pith. machine review for the scientific record. sign in

arxiv: 2604.26238 · v1 · submitted 2026-04-29 · 💻 cs.CV

Recognition: unknown

EnerGS: Energy-Based Gaussian Splatting with Partial Geometric Priors

Rui Song , Tianhui Cai , Markus Gross , Yun Zhang , Walter Zimmer , Zhiyu Huang , Olaf Wysocki , Jiaqi Ma
Authors on Pith no claims yet

Pith reviewed 2026-05-07 13:53 UTC · model grok-4.3

classification 💻 cs.CV
keywords 3D Gaussian Splattingenergy-based optimizationpartial geometric priorsscene reconstructionoutdoor environmentssoft constraintsphotometric qualitygeometric stability
0
0 comments X

The pith

Modeling partial geometry as a continuous energy field lets EnerGS supply soft guidance to Gaussian primitive optimization instead of hard constraints.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

3D Gaussian Splatting faces a coupled non-convex optimization problem when reconstructing large outdoor scenes, where geometric measurements such as LiDAR points are typically incomplete and uneven. The paper shows that turning this partial evidence into a continuous energy field supplies soft directional influence on the Gaussians during training. This approach matters because it aims to raise image fidelity and geometric consistency while curbing overfitting, especially in sparse multi-view or single-camera capture settings. A reader would care if the method succeeds in letting available geometry steer the solution without closing off better photometric arrangements that the data alone cannot fully specify.

Core claim

The paper claims that representing partially observable geometry as a continuous energy field induced by the available geometric evidence supplies soft guidance to the optimization of Gaussian primitives. This steers the 3DGS training process without imposing direct restrictions on the solution space, yielding higher photometric quality, greater geometric stability, and reduced overfitting on large-scale outdoor scenes under both sparse multi-view and monocular conditions.

What carries the argument

The continuous energy field induced by geometric evidence, which supplies soft directional guidance to the placement and attributes of Gaussian primitives.

If this is right

  • Photometric quality improves consistently under sparse multi-view and monocular capture.
  • Geometric stability of the reconstructed scene increases.
  • Overfitting during 3DGS training is reduced on large outdoor environments.
  • The method remains effective even when geometric evidence is spatially incomplete and uneven.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same soft-energy formulation could be tested on indoor scenes or other reconstruction primitives to check whether the benefit is specific to outdoor scale.
  • Energy-based soft guidance might extend to additional uncertain inputs such as noisy depth maps or semantic labels without requiring new hard constraints.
  • If the energy field is made differentiable with respect to sensor noise levels, it could support uncertainty-aware training loops in future pipelines.

Load-bearing premise

That a continuous energy field built from incomplete and uneven geometric evidence can steer Gaussian optimization effectively without shrinking the solution space in ways that lower final image quality.

What would settle it

A controlled test on scenes with highly incomplete LiDAR coverage in which the energy-guided model produces lower PSNR or more visible artifacts than plain 3DGS would falsify the claim of reliable improvement.

Figures

Figures reproduced from arXiv: 2604.26238 by Jiaqi Ma, Markus Gross, Olaf Wysocki, Rui Song, Tianhui Cai, Walter Zimmer, Yun Zhang, Zhiyu Huang.

