3D Gaussian Splatting (3DGS) provides high-quality real-time novel-view synthesis, but deployment budgets vary across devices and network conditions. Most compact methods target one operating point, so another budget requires a separate representation. SHIN is a scalable 3DGS hierarchy with a reusable Base and additive enhancement layers, so each higher-quality operating point extends rather than replaces the decoded representation. Constructing this hierarchy requires deciding which Gaussians each stage adds and how much spherical-harmonic (SH) appearance each retains, because these choices remain fixed in later prefixes. SHIN derives one shared Gaussian order from multi-scale position-scale sensitivity and divides each stage budget between adding Gaussians and retaining SH bands using calibrated gradient-based scores per byte. Lightweight opacity updates adjust compositing while transmitted geometry and SH coefficients remain unchanged. Across three benchmarks, SHIN progressively improves one cumulative representation and achieves the best PSNR-size efficiency in every overlapping comparison with the multi-rate and progressive baselines.
Figures & tables
Figure 1: Scalable hierarchy versus LoD. (a) SHIN cumulatively extends a decoded Base with enhancement layers. (b) LoD selects operating states from a shared representation without requiring cumulative transmission.
Figure 2: SHIN pipeline. (a) Multi-scale position–scale sensitivity defines a Gaussian order shared across stages. (b) Layer–Degree Allocation divides each stage budget between adding Gaussians and retaining SH bands using calibrated scores per byte. (c) Each layer is appended while transmitted geometry and SH remain fixed, with opacity updates applied to the existing prefix.
Figure 3: Disjoint layer contents and cumulative prefixes on Bicycle. Top: individual layer contents ( H0,H1,H2 ) rendered in isolation. Bottom: cumulative prefixes ( P0,P1,P2 ).
Figure 4: PSNR–model-size trade-offs. Curves denote multi-rate representations and markers denote independently optimized compact models; SHIN stages form cumulative prefixes.
Stage
Method
Mip-NeRF360
Tanks&Temples
Deep Blending
PSNR ↑
SSIM ↑
LPIPS ↓
Size (MiB) ↓
FPS ↑
PSNR ↑
SSIM ↑
LPIPS ↓
Size (MiB) ↓
FPS ↑
PSNR ↑
SSIM ↑
LPIPS ↓
Size (MiB) ↓
FPS ↑
Base
SHIN
27.08
.8032
.2396
61
326
22.63
.7984
.2807
43
219
29.72
.9029
.2491
59
448
MGS [ 9 ]
24.24
.6531
.3839
56
529
21.42
.7371
.3661
56
394
27.86
.8788
.2906
56
602
GoDe [ 6 ]
26.90
.7842
.2738
56
369
22.82
.7868
.3068
49
195
29.52
.8950
.2757
51
497
FLoD [ 23 ]
24.22
.6389
.4358
119
444
21.15
.7085
.4326
85
292
28.15
.8689
.3326
60
661
CLoD-GS [ 5 ]
26.92
.7984
.1998
450
198
23.19
.8114
.2256
237
234
28.68
.8960
.1882
369
260
Table 1: Quality, model-size, and rendering-speed comparison. Size denotes cumulative model size; parentheses show the additional size from the preceding operating point for SHIN, MGS, and LapisGS. Other methods report total model size at each operating point.
Figure 5: Qualitative comparison on Garden and Train. Insets highlight fine structures; labels report PSNR and model size.
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
Measurement
Value
Multi-scale ordering
14.2 s
SH-band scoring
10.5 s
Layer–Degree Allocation
104.8 s
Base refinement
139.3 s
E1 refinement
263.1 s
E2 refinement
359.4 s
Appendix
Table 7: Construction profile for Garden on an RTX 6000 Ada.
Budget
3-stage
5-stage
Δ PSNR
MiB
10%
27.084
27.085
+0.001
61
14%
–
27.503
–
87
18%
27.685
27.648
−0.037
113
25%
–
27.717
–
158
32%
27.753
27.734
−0.019
204
Appendix
Table 10: Three- versus five-stage hierarchy on Mip-NeRF360. Sizes are five-stage cumulative footprints.
