Pith. sign in

REVIEW 3 major objections 5 minor 36 references

Frame caching and tile splitting make 3DGS streaming up to 10× faster

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 16:47 UTC pith:CBGWGUOJ

load-bearing objection Solid engineering with a real coverage gap in the speculative cache; worth refereeing but needs code, data, and a corrected proof or a caveat. the 3 major comments →

arxiv 2607.17842 v1 pith:CBGWGUOJ submitted 2026-07-20 cs.CV cs.AI

CaT-GS: Efficient 3DGS Rendering for Large Scale Scenes via Inter-frame Caching and Tile Scheduling

Tingjia Zhang , Bo Chen , Shengzhong Liu , Fan Wu , Guihai Chen This is my paper
classification cs.CV cs.AI
keywords 3D Gaussian Splattingreal-time renderinginter-frame redundancyspeculative pre-processingtile load balancingGPU rasterizationlarge-scale scenesneural rendering
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that the standard 3D Gaussian Splatting (3DGS) rendering pipeline wastes most of its time on nearly identical work across consecutive frames: frustum culling, sorting, and tile assignment barely change when the camera moves slightly. It proposes CaT-GS, a pipeline that speculatively preprocesses a small batch of frames at once, caches the culled and sorted Gaussian list across frames, and splits overloaded tiles into smaller tasks so GPU units stay busy. The result, on large city-scale scenes, is up to 10× faster rendering than the original pipeline and up to 70% faster than previous acceleration methods, with visual quality nearly unchanged. If true, this makes real-time interactive rendering of large 3DGS scenes practical on a single GPU without retraining.

Core claim

The paper's central claim is that inter-frame redundancy—not just intra-frame redundancy—is a dominant cost in streaming 3DGS, and it can be exploited without retraining. The authors identify three bottlenecks: redundant pre-processing across consecutive frames, viewpoint-redundant rendering stages, and severe tile-level load imbalance. Their speculative multi-frame pre-processing predicts camera motion, computes the swept 'Gaussian trail' of each Gaussian across the frame batch, and generates a render list guaranteed to cover all needed Gaussians; inter-frame caching then lets sub-frames skip frustum culling and sorting entirely. A refactored rasterization kernel splits any tile whose Gauss

What carries the argument

The Gaussian trail: the region swept by a Gaussian's projected ellipse as the camera moves by a predicted (Δu, Δv). The method expands each Gaussian's bounding window by the motion vector, tests tile intersection against the trail boundary (two half-ellipses and two tangent lines), and thereby produces a render list that remains valid for an entire frame batch. Two supporting mechanisms carry the argument: inter-frame caching (storing hash indices of culled Gaussians and the sorted render list from the key frame, so sub-frames skip culling and sorting) and load-aware task splitting (a batched alpha-blending reformulation C = Σ (slice color)·(residual transmittance) that lets a heavy tile be

Load-bearing premise

The design assumes that consecutive frames in the target deployment are view-coherent enough that a 4-frame speculation window always covers the Gaussians actually needed; if camera jumps, occlusion flips, or very close fast motion occur, the speculation is skipped and the speedup disappears for those frames.

What would settle it

Record a camera trace with deliberate abrupt jumps or fast close-up lateral movements and measure average FPS: if the motion-adaptive scheduler repeatedly shrinks the speculation window to zero or one frame, CaT-GS's advantage over a comparable intra-frame-optimized pipeline should collapse toward the no-caching variant (roughly 10–20% instead of 50–80%). Alternatively, render a scene with large depth discontinuities and compare sub-frame outputs against ground truth; depth-order flips inside the cached sorted list would produce visible artifacts.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • On large scenes with 5–8 million Gaussians, CaT-GS sustains over 200 FPS at 1080p, where the vanilla pipeline runs at 23–54 FPS.
  • The speedup grows with scene scale: improvements over previous state-of-the-art software optimization rise from about 50% on standard scenes to 60–80% on UAV city-scale models, because pre-processing and sorting overhead scale with model size.
  • Rendering quality of sub-frames stays within about 0.04 dB PSNR and 0.005 SSIM of the key-frame pipeline, making the caching visually indistinguishable in practice.
  • The method composes with model pruning: on pruned large scenes, CaT-GS reaches roughly 448–746 FPS, keeping a large margin over baselines.
  • No retraining is needed, so the speedup applies to already-trained 3DGS models, which is what deployment requires.