Figure 1
Figure 1. Figure 1: Accurate geometric priors can significantly improve Gaussian initialization and optimization (e.g., via point clouds from LiDAR). However, in large-scale outdoor scenes, such priors are often spatially incomplete. EnerGS addresses this limitation by partitioning space into occupied, free, and unknown regions and guiding the spatial distribution of Gaussians with a geometric energy field, enabling more stab… view at source ↗
Figure 2
Figure 2. Figure 2: Visual Comparison on KITTI and Waymo Open Dataset. Our EnerGS achieves superior novel view synthesis, particularly in geometrically unobserved regions (e.g., upper structures beyond LiDAR coverage). Our method renders significantly finer details in these areas compared to baselines, aligning with our theoretical expectation that the adaptive energy field facilitates robust reconstruction in sensor blind sp… view at source ↗
Figure 3
Figure 3. Figure 3: The Gap between Train and Test PSNR with Training Iteration. Our EnerGS consistently maintains a smaller train–test gap throughout training, indicating that our method encourages the model to learn multi-view consistent geometry rather than view￾point memorization and ultimately achieves the best performance, as reported in Tab. 1. 6. Conclusion We propose EnerGS, an energy-based framework for 3D Gaussian … view at source ↗
Figure 4
Figure 4. Figure 4: Random Initialization Experiment. We randomly initialize 50,000 Gaussian primitives (with 500 of them tracked and recorded) in free space and optimize them using vanilla 3D Gaussian Splatting (left) and EnerGS (right), respectively. The figure illustrates the initial and final positions of these Gaussians. Note that this visualization shows a slice at the camera height; therefore, some Gaussians may appear… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of Gradient Norm between Vanilla 3DGS and EnerGS. 12 view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of rendering results in LiDAR blind-spot regions (unobservable geometry), highlighting the effect of enabling the UNK field. B. Derivation and Interpretation of Equations (6) and (8) This section provides the rigorous mathematical derivation and physical interpretation of the energy potentials used in our optimization framework. We specifically detail the gradients for the occupancy energy (attr… view at source ↗
read the original abstract

3D Gaussian Splatting (3DGS) has been widely adopted for scene reconstruction, where training inherently constitutes a highly coupled and non-convex optimization problem. Recent works commonly incorporate geometric priors, such as LiDAR measurements, either for initialization or as training constraints, with the goal of improving photometric reconstruction quality. However, in large-scale outdoor scenarios, such geometric supervision is often spatially incomplete and uneven, which limits its effectiveness as a reliable prior and can even be detrimental to the final reconstruction. To address this challenge, we model partially observable geometry as a continuous energy field induced by geometric evidence and propose EnerGS. Rather than enforcing geometry as a hard constraint, EnerGS provides a soft geometric guidance for the optimization of Gaussian primitives, allowing geometric information to steer the optimization process without directly restricting the solution space. Extensive experiments on large-scale outdoor scenes demonstrate that, under both sparse multi-view and monocular settings, EnerGS consistently improves photometric quality and geometric stability, while effectively mitigating overfitting during 3DGS training.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces EnerGS, which models partially observable geometry (e.g., from sparse LiDAR or monocular depth) as a continuous energy field that supplies soft guidance during 3D Gaussian Splatting optimization. Rather than imposing hard geometric constraints, the energy field steers Gaussian primitive placement and attributes to improve photometric reconstruction quality, geometric stability, and overfitting resistance in large-scale outdoor scenes under sparse multi-view and monocular settings.

Significance. If validated, the method would offer a practical way to incorporate incomplete geometric priors without restricting the solution space, addressing a common limitation in 3DGS for real-world outdoor reconstruction. This could improve robustness in applications like autonomous driving or mapping where geometric evidence is uneven.

major comments (2)
  1. [Abstract] The central claim that the continuous energy field provides effective soft guidance without introducing biased gradients in low-evidence regions is load-bearing but unsupported by the provided text. The abstract asserts consistent improvements in photometric quality and overfitting mitigation, yet supplies no quantitative metrics, ablation studies, baseline comparisons, or error analysis (e.g., PSNR/SSIM deltas or per-region breakdowns), preventing assessment of whether the field harms fidelity where geometric coverage is zero or near-zero.
  2. [Method (energy field definition)] The construction of the energy field from partial observations is not detailed enough to evaluate the skeptic concern. If the field relies on interpolation, propagation, or a learned model, regions with spatially incomplete evidence (norm in large-scale outdoor scenes) could receive extrapolated values whose gradients bias optimization; the manuscript must specify the field's functional form, how it is induced by evidence, and any regularization to avoid restricting the Gaussian solution space.
minor comments (2)
  1. Clarify the exact form of the energy function and its integration into the 3DGS loss (e.g., as an additive term or modulation of covariance/position updates) to make the soft-guidance mechanism reproducible.
  2. [Abstract] The abstract's phrasing 'consistently improves' should be qualified with the specific settings (sparse multi-view vs. monocular) and any failure cases observed.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and have revised the manuscript to provide additional details and clarifications.

