Library

LP-3DGS: Learning to Prune 3D Gaussian Splatting

Recently, 3D Gaussian Splatting (3DGS) has become one of the mainstream methodologies for novel view synthesis (NVS) due to its high quality and fast rendering speed. However, as a point-based scene representation, 3DGS potentially generates a large number of Gaussians to fit the scene, leading to high memory usage. Improvements that have been proposed require either an empirical and preset pruning ratio or importance score threshold to prune the point cloud. Such hyperparamter requires multiple rounds of training to optimize and achieve the maximum pruning ratio, while maintaining the rendering quality for each scene. In this work, we propose learning-to-prune 3DGS (LP-3DGS), where a trainable binary mask is applied to the importance score that can find optimal pruning ratio automatically. Instead of using the traditional straight-through estimator (STE) method to approximate the binary mask gradient, we redesign the masking function to leverage the Gumbel-Sigmoid method, making it differentiable and compatible with the existing training process of 3DGS. Extensive experiments have shown that LP-3DGS consistently produces a good balance that is both efficient and high quality.

cs.CVComputer Science
Source

Paper

References (26)

01Mini-Splatting: Representing Scenes with a Constrained Number of GaussiansGuangchi Fang, Bing Wang2024 · European Conference on Computer Vision · 290 citations In Library
02RadSplat: Radiance Field-Informed Gaussian Splatting for Robust Real- Time Rendering with 900+ FPSMichael Niemeyer, Fabian Manhardt, Marie-Julie Rakotosaona et al.2024 · International Conference on 3D Vision · 140 citations In Library
03GeoGaussian: Geometry-aware Gaussian Splatting for Scene RenderingYanyan Li, Chen Lyu, Yan Di et al.2024 · European Conference on Computer Vision · 71 citations In Library
04FreGS: 3D Gaussian Splatting with Progressive Frequency RegularizationJiahui Zhang, Fangneng Zhan, Muyu Xu et al.2024 · Computer Vision and Pattern Recognition · 159 citations In Library
05DNGaussian: Optimizing Sparse-View 3D Gaussian Radiance Fields with Global-Local Depth NormalizationJiahe Li, J. Zhang, Xiao Bai et al.2024 · Computer Vision and Pattern Recognition · 337 citations In Library
06LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPSZhiwen Fan, Kevin Wang, Kairun Wen et al.2023 · Neural Information Processing Systems · 600 citations In Library
07Compact 3D Gaussian Representation for Radiance FieldJ. Lee, Daniel Rho, Xiangyu Sun et al.2023 · Computer Vision and Pattern Recognition · 483 citations In Library
08Depth-Regularized Optimization for 3D Gaussian Splatting in Few-Shot ImagesJaeyoung Chung, Jeongtaek Oh, Kyoung Mu Lee2023 · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · 250 citations In Library
09SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh RenderingAntoine Guédon, Vincent Lepetit2023 · Computer Vision and Pattern Recognition · 875 citations In Library
10PDF: Point Diffusion Implicit Function for Large-scale Scene Neural RepresentationYuhan Ding, Fukun Yin, Jiayuan Fan et al.2023 · Neural Information Processing Systems · 7 citations In Library
113D Gaussian Splatting for Real-Time Radiance Field RenderingBernhard Kerbl, Georgios Kopanas, Thomas Leimkuehler et al.2023 · ACM Transactions on Graphics · 9.6k citations In Library
12TensoRF: Tensorial Radiance FieldsAnpei Chen, Zexiang Xu, Andreas Geiger et al.2022 · European Conference on Computer Vision · 1.8k citations In Library