Calibration
Base
+E1
+E2
BD-rate ↓
Q50
27.09
27.67
27.75
+2.18%
Q90
27.06
27.68
27.76
+1.06%
Q95
27.08
27.68
27.75
0.00%
Q99
27.16
27.67
27.71
+0.08%
Appendix
Table 11: Calibration-percentile robustness on Mip-NeRF360. PSNR is in dB; Q95 is the default.
3D Gaussian Splatting (3DGS) has become the standard for real-time novel view synthesis on commodity GPUs. Its pipeline ties spatial partitioning and rasterization to one tile size, yet the two pull in opposite directions: partitioning, which bins and depth-sorts gaussians, grows cheaper with larger tiles, while rasterization gets cheaper with smaller ones. Prior acceleration work reduces the cost of individual stages but keeps both locked to that single scale, where a few dense tiles dominate frame time. We present Hierarchically Tiled Gaussian Splatting (HiGS), which gives each its own scale: partitioning runs over coarse macro-tiles, while rasterization runs over the fine render tiles within them. Rasterization work is then issued in proportion to the gaussians in each macro-tile rather than per tile, so dense regions spread across many parallel units instead of serializing through one. Across tested scenes, HiGS renders up to 15.8x faster than the original 3DGS and outperforms every other rasterizer we evaluate, while preserving exact front-to-back alpha compositing.
3D Gaussian Splatting (3DGS) enables real-time novel view synthesis, but existing general-purpose acceleration methods suffer severe rendering quality degradation when extended to more complex, large-scale scenes. To address this issue, we propose EffGS, a more general acceleration framework that improves training and rendering efficiency while maintaining reconstruction quality comparable to or better than vanilla 3DGS across bounded and large-scale scenes. EffGS combines frequency-aware guidance, localized density control, and adaptive primitive scale modulation. First, an importance scoring mechanism combines pixel-wise reconstruction errors with a difference-of-Gaussians mask scheduled over training to provide stage-dependent spatial guidance. Second, localized densification and pruning restricts density modifications to Gaussians with valid projected footprints in the sampled views. Third, learnable per-Gaussian scale modulation adjusts effective primitive extent during optimization while retaining the Compact Box rasterization rule. Extensive experiments on bounded and large-scale scene datasets demonstrate a favorable balance between reconstruction quality, training time, and primitive count. Component ablations and matched-primitive-budget comparisons further support the effectiveness of the framework.
Changbai Li, Shuo Yang, Yichen Yang +2
Beihang University · Nanyang Technological University
Single-image feed-forward 3D Gaussian Splatting (3DGS) aims to directly generate a renderable 3D scene representation from one input image, avoiding the cost of multi-view capture and per-scene optimization. However, existing methods are often constrained by a pixel-aligned representation, where Gaussians are predicted from fixed image-grid locations. Such pixel-aligned primitives can produce promising nearby-view renderings, but they remain weakly coupled to underlying scene surfaces and struggle to preserve coherent structures under large viewpoint shifts. We present InfiniSplat, a feed-forward single-image 3DGS framework that moves from a pixel-aligned representation toward a surface-aligned representation. InfiniSplat constructs this representation by first using geometry-guided sampling to place 2D supports according to depth-induced local surface structure, and then applying a query-conditioned implicit decoder to predict Gaussian attributes from the image features queried at these supports. By grounding support locations in geometry while decoupling Gaussian prediction from fixed pixel centers, InfiniSplat produces Gaussian layouts that better follow scene surfaces and reduce scattered primitives caused by grid discretization. Across multiple cross-dataset NVS evaluations, InfiniSplat achieves state-of-the-art performance compared with single-image feed-forward baselines, and demonstrates zero-shot generalization from Hypersim indoor synthetic training to complex open-world scenes. Project page: https://zju3dv.github.io/InfiniSplat.
Jiawei Wang, Hao Yu, Yongzhen Hu +7
State Key Laboratory of CAD&CG, Zhejiang University, China · Zhejiang University, China · Udeer.ai, China +1