- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper introduces SampleSplat, a feed-forward 3DGS framework from uncalibrated input images. Instead of predicting fixed number of Gaussians, SampleSplat predicts a categorical smapling probability and a local 3D Gaussian offset distribution per anchor, drawing anchors with replacement to place primitives where scene complexity is high. SampleSplat achieves state-of-the-art performance.
Critical strengths
[S1] The visualizations of the learned sampling probability are intuitive and show that the model naturally learns where to concentrate on without heuristics.
[S2] The results are strong and the qualitative result shows sharp novel-view synthesis compare to other baselines.
Critical weaknesses
[W1] Missing adaptive / compact 3DGS baselines: Figure 2 highlights F4Splat and Off The Grid, but they are not included in the related work or compared in any experiment. The paper also misses C3G [1], which compresses pose-free feed-forward 3DGS to ~2K Gaussians, and closely related designs like Scaffold-GS (anchor + offset decoding) [2].
[W2] In Table 4, learned sampling improves PSNR by only about 0.6, whereas GS Depth head and point transformer drive most of the gain.
[W3] Because the sampling probability receives its gradient through each Gaussian's opacity, an anchor's probability only increases when making its existing Gaussians more opaque lowers the rendering error. However, making one Gaussian more opaque is not the same as needing more Gaussians in that location (for example, a blurry edge covered by an already-opaque Gaussian needs to be split into several smaller Gaussians, not made more opaque). Also, anchors that are not sampled in a step get no gradient at all even if their rendering error is huge, while anchors sampled multiple times accumulate larger gradients.
Questions for authors
[Q1] How does SampleSplat performs against F4Splat, Off The Grid, and C3G under matched # Gaussians budgets?
3 comments
Thank you for the detailed rebuttal and for running the additional experiments. However, two concerns still remain regarding the comparisons.
- In the 256x256 matched comparison, SampleSplat degrades monotonically as the number of Gaussians increases, with a similar degradation at 518x518 comparison. Nearly 1dB drop suggests that SampleSplat does not scale to larger Gaussian budgets.
- Although C3G cannot retrained, comparing it with SampleSplat at 2K Gaussians would be a fair comparison. Since SampleSplat supports a controlalble Gaussian budget, a direct evaluation at a matched 2K budget is still missing.
Given these remaining concerns, I will maintain my score.
[Q1] Reconstruction quality as the Gaussian budget increases
Thank you for pointing out how reconstruction quality changes as the Gaussian budget increases.
- At 256×256, PSNR decreases at the highest tested budget, while SampleSplat retains its advantage at every matched budget. From the best-performing budget to the largest tested budget, PSNR decreases by 0.98 dB for two input views and 0.90 dB for six input views. At the largest tested budget, SampleSplat still leads the strongest baseline at that budget by 2.54 dB and 4.57 dB, respectively. Across the comparisons with matched input FOV, resolution, and Gaussian count, SampleSplat outperforms the corresponding baselines on all four quality metrics, with PSNR gains of 2.37–6.23 dB.
- At 518×518, increasing the budget improves reconstruction quality over part of the tested range. With two input views, increasing the budget from 44K to 77K raises PSNR from 25.9987 dB to 26.7233 dB and improves the other three metrics. Increasing the budget further to 88K lowers PSNR by only 0.025 dB. With six input views, PSNR likewise peaks at the intermediate tested budget of 230K. The best-performing budget therefore depends on the input configuration.
- The other models have a fixed upper limit on their Gaussian budgets. For a given input configuration, EcoSplat, F4Splat, and OffTheGrid each have a fixed maximum Gaussian budget, and their training covers the corresponding full-budget settings.
- SampleSplat is trained with a single Gaussian budget setting, and its performance at all other budgets in the reported sweeps reflects test-time generalization from the same checkpoint. Its sampling-based design imposes no fixed architectural cap on the Gaussian count, so users can choose a budget at inference based on the trade-off between reconstruction quality and computational cost.
Nevertheless, SampleSplat outperforms the evaluated baselines on all four reconstruction metrics at matched Gaussian budgets representative of those commonly used by existing models. A single SampleSplat checkpoint supports different input resolutions, aspect ratios, and view counts, while allowing users to specify the Gaussian count at inference to fit their resource constraints without retraining.