Scroll for more · 14 remaining

Similar papers

PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian SplattingAlex Hanson, Allen Tu, Vasu Singla et al.2024 · Computer Vision and Pattern Recognition · 95 citations In Library
GPGS: Geometric Priors for 3D Gaussian Splatting in Structural EnvironmentsZiwei Xu, Wen Chen, Shilong Wang et al.2025 · IEEE/RJS International Conference on Intelligent RObots and Systems · 0 citations In Library
High-Fold 3D Gaussian Splatting Model Pruning Method Assisted by OpacityShiyu Qiu, Chunlei Wu, Zhe Wan et al.2025 · Applied Sciences In Library
EfficientGS: Streamlining Gaussian Splatting for Large-Scale High-Resolution Scene RepresentationWenkai Liu, Tao Guan, Bin Zhu et al.2024 · arXiv (Cornell University) In Library
LargeSceneGaussian: High-Efficiency 3D Gaussian Splatting for Large-Scale Scene ReconstructionYongxin Ge, Junqi Liao, Li Li et al.2025 In Library
SparseGS: Sparse View Synthesis using 3D Gaussian SplattingHaolin Xiong, Sairisheek Muttukuru, Xiao, Hanyuan et al.2023 · arXiv (Cornell University) In Library
Adversarial Pruning Networks for Compact 3D Gaussian SplattingHui Shuai, Yucheng Shi, Yubao Sun et al.2025 · IEEE Transactions on Multimedia · 3 citations In Library
GP-GS: Gaussian Processes Densification for 3D Gaussian SplattingZhihao Guo, Jingxuan Su, Shenglin Wang et al.2025 · 4 citations In Library
Trimming the Fat: Efficient Compression of 3D Gaussian Splats through PruningMuhammad Salman Ali, Maryam Qamar, Sung-Ho Bae et al.2024 · British Machine Vision Conference · 57 citations In Library
DP-GS: Depth-prior & Perception-guided Gaussian Splatting for Sparse-view Novel View SynthesisBo Gao, Zhao Lu, Mingyi He et al.2025 In Library

Peer review

Accept (poster)NeurIPS 2024 · via OpenReview
Reviewer 1UP25/10 · confidence 4/52024-07-10

Summary

The paper proposed a pruning method for Gaussian Splatting training, which employ a learnable mask to find optimal pruning rate.

Strengths

The paper is well-written and the proposed method is easy to understand.

Weaknesses

* The proposed method use a regularization loss to encourage the model to prune, so that the weight parameter is an important balancing factor. The discussion on how to set such parameter is missing. * Figure 2 is not clear enough for me to well understand the proposed idea. In Figure 3, why pattern like (b) is better than (a)? * Experiments section use existing works for comparison, but they are all self-implemented. The performance in their original paper should be included for reference. * The author claim the proposed method can find optimal pruning ratio. Based on my observation, the conclusion is not solid. First, what is definition of optimal? Second, not all the subfigures in Fig. 4 & 6 support this conclusion.

Questions

Please see the "weaknesses".

Rating

5

Confidence

4

Soundness

2

Presentation

2

Contribution

2

Limitations

No potential negative social impact.

Authorsrebuttal2024-08-11

Dear reviewer, We hope our rebuttal can address your comments and concern. As the deadline for the author-reviewer discussion is approaching, would you please share your thoughts on our rebuttal and reconsider the review score? thanks a lot