read point-by-point responses

  1. Referee: [Abstract] The central claim that the continuous energy field provides effective soft guidance without introducing biased gradients in low-evidence regions is load-bearing but unsupported by the provided text. The abstract asserts consistent improvements in photometric quality and overfitting mitigation, yet supplies no quantitative metrics, ablation studies, baseline comparisons, or error analysis (e.g., PSNR/SSIM deltas or per-region breakdowns), preventing assessment of whether the field harms fidelity where geometric coverage is zero or near-zero.

    Authors: The abstract is intentionally concise as a high-level summary. The full manuscript (Section 4) contains quantitative results, including PSNR/SSIM/LPIPS tables, baseline comparisons, and ablations on sparse outdoor scenes that demonstrate consistent photometric gains and reduced overfitting. To directly address the concern, we have revised the abstract to include key quantitative deltas (e.g., average PSNR improvement) and a brief statement on the energy field's zero-influence design in uncovered regions, supported by per-region analysis in the experiments. revision: yes

  2. Referee: [Method (energy field definition)] The construction of the energy field from partial observations is not detailed enough to evaluate the skeptic concern. If the field relies on interpolation, propagation, or a learned model, regions with spatially incomplete evidence (norm in large-scale outdoor scenes) could receive extrapolated values whose gradients bias optimization; the manuscript must specify the field's functional form, how it is induced by evidence, and any regularization to avoid restricting the Gaussian solution space.

    Authors: We agree that the energy field construction requires more explicit specification. In the revised manuscript, we expand Section 3.2 with the precise functional form (a distance-decay formulation induced directly from partial observations such as LiDAR points or monocular depth), the induction mechanism (nearest-neighbor propagation with exponential falloff), and the regularization (a hard zero-threshold beyond a coverage radius) that ensures no gradient bias or solution-space restriction in low-evidence regions. revision: yes

Circularity Check

0 steps flagged

No significant circularity; new modeling construct is independent of inputs.

full rationale

The paper introduces a continuous energy field induced by partial geometric evidence as soft guidance for 3DGS optimization, explicitly contrasting it with hard constraints. This is presented as a novel proposal whose value is shown via experiments on photometric quality and overfitting mitigation in incomplete outdoor scenes. No equations or steps in the provided abstract reduce the energy field or guidance mechanism to a fitted parameter, self-definition, or prior result by construction. The derivation chain remains self-contained, with the central claim resting on the proposed soft-guidance framework rather than any load-bearing self-citation or renaming of known patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that geometry can be usefully represented as a continuous energy field that provides beneficial soft guidance.

axioms (1)
  • domain assumption Partially observable geometry can be modeled as a continuous energy field induced by geometric evidence to provide soft guidance
    This is the core modeling choice stated in the abstract for handling incomplete priors.
invented entities (1)
  • Continuous energy field for geometry no independent evidence
    purpose: To supply soft geometric guidance during Gaussian optimization without hard constraints
    New construct introduced to address the stated limitations of incomplete and uneven priors.

pith-pipeline@v0.9.0 · 5491 in / 1249 out tokens · 44710 ms · 2026-05-07T13:53:30.315197+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. MDrive: Benchmarking Closed-Loop Cooperative Driving for End-to-End Multi-agent Systems

    cs.RO 2026-05 unverdicted novelty 7.0

    MDrive benchmark shows multi-agent cooperative driving systems generally outperform single-agent ones in closed-loop settings but perception sharing does not always improve planning and negotiation can harm performanc...

