arxiv: 2604.21400 · v2 · submitted 2026-04-23 · 💻 cs.CV

Recognition: unknown

You Only Gaussian Once: Controllable 3D Gaussian Splatting for Ultra-Densely Sampled Scenes

Jinrang Jia , Zhenjia Li , Yifeng Shi

Authors on Pith no claims yet

Pith reviewed 2026-05-09 22:26 UTC · model grok-4.3

classification 💻 cs.CV

keywords 3D Gaussian Splattingneural renderingdeterministic reconstructionultra-dense datasetmulti-sensor fusionbudget controlindoor scene reconstructionproduction rendering

0 comments

The pith

YOGO converts stochastic 3D Gaussian growth into a deterministic budget-controlled process that reaches state-of-the-art quality.

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

The paper seeks to make 3D Gaussian Splatting suitable for production by eliminating unpredictable growth in the number of Gaussians and by removing reliance on sparse test scenes that reward interpolation tricks. It achieves this through a budget controller that caps resource use according to hardware limits and a registration protocol that fuses multi-sensor inputs consistently. The authors also supply an ultra-dense indoor dataset with saturated camera coverage so that success must come from accurate physical modeling rather than viewpoint gaps. A reader would care because existing methods often consume variable memory and compute or produce results that fail when viewed from new positions, which blocks reliable use in robotics or virtual environments.

Core claim

YOGO reformulates the stochastic growth process of 3D Gaussians into a deterministic, budget-aware equilibrium. It integrates a novel budget controller for hardware-constrained resource allocation and an availability-registration protocol for robust multi-sensor fusion. Paired with the Immersion v1.0 ultra-dense indoor dataset that supplies saturated viewpoint coverage, the system delivers state-of-the-art visual quality while remaining strictly deterministic.

What carries the argument

The budget controller and availability-registration protocol, which convert heuristic Gaussian growth into a controllable equilibrium for resource allocation and sensor fusion.

If this is right

Resource consumption becomes fixed and predictable, allowing safe deployment under hardware limits.
Reconstruction quality must derive from physical scene properties rather than sparse-view interpolation.
Multi-sensor inputs fuse without introducing data pollution or inconsistencies.
A reproducible baseline exists for deterministic, high-fidelity 3D Gaussian Splatting in production settings.

Where Pith is reading between the lines

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

Ultra-dense sampling could expose that many reported gains in prior work depend on the sparsity of existing benchmarks.
Fixed-budget control may enable direct integration of 3DGS into real-time pipelines where variable compute is disallowed.
The same controller pattern could be tested on other scene representations to check whether determinism generalizes beyond Gaussians.

Load-bearing premise

The budget controller and registration protocol must preserve or raise reconstruction quality without heuristic growth, and the ultra-dense dataset must compel physical accuracy rather than permit new overfitting.

What would settle it

On Immersion v1.0, if YOGO produces lower visual quality than prior heuristic methods or exceeds its allocated Gaussian budget while quality holds on sparse tests, the central claims are falsified.

Figures

Figures reproduced from arXiv: 2604.21400 by Jinrang Jia, Yifeng Shi, Zhenjia Li.

**Figure 1.** Figure 1: (A) Vanilla 3DGS suffers from uncontrollable growth and OOM risks on our challenging Immersion dataset. (B) YOGO ensures high-fidelity reconstruction under a deterministic budget (e.g., 1.5M points) via robust multi-sensor fusion. (C) Unlike sparse conventional benchmarks (left), Immersion provides ultra-dense saturation (right), breaking the sparsity shield to force true physical fidelity. ⋆ Equal contri… view at source ↗

**Figure 2.** Figure 2: Overview of the YOGO Framework. The pipeline begins with multi-sensor data undergoing Availability-Registration Multi-Sensor Fusion (Sec. 3.2) to filter polluted inputs. Under the deterministic budget controller (Sec. 3.1), the number of Gaussian points at each stage is strictly controlled, which regulates growth based on preset constraints and Polygon regions. The process is enhanced by the Solid Optimi… view at source ↗

**Figure 3.** Figure 3: Characteristics of the Immersion Dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Qalign correlation with PSNR, SSIM, and LPIPS (95% confidence intervals, p<0.001) alongside example renderings sorted by increasing Qalign. Higher Qalign values consistently correspond to improved perceptual quality, demonstrating its reliability as a no-reference 3DGS metric. 5.2 Reliability of the Qalign Metric As evidenced in [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗

**Figure 5.** Figure 5: Qualitative comparison across Immersion v1.0 dataset. Fusion Method S20 Images Point Evalutation Test PSNR SSIM LPIPS Qalign Qalign (a) Single Sensor 0 1.49M 27.73 0.8870 0.2681 3.6839 3.7142 (b) Direct Fusion 9769 1.39M 27.34 0.8855 0.2672 3.6934 3.7215 (c) Random Sampling 3884 1.45M 27.45 0.8859 0.2685 3.6717 3.7249 (d) A.R. with τ=0.1 1995 1.47M 27.69 0.8865 0.2683 3.6801 3.7368 (e) A.R. with τ=0.15 388… view at source ↗

read the original abstract

3D Gaussian Splatting (3DGS) has revolutionized neural rendering, yet existing methods remain predominantly research prototypes ill-suited for production-level deployment. We identify a critical "Industry-Academia Gap" hindering real-world application: unpredictable resource consumption from heuristic Gaussian growth, the "sparsity shield" of current benchmarks that rewards hallucination over physical fidelity, and severe multi-sensor data pollution. To bridge this gap, we propose YOGO (You Only Gaussian Once), a system-level framework that reformulates the stochastic growth process into a deterministic, budget-aware equilibrium. YOGO integrates a novel budget controller for hardware-constrained resource allocation and an availability-registration protocol for robust multi-sensor fusion. To push the boundaries of reconstruction fidelity, we introduce Immersion v1.0, the first ultra-dense indoor dataset specifically designed to break the "sparsity shield." By providing saturated viewpoint coverage, Immersion v1.0 forces algorithms to focus on extreme physical fidelity rather than viewpoint interpolation, and enables the community to focus on the upper limits of high-fidelity reconstruction. Extensive experiments demonstrate that YOGO achieves state-of-the-art visual quality while maintaining a strictly deterministic profile, establishing a new standard for production-grade 3DGS. To facilitate reproducibility, part scenes of Immersion v1.0 dataset and source code of YOGO has been publicly released. The project link is https://jjrcn.github.io/yogo-project-home/

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.

Desk Editor's Note private letter to a colleague

YOGO turns 3DGS growth deterministic via a budget controller and adds an ultra-dense dataset, but the fidelity claims hinge on experiments that isolate the controller from the dense views.

read the letter

The core move here is replacing the usual stochastic densification in 3D Gaussian Splatting with a budget controller that enforces a fixed resource envelope and an availability-registration step for cleaning up multi-sensor inputs. They pair that with Immersion v1.0, a new indoor dataset that saturates viewpoints so methods cannot rely on interpolation tricks. Those two pieces are what the paper actually contributes beyond prior 3DGS work. The practical framing around unpredictable compute and the sparsity shield in existing benchmarks is useful; it names deployment problems that matter for robotics or AR pipelines. The dataset release is also a concrete step that could let others test upper limits on fidelity. The soft spot is exactly the one the stress-test flags. The abstract states SOTA visual quality under strict determinism, yet the argument rests on whether the controller reaches an equilibrium that matches the expressive range of heuristic splitting without the dense sampling doing most of the work. If the experiments do not show ablations that hold total Gaussian count fixed while varying the controller, or compare against baselines on the same ultra-dense data, the production-grade claim stays unproven. The dataset itself could also reward fitting to capture-specific noise rather than true geometry, and that risk is not obviously ruled out. This is for people who build controllable rendering systems and need predictable memory and compute. It is grounded enough in a real problem and has enough new machinery that it deserves a serious referee to check the isolation of the controller and the dataset results. I would send it to review.

Referee Report

3 major / 3 minor

Summary. The paper proposes YOGO, a deterministic reformulation of 3D Gaussian Splatting that replaces heuristic Gaussian growth with a budget controller for resource allocation and an availability-registration protocol for multi-sensor fusion. It introduces Immersion v1.0, an ultra-dense indoor dataset designed to enforce physical fidelity by providing saturated viewpoint coverage. The central claim is that this system achieves state-of-the-art visual quality while remaining strictly deterministic and suitable for production deployment, with partial dataset and code released for reproducibility.