Authorsrebuttal2024-08-13

Thanks for your comments ### For "optimal" wording As for the term of "optimal pruning ratio", in our response to the common question, we already claimed that we will not use it and change to "learned pruning ratio" with more detailed definition. Please refer to above for our detailed clarification. The reviewer o7SR suggested us to change to "learned effective pruning ratio" and we plan to use such in our updated manuscript. For the sensitivity analysis of rendering image quality against pruning ratios, as we show in our experiment results (figure 4 and 6), each scene has different sensitivities against pruning and the "band" is in general within around 0.5 to 0.8, rather than 0.7 to 0.8 or 0.6 to 0.8, where the rendering image quality starts to drop with increasing pruning ratio. In order to find such "band" for different scenes, a manual tuning of pruning ratios with multiple rounds of training is still needed in prior works, which is costly and not an "easy approach". We kindly disagree that just setting the pruning ratio to 0.75 will be just fine. To justify such, we conducted extra experiments based on the pruning ratio of 0.75 based on the Radsplat score for kitchen and room scenes, due to the limited rebuttal time left. Note that, table 1 in our manuscript reported the complete set of our method learned pruning ratio and its corresponding image quality of each scene. From table 1, it can be seen that, in average, our model with learned pruning ratio has almost the same rendering image quality as the unpruned version, which is what we defined in the loss function and the experiment results support that well. In particular, the learned pruning ratio is 0.58 for kitchen scene and 0.74 for room scene. It can be seen from the extra experiments in below table that the kitchen scene has a significant quality drop with a random 0.75 pruning ratio than that of learned value of 0.58. While, since the randomly chosen 0.75 pruning ratio is close to the learned ratio value of 0.74 for Room scene, the rendering image quality is very similar. Furthermore, such 0.75 pruning ratio is actually not "random", but from the thorough complete pruning sensitivity analysis reported in our submitted manuscript. We believe these ablation study could further prove the motivation of this work and effectiveness of our proposed method. To summarize, for scientific research and contributions, we believe our work discovered the pattern of pruning sensitivity in 3DGS and further propose a general learning method that could be applied to different types of pruning importance scores with a target of automatic pruning, which is a solid contribution to the community and will inspire more research in the future. | | Pruning Raio | Kitchen | Room | |:-----:|:------------:|:-------:|:------:| | | 0 | 31.581 | 31.500 | | PSRN | Learned | 31.515 | 31.490 | | | 0.75 | 31.187 | 31.420 | || | | 0 | 0.9230 | 0.9249 | | SSIM | Learned | 0.9311 | 0.9243 | | | 0.75 | 0.9270 | 0.9235 | || | | 0 | 0.1164 | 0.1978 | | LPIPS | Learned | 0.1194 | 0.2032 | | | 0.75 | 0.1293 | 0.2045 | ### We are not sure whether the reviewers are criticising the effectiveness or efficiency. We discussed about both in below to clarify. Method effectiveness: For the Bicycle-Radsplat subfigure of Figure 6, all the quantitative metrics, i.e., PSNR/SSIM/LPIPS, that are very popularly used in this community to evaluate the rendering image quality are starting to become worse beyond 0.6 pruning ratio. It actually shows the effectiveness of our method that could learn such pruning ratio with only one time training. The pruning is always a tradeoff process between compression ratio and rendering quality, which is the nature of this problem. We never claimed our learned pruning ratio is always better than other manual pruning ratio settings, but the key point is our method is learnable and automatic that could efficiently and effectively achieve a pruning ratio that is always within the "band" mentioned by the reviewer, rather than costly manual tuning. Method efficiency: We conducted comprehensive training cost analysis in table 2. The actual training time compared with vanilla 3DGS reduces even for single round of training. Compared with other manual pruning method, as we discussed before, our method could achieve a learned effective pruning ratio, embedding with the 3DGS model construction learning process, without the costly pruning sensitivity analysis shown in the figure 4 and 6 that require multiple rounds of training with preset pruning ratio. We believe our method could improve both the 3DGS learning efficiency and pruning efficiency significantly compared with prior works. We kindly request the reviewer to reconsider our technical contributions to the community, as well as the final review score of our work.

Reviewer 1UP22024-08-13

Maybe the authors misunderstand my first point. I'm not criticizing the wording. In the comment the authors claimed that "the "band" is in general within around 0.5 to 0.8, rather than 0.7 to 0.8 or 0.6 to 0.8". I do not think there is a substantial difference between them. However, I would like to provide some advice to evaluate your method more scientifically. To prove that your method is more effective than randomly selecting the prune ratio in the empirical interval e.g. 0.5-0.8, you can randomly select many points from the interval to get the expectation of these metrics including PSNR/LPIPS/SSIM. Compared with LP-3DGS, if there is a significant improvement in each indicator and scenario, it indicates that your method is effective. For example, I would expect that LP-3DGS outperforms the random strategy in 80%-90% scenes. As for now, the evaluation results are still vague. As for the presentation, you need to demonstrate the strong correlation between the selected prune ratio and these metrics, in the empirical interval of 0.5-0.8 rather than the 0.0-1.0 shown in your figures. The big interval demonstrated(0.0-1.0) in Fig4&6 contains too many impossible value while overwhelming the effectiveness of the proposed methods.