  2. ConFixGS: Learning to Fix Feedforward 3D Gaussian Splatting with Confidence-Aware Diffusion Priors in Driving Scenes

    cs.CV 2026-05 unverdicted novelty 7.0

    ConFixGS repairs feedforward 3D Gaussian Splatting with confidence-aware diffusion priors, delivering up to 3.68 dB PSNR gains and halved FID scores on Waymo, nuScenes, and KITTI novel view synthesis tasks.

Reference graph

Works this paper leans on

52 extracted references · 16 canonical work pages · cited by 2 Pith papers

  1. [1]

    Barron, J. T. A general and adaptive robust loss func- tion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 4331– 4339, 2019

    2019

  2. [2]

    T., Mildenhall, B., Tancik, M., Hedman, P., Martin-Brualla, R., and Srinivasan, P

    Barron, J. T., Mildenhall, B., Tancik, M., Hedman, P., Martin-Brualla, R., and Srinivasan, P. P. Mip-nerf: A multiscale representation for anti-aliasing neural radi- ance fields. InProceedings of the IEEE/CVF interna- tional conference on computer vision, pp. 5855–5864, 2021

    2021

  3. [3]

    T., Mildenhall, B., Verbin, D., Srinivasan, P

    Barron, J. T., Mildenhall, B., Verbin, D., Srinivasan, P. P., and Hedman, P. Mip-nerf 360: Unbounded anti- aliased neural radiance fields. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 5470–5479, 2022

    2022

  4. [4]

    Tackling view-dependent semantics in 3d language gaussian splatting.arXiv preprint arXiv:2505.24746, 2025

    Cen, J., Zhou, X., Fang, J., Wen, C., Xie, L., Zhang, X., Shen, W., and Tian, Q. Tackling view-dependent semantics in 3d language gaussian splatting.arXiv preprint arXiv:2505.24746, 2025

    work page arXiv 2025

  5. [5]

    Pgsr: Planar-based gaussian splatting for efficient and high- fidelity surface reconstruction.IEEE Transactions on Visualization and Computer Graphics, 2024

    Chen, D., Li, H., Ye, W., Wang, Y ., Xie, W., Zhai, S., Wang, N., Liu, H., Bao, H., and Zhang, G. Pgsr: Planar-based gaussian splatting for efficient and high- fidelity surface reconstruction.IEEE Transactions on Visualization and Computer Graphics, 2024

    2024

  6. [6]

    Periodic vibration gaussian: Dynamic urban scene reconstruction and real-time rendering.International Journal of Computer Vision, 134(3):83, 2026

    Chen, Y ., Gu, C., Jiang, J., Zhu, X., and Zhang, L. Periodic vibration gaussian: Dynamic urban scene reconstruction and real-time rendering.International Journal of Computer Vision, 134(3):83, 2026

    2026

  7. [7]

    Gaussianpro: 3d gaussian splatting with progressive propagation

    Cheng, K., Long, X., Yang, K., Yao, Y ., Yin, W., Ma, Y ., Wang, W., and Chen, X. Gaussianpro: 3d gaussian splatting with progressive propagation. InForty-first International Conference on Machine Learning, 2024

    2024

  8. [8]

    Chung, J., Oh, J., and Lee, K. M. Depth-regularized optimization for 3d gaussian splatting in few-shot im- ages. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 811– 820, 2024

    2024

  9. [9]

    Streetsurfgs: Scalable urban street surface reconstruction with planar-based gaussian splatting

    Cui, X., Ye, W., Wang, Y ., Zhang, G., Zhou, W., He, T., and Li, H. Streetsurfgs: Scalable urban street surface reconstruction with planar-based gaussian splatting. IEEE Transactions on Circuits and Systems for Video Technology, 2025

    2025

  10. [10]

    Nerf is a valuable assistant for 3d gaussian splatting

    Fang, S., Shen, I., Igarashi, T., Wang, Y ., Wang, Z., Yang, Y ., Ding, W., Zhou, S., et al. Nerf is a valuable assistant for 3d gaussian splatting. InProceedings of the IEEE/CVF International Conference on Computer Vision, pp. 26230–26240, 2025