Significance. If the claims are substantiated, the work could help close the industry-academia gap in 3DGS by enabling predictable resource use and robust handling of multi-sensor data. The ultra-dense dataset may encourage the field to prioritize high-fidelity reconstruction over interpolation, and the public release of code and data supports reproducibility.

major comments (3)

[§3.1] §3.1 (Budget Controller): The description states that the controller reaches a deterministic equilibrium matching the expressive power of prior stochastic densification, but no equations, convergence analysis, or proof sketch is provided to show how the fixed budget avoids under- or over-expression; this is load-bearing for the strict-determinism claim.
[§5.2] §5.2 and Table 3 (Ablation on Immersion v1.0): The reported PSNR/SSIM gains for YOGO are presented without an ablation that disables the availability-registration protocol while keeping the budget controller; it is therefore unclear whether fidelity improvements stem from the new components or simply from the saturated coverage of the new dataset.
[§4.3] §4.3 (Multi-sensor fusion): The availability-registration protocol is claimed to mitigate data pollution, yet no quantitative metric (e.g., cross-sensor consistency error or hallucination rate) is reported on the multi-sensor subsets of Immersion v1.0; this directly affects the production-grade robustness claim.

minor comments (3)

[Abstract] The abstract and §1 repeatedly use the phrase 'strictly deterministic profile' without defining the precise scope (e.g., whether randomness in optimization or initialization is still permitted).
[Figure 4] Figure 4 caption refers to 'qualitative results' but the figure itself lacks scale bars or viewpoint labels, making direct visual comparison to baselines difficult.
[§2] The related-work section cites several 3DGS variants but omits discussion of recent controllable or memory-bounded variants (e.g., those using explicit pruning schedules).

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments that help clarify the contributions and strengthen the determinism and robustness claims. We address each major point below and commit to the necessary revisions.

read point-by-point responses

Referee: [§3.1] §3.1 (Budget Controller): The description states that the controller reaches a deterministic equilibrium matching the expressive power of prior stochastic densification, but no equations, convergence analysis, or proof sketch is provided to show how the fixed budget avoids under- or over-expression; this is load-bearing for the strict-determinism claim.

Authors: We agree that a formal treatment is required to support the strict-determinism claim. In the revised manuscript we will insert the governing equations of the budget controller, a convergence analysis establishing that the fixed-budget iteration reaches a unique equilibrium, and a short proof sketch demonstrating that this equilibrium matches the expressive capacity of prior stochastic densification while preventing both under- and over-expression. Supporting empirical traces will also be added. revision: yes
Referee: [§5.2] §5.2 and Table 3 (Ablation on Immersion v1.0): The reported PSNR/SSIM gains for YOGO are presented without an ablation that disables the availability-registration protocol while keeping the budget controller; it is therefore unclear whether fidelity improvements stem from the new components or simply from the saturated coverage of the new dataset.

Authors: The referee is correct that the current ablation does not isolate the contribution of the availability-registration protocol. We will add a new ablation row (and corresponding text) that disables only the registration protocol while retaining the budget controller, thereby clarifying whether the observed gains arise from the protocol, the budget controller, or the dataset density itself. revision: yes
Referee: [§4.3] §4.3 (Multi-sensor fusion): The availability-registration protocol is claimed to mitigate data pollution, yet no quantitative metric (e.g., cross-sensor consistency error or hallucination rate) is reported on the multi-sensor subsets of Immersion v1.0; this directly affects the production-grade robustness claim.

Authors: We acknowledge the absence of quantitative support for the data-pollution mitigation claim. In the revision we will report cross-sensor consistency error and hallucination rate on the multi-sensor subsets of Immersion v1.0, providing direct numerical evidence for the protocol’s effectiveness. revision: yes

Circularity Check

0 steps flagged

No significant circularity; framework and claims rest on new components and experiments rather than self-referential reductions.

full rationale

The paper introduces YOGO as a novel system-level reformulation with a budget controller and availability-registration protocol, plus the new Immersion v1.0 dataset. No equations, derivations, or load-bearing predictions appear in the abstract or described structure that reduce by construction to fitted inputs or prior self-citations. The central claims of SOTA quality and determinism are presented as outcomes of these independent components and extensive experiments, without visible self-definitional loops or renaming of known results. This matches the default expectation of a non-circular paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no technical sections, equations, or implementation details, preventing identification of specific free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5567 in / 1172 out tokens · 26392 ms · 2026-05-09T22:26:37.104345+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

42 extracted references · 19 canonical work pages

[1]