Authorsrebuttal2024-08-13

Dear reviewer, 1. We did not misunderstand your first comment. In the previous response, we just used one sentence to clarify the word of "optimal". While, the majority rest is about the band interval discussion and clarification of our method. please re-evaluate. 2. In our reported experiments and our several rounds of discussion above, we already stated very clearly that we never argued our learned pruning ratio is always 'better' than random chosen pruning ratio, since there is no 'better' pruning ratio in model pruning problem. It is always a tradeoff between pruning/compression ratio and quality, which is the nature of this problem. But the key point is our method is learnable and automatic that could efficiently and effectively achieve a pruning ratio that is always within the "band" mentioned by the reviewer, ''with only one round of training and embedding with the 3DGS model learning phase, rather than costly manual tuning''. From this point of view, we strongly believe all the experiment results and extra results provided during this rebuttal period to all reviewers are sufficient and strong enough to prove the motivation and effectiveness of our proposed method. We kindly recommend the reviewer to check our discussion with the reviewers o7SR and fvLt above. Both agreed the community will benefit from our proposed automatic pruning and it is more efficient than running entire pruning sensitivity analysis for every scene when compressing the large size of 3DGS model. 3. As for the suggestions of adding more pruning results during 0.5 to 0.8 interval. we will start running more experiments as suggested. However, we do not see the point why it is helping for our method, since our method already learned a large pruning ratio that has almost no quality drop for every unique scene, with only one round of training. More points within the 0.5 to 0.8 interval means more training cost for manual pruning ratio tuning, it will only make the benchmarking baseline worse. 4. Again, we want to highlight that such 0.5 to 0.8 pruning ratio interval "found in this work" is based on the vanilla and default 3DGS training setup. It reveals certain redundancy during the Gaussian densification process. Different 3DGS training hyperparameter settings, for example, the gradient threshold of densification process, will significantly change the number of Gaussian of the final trained model. Then, it is obvious that such "empirical 0.5 to 0.8 interval" will change based on the model redundancy. In this case, even with prior general pruning experience, the cutoff band would change accordingly. To optimizing the tradeoff between compression ratio and rendering quality, new costly pruning sensitivity analysis will be required to get such new cutoff band. we already open-sourced all our codes and results in the github for the community to peer-review and test on their different pruning importance score. We appreciate the reviewer's comments and discussion. But for scientific research, we want to make sure the major contribution of this work is highlighted and understood well.

Reviewer 1UP22024-08-13

Dear authors, I do agree that "our learned pruning ratio is always 'better' than random chosen pruning ratio, since there is no 'better' pruning ratio in model pruning problem". However, in my previous comment, I only expect that your method is better in over 80% scenes. I do not require an absolute better or optimal. At least a number larger than 50% can prove that the proposed method is effective. I do not think a successful rate of 80% is a high standard for a publishable paper. Learnable and automatic method is not preferred if just a random manual number setting can solve the problem comparably. This is the most important point in my opinion. If the proposed method can not demonstrate enough effectiveness, I can only regard it as an engineering trick in GS reconstruction. As for the comment "such 0.5 to 0.8 pruning ratio interval "found in this work" is based on the vanilla and default 3DGS training setup, ... Different 3DGS training hyperparameter settings...". I want to clarify that: If you want to prove your method is effective on other variants of GS, please just show evidence. If you want to prove that finding empirical prune ratio of other GS is more difficult compared with adopting your method, please show evidence. Finally, as for your comment "We kindly recommend the reviewer to check our discussion with the reviewers o7SR and fvLt above. Both agreed the community will benefit from our proposed automatic pruning", I do have seen all the discussions above and **I will keep my independent thinking**. I've spent lots of time on this paper and I am not an outsider of GS. I kindly recommend the authors repect every review equally no matter they are positive or negative.

