Feng, and Anima Anandkumar. V oxformer: Sparse voxel transformer for camera-based 3d semantic scene completion. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern ...

    2023

  33. [41]

    Sscbench: A large-scale 3d semantic scene completion benchmark for autonomous driving

    Yiming Li, Sihang Li, Xinhao Liu, Moonjun Gong, Ke- nan Li, Nuo Chen, Zijun Wang, Zhiheng Li, Tao Jiang, Fisher Yu, et al. Sscbench: A large-scale 3d semantic scene completion benchmark for autonomous driving. In 2024 IEEE/RSJ International Conference on Intelligent Robots and...

    2024

  34. [42]

    Fb-occ: 3d occu- pancy prediction based on forward-backward view trans- formation

    Zhiqi Li, Zhiding Yu, David Austin, Mingsheng Fang, Shiyi Lan, Jan Kautz, and Jose M Alvarez. Fb-occ: 3d occu- pancy prediction based on forward-backward view trans- formation. arXiv preprint arXiv:2307.01492, 2023. 3

    2023 arXiv

  35. [43]

    Bevformer: learning bird’s-eye-view representation from lidar-camera via spatiotemporal transformers.IEEE Transactions on Pat- tern Analysis and Machine Intelligence, 2024

    Zhiqi Li, Wenhai Wang, Hongyang Li, Enze Xie, Chong- hao Sima, Tong Lu, Qiao Yu, and Jifeng Dai. Bevformer: learning bird’s-eye-view representation from lidar-camera via spatiotemporal transformers.IEEE Transactions on Pat- tern Analysis and Machine Intelligence, 2024. 3

    2024

  36. [44]

    Kitti-360: A novel dataset and benchmarks for urban scene understanding in 2d and 3d

    Yiyi Liao, Jun Xie, and Andreas Geiger. Kitti-360: A novel dataset and benchmarks for urban scene understanding in 2d and 3d. IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, 45(3):3292–3310, 2022. 3

    2022

  37. [45]

    Vastgaussian: Vast 3d gaussians for large scene reconstruction

    Jiaqi Lin, Zhihao Li, Xiao Tang, Jianzhuang Liu, Shiyong Liu, Jiayue Liu, Yangdi Lu, Xiaofei Wu, Songcen Xu, You- liang Yan, et al. Vastgaussian: Vast 3d gaussians for large scene reconstruction. arXiv preprint arXiv:2402.17427 ,

    arXiv

  38. [46]

    V oteflow: Enforcing local rigidity in self-supervised scene flow, 2025

    Yancong Lin, Shiming Wang, Liangliang Nan, Julian Kooij, and Holger Caesar. V oteflow: Enforcing local rigidity in self-supervised scene flow, 2025. 2

    2025

  39. [47]

    Lidar-based 4d occu- pancy completion and forecasting

    Xinhao Liu, Moonjun Gong, Qi Fang, Haoyu Xie, Yim- ing Li, Hang Zhao, and Chen Feng. Lidar-based 4d occu- pancy completion and forecasting. In 2024 IEEE/RSJ In- ternational Conference on Intelligent Robots and Systems (IROS), pages 11102–11109. IEEE, 2024. 1, 3

    2024

  40. [48]

    Citygaussian: Real-time high-quality large-scale scene rendering with gaussians

    Yang Liu, Chuanchen Luo, Lue Fan, Naiyan Wang, Jun- ran Peng, and Zhaoxiang Zhang. Citygaussian: Real-time high-quality large-scale scene rendering with gaussians. In European Conference on Computer Vision, pages 265–282. Springer, 2024. 2

    2024

  41. [49]

    Let occ flow: Self- supervised 3d occupancy flow prediction

    Yili Liu, Linzhan Mou, Xuan Yu, Chenrui Han, Sitong Mao, Rong Xiong, and Yue Wang. Let occ flow: Self- supervised 3d occupancy flow prediction. arXiv preprint arXiv:2407.07587, 2024. 3