Where Pith is reading between the lines

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

  • The measured speedup is tied to the 120 FPS, viewer-style camera traces used for evaluation; at lower frame rates or with erratic camera paths, inter-frame redundancy shrinks and the caching benefit should diminish, so the method deserves re-benchmarking across a range of frame rates.
  • The batched alpha-blending split suggests a general recipe for parallelizing ordered per-pixel work: split the ordered list, compute partial sums, and recombine with transmittance residuals—an identity that could accelerate other splatting or volumetric compositing pipelines, not just 3DGS.
  • A fully adaptive speculation window, chosen per frame from the motion prediction rather than the current fixed initial window of 4, could push the speed-versus-robustness tradeoff further; the paper itself flags this as future work.
  • Sub-frame quality loss is likely concentrated where Gaussians move fastest relative to the camera; a testable extension would be to correlate sub-frame error maps with per-Gaussian motion magnitude to validate the motion-thresholding design.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes CaT-GS, a 3D Gaussian Splatting rendering pipeline that targets interactive large-scale scenes. It processes frames in groups: a key frame performs speculative multi-frame pre-processing that predicts camera motion and builds a cached render list, and subsequent sub-frames reuse this list to skip frustum culling, sorting, and tile identification. To address tile load imbalance, the paper also introduces a load-aware task-splitting strategy and a modified CUDA rasterization kernel. Experiments on standard datasets and a self-collected UAV dataset report up to 10× speedup over vanilla 3DGS and up to ~70% over Flash-GS/ADR-GS, with approximately equal PSNR/SSIM.

Significance. The paper addresses a practically important bottleneck: the full rendering pipeline of 3DGS in interactive, high-frame-rate settings, rather than only the rasterization stage. The task-splitting formulation for alpha blending is an interesting contribution, and the ablations isolate the effects of inter-frame caching and tile splitting. The self-collected UAV dataset and the use of recorded SIBR viewer traces are valuable for reproducibility, assuming the dataset is released. However, the central correctness claim—that the speculative cached list always contains all Gaussians needed by the sub-frames—is not rigorously established, and the quality evaluation protocol may not detect violations of this claim. The speedup numbers are internally consistent, but the validity of the quality-equivalence claim is the main risk.

major comments (3)
  1. [§3.2.2, Eqs. (5)–(6)] The trail intersection test is claimed to ensure that the cached render list contains all Gaussians needed for the sub-frames, but the acceptance criterion is not conservative. A tile lying wholly inside one of the half-ellipse caps (and thus inside the trail) but whose center is outside the small central rectangle, and that does not touch the boundary, will be rejected. More fundamentally, the trail itself assumes a constant 2D covariance and a single pixel displacement (Δu,Δv) for the whole Gaussian; under rotation and perspective the projected ellipse changes and the screen-space displacement varies across the footprint, so the Minkowski sum of the key-frame ellipse and a line segment need not contain the true swept region. Without a proof of conservatism or a systematic coverage check, the central "no missing Gaussians" guarantee is unsubstantiated; the supplementary Fig. 7 is one ex
  2. [§3.2.3] The motion-adaptive adjustment checks only Gaussians with depth larger than the threshold d (=0.4) for excessive motion. Closer Gaussians can move by arbitrarily large pixel amounts under the camera motions considered, and these are exactly the Gaussians that can produce large missing-splat regions if they are incorrectly excluded from the cached list. The paper provides no argument that ignoring Gaussians closer than d preserves the coverage guarantee; this is a direct gap in the correctness argument.
  3. [§5.3, Table 2] The 120 FPS ground-truth frames are obtained by interpolating 30 FPS video for the Tanks&Temples and UAV datasets. This protocol cannot validate that sub-frames are artifact-free: temporal interpolation smooths over exactly the missing-splat and flicker artifacts that a non-conservative cache would produce. The authors should compare Ours-Sub against a full per-frame render (e.g., 3DGS or Flash-GS) on the actual trace frames, reporting max or percentile error or counts of pixels with missing Gaussian contributions, and ideally collect real high-FPS ground truth for at least a subset of traces.
minor comments (5)
  1. [Tables 1 and 3] The variant name is inconsistent: Table 1 uses "Ours-w/o" while Table 3 uses "Ours-w/cache" for the same no-inter-frame-caching condition. Use one name throughout.
  2. [Eqs. (7)–(8)] T_j is not defined in the split alpha-blending formulation. Specify that T_j is the intra-segment transmittance and clarify how it relates to the product in Eq. (2).
  3. [Figure 5] The tangent points (x_t, y_t) used to define boundary lines C and D are not derived. Please provide the derivation or a reference.
  4. [Eq. (3)] "inner parameter matrix" should be "intrinsic parameter matrix." Also, the mapping in Eq. (3) is projective, not affine; the affine approximation should be stated explicitly together with its validity conditions.
  5. [Supplementary Fig. 7] The columns "w/o-spec" and "w-spec" are not explained in the caption or text; define these terms.