Authorsrebuttal2024-08-13

Dear reviewer 1UP2 thanks for the comments. 1. If you agree there is no 'better' pruning ratio, but how to guarantee our proposed method could learn the pruning ratio located within the cutoff band in the pruning sensitivity analysis plot shown in figure 4 and 6, what does that mean "At least a number larger than 50% can prove that the proposed method is effective." ? We believe all our learned pruning ratios for every scene is already located within the cutoff band of pruning sensitivity analysis plot. the table 1 also reports, with our learned pruning ratio, almost every scene has the same rendering image quality with the vanilla 3DGS. The average rendering image quality is also the same as the unpruned version with an average of 63% pruning ratio. We do not quite understand what are the requested extra experiments to prove we are at least 50% better. if you mean rendering quality, the average rendering quality of the pruned model is already the same as unpruned version, but with 2.7X smaller model size (i.e. average 63% pruning ratio) 2. We already setup the experiments with different vanilla 3DGS parameters, which will lead to different initial model redundancy. Hope we could get the results before rebuttal deadline. We will post the results once we get it

Reviewer 1UP22024-08-13

I understand that you are talking about the trade-off: better compression ratio means lower rendering quality. However, to demonstrate the proposed method is better than random selection in empirical interval is necessary. The metric that falling into the "band" is so loose. How about defining a metric for such trade-off task? Take PSNR as example. In the case of maximum loss of 1dB(example value), the higher the compression rate, the better.

Authorsrebuttal2024-08-13

Dear reviewer 1UP2 thanks for the comments. We did not only use metric of 'falling into the band' as the evaluation, this is what we want to show the motivation of this work. we actually reported quantitive results of pruning performance in table 1 of our original manuscript. Please refer to that. We did not even set 1db drop as the bar, but targeting to achieve almost no rendering quality drop and learn the corresponding pruning ratio. it is proven in the quantitive results in table 1. For the MipNeRF360 dataset, the average PSNR is 27.48 for unpruned 3DGS model (baseline in table 1). our learned pruned model based on Radsplat score has an average PSNR of 27.47 with average pruning ratio of 0.63. the learned pruned model based on mini-splatting score has an average PSNR of 27.12 with average pruning ratio of 0.6. This difference comes with different definitions of pruning importance scores in both methods, but showing similar performance. Also, we want to note that we only use the pruning importance scores defined in Radsplat and Mini-Splatting as the pruning metric from baseline 3DGS model, without any other tricks in those works to further improve the quality, for fair comparison. Therefore, the theoretical performance ceiling is the unpruned model. For comparing with random pruning ratio selection, the pruning sensitivity analysis in figure 4 and 6 already shows the rendering quality remains almost the same when it is smaller than our learned pruning ratio for each scene. The rendering quality starts to drop when it is larger than our learned pruning ratio. Hope this is what the reviewer requested for proving the 'effectiveness' of our method. thanks prune the vanilla

Reviewer 1UP22024-08-13

"Almost no rendering quality drop" is vague. I recommend you specify the value range and the compression rate in your Tables. The most controversial point we have discussed origins from the task itself. I will raise my score.

Authorsrebuttal2024-08-13

Dear Reviewer 1UP2, Thanks for the suggestion and we will update the manuscript as suggested. Appreciate the discussion and consideration for our work

Reviewer CWFJ4/10 · confidence 3/52024-07-13

Summary

This work introduces LP-3DGS, aiming to compress 3DGS by replacing the previously manually set threshold with a learning-to-prune scheme. In particular, LP-3DGS learns a binary mask to automatically prune 3DGS, where such a mask is regularized with the Gumbel-Sigmoid technique. Experiments on three benchmarks showcase the soundness of the introduced methods.

Strengths

* Applying Gumbel-Sigmoid to regularize the binary mask is reasonable. * Extensive experiments are conducted on several existing benchmarks. * The paper is well-structured and easy to follow.