[Q2] Direct comparison with C3G at a matched 2K Gaussian budget
To address the reviewer’s suggestion, we fine-tuned the existing SampleSplat checkpoint for only 0.4K steps at C3G’s input resolutions, setting the total number of sampled Gaussians to 2K to match C3G’s budget. With this brief fine-tuning, SampleSplat outperforms C3G on all four reconstruction metrics at every tested view count. C3G uses a dedicated checkpoint for inference at , while SampleSplat uses the same fine-tuned checkpoint for all view counts reported below.
| 2 views | 6 views | 8 views | Average | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Method | Input Resolution | PSNR↑ | SSIM↑ | LPIPS↓ | DISTS↓ | #GS | PSNR↑ | SSIM↑ | LPIPS↓ | DISTS↓ | #GS | PSNR↑ | SSIM↑ | LPIPS↓ | DISTS↓ | #GS | PSNR↑ | SSIM↑ | LPIPS↓ | DISTS↓ | #GS |
| C3G (GT Tgt. Pose) | 224×224/256×256 | 15.5336 | 0.4331 | 0.5663 | 0.2855 | 2K | 15.1811 | 0.4433 | 0.6224 | 0.3711 | 2K | 15.2102 | 0.4437 | 0.6211 | 0.3703 | 2K | 15.3083 | 0.4400 | 0.6033 | 0.3423 | 2K |
| SampleSplat (GT Tgt. Pose) | 224×224/256×256 | 19.2920 | 0.5124 | 0.4093 | 0.2443 | 2K | 17.9080 | 0.4567 | 0.4991 | 0.2955 | 2K | 17.9880 | 0.4617 | 0.4953 | 0.2941 | 2K | 18.3960 | 0.4769 | 0.4679 | 0.2780 | 2K |
| SampleSplat (GT Tgt. Pose) | 540×960 | 25.9471 | 0.8088 | 0.2507 | 0.1015 | 113K | 22.5389 | 0.6994 | 0.3071 | 0.1121 | 259K | 22.8296 | 0.7089 | 0.2930 | 0.1055 | 324K | 23.7719 | 0.7390 | 0.2836 | 0.1064 | 232K |
| 16 views | 24 views | 32 views | Average | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Method | Input Resolution | PSNR↑ | SSIM↑ | LPIPS↓ | DISTS↓ | #GS | PSNR↑ | SSIM↑ | LPIPS↓ | DISTS↓ | #GS | PSNR↑ | SSIM↑ | LPIPS↓ | DISTS↓ | #GS | PSNR↑ | SSIM↑ | LPIPS↓ | DISTS↓ | #GS |
| C3G (GT Tgt. Pose) | 256×256 | 14.6760 | 0.4392 | 0.6444 | 0.4166 | 2K | 13.4284 | 0.4085 | 0.6816 | 0.5105 | 2K | 13.1898 | 0.4030 | 0.6883 | 0.5269 | 2K | 13.7647 | 0.4169 | 0.6714 | 0.4847 | 2K |
| SampleSplat (GT Tgt. Pose) | 256×256 | 17.6310 | 0.4437 | 0.5154 | 0.3093 | 2K | 17.4710 | 0.4370 | 0.5475 | 0.3384 | 2K | 16.6940 | 0.4044 | 0.5571 | 0.3371 | 2K | 17.2653 | 0.4284 | 0.5400 | 0.3283 | 2K |
| SampleSplat (GT Tgt. Pose) | 540×960 | 22.5594 | 0.6918 | 0.2730 | 0.0962 | 648K | 21.1138 | 0.6432 | 0.3120 | 0.1133 | 972K | 20.9945 | 0.6403 | 0.3124 | 0.1133 | 1296K | 21.5559 | 0.6585 | 0.2991 | 0.1076 | 972K |
We hope the additional experiments and clarifications address your concerns, and we would appreciate it if you would consider revisiting your score in light of these results.
Dear Reviewer GscZ,
Thank you for taking the time to review our work. Your comments helped us clarify what learned sampling contributes and how the opacity surrogate works. We have marked all revisions in blue.
We would welcome any further feedback. If you feel that our response and revisions address your concerns, we would be grateful if you would reconsider your rating.
Best regards,
The Authors