Circularity Check

0 steps flagged

No significant circularity; speedups are benchmarked against external baselines and the central caching/speculation design is an approximation, not a self-referential derivation. One minor non-load-bearing self-citation (Ref. [33]).

full rationale

CaT-GS's headline speedups (Table 1) are empirical FPS measurements against external baselines (vanilla 3DGS, ADR-GS, Flash-GS) on public datasets plus a self-collected UAV dataset; no speedup or quality metric is the input to the computation that is then reported as a prediction. The speculative trail construction (Section 3.2.2) is an approximation: it treats the projected ellipse shape as constant and moves it along a single (Δu,Δv) vector. That may be non-conservative, but if it fails the result is missing-splat artifacts, not a definitional identity between input and output. The alpha-blending split in Section 3.4 is a restructuring of the standard blending sum, and the load-balancing and caching ablations are measured, not assumed. The only self-citation found is Ref. [33] (co-author Bo Chen), cited as a domain pointer for immersive streaming; it is not load-bearing for the central derivation. The interpolated-120FPS quality protocol is a potential validity weakness but not circularity. Hence no circular step can be quoted.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 2 invented entities

The central method is an engineering pipeline, so the 'axiom load' is mostly domain assumptions about motion coherence, shape stability, sort order stability, and a few tuning knobs (d, W_initial, F_target, k). The paper does not derive parameters from first principles; it tunes them on its own traces. The core geometric/trail derivations are not fully shown, making the mathematical burden larger than the text admits.

free parameters (4)
  • d (speculation depth threshold) = 0.4
    Threshold for when Gaussians deeper than d are eligible for motion speculation (§3.2.3, stated in §4 implementation). No sensitivity analysis for d is given.
  • W_initial (initial speculation window) = 4 frames
    Initial guess of frames to cache from one key frame (§3.2.3, §4). Supplementary Table 6 shows speedup changes with W; 4 is justified only by that ablation.
  • F_target (target frame rate) = 120 FPS
    Notional target used in motion-adaptive scheduling to assess viewpoint-similarity (§3.2.3, §4). It aligns with the 120 FPS trace capture and can be seen as a design setting.
  • k (tile-splitting Gaussian-count threshold) = not stated numerically (implicitly 2l per §3.4, but Algorithm 1 uses k without a formula)
    Decision threshold for splitting a tile's task; the paper says 'more than 2l Gaussians' in §3.4, but Algorithm 1 leaves k abstract. Whichever constant is used, it is a hand-picked tuning parameter.
axioms (5)
  • domain assumption Within a short time interval camera motion can be approximated as constant and the affine map from Eq. (3) captures the Gaussian pixel displacement across the whole speculation window.
    Section 3.2.1: 'Within this short time interval, the motion can be considered nearly constant and unique.' This is the load-bearing premise for speculative multi-frame preprocessing.
  • domain assumption The projected 2D Gaussian shape remains approximately constant along the motion trail (only its center moves).
    Section 3.2.1: 'Although the projected shape of the Gaussian changes slightly due to rotation, we treat it as constant since its variation is negligible.' This premise underpins the ellipse-boundary trail intersection tests.
  • domain assumption The depth-sorted order of Gaussians per tile remains mostly unchanged across a 4-frame speculation window.
    Section 3.3.2: 'depth order of Gaussians remains mostly consistent across frames within a short time.' Sort caching depends on this.
  • standard math The split alpha-blending factorization Eq. (7)/(8) is mathematically valid given per-slice accumulated alpha.
    The factorization C = sum_i (partial products) with R_i = prod(1-A_m) is algebraically standard for alpha compositing; no separate derivation is shown, but cross-checking the equations shows a correct associativity argument.
  • domain assumption The 30 FPS ground-truth video can be interpolated to 120 FPS without changing which frame corresponds to key vs. sub-frame quality evaluation.
    Section 5.3: they interpolate Truck/Train and UAV video to 120 FPS for frame-accurate quality evaluation. The interpolation scheme is not described.
invented entities (2)
  • Gaussian trail no independent evidence
    purpose: Swept region of a Gaussian center along the predicted (Δu, Δv) motion; used to expand the tile-intersection test so that one key-frame render list covers all Gaussians for the batch.
    The trail is an internal computational abstraction, not a physical/measurable entity. Its correctness is evidenced only by the paper's own quality tables and ablation.
  • Inter-frame cached render list (frustum/sort cache) no independent evidence
    purpose: Reused pre-processing artifacts (visible Gaussian set, sorted per-tile list) shared by key and sub-frames.
    Internal caching of existing pipeline data; no independent handle outside the paper.