    2024 arXiv

  42. [50]

    Bevfusion: Multi-task multi-sensor fusion with unified bird’s-eye view representation

    Zhijian Liu, Haotian Tang, Alexander Amini, Xinyu Yang, Huizi Mao, Daniela L Rus, and Song Han. Bevfusion: Multi-task multi-sensor fusion with unified bird’s-eye view representation. In 2023 IEEE international conference on robotics and automation (ICRA), pages 2774–2781. IEEE,

    2023

  43. [51]

    Urban radiance field represen- tation with deformable neural mesh primitives

    Fan Lu, Yan Xu, Guang Chen, Hongsheng Li, Kwan-Yee Lin, and Changjun Jiang. Urban radiance field represen- tation with deformable neural mesh primitives. In ICCV, pages 465–476, 2023. 2

    2023

  44. [52]

    Scaffold-gs: Structured 3d gaussians for view-adaptive rendering

    Tao Lu, Mulin Yu, Linning Xu, Yuanbo Xiangli, Limin Wang, Dahua Lin, and Bo Dai. Scaffold-gs: Structured 3d gaussians for view-adaptive rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20654–20664, 2024. 3

    2024

  45. [53]

    Infinicube: Un- bounded and controllable dynamic 3d driving scene gen- eration with world-guided video models, 2024

    Yifan Lu, Xuanchi Ren, Jiawei Yang, Tianchang Shen, Zhangjie Wu, Jun Gao, Yue Wang, Siheng Chen, Mike Chen, Sanja Fidler, and Jiahui Huang. Infinicube: Un- bounded and controllable dynamic 3d driving scene gen- eration with world-guided video models, 2024. 2

    2024

  46. [54]

    Zopp: A framework of zero-shot offboard panoptic perception for autonomous driving, 2024

    Tao Ma, Hongbin Zhou, Qiusheng Huang, Xuemeng Yang, Jianfei Guo, Bo Zhang, Min Dou, Yu Qiao, Botian Shi, and Hongsheng Li. Zopp: A framework of zero-shot offboard panoptic perception for autonomous driving, 2024. 1

    2024

  47. [55]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. 65(1):99–106, 2021. 2

    2021

  48. [56]

    Instant neural graphics primitives with a multiresolution hash encoding

    Thomas Müller, Alex Evans, Christoph Schied, and Alexander Keller. Instant neural graphics primitives with a multiresolution hash encoding. ACM TOG, 2022. 2

    2022

  49. [57]

    Scaling diffusion models to real-world 3d lidar scene completion

    Lucas Nunes, Rodrigo Marcuzzi, Benedikt Mersch, Jens Behley, and Cyrill Stachniss. Scaling diffusion models to real-world 3d lidar scene completion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14770–14780, 2024. 2

    2024

  50. [58]

    Neural scene graphs for dynamic scenes

    Julian Ost, Fahim Mannan, Nils Thuerey, Julian Knodt, and Felix Heide. Neural scene graphs for dynamic scenes. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 2856–2865, 2021. 2

    2021

  51. [59]

    Co-occ: Coupling explicit feature fusion with volume rendering regularization for multi-modal 3d semantic occupancy prediction

    Jingyi Pan, Zipeng Wang, and Lin Wang. Co-occ: Coupling explicit feature fusion with volume rendering regularization for multi-modal 3d semantic occupancy prediction. IEEE Robotics and Automation Letters, 2024. 3

    2024

  52. [60]

    Renderocc: Vision-centric 3d occupancy pre- diction with 2d rendering supervision

    Mingjie Pan, Jiaming Liu, Renrui Zhang, Peixiang Huang, Xiaoqi Li, Hongwei Xie, Bing Wang, Li Liu, and Shang- hang Zhang. Renderocc: Vision-centric 3d occupancy pre- diction with 2d rendering supervision. In 2024 IEEE Inter- national Conference on Robotics and Automation (ICRA...