    2025

  11. [11]

    Felzenszwalb, P. F. and Huttenlocher, D. P. Distance transforms of sampled functions.Theory of Comput- ing, 8(1):415–428, sep 2012

    2012

  12. [12]

    B., Fahmy, A., Song, R., Cremers, D., and Meess, H

    Gross, M., Matha, S. B., Fahmy, A., Song, R., Cremers, D., and Meess, H. Occufly: A 3d vision benchmark for semantic scene completion from the aerial perspec- tive. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026

    2026

  13. [13]

    and Lepetit, V

    Gu´edon, A. and Lepetit, V . Sugar: Surface-aligned gaussian splatting for efficient 3d mesh reconstruction and high-quality mesh rendering. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 5354–5363, 2024

    2024

  14. [14]

    Streetsurf: Extending multi-view implicit surface reconstruction to street views.arXiv preprint arXiv:2306.04988, 2023

    Guo, J., Deng, N., Li, X., Bai, Y ., Shi, B., Wang, C., Ding, C., Wang, D., and Li, Y . Streetsurf: Extending multi-view implicit surface reconstruction to street views.arXiv preprint arXiv:2306.04988, 2023

    work page arXiv 2023

  15. [15]

    Refersplat: Referring segmentation in 3d gaussian splatting.arXiv preprint arXiv:2508.08252, 2025

    He, S., Jie, G., Wang, C., Zhou, Y ., Hu, S., Li, G., and Ding, H. Refersplat: Referring segmentation in 3d gaussian splatting.arXiv preprint arXiv:2508.08252, 2025

    work page arXiv 2025

  16. [16]

    Splatad: Real-time lidar and camera rendering with 3d gaussian splatting for autonomous driving

    Hess, G., Lindstr ¨om, C., Fatemi, M., Petersson, C., and Svensson, L. Splatad: Real-time lidar and camera rendering with 3d gaussian splatting for autonomous driving. InProceedings of the Computer Vision and Pattern Recognition Conference, pp. 11982–11992, 2025

    2025

  17. [17]

    arXiv preprint arXiv:2410.22128 (2024) 3

    Hong, S., Jung, J., Shin, H., Han, J., Yang, J., Luo, C., and Kim, S. Pf3plat: Pose-free feed-forward 3d gaussian splatting.arXiv preprint arXiv:2410.22128, 2024

    work page arXiv 2024

  18. [18]

    and Han, Z

    Hu, P. and Han, Z. Vtgaussian-slam: Rgbd slam for large scale scenes with splatting view-tied 3d gaus- sians.arXiv preprint arXiv:2506.02741, 2025

    work page arXiv 2025

  19. [19]

    2d gaussian splatting for geometrically accurate radi- ance fields

    Huang, B., Yu, Z., Chen, A., Geiger, A., and Gao, S. 2d gaussian splatting for geometrically accurate radi- ance fields. InSIGGRAPH 2024 Conference Papers. Association for Computing Machinery, 2024. doi: 10.1145/3641519.3657428

    work page doi:10.1145/3641519.3657428 2024

  20. [20]

    Fatesgs: Fast and accurate sparse-view surface reconstruction using gaussian splatting with depth-feature consistency

    Huang, H., Wu, Y ., Deng, C., Gao, G., Gu, M., and Liu, Y .-S. Fatesgs: Fast and accurate sparse-view surface reconstruction using gaussian splatting with depth-feature consistency. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pp. 3644–3652, 2025

    2025

  21. [21]

    S3gaussian: Self-supervised street gaussians for 9 EnerGS: Energy-Based Gaussian Splatting with Partial Geometric Priors autonomous driving

    Huang, N., Wei, X., Zheng, W., An, P., Lu, M., Zhan, W., Tomizuka, M., Keutzer, K., and Zhang, S. S3gaussian: Self-supervised street gaussians for 9 EnerGS: Energy-Based Gaussian Splatting with Partial Geometric Priors autonomous driving. InProceedings of the IEEE In- ternational Conference on Robotics and Automation (ICRA), 2026