pith-pipeline@v1.3.0-alltime-deepseek · 15507 in / 9257 out tokens · 66804 ms · 2026-08-01T16:47:19.713014+00:00 · methodology

0 comments
read the original abstract

Recent breakthroughs in 3D Gaussian Splatting (3DGS) have advanced neural rendering with high fidelity and speed. However, its performance degrades significantly in large-scale scenes due to the computational burden of tile-based rasterization. Existing optimization efforts either require costly scene re-training or focus on narrow aspects of the pipeline, overlooking critical inefficiencies in real-world deployments. Through a comprehensive analysis, we identify three primary sources of redundancy and low GPU utilization: redundant inter-frame pre-processing, viewpoint-based occlusion redundancy, and severe tile-level load imbalance. To address these issues, we propose CaT-GS, a novel and efficient 3DGS rendering pipeline. CaT-GS introduces a speculative multi-frame preprocessing method to eliminate redundant computations across consecutive frames, and an inter-frame caching mechanism to eliminate viewpoint redundant rendering stages. Furthermore, it refactors rasterization tasks with a dedicated kernel to mitigate tile load imbalance, significantly boosting GPU utilization. Extensive experiments demonstrate that CaT-GS achieves a speedup of up to 10 times over the original 3DGS and up to 70% over previous state-of-the-art methods, establishing a new benchmark for high-fidelity, real-time rendering of large-scale scenes.

Figures

Figures reproduced from arXiv: 2607.17842 by Bo Chen, Fan Wu, Guihai Chen, Shengzhong Liu, Tingjia Zhang.

Figure 1
Figure 1. Figure 1: General 3DGS Rendering Pipeline, composed by pre [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The ratio of rendering time spent on pre-processing, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Overview of CaT-GS. 3.2.1. Motion Prediction To extend the validity of pre-processing results across mul￾tiple adjacent frames, we begin by analyzing how frame contents change across frames, as this is determined by the input viewpoint’s movement. We first establish a model to quantify the effect of viewpoint movement. We define three types of viewpoint motions that can occur in any instan￾taneous movement… view at source ↗
Figure 5
Figure 5. Figure 5: Steps to apply fine-grained intersection detection to a [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Task redistribution and efficient kernel improve raster [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The speculative process includes the needed Gaussian in [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

36 extracted references · 7 linked inside Pith

  1. [1]

    Compression in 3d gaussian splatting: A survey of methods, trends, and future directions.arXiv preprint arXiv:2502.19457, 2025

    Muhammad Salman Ali, Chaoning Zhang, Marco Cagnazzo, Giuseppe Valenzise, Enzo Tartaglione, and Sung-Ho Bae. Compression in 3d gaussian splatting: A survey of methods, trends, and future directions.arXiv preprint arXiv:2502.19457, 2025. 8

    Pith/arXiv arXiv 2025

  2. [2]

    Mip-nerf 360: Unbounded anti-aliased neural radiance fields

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5470–5479, 2022. 3, 6

    2022

  3. [3]

    You only render once: En- hancing energy and computation efficiency of mobile virtual reality

    Xingyu Chen, Xinmin Fang, Shuting Zhang, Xinyu Zhang, Liang He, and Zhengxiong Li. You only render once: En- hancing energy and computation efficiency of mobile virtual reality. InProceedings of the 23rd Annual International Con- ference on Mobile Systems, Applications and Services, pages 263–276, 2025. 8