In: 2016 IEEE Inter- national Conference on Automatic Control and Intelligent Systems (I2CACIS)

Ann, N.Q., Achmad, M.H., Bayuaji, L., Daud, M.R., Pebrianti, D.: Study on 3d scene reconstruction in robot navigation using stereo vision. In: 2016 IEEE Inter- national Conference on Automatic Control and Intelligent Systems (I2CACIS). pp. 72–77 (2016).https://doi.org/10.1109/I2CACIS.2016.7885292

work page doi:10.1109/i2cacis.2016.7885292 2016
[2]

CVPR (2022) YOGO 15

Barron, J.T., Mildenhall, B., Verbin, D., Srinivasan, P.P., Hedman, P.: Mip-nerf 360: Unbounded anti-aliased neural radiance fields. CVPR (2022) YOGO 15

2022
[3]

Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction.arxiv preprint arxiv:2406.06521, 2024

Chen, D., Li, H., Ye, W., Wang, Y., Xie, W., Zhai, S., Wang, N., Liu, H., Bao, H., Zhang, G.: Pgsr: Planar-based gaussian splatting for efficient and high-fidelity surface reconstruction. arXiv preprint arXiv:2406.06521 (2024)

work page arXiv 2024
[4]

Chen, K., Mao, T., Ma, Z., Jiang, H., Li, Z., Liu, Z., Gao, S., Zhao, H., Dai, F., Zhang, Y., Wang, Z.: Metrogs: Efficient and stable reconstruction of geometrically accurate high-fidelity large-scale scenes (2025),https://arxiv.org/abs/2511. 19172

2025
[5]

Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images.arXiv preprint arXiv:2403.14627, 2024

Chen, Y., Xu, H., Zheng, C., Zhuang, B., Pollefeys, M., Geiger, A., Cham, T.J., Cai, J.: Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images. arXiv preprint arXiv:2403.14627 (2024)

work page arXiv 2024
[6]

arXiv preprint arXiv:2311.13398 (2023)

Chung, J., Oh, J., Lee, K.M.: Depth-regularized optimization for 3d gaussian splat- ting in few-shot images. arXiv preprint arXiv:2311.13398 (2023)

work page arXiv 2023
[7]

In: Proc

Dai,A.,Chang,A.X.,Savva,M.,Halber,M.,Funkhouser,T.,Nießner,M.:Scannet: Richly-annotated 3d reconstructions of indoor scenes. In: Proc. Computer Vision and Pattern Recognition (CVPR), IEEE (2017)

2017
[8]

Fang,G.,Wang,B.:Mini-splatting:Representingsceneswithaconstrainednumber of gaussians (2024),https://arxiv.org/abs/2403.14166

work page arXiv 2024
[9]

CVPR (2024)

Guédon, A., Lepetit, V.: Sugar: Surface-aligned gaussian splatting for efficient 3d mesh reconstruction and high-quality mesh rendering. CVPR (2024)

2024
[10]

2d gaussian splatting for geometrically accurate radiance fields

Huang,B.,Yu,Z.,Chen,A.,Geiger,A.,Gao,S.:2dgaussiansplattingforgeometri- cally accurate radiance fields. In: SIGGRAPH 2024 Conference Papers. Association for Computing Machinery (2024).https://doi.org/10.1145/3641519.3657428

work page doi:10.1145/3641519.3657428 2024
[11]

Huynh, A., Silva, J.M., Caesar, H., Son, T.D.: Material-informed gaussian splatting for 3d world reconstruction in a digital twin (2026),https://arxiv.org/abs/ 2511.20348

work page arXiv 2026
[12]

In: Thirty-seventh Con- ference on Neural Information Processing Systems (2023),https://openreview

Jia,J.,Li,Z.,Shi,Y.:Monouni:Aunifiedvehicleandinfrastructure-sidemonocular 3d object detection network with sufficient depth clues. In: Thirty-seventh Con- ference on Neural Information Processing Systems (2023),https://openreview. net/forum?id=v2oGdhbKxi

2023
[13]

ACM Transactions on Graphics42(4) (July 2023),https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/

Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics42(4) (July 2023),https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/

2023
[14]

Kerbl, B., Meuleman, A., Kopanas, G., Wimmer, M., Lanvin, A., Drettakis, G.: A hierarchical 3d gaussian representation for real-time rendering of very large datasets (2024),https://arxiv.org/abs/2406.12080

work page arXiv 2024
[15]