    2026

  22. [22]

    Det-gs: Depth-and edge-aware regularization for high-fidelity 3d gaussian splatting.arXiv preprint arXiv:2508.04099, 2025

    Huang, Z., Xu, M., and Perry, S. Det-gs: Depth-and edge-aware regularization for high-fidelity 3d gaussian splatting.arXiv preprint arXiv:2508.04099, 2025

    work page arXiv 2025

  23. [23]

    Li-gs: Gaussian splatting with lidar incorporated for accurate large-scale reconstruction

    Jiang, C., Gao, R., Shao, K., Wang, Y ., Xiong, R., and Zhang, Y . Li-gs: Gaussian splatting with lidar incorporated for accurate large-scale reconstruction. IEEE Robotics and Automation Letters, 2024

    2024

  24. [24]

    3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics, 42(4), July 2023

    Kerbl, B., Kopanas, G., Leimk¨uhler, T., and Drettakis, G. 3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics, 42(4), July 2023. URL https://repo-sam.inria. fr/fungraph/3d-gaussian-splatting/

    2023

  25. [25]

    Geomgs: Lidar-guided geometry-aware gaussian splatting for robot localization.arXiv preprint arXiv:2501.13417, 2025

    Lee, J., Kong, M., Park, M., and Kim, E. Geomgs: Lidar-guided geometry-aware gaussian splatting for robot localization.arXiv preprint arXiv:2501.13417, 2025

    work page arXiv 2025

  26. [26]

    Ms-gs: Multi-appearance sparse-view 3d gaussian splatting in the wild

    Li, D., Jiang, K., Tang, Y ., Ramamoorthi, R., Chel- lappa, R., and Peng, C. Ms-gs: Multi-appearance sparse-view 3d gaussian splatting in the wild. InThe Thirty-ninth Annual Conference on Neural Informa- tion Processing Systems, 2025

    2025

  27. [27]

    Dngaussian: Optimizing sparse-view 3d gaussian radiance fields with global-local depth normalization

    Li, J., Zhang, J., Bai, X., Zheng, J., Ning, X., Zhou, J., and Gu, L. Dngaussian: Optimizing sparse-view 3d gaussian radiance fields with global-local depth normalization. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pp. 20775–20785, 2024

    2024

  28. [28]

    H., and Tombari, F

    Li, Y ., Lyu, C., Di, Y ., Zhai, G., Lee, G. H., and Tombari, F. Geogaussian: Geometry-aware gaussian splatting for scene rendering. InEuropean conference on computer vision, pp. 441–457. Springer, 2024

    2024

  29. [29]

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

    Lu, T., Yu, M., Xu, L., Xiangli, Y ., Wang, L., Lin, D., and Dai, B. Scaffold-gs: Structured 3d gaussians for view-adaptive rendering. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pp. 20654–20664, 2024

    2024

  30. [30]

    S., Goel, R., Kerbl, B., Steinberger, M., Carrasco, F

    Mallick, S. S., Goel, R., Kerbl, B., Steinberger, M., Carrasco, F. V ., and De La Torre, F. Taming 3dgs: High-quality radiance fields with limited resources. In SIGGRAPH Asia 2024 Conference Papers, pp. 1–11, 2024

    2024

  31. [31]

    Evolsplat: Effi- cient volume-based gaussian splatting for urban view synthesis

    Miao, S., Huang, J., Bai, D., Yan, X., Zhou, H., Wang, Y ., Liu, B., Geiger, A., and Liao, Y . Evolsplat: Effi- cient volume-based gaussian splatting for urban view synthesis. InProceedings of the Computer Vision and Pattern Recognition Conference, pp. 11286–11296, 2025

    2025

  32. [32]