    2024

  53. [61]

    Time will tell: New outlooks and a baseline for temporal multi- view 3d object detection

    Jinhyung Park, Chenfeng Xu, Shijia Yang, Kurt Keutzer, Kris Kitani, Masayoshi Tomizuka, and Wei Zhan. Time will tell: New outlooks and a baseline for temporal multi- view 3d object detection. arXiv preprint arXiv:2210.02443,

    arXiv

  54. [62]

    Octree-gs: Towards consistent real-time rendering with lod-structured 3d gaussians

    Kerui Ren, Lihan Jiang, Tao Lu, Mulin Yu, Linning Xu, Zhangkai Ni, and Bo Dai. Octree-gs: Towards consistent real-time rendering with lod-structured 3d gaussians. arXiv preprint arXiv:2403.17898, 2024. 3

    2024 arXiv

  55. [63]

    Lmscnet: Lightweight multiscale 3d semantic completion

    Luis Roldao, Raoul De Charette, and Anne Verroust- Blondet. Lmscnet: Lightweight multiscale 3d semantic completion. In 2020 International Conference on 3D Vi- sion (3DV), pages 111–119. IEEE, 2020. 3

    2020

  56. [64]

    Gvkf: Gaus- sian voxel kernel functions for highly efficient surface re- construction in open scenes

    Gaochao Song, Chong Cheng, and Hao Wang. Gvkf: Gaus- sian voxel kernel functions for highly efficient surface re- construction in open scenes. Advances in Neural Informa- tion Processing Systems, 37:104792–104815, 2025. 2, 5, 6, 8

    2025

  57. [65]

    Coda-4dgs: Dynamic gaussian splatting with context and deformation awareness for autonomous driving

    Rui Song, Chenwei Liang, Yan Xia, Walter Zimmer, Hu Cao, Holger Caesar, Andreas Festag, and Alois Knoll. Coda-4dgs: Dynamic gaussian splatting with context and deformation awareness for autonomous driving. arXiv preprint arXiv:2503.06744, 2025. 2

    2025 arXiv

  58. [66]

    Semantic scene completion from a single depth image

    Shuran Song, Fisher Yu, Andy Zeng, Angel X Chang, Manolis Savva, and Thomas Funkhouser. Semantic scene completion from a single depth image. In Proceedings of the IEEE conference on computer vision and pattern recog- nition, pages 1746–1754, 2017. 3

    2017

  59. [67]

    Don’t shake the wheel: Momentum- aware planning in end-to-end autonomous driving

    Ziying Song, Caiyan Jia, Lin Liu, Hongyu Pan, Yongchang Zhang, Junming Wang, Xingyu Zhang, Shaoqing Xu, Lei Yang, and Yadan Luo. Don’t shake the wheel: Momentum- aware planning in end-to-end autonomous driving. In Pro- ceedings of the Computer Vision and Pattern Recognition Co...

    2025

  60. [68]

    Loftr: Detector-free local feature matching with transformers

    Jiaming Sun, Zehong Shen, Yuang Wang, Hujun Bao, and Xiaowei Zhou. Loftr: Detector-free local feature matching with transformers. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 8922–8931, 2021. 3

    2021

  61. [69]

    Scalability in perception for autonomous driving: Waymo open dataset

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aure- lien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, et al. Scalability in perception for autonomous driving: Waymo open dataset. In Proceedings of the IEEE/CVF conference on computer ...