ACM Transactions on Graphics36(4) (2017)

Knapitsch, A., Park, J., Zhou, Q.Y., Koltun, V.: Tanks and temples: Benchmarking large-scale scene reconstruction. ACM Transactions on Graphics36(4) (2017)

2017
[16]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Li, J., Zhang, J., Bai, X., Zheng, J., Ning, X., Zhou, J., Gu, L.: Dngaussian: Op- timizing sparse-view 3d gaussian radiance fields with global-local depth normal- ization. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 20775–20785 (June 2024)

2024
[17]

Li, Y., Jiang, L., Xu, L., Xiangli, Y., Wang, Z., Lin, D., Dai, B.: Matrixcity: A large-scale city dataset for city-scale neural rendering and beyond (2023),https: //arxiv.org/abs/2309.16553

work page arXiv 2023
[18]

Li, Z., Jia, J., Shi, Y.: Monolss: Learnable sample selection for monocular 3d de- tection (2023)

2023
[19]

In: ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPatternRecog- nition (CVPR)

Lin, J., Li, Z., Tang, X., Liu, J., Liu, S., Liu, J., Lu, Y., Wu, X., Xu, S., Yan, Y., Yang, W.: Vastgaussian: Vast 3d gaussians for large scene reconstruction. In: ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPatternRecog- nition (CVPR). pp. 5166–5175 (June 2024) 16 J. Jia et al

2024
[20]

In: Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part VIII

Lin,L.,Liu,Y.,Hu,Y.,Yan,X.,Xie,K.,Huang,H.:Capturing,reconstructing,and simulating: The urbanscene3d dataset. In: Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part VIII. p. 93–109. Springer-Verlag, Berlin, Heidelberg (2022).https://doi.org/10. 1007/978-3-031-20074-8_6,https://doi.org/10.1007/97...

work page doi:10.1007/978-3-031-20074-8_6 2022
[21]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Ling, L., Sheng, Y., Tu, Z., Zhao, W., Xin, C., Wan, K., Yu, L., Guo, Q., Yu, Z., Lu, Y., et al.: Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 22160–22169 (2024)

2024
[22]

In: Proceedings of the IEEE/CVF International Conference on Com- puter Vision (ICCV)

Liu, S., Tang, X., Li, Z., He, Y., Ye, C., Liu, J., Huang, B., Zhou, S., Wu, X.: Oc- clugaussian: Occlusion-aware gaussian splatting for large scene reconstruction and rendering. In: Proceedings of the IEEE/CVF International Conference on Com- puter Vision (ICCV). pp. 26643–26652 (October 2025)

2025
[23]

Liu, Z., Li, Z., Shi, Y., Li, X.: Attentiongs: Towards initialization-free 3d gaussian splatting via structural attention (2025),https://arxiv.org/abs/2506.23611

work page arXiv 2025
[24]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Lu, T., Yu, M., Xu, L., Xiangli, Y., Wang, L., Lin, D., Dai, B.: Scaffold-gs: Struc- tured 3d gaussians for view-adaptive rendering. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 20654–20664 (2024)

2024
[25]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

Lyu, Y., Cheng, K., Kang, X., Chen, X.: Resgs: Residual densification of 3d gaus- sian for efficient detail recovery. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 28093–28102 (October 2025)

2025
[26]

Mildenhall, B., Srinivasan, P.P., Tancik, M., Barron, J.T., Ramamoorthi, R., Ng, R.: Nerf: Representing scenes as neural radiance fields for view synthesis (2020), https://arxiv.org/abs/2003.08934

work page arXiv 2020
[27]

URL https://doi.org/10.1145/3528223

Müller, T., Evans, A., Schied, C., Keller, A.: Instant neural graphics primitives with a multiresolution hash encoding. ACM Trans. Graph.41(4), 102:1–102:15 (Jul 2022).https://doi.org/10.1145/3528223.3530127,https://doi.org/10. 1145/3528223.3530127

work page doi:10.1145/3528223.3530127 2022
[28]

Niedermayr, S., Stumpfegger, J., Westermann, R.: Compressed 3d gaussian splat- ting for accelerated novel view synthesis (2023)

2023
[29]

Octree-gs: Towards consistent real-time rendering with lod-structured 3d gaussians.arXiv preprint arXiv:2403.17898, 2024