Weaknesses

* Limited technical contributions. This work seems to be a direct application of Gumbel-Sigmoid, which is a general engineering trick (https://github.com/AngelosNal/PyTorch-Gumbel-Sigmoid) to approximate discrete sampling in a differentiable manner. It needs to be clarified what the main technical insights this work aims to bring to the 3DGS compression community. To me, replacing the threshold setting with the Gumbel-Sigmoid is too intuitive to hit the standard of a top-tier conference. * It is difficult to know whether the learned binary mask achieves the “optimal Gaussian point size”. This work demonstrates that it learns the “optimal” solutions by plotting several line charts without any theoretical analysis. In some cases, LP-3DGS did not find the best balance, e.g., Fig. 4 Mini-Splatting on Kitchen. Besides, LP-3DGS seems to lose the flexibility of finding better solutions compared to its baseline, e.g., RadSplat can set a smaller ratio to compress further while maintaining reasonable visual quality. * Lack of discussions with some recent related works. Several more recent approaches in the 3DGS compression community need discussion, and it would be better to add some comparisons with them to understand the advantages of LP-3DGS better. * Girish, Sharath, Kamal Gupta, and Abhinav Shrivastava. "Eagles: Efficient accelerated 3d gaussians with lightweight encodings." arXiv:2312.04564. * Lu, Tao, et al. "Scaffold-gs: Structured 3d gaussians for view-adaptive rendering." CVPR2024. * Chen, Yihang, et al. "HAC: Hash-grid Assisted Context for 3D Gaussian Splatting Compression." arXiv:2403.14530.

Questions

Please kindly refer to the [Weaknesses].

Rating

4

Confidence

3

Soundness

2

Presentation

2

Contribution

1

Limitations

Please kindly refer to the [Weaknesses].

Authorsrebuttal2024-08-11

Dear reviewer, We hope our rebuttal can address your comments and concern. As the deadline for the author-reviewer discussion is approaching, would you please share your thoughts on our rebuttal and reconsider the review score? thanks a lot

Reviewer fvLt6/10 · confidence 3/52024-07-13

Summary

This paper proposes a method to optimally prune Gaussians that do not participate in the rendering or optimizing process of the 3D Gaussian Splatting 3D reconstruction algorithm. The key idea in this paper is to use a Gumbel-sigmoid function instead of the standard Sigmoid function to optimize a binary mask based on transmittance of each splatted Gaussian. Experimental results show that by doing so the number of Gaussians can be reduced without losing reconstruction quality, keeping a near optimal balance between quality and computational speed. As a result, higher frame rate can be obtained for rendering.

Strengths

- The idea to use the Gumbel-function is interesting because it allows to naturally binarize the input values while keeping differentiability. - Results show that higher frame rate can be achieved without losing accuracy or training time. The ablation study is also good because it shows the advantage of optimizing the mask values compared to hard thresholding.

Weaknesses

- Except for replacing the sigmoid function with the Gumbel-Sigmoid function it seems that there is no other novel technical point in the paper. From the results it seems that a threshold value of 0.6 would be better than the naïve 0.5. I am wondering how the ablation study changes if using different threshold. - In the experiments, no comparisons to other methods that focus on optimizing number compressing the model. For example, comparison with LightGaussian would be interesting. It seems from the Figure 4 that the proposed method performs slightly worse than mini splatting. - What is the meaning of the “pruning ratio” for the proposed method in Figure 4? It seems to be different for the different datasets. However, as far as I understood, the proposed method does not use a pruning ratio: l. 194 => pruning […] points with mask values of 0. This is confusing. - There is no significant gain in accuracy, training time or memory consumption. The only significant gain is in rendering speed. This is interesting but because original 3DGS is already quite fast, in practice I am not sure the impact is high. For example if using a simple threshold of 0.5 in Compact3D how does frame rate compares? - Writing can be brushed up. o L.176-178 is not correct. o In figure 4, it seems legends are wrong in the lines for Mini-splatting => blue is described as RadSplat. o Tables are not easy to read. Add some special fonts like bold or underline to emphasize best and second-best results.

Questions

- Clarify Figure 4 and explain what the meaning of the pruning ratio for the proposed method is. - Add more comparative experiments with other compact 3DGS methods.

Rating

6

Confidence

3

Soundness

3

Presentation

2

Contribution

2

Limitations

- Limitations are properly discussed in the conclusion section.

Authorsrebuttal2024-08-11

Dear reviewer, We hope our rebuttal can address your comments and concern. As the deadline for the author-reviewer discussion is approaching, would you please share your thoughts on our rebuttal and reconsider the review score? thanks a lot

Reviewer fvLt2024-08-12

I have read the authors' response and the other reviews. Thanks for the detailed answers. I had similar question as reviewer o7SR about W2. Figure 1-4 kind of gave the impression that the pruning ratio can be decided. It is now clarified that the parameter is the importance score, which results in some pruning ratio. And the proposed method allows to find the score to achieve desired pruning ratio. Other comments have also been addressed the rebuttal. I raise my rating to 6.