    2025

  4. [4]

    3dgv: 3d gaussian splatting-based holographic video streaming over wireless networks

    Zihao Chen, Longhao Zou, and Xiaofeng Tao. 3dgv: 3d gaussian splatting-based holographic video streaming over wireless networks. In2025 IEEE International Sympo- sium on Broadband Multimedia Systems and Broadcasting (BMSB), pages 1–6. IEEE, 2025. 1

    2025

  5. [5]

    Flashgs: Efficient 3d gaussian splatting for large-scale and high-resolution rendering

    Guofeng Feng, Siyan Chen, Rong Fu, Zimu Liao, Yi Wang, Tao Liu, Boni Hu, Linning Xu, Zhilin Pei, Hengjie Li, et al. Flashgs: Efficient 3d gaussian splatting for large-scale and high-resolution rendering. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 26652– 26662, 2025. 1, 3, 7, 8

    2025

  6. [6]

    Lu- mina: Real-time neural rendering by exploiting computa- tional redundancy

    Yu Feng, Weikai Lin, Yuge Cheng, Zihan Liu, Jingwen Leng, Minyi Guo, Chen Chen, Shixuan Sun, and Yuhao Zhu. Lu- mina: Real-time neural rendering by exploiting computa- tional redundancy. InProceedings of the 52nd Annual Inter- national Symposium on Computer Architecture, pages 1925– 1939, 2025. 8

    1925

  7. [7]

    Ggs: Generalizable gaussian splatting for lane switching in autonomous driving

    Huasong Han, Kaixuan Zhou, Xiaoxiao Long, Yusen Wang, and Chunxia Xiao. Ggs: Generalizable gaussian splatting for lane switching in autonomous driving. InProceedings of the AAAI Conference on Artificial Intelligence, pages 3329– 3337, 2025. 8

    2025

  8. [8]

    Speedy-splat: Fast 3d gaus- sian splatting with sparse pixels and sparse primitives

    Alex Hanson, Allen Tu, Geng Lin, Vasu Singla, Matthias Zwicker, and Tom Goldstein. Speedy-splat: Fast 3d gaus- sian splatting with sparse pixels and sparse primitives. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 21537–21546, 2025. 8

    2025

  9. [9]

    Pup 3d-gs: Principled uncertainty pruning for 3d gaussian splatting

    Alex Hanson, Allen Tu, Vasu Singla, Mayuka Jayawardhana, Matthias Zwicker, and Tom Goldstein. Pup 3d-gs: Principled uncertainty pruning for 3d gaussian splatting. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 5949–5958, 2025. 8

    2025

  10. [10]

    Deep blending for free-viewpoint image-based rendering.ACM Transactions on Graphics (ToG), 37(6):1–15, 2018

    Peter Hedman, Julien Philip, True Price, Jan-Michael Frahm, George Drettakis, and Gabriel Brostow. Deep blending for free-viewpoint image-based rendering.ACM Transactions on Graphics (ToG), 37(6):1–15, 2018. 6

    2018

  11. [11]

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

    Georg Hess, Carl Lindstr ¨om, Maryam Fatemi, Christoffer Petersson, and Lennart Svensson. Splatad: Real-time li- dar and camera rendering with 3d gaussian splatting for au- tonomous driving. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 11982–11992,

  12. [12]

    Sort-free gaussian splatting via weighted sum rendering

    Qiqi Hou, Randall Rauwendaal, Zifeng Li, Hoang Le, Farzad Farhadzadeh, Fatih Porikli, Alexei Bourd, and Amir Said. Sort-free gaussian splatting via weighted sum rendering. arXiv preprint arXiv:2410.18931, 2024. 8

    Pith/arXiv arXiv 2024

  13. [13]

    3d gaussian splatting for real-time radiance field rendering.ACM Trans

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

  14. [14]

    Stochasticsplats: Stochastic rasteriza- tion for sorting-free 3d gaussian splatting.arXiv preprint arXiv:2503.24366, 2025

    Shakiba Kheradmand, Delio Vicini, George Kopanas, Dmitry Lagun, Kwang Moo Yi, Mark Matthews, and An- drea Tagliasacchi. Stochasticsplats: Stochastic rasteriza- tion for sorting-free 3d gaussian splatting.arXiv preprint arXiv:2503.24366, 2025. 8