    2020

  62. [70]

    Block-nerf: Scalable large scene neural view synthesis

    Matthew Tancik, Vincent Casser, Xinchen Yan, Sabeek Pradhan, Ben Mildenhall, Pratul P Srinivasan, Jonathan T Barron, and Henrik Kretzschmar. Block-nerf: Scalable large scene neural view synthesis. In CVPR, pages 8248– 8258, 2022. 2

    2022

  63. [71]

    Occ3d: A large-scale 3d occupancy prediction benchmark for autonomous driving

    Xiaoyu Tian, Tao Jiang, Longfei Yun, Yucheng Mao, Huitong Yang, Yue Wang, Yilun Wang, and Hang Zhao. Occ3d: A large-scale 3d occupancy prediction benchmark for autonomous driving. Advances in Neural Information Processing Systems, 36:64318–64330, 2023. 1, 3, 5, 6

    2023

  64. [72]

    Neurad: Neural rendering for autonomous driving

    Adam Tonderski, Carl Lindström, Georg Hess, William Ljungbergh, Lennart Svensson, and Christoffer Petersson. Neurad: Neural rendering for autonomous driving. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14895–14904, 2024. 2

    2024

  65. [73]

    Mega-nerf: Scalable construction of large-scale nerfs for virtual fly-throughs

    Haithem Turki, Deva Ramanan, and Mahadev Satya- narayanan. Mega-nerf: Scalable construction of large-scale nerfs for virtual fly-throughs. In CVPR, pages 12922– 12931, 2022. 2

    2022

  66. [74]

    Suds: Scalable urban dynamic scenes

    Haithem Turki, Jason Y Zhang, Francesco Ferroni, and Deva Ramanan. Suds: Scalable urban dynamic scenes. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition , pages 12375–12385, 2023. 2

    2023

  67. [75]

    Dn-splatter: Depth and normal priors for gaussian splatting and meshing.arXiv preprint arXiv:2403.17822, 2024

    Matias Turkulainen, Xuqian Ren, Iaroslav Melekhov, Otto Seiskari, Esa Rahtu, and Juho Kannala. Dn-splatter: Depth and normal priors for gaussian splatting and meshing.arXiv preprint arXiv:2403.17822, 2024. 2

    2024 arXiv

  68. [76]

    Vggt: Visual geometry grounded transformer

    Jianyuan Wang, Minghao Chen, Nikita Karaev, Andrea Vedaldi, Christian Rupprecht, and David Novotny. Vggt: Visual geometry grounded transformer. In Proceedings of the Computer Vision and Pattern Recognition Conference , pages 5294–5306, 2025. 8

    2025

  69. [77]

    Unifying appearance codes and bilateral grids for driving scene gaussian splatting, 2025

    Nan Wang, Yuantao Chen, Lixing Xiao, Weiqing Xiao, Bo- han Li, Zhaoxi Chen, Chongjie Ye, Shaocong Xu, Saining Zhang, Ziyang Yan, Pierre Merriaux, Lei Lei, Tianfan Xue, and Hao Zhao. Unifying appearance codes and bilateral grids for driving scene gaussian splatting, 2025. 2

    2025

  70. [78]

    Neus: Learning neural implicit surfaces by volume rendering for multi-view re- construction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view re- construction. arXiv preprint arXiv:2106.10689 , 2021. 5, 6

    2021 arXiv

  71. [79]

    F2-nerf: Fast neural radiance field training with free camera trajectories

    Peng Wang, Yuan Liu, Zhaoxi Chen, Lingjie Liu, Ziwei Liu, Taku Komura, Christian Theobalt, and Wenping Wang. F2-nerf: Fast neural radiance field training with free camera trajectories. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages ...

    2023

  72. [80]

    Dust3r: Geometric 3d vision made easy

    Shuzhe Wang, Vincent Leroy, Yohann Cabon, Boris Chidlovskii, and Jerome Revaud. Dust3r: Geometric 3d vision made easy. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition , pages 20697–20709, 2024. 8

    2024

  73. [81]

    Openoccupancy: A large scale benchmark for surrounding semantic occupancy perception

    Xiaofeng Wang, Zheng Zhu, Wenbo Xu, Yunpeng Zhang, Yi Wei, Xu Chi, Yun Ye, Dalong Du, Jiwen Lu, and Xin- gang Wang. Openoccupancy: A large scale benchmark for surrounding semantic occupancy perception. In Proceed- ings of the IEEE/CVF International Conference on Com- puter Vis...