Reviewer o7SR6/10 · confidence 4/52024-07-14

Summary

Whereas other works set a fixed pruning ratio or threshold, this paper introduces a method to learn how to prune Gaussians from a scene while retaining high image fidelity. Different scenes have a drop in fidelity at different pruning percentages and the goal of this work is to remove the need to train the model multiple times while searching for a pruning percentage near, but before, that drop. Specifically this work uses a Gumbel-Sigmoid — which is a 2-class Gumbel softmax followed by a Sigmoid function — to enable a differentiable mask that can learn to prune Gaussians during training. This mask is then applied to set the opacity of each Gaussian. Gaussians where the mask goes to 0 are pruned. Evaluations are performed on 2 real world datasets — MipNerf 360 and Tanks & Temples — and one synthetic dataset — NeRF-Sythetic.

Strengths

S1: The writing in this work is quite clear. The methodology is straightforward to understand and the explanations appear to be detailed enough to be able to replicate experiments S2: This method seems useful to practitioners who are low on resources and want to quickly fit and compress a 3D-GS model.

Weaknesses

**Main Weaknesses:** W1: This paper uses the terms “optimal” and “best” to describe the pruning ratio found by learning. Specifically a soft definition of the optimal pruning ratio is given in line 41: “minimize the number of Gaussian points while keeping the rendering quality.” But this method does not “keep” the rendering quality — image quality metrics decrease after pruning. Generally the terms “optimal” and “best” are used to describe maximal or minimal points in some meaningful measure. In the domain of pruning 3D-GS models, a more appropriate definition of optimal pruning would be the highest possible compression that can be applied to the model while retaining rendering quality to some percent of the original rendering quality. W2: It would be helpful to also provide analysis on pruning threshold vs. image quality metrics. It’s possible that there exists a pruning threshold for RadSplat and/or Mini-Splatting that prunes a large percent of Gaussians while retaining image quality. W3: Table 2: Minutes isn’t a great metric here. FLOPs or MACs over training would be a lot better. Other works that want to compare to this method will need to use exactly this manuscript’s architecture to replicate its results. **Other minor considerations:** MC1: There are still several hyper parameters that haven’t been ablated. This is mostly fine, but it would be good to know how big of an impact these parameters have: - Gumbel $\tau$ - $\lambda_m$ - Applying the pruning earlier/later than 20000 epochs MC2: It would be interesting to ablate equation 8 by removing $S_i$. Can you simply just learn $m_i$ parameter without computing an importance score? **Cosmetic issues:** C1: line 46, line 50, line 62: could -> can C2: $T$ in eq. 7 should be a $\tau$ C3: Tables 1 and 2, move the metric to the first column for clarity (like tables 4 and 5)

Questions

My biggest concern for this manuscript is that the use of terms “optimal” and “best” is misleading. This method can find a good pruning ratio in one-shot, but it’s not clear that this ratio is optimal. I’m rating this manuscript a borderline reject for now because I strongly think this wording needs to change. I will raise my score if the authors can provide a detailed description of how they will update this wording in their rebuttal. I would also like to see W2 and W3 addressed, but these are likely computationally expensive and aren’t as important to me as W1.

Rating

6

Confidence

4

Soundness

3

Presentation

3

Contribution

3

Limitations

Once the wording in this manuscript is changed, I think the main limitations will be addressed.