    P., Tancik, M., Barron, J

    Mildenhall, B., Srinivasan, P. P., Tancik, M., Barron, J. T., Ramamoorthi, R., and Ng, R. Nerf: Representing scenes as neural radiance fields for view synthesis. Communications of the ACM, 65(1):99–106, 2021

    2021

  33. [33]

    Lidar-enhanced 3d gaussian splatting mapping.arXiv preprint arXiv:2503.05425, 2025

    Shen, J., Yu, H., Wu, J., Yang, W., and Xia, G.-S. Lidar-enhanced 3d gaussian splatting mapping.arXiv preprint arXiv:2503.05425, 2025

    work page arXiv 2025

  34. [34]

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

    Song, R., Liang, C., Xia, Y ., Zimmer, W., Cao, H., Caesar, H., Festag, A., and Knoll, A. Coda-4dgs: Dy- namic gaussian splatting with context and deformation awareness for autonomous driving. InProceedings of the IEEE/CVF International Conference on Computer Vision, pp. 28031–28041, 2025

    2025

  35. [35]

    Scalability in perception for autonomous driving: Waymo open dataset

    Sun, P., Kretzschmar, H., Dotiwalla, X., Chouard, A., Patnaik, V ., Tsui, P., Guo, J., Zhou, Y ., Chai, Y ., Caine, B., et al. Scalability in perception for autonomous driving: Waymo open dataset. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 2446–2454, 2020

    2020

  36. [36]

    Dn-splatter: Depth and normal priors for gaussian splatting and meshing

    Turkulainen, M., Ren, X., Melekhov, I., Seiskari, O., Rahtu, E., and Kannala, J. Dn-splatter: Depth and normal priors for gaussian splatting and meshing. In 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pp. 2421–2431. IEEE, 2025

    2025

  37. [37]

    V oigtlaender, P., Krause, M., Osep, A., Luiten, J., Sekar, B. B. G., Geiger, A., and Leibe, B. Mots: Multi- object tracking and segmentation. InConference on Computer Vision and Pattern Recognition (CVPR), 2019

    2019

  38. [38]

    Generative ai for autonomous driving: Frontiers and opportunities

    Wang, Y ., Xing, S., Can, C., Li, R., Hua, H., Tian, K., Mo, Z., Gao, X., Wu, K., Zhou, S., et al. Generative ai for autonomous driving: Frontiers and opportunities. arXiv preprint arXiv:2505.08854, 2025

    work page arXiv 2025

  39. [39]

    4d gaussian splatting for real-time dynamic scene rendering

    Wu, G., Yi, T., Fang, J., Xie, L., Zhang, X., Wei, W., Liu, W., Tian, Q., and Wang, X. 4d gaussian splatting for real-time dynamic scene rendering. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 20310–20320, 2024

    2024

  40. [40]

    D2gs: Dense depth regularization for lidar-free urban scene reconstruction.arXiv preprint arXiv:2510.25173, 2025

    Xia, K., Jia, J., Jin, K., Bai, Y ., Sun, L., Tao, D., and Zhang, Y . D2gs: Dense depth regularization for lidar-free urban scene reconstruction.arXiv preprint arXiv:2510.25173, 2025

    work page arXiv 2025

  41. [41]

    Depthsplat: Connecting gaussian splatting and depth

    Xu, H., Peng, S., Wang, F., Blum, H., Barath, D., Geiger, A., and Pollefeys, M. Depthsplat: Connecting gaussian splatting and depth. InProceedings of the 10 EnerGS: Energy-Based Gaussian Splatting with Partial Geometric Priors Computer Vision and Pattern Recognition Conference, pp. 16453–16463, 2025

    2025

  42. [42]

    Ad-gs: Object-aware b-spline gaussian splatting for self-supervised autonomous driving

    Xu, J., Deng, K., Fan, Z., Wang, S., Xie, J., and Yang, J. Ad-gs: Object-aware b-spline gaussian splatting for self-supervised autonomous driving. InProceedings of the IEEE/CVF International Conference on Computer Vision, pp. 24770–24779, 2025