    Pith/arXiv arXiv 2025

  15. [15]

    Tanks and temples: Benchmarking large-scale scene reconstruction.ACM Transactions on Graphics, 36(4), 2017

    Arno Knapitsch, Jaesik Park, Qian-Yi Zhou, and Vladlen Koltun. Tanks and temples: Benchmarking large-scale scene reconstruction.ACM Transactions on Graphics, 36(4), 2017. 6

    2017

  16. [16]

    Realistic and interactive virtual museum representation using 3d gaussian splatting

    Ohyang Kwon and Jeongmin Yu. Realistic and interactive virtual museum representation using 3d gaussian splatting. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, pages 185–192, 2025. 1

    2025

  17. [17]

    Safe- guardgs: 3d gaussian primitive pruning while avoiding catas- trophic scene destruction.arXiv preprint arXiv:2405.17793,

    Yongjae Lee, Zhaoliang Zhang, and Deliang Fan. Safe- guardgs: 3d gaussian primitive pruning while avoiding catas- trophic scene destruction.arXiv preprint arXiv:2405.17793,

    Pith/arXiv arXiv

  18. [18]

    Gaurast: Enhancing gpu triangle rasteriz- ers to accelerate 3d gaussian splatting.arXiv preprint arXiv:2503.16681, 2025

    Sixu Li, Ben Keller, Yingyan Celine Lin, and Brucek Khailany. Gaurast: Enhancing gpu triangle rasteriz- ers to accelerate 3d gaussian splatting.arXiv preprint arXiv:2503.16681, 2025. 8

    Pith/arXiv arXiv 2025

  19. [19]

    Streamgs: Online general- izable gaussian splatting reconstruction for unposed image streams.arXiv preprint arXiv:2503.06235, 2025

    Yang Li, Jinglu Wang, Lei Chu, Xiao Li, Shiu-hong Kao, Ying-Cong Chen, and Yan Lu. Streamgs: Online general- izable gaussian splatting reconstruction for unposed image streams.arXiv preprint arXiv:2503.06235, 2025. 8

    Pith/arXiv arXiv 2025

  20. [20]

    Omnireason: A temporal-guided vision-language- action framework for autonomous driving.arXiv preprint arXiv:2509.00789, 2025

    Pei Liu, Qingtian Ning, Xinyan Lu, Haipeng Liu, Weil- iang Ma, Dangen She, Peng Jia, Xianpeng Lang, and Jun Ma. Omnireason: A temporal-guided vision-language- action framework for autonomous driving.arXiv preprint arXiv:2509.00789, 2025. 1

    Pith/arXiv arXiv 2025

  21. [21]

    Compgs: Efficient 3d scene representa- tion via compressed gaussian splatting

    Xiangrui Liu, Xinju Wu, Pingping Zhang, Shiqi Wang, Zhu Li, and Sam Kwong. Compgs: Efficient 3d scene representa- tion via compressed gaussian splatting. InProceedings of the 32nd ACM International Conference on Multimedia, pages 2936–2944, 2024. 1, 8

    2024

  22. [22]

    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. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20654–20664, 2024. 1

    2024

  23. [23]

    Taming 3dgs: High-quality radiance fields with limited resources

    Saswat Subhajyoti Mallick, Rahul Goel, Bernhard Kerbl, Markus Steinberger, Francisco Vicente Carrasco, and Fer- nando De La Torre. Taming 3dgs: High-quality radiance fields with limited resources. InSIGGRAPH Asia 2024 Con- ference Papers, pages 1–11, 2024. 1

    2024

  24. [24]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021

    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.Communications of the ACM, 65(1):99–106, 2021. 1

    2021

  25. [25]

    Radsplat: Radiance field-informed gaussian splat- ting for robust real-time rendering with 900+ fps

    Michael Niemeyer, Fabian Manhardt, Marie-Julie Rakoto- saona, Michael Oechsle, Daniel Duckworth, Rama Gosula, Keisuke Tateno, John Bates, Dominik Kaeser, and Federico Tombari. Radsplat: Radiance field-informed gaussian splat- ting for robust real-time rendering with 900+ fps. In2025 International Conference on 3D Vision (3DV), pages 134–

  26. [26]

    Pytorch: An im- perative style, high-performance deep learning library, 2019

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas K ¨opf, Edward Yang, Zach DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, and Soumith Chintala. Pytorch: An im- perative style, high-perf...