Ren, K., Jiang, L., Lu, T., Yu, M., Xu, L., Ni, Z., Dai, B.: Octree-gs: Towards consistent real-time rendering with lod-structured 3d gaussians. arXiv preprint arXiv:2403.17898 (2024)

work page arXiv 2024
[30]

Spotlesssplats: Ignoring distractors in 3d gaussian splatting.arXiv preprint arXiv:2406.20055, 2024

Sabour, S., Goli, L., Kopanas, G., Matthews, M., Lagun, D., Guibas, L., Jacobson, A.,Fleet,D.J.,Tagliasacchi,A.:SpotLessSplats:Ignoringdistractorsin3dgaussian splatting. arXiv:2406.20055 (2024)

work page arXiv 2024
[31]

arXiv preprint arXiv:2312.17090 (2023)

Wu, H., Zhang, Z., Zhang, W., Chen, C., Li, C., Liao, L., Wang, A., Zhang, E., Sun, W., Yan, Q., Min, X., Zhai, G., Lin, W.: Q-align: Teaching lmms for visual scoring via discrete text-defined levels. arXiv preprint arXiv:2312.17090 (2023), equal Contribution by Wu, Haoning and Zhang, Zicheng. Project Lead by Wu, Haoning. Corresponding Authors: Zhai, Guan...

work page arXiv 2023
[32]

In: 2025 IEEE International Conference on Robotics and Automation (ICRA)

Xiang, H., Li, X., Cheng, K., Lai, X., Zhang, W., Liao, Z., Zeng, L., Liu, X.: Gaussianroom: Improving 3d gaussian splatting with sdf guidance and monocular cues for indoor scene reconstruction. In: 2025 IEEE International Conference on Robotics and Automation (ICRA). pp. 2686–2693. IEEE (2025)

2025
[33]

Xu, H., Zhang, S., Li, P., Ye, B., Chen, X., ang Gao, H., Zheng, J., Song, X., Peng, Z., Miao, R., Jia, J., Shi, Y., Yi, G., Zhao, H., Tang, H., Li, H., Yu, K., Zhao, H.: Cruise: Cooperative reconstruction and editing in v2x scenarios using gaussian splatting (2025),https://arxiv.org/abs/2507.18473 YOGO 17

work page arXiv 2025
[34]

In: 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Xu, H., Zhang, S., Li, P., Ye, B., Chen, X., Gao, H.A., Zheng, J., Song, X., Peng, Z., Miao, R., Jia, J., Shi, Y., Yi, G., Zhao, H., Tang, H., Li, H., Yu, K., Zhao, H.: Cruise: Cooperative reconstruction and editing in v2x scenarios using gaussian splatting. In: 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). pp. 12518–125...

work page doi:10.1109/iros60139 2025
[35]

In: ECCV (2024)

Yan, Y., Lin, H., Zhou, C., Wang, W., Sun, H., Zhan, K., Lang, X., Zhou, X., Peng, S.: Street gaussians: Modeling dynamic urban scenes with gaussian splatting. In: ECCV (2024)

2024
[36]

Ye, Z., Li, W., Liu, S., Qiao, P., Dou, Y.: Absgs: Recovering fine details for 3d gaussian splatting (2024)

2024
[37]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision

Yeshwanth, C., Liu, Y.C., Nießner, M., Dai, A.: Scannet++: A high-fidelity dataset of 3d indoor scenes. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 12–22 (2023)

2023
[38]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Yu, Z., Chen, A., Huang, B., Sattler, T., Geiger, A.: Mip-splatting: Alias-free 3d gaussian splatting. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 19447–19456 (June 2024)

2024
[39]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

Yuan, Z., Huang, H., Xiong, Z., Wang, D., Yang, G.: Robust and efficient 3d gaus- sian splatting for urban scene reconstruction. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 26209–26219 (October 2025)

2025
[40]

In: CVPR (2018)

Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The unreasonable effectiveness of deep features as a perceptual metric. In: CVPR (2018)

2018
[41]

In: ECCV (2024)

Zhang, Z., Hu, W., Lao, Y., He, T., Zhao, H.: Pixel-gs: Density control with pixel- aware gradient for 3d gaussian splatting. In: ECCV (2024)

2024
[42]

Zhou, H., Ni, Z.: Perceptual-gs: Scene-adaptive perceptual densification for gaus- sian splatting (2025),https://arxiv.org/abs/2506.12400

work page arXiv 2025