Authorsrebuttal2024-08-12

# W1 We appreciate your comment. we will change to learned effective pruning ratio. # W2 We thank you for your further clarification. Please see our detailed response below: ***"It’s possible that for one scene a score may be at the 40% pruning ratio whereas that same score may be at the 60% pruning ratio in a different scene."*** We totally agree with the reviewer. To justify such, we conducted extra experiments to get the pruning ratio while fixing the pruning threshold of RadSplat defined importance score. The results are listed in below table. It can be clearly seen that the same threshold score for different scenes lead to different pruning ratios for different scenes. This actually further justifies the importance of our work, as it implies that different scenes indeed need the "learned effective pruning ratio'' to optimize tradeoff between compression ratio and rendering quality. Due to the fact that it is a most common setting in prior works to measure the rendering image quality with target pruning ratios, for fair comparison, we design our framework to learn an effective pruning ratio for each scene, without multiple rounds of manual tuning of pruning ratios as used in prior works | Threhold | Bicycle | Bonsai | Counter | Kitchen | Room | Stump | Garden | Flowers | Treehill | |:-----:|:-------:|:------:|:-------:|:-------:|:------:|:------:|:------:|:-------:|:--------:| | 0.025 | 50.88% | 43.79% | 47.86% | 38.46% | 52.73% | 51.07% | 48.05% | 50.65% | 41.26% | ***"It’s entirely possible that a specific score may provide a better threshold across scenes because it likely will not corespondent to a specific pruning ratio, but will be a measure of the confidence in a particular Gaussian."*** This is a very interesting point. We conducted such experiments on different scenes using different threshold values, based on RadSplat pruning importance score and the results are shown in figure below. It can be seen that, in the kitchen scene, the PSNR starts to drop significantly when the threshold is larger than 0.2. However, in the room scene, such cutoff point is around 0.15. It shows different scenes have different sensitivity against pruning threshold. [Kitchen](https://drive.google.com/uc?export=view&id=1dTP7mA_MIw_ROhxstItbThJyI6m-Sr3L) [Room](https://drive.google.com/uc?export=view&id=1qU1gwcNFo6hwcDOC7qch71YZcTcET0rU) More interestingly, since our proposed method learns the effective pruning ratio for different scenes during 3DGS model construction, we discovered that it naturally captures such unique cutoff threshold score for different scenes. # MC2 We appreciate you for sharing this valuable insight. In equation 8, if we only use $m_i$, then we could regard opacity ($o$) as the importance score. It is also true that the opacity itself could be one type of importance score. Our objective is to propose a method that is general, learnable, and orthogonal to other prior or potentially future works exploring different definitions of effective 3DGS pruning importance scores.

Reviewer o7SR2024-08-08

W1: Thanks for the detailed reply. I think it’s fine if you include a different adjective to sell this method, just not “optimal”. For instance I think it’s totally fine to claim this is a “learned *effective* pruning ratio”. W2: I think there may be some confusion here. I’m referring to the actual score computed by these methods, $s \in \mathbb{R}$. A score is computed per Gaussian, then these scores are ordered to prescribe, for a given pruning ratio $x$%, which of them should be pruned. It’s not possible, given only the Gaussian’s ordering (i.e., the pruning ratio) to obtain the original scores. This criticism was meant to call out that all plots are relative to this percentage, but not the actual scores. It’s possible that for one scene a score $s$ may be at the $40$% pruning ratio whereas that *same* score may be at the $60$% pruning ratio in a different scene. It’s entirely possible that a specific score may provide a better threshold across scenes because it likely will not corespondent to a specific pruning ratio, but will be a measure of the confidence in a particular Gaussian. W3: I agree other methods report minutes. I am not overly concerned by this and don’t require the authors to run this. Note: by FLOPs above I mean “floating point operations”, not FLOPS “floating point operations per second”. MC2: This comment was that it would be interesting to ablate how well this method performs *without* the use of an auxiliary pruning score. It’s possible that $m_i$ may prove to be a good pruning score on its own. I suspect that $S_i$ may be only providing a good “initialization”. Overall: I still find this method interesting and my primary concern, the wording of the manuscript, has been addressed.

Reviewer o7SR2024-08-11

Apologies to the authors, I meant to raise my score to a 5 earlier based on the rebuttal. I'm happy to raise my score to a 6 if my concern regarding W2 can be addressed.

Authorsrebuttal2024-08-12

We appreciate all the reviewer’s valuable comments and insights, which have genuinely helped improve our work. We would be grateful if the reviewer could consider raising the final score.

Reviewer 1UP22024-08-12

While the authors addressed the most of my minor concerns, I still insist on my opinion about the key point about the "optimal prune ratio". First, according to the experiments, most of the "optimal" prune ratios are similar, around