    2023

  74. [82]

    Uniocc: A unified benchmark for occupancy forecasting and prediction in autonomous driving

    Yuping Wang, Xiangyu Huang, Xiaokang Sun, Mingx- uan Yan, Shuo Xing, Zhengzhong Tu, and Jiachen Li. Uniocc: A unified benchmark for occupancy forecasting and prediction in autonomous driving. arXiv preprint arXiv:2503.24381, 2025. 3

    2025 arXiv

  75. [83]

    Surroundocc: Multi-camera 3d occu- pancy prediction for autonomous driving

    Yi Wei, Linqing Zhao, Wenzhao Zheng, Zheng Zhu, Jie Zhou, and Jiwen Lu. Surroundocc: Multi-camera 3d occu- pancy prediction for autonomous driving. InProceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 21729–21740, 2023. 3

    2023

  76. [84]

    Mars: An instance-aware, modular and realistic simulator for au- tonomous driving

    Zirui Wu, Tianyu Liu, Liyi Luo, Zhide Zhong, Jianteng Chen, Hongmin Xiao, Chao Hou, Haozhe Lou, Yuan- tao Chen, Runyi Yang, Yuxin Huang, Xiaoyu Ye, Zike Yan, Yongliang Shi, Yiyi Liao, and Hao Zhao. Mars: An instance-aware, modular and realistic simulator for au- tonomous drivi...

    2023

  77. [85]

    Cruise: Cooperative reconstruction and editing in v2x scenarios using gaussian splatting, 2025

    Haoran Xu, Saining Zhang, Peishuo Li, Baijun Ye, Xiaoxue Chen, Huan ang Gao, Jv Zheng, Xiaowei Song, Ziqiao Peng, Run Miao, Jinrang Jia, Yifeng Shi, Guangqi Yi, Hang Zhao, Hao Tang, Hongyang Li, Kaicheng Yu, and Hao Zhao. Cruise: Cooperative reconstruction and editing in v2x s...

    2025

  78. [86]

    Grid-guided neural radiance fields for large urban scenes

    Linning Xu, Yuanbo Xiangli, Sida Peng, Xingang Pan, Nanxuan Zhao, Christian Theobalt, Bo Dai, and Dahua Lin. Grid-guided neural radiance fields for large urban scenes. In CVPR, pages 8296–8306, 2023. 2

    2023

  79. [87]

    Street gaussians for modeling dynamic ur- ban scenes

    Yunzhi Yan, Haotong Lin, Chenxu Zhou, Weijie Wang, Haiyang Sun, Kun Zhan, Xianpeng Lang, Xiaowei Zhou, and Sida Peng. Street gaussians for modeling dynamic ur- ban scenes. arXiv preprint arXiv:2401.01339, 2024. 2, 4

    2024 arXiv

  80. [88]

    Emernerf: Emergent spatial-temporal scene decomposition via self-supervision

    Jiawei Yang, Boris Ivanovic, Or Litany, Xinshuo Weng, Se- ung Wook Kim, Boyi Li, Tong Che, Danfei Xu, Sanja Fi- dler, Marco Pavone, and Yue Wang. Emernerf: Emergent spatial-temporal scene decomposition via self-supervision. In International Conference on Learning Representations ,

  81. [89]

    Spectrally pruned gaussian fields with neural com- pensation, 2024

    Runyi Yang, Zhenxin Zhu, Zhou Jiang, Baijun Ye, Xiaoxue Chen, Yifei Zhang, Yuantao Chen, Jian Zhao, and Hao Zhao. Spectrally pruned gaussian fields with neural com- pensation, 2024. 2

    2024

  82. [90]

    X-scene: Large-scale driving scene gen- eration with high fidelity and flexible controllability

    Yu Yang, Alan Liang, Jianbiao Mei, Yukai Ma, Yong Liu, and Gim Hee Lee. X-scene: Large-scale driving scene gen- eration with high fidelity and flexible controllability. arXiv preprint arXiv:2506.13558, 2025. 2

    2025

  83. [91]

    Unisim: A neural closed-loop sensor simulator

    Ze Yang, Yun Chen, Jingkang Wang, Sivabalan Mani- vasagam, Wei-Chiu Ma, Anqi Joyce Yang, and Raquel Ur- tasun. Unisim: A neural closed-loop sensor simulator. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 1389–1399, 2023. 2

    2023

  84. [92]

    Blending distributed nerfs with tri-stage robust pose optimization

    Baijun Ye, Caiyun Liu, Xiaoyu Ye, Yuantao Chen, Yuhai Wang, Zike Yan, Yongliang Shi, Hao Zhao, and Guyue Zhou. Blending distributed nerfs with tri-stage robust pose optimization. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 7975–

    2024

  85. [93]

    Bevdiffuser: Plug-and-play dif- fusion model for bev denoising with ground-truth guidance

    Xin Ye, Burhaneddin Yaman, Sheng Cheng, Feng Tao, Ab- hirup Mallik, and Liu Ren. Bevdiffuser: Plug-and-play dif- fusion model for bev denoising with ground-truth guidance. In Proceedings of the Computer Vision and Pattern Recog- nition Conference, pages 1495–1504, 2025. 2

    2025

  86. [94]

    Cvt-occ: Cost volume temporal fusion for 3d occupancy prediction

    Zhangchen Ye, Tao Jiang, Chenfeng Xu, Yiming Li, and Hang Zhao. Cvt-occ: Cost volume temporal fusion for 3d occupancy prediction. In European Conference on Com- puter Vision, pages 381–397. Springer, 2024. 3, 6, 7

    2024

  87. [95]

    Gsdf: 3dgs meets sdf for improved render- ing and reconstruction

    Mulin Yu, Tao Lu, Linning Xu, Lihan Jiang, Yuanbo Xian- gli, and Bo Dai. Gsdf: 3dgs meets sdf for improved render- ing and reconstruction. arXiv preprint arXiv:2403.16964,

    arXiv

  88. [96]

    Gaussian opacity fields: Efficient and compact surface reconstruction in unbounded scenes

    Zehao Yu, Torsten Sattler, and Andreas Geiger. Gaussian opacity fields: Efficient and compact surface reconstruction in unbounded scenes. arXiv preprint arXiv:2404.10772 ,

    arXiv

  89. [97]

    Presight: Enhancing au- tonomous vehicle perception with city-scale nerf priors

    Tianyuan Yuan, Yucheng Mao, Jiawei Yang, Yicheng Liu, Yue Wang, and Hang Zhao. Presight: Enhancing au- tonomous vehicle perception with city-scale nerf priors. In European Conference on Computer Vision, pages 323–339. Springer, 2024. 2

    2024

  90. [98]

    Futuresight- drive: Thinking visually with spatio-temporal cot for au- tonomous driving

    Shuang Zeng, Xinyuan Chang, Mengwei Xie, Xinran Liu, Yifan Bai, Zheng Pan, Mu Xu, and Xing Wei. Futuresight- drive: Thinking visually with spatio-temporal cot for au- tonomous driving. arXiv preprint arXiv:2505.17685, 2025. 2

    2025 arXiv

  91. [99]

    Rade-gs: Ras- terizing depth in gaussian splatting

    Baowen Zhang, Chuan Fang, Rakesh Shrestha, Yixun Liang, Xiaoxiao Long, and Ping Tan. Rade-gs: Ras- terizing depth in gaussian splatting. arXiv preprint arXiv:2406.01467, 2024. 2

    2024 arXiv

  92. [100]

    Occnerf: Advancing 3d occupancy prediction in lidar-free environ- ments

    Chubin Zhang, Juncheng Yan, Yi Wei, Jiaxin Li, Li Liu, Yansong Tang, Yueqi Duan, and Jiwen Lu. Occnerf: Advancing 3d occupancy prediction in lidar-free environ- ments. arXiv preprint arXiv:2312.09243, 2023. 3

    2023 arXiv

  93. [101]

    Drone-assisted road gaussian splatting with cross- view uncertainty

    Saining Zhang, Baijun Ye, Xiaoxue Chen, Yuantao Chen, Zongzheng Zhang, Cheng Peng, Yongliang Shi, and Hao Zhao. Drone-assisted road gaussian splatting with cross- view uncertainty. arXiv preprint arXiv:2408.15242, 2024. 2

    2024 arXiv

  94. [102]

    Occformer: Dual-path transformer for vision-based 3d semantic occu- pancy prediction

    Yunpeng Zhang, Zheng Zhu, and Dalong Du. Occformer: Dual-path transformer for vision-based 3d semantic occu- pancy prediction. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision , pages 9433–9443,

  95. [103]

    Veon: V ocabulary-enhanced occupancy prediction

    Jilai Zheng, Pin Tang, Zhongdao Wang, Guoqing Wang, Xiangxuan Ren, Bailan Feng, and Chao Ma. Veon: V ocabulary-enhanced occupancy prediction. In European Conference on Computer Vision , pages 92–108. Springer,

  96. [104]

    Gaussianad: Gaussian- centric end-to-end autonomous driving

    Wenzhao Zheng, Junjie Wu, Yao Zheng, Sicheng Zuo, Zixun Xie, Longchao Yang, Yong Pan, Zhihui Hao, Peng Jia, Xianpeng Lang, et al. Gaussianad: Gaussian- centric end-to-end autonomous driving. arXiv preprint arXiv:2412.10371, 2024. 1, 3

    2024 arXiv

  97. [105]

    World4drive: End-to-end autonomous driving via intention-aware physical latent world model, 2025

    Yupeng Zheng, Pengxuan Yang, Zebin Xing, Qichao Zhang, Yuhang Zheng, Yinfeng Gao, Pengfei Li, Teng Zhang, Zhongpu Xia, Peng Jia, and Dongbin Zhao. World4drive: End-to-end autonomous driving via intention-aware physical latent world model, 2025. 2

    2025

  98. [106]

    Hugsim: A real-time, photo-realistic and closed-loop simulator for autonomous driving, 2024

    Hongyu Zhou, Longzhong Lin, Jiabao Wang, Yichong Lu, Dongfeng Bai, Bingbing Liu, Yue Wang, Andreas Geiger, and Yiyi Liao. Hugsim: A real-time, photo-realistic and closed-loop simulator for autonomous driving, 2024. 2

    2024

  99. [107]

    Hugs: Holistic urban 3d scene understanding via gaussian splatting

    Hongyu Zhou, Jiahao Shao, Lu Xu, Dongfeng Bai, We- ichao Qiu, Bingbing Liu, Yue Wang, Andreas Geiger, and Yiyi Liao. Hugs: Holistic urban 3d scene understanding via gaussian splatting. arXiv preprint arXiv:2403.12722, 2024. 2

    2024 arXiv

  100. [108]

    Driving- gaussian: Composite gaussian splatting for surrounding dynamic autonomous driving scenes

    Xiaoyu Zhou, Zhiwei Lin, Xiaojun Shan, Yongtao Wang, Deqing Sun, and Ming-Hsuan Yang. Driving- gaussian: Composite gaussian splatting for surrounding dynamic autonomous driving scenes. arXiv preprint arXiv:2312.07920, 2023. 2

    2023 arXiv

  101. [109]

    Occgs: Zero-shot 3d oc- cupancy reconstruction with semantic and geometric-aware gaussian splatting

    Xiaoyu Zhou, Jingqi Wang, Yongtao Wang, Yufei Wei, Nan Dong, and Ming-Hsuan Yang. Occgs: Zero-shot 3d oc- cupancy reconstruction with semantic and geometric-aware gaussian splatting. arXiv preprint arXiv:2502.04981, 2025. 3

    2025 arXiv

Pith tools

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