    2025

  43. [43]

    Street gaussians: Modeling dynamic urban scenes with gaussian splat- ting

    Yan, Y ., Lin, H., Zhou, C., Wang, W., Sun, H., Zhan, K., Lang, X., Zhou, X., and Peng, S. Street gaussians: Modeling dynamic urban scenes with gaussian splat- ting. InEuropean Conference on Computer Vision, pp. 156–173. Springer, 2024

    2024

  44. [44]

    Storm: Spatio-temporal reconstruction model for large-scale outdoor scenes.arXiv preprint arXiv:2501.00602, 2024

    Yang, J., Huang, J., Chen, Y ., Wang, Y ., Li, B., You, Y ., Sharma, A., Igl, M., Karkus, P., Xu, D., et al. Storm: Spatio-temporal reconstruction model for large-scale outdoor scenes.arXiv preprint arXiv:2501.00602, 2024

    work page arXiv 2024

  45. [45]

    Hybridgs: High-efficiency gaussian splatting data com- pression using dual-channel sparse representation and point cloud encoder.arXiv preprint arXiv:2505.01938, 2025

    Yang, Q., Yang, L., Van Der Auwera, G., and Li, Z. Hybridgs: High-efficiency gaussian splatting data com- pression using dual-channel sparse representation and point cloud encoder.arXiv preprint arXiv:2505.01938, 2025

    work page arXiv 2025

  46. [46]

    Mip-splatting: Alias-free 3d gaussian splatting

    Yu, Z., Chen, A., Huang, B., Sattler, T., and Geiger, A. Mip-splatting: Alias-free 3d gaussian splatting. InProceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, pp. 19447–19456, 2024

    2024

  47. [47]

    Nerf++: Analyzing and improving neural radiance fields

    Zhang, K., Riegler, G., Snavely, N., and Koltun, V . Nerf++: Analyzing and improving neural radiance fields.arXiv preprint arXiv:2010.07492, 2020

    work page arXiv 2010

  48. [48]

    Gs-lrm: Large reconstruc- tion model for 3d gaussian splatting

    Zhang, K., Bi, S., Tan, H., Xiangli, Y ., Zhao, N., Sunkavalli, K., and Xu, Z. Gs-lrm: Large reconstruc- tion model for 3d gaussian splatting. InEuropean Conference on Computer Vision, pp. 1–19. Springer, 2024

    2024

  49. [49]

    Cdgs: Confidence-aware depth regularization for 3d gaussian splatting.arXiv preprint arXiv:2502.14684, 2025

    Zhang, Q., Wysocki, O., Urban, S., and Jutzi, B. Cdgs: Confidence-aware depth regularization for 3d gaussian splatting.arXiv preprint arXiv:2502.14684, 2025

    work page arXiv 2025

  50. [50]

    and Ni, Z

    Zhou, H. and Ni, Z. Perceptual-gs: Scene-adaptive perceptual densification for gaussian splatting.arXiv preprint arXiv:2506.12400, 2025

    work page arXiv 2025

  51. [51]

    Drivinggaussian: Composite gaussian splatting for surrounding dynamic autonomous driving scenes

    Zhou, X., Lin, Z., Shan, X., Wang, Y ., Sun, D., and Yang, M.-H. Drivinggaussian: Composite gaussian splatting for surrounding dynamic autonomous driving scenes. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 21634– 21643, 2024

    2024

  52. [52]

    blue holes

    Zhou, Z., Xiang, H., Zheng, Z., Zhao, S. Z., Lei, M., Zhang, Y ., Cai, T., Liu, X., Liu, J., Bajji, M., et al. V2xpnp: Vehicle-to-everything spatio-temporal fusion for multi-agent perception and prediction. InProceed- ings of the IEEE/CVF International Conference on Computer Vision, pp. 25399–25409, 2025. 11 EnerGS: Energy-Based Gaussian Splatting with Pa...

    2025