    2019

  27. [27]

    3dgs-avatar: Animatable avatars via deformable 3d gaussian splatting

    Zhiyin Qian, Shaofei Wang, Marko Mihajlovic, Andreas Geiger, and Siyu Tang. 3dgs-avatar: Animatable avatars via deformable 3d gaussian splatting. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5020–5030, 2024. 8

    2024

  28. [28]

    Gaussianshopvr: Facilitating immersive 3d authoring using gaussian splatting in vr

    Yulin Shen, Boyu Li, Jiayang Huang, David Yip, and Zeyu Wang. Gaussianshopvr: Facilitating immersive 3d authoring using gaussian splatting in vr. InProceedings of the 38th An- nual ACM Symposium on User Interface Software and Tech- nology, pages 1–14, 2025. 8

    2025

  29. [29]

    3d gaussian-based immersive media streaming in networked extended reality

    Yuang Shi. 3d gaussian-based immersive media streaming in networked extended reality. InProceedings of the 16th ACM Multimedia Systems Conference, pages 356–360, 2025. 1

    2025

  30. [30]

    F-3dgs: Factorized co- ordinates and representations for 3d gaussian splatting

    Xiangyu Sun, Joo Chan Lee, Daniel Rho, Jong Hwan Ko, Usman Ali, and Eunbyung Park. F-3dgs: Factorized co- ordinates and representations for 3d gaussian splatting. In Proceedings of the 32nd ACM International Conference on Multimedia, pages 7957–7965, 2024. 8

    2024

  31. [31]

    Streaming 3dgs virtual worlds in 6dof over next-generation networks

    Yuan-Chun Sun. Streaming 3dgs virtual worlds in 6dof over next-generation networks. InProceedings of the 33rd ACM International Conference on Multimedia, pages 13561–13565, 2025. 1

    2025

  32. [32]

    Adr-gaussian: Ac- celerating gaussian splatting with adaptive radius

    Xinzhe Wang, Ran Yi, and Lizhuang Ma. Adr-gaussian: Ac- celerating gaussian splatting with adaptive radius. InSIG- GRAPH Asia 2024 Conference Papers, pages 1–10, 2024. 1, 2, 3, 7, 8

    2024

  33. [33]

    Advancing immersive content delivery with dynamic 3d gaussian splatting

    Nan Wu, Weikai Lin, Ruizhi Cheng, Bo Chen, Yuhao Zhu, Klara Nahrstedt, and Bo Han. Advancing immersive content delivery with dynamic 3d gaussian splatting. InProceed- ings of the 26th International Workshop on Mobile Comput- ing Systems and Applications, pages 109–114, 2025. 1

    2025

  34. [34]

    gsplat: An open-source library for gaussian splatting.Journal of Machine Learning Research, 26(34):1–17, 2025

    Vickie Ye, Ruilong Li, Justin Kerr, Matias Turkulainen, Brent Yi, Zhuoyang Pan, Otto Seiskari, Jianbo Ye, Jeffrey Hu, Matthew Tancik, et al. gsplat: An open-source library for gaussian splatting.Journal of Machine Learning Research, 26(34):1–17, 2025. 8

    2025

  35. [35]

    Lp-3dgs: Learning to prune 3d gaussian splatting.Advances in Neural Information Processing Systems, 37:122434–122457, 2024

    Zhaoliang Zhang, Tianchen Song, Yongjae Lee, Li Yang, Cheng Peng, Rama Chellappa, and Deliang Fan. Lp-3dgs: Learning to prune 3d gaussian splatting.Advances in Neural Information Processing Systems, 37:122434–122457, 2024. 8

    2024

  36. [36]

    Drivinggaussian: Composite gaussian splatting for surrounding dynamic au- tonomous driving scenes

    Xiaoyu Zhou, Zhiwei Lin, Xiaojun Shan, Yongtao Wang, Deqing Sun, and Ming-Hsuan Yang. Drivinggaussian: Composite gaussian splatting for surrounding dynamic au- tonomous driving scenes. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 21634–21643, 2024. 1 CaT-GS: Efficient 3DGS Rendering for Large-Scale Scenes with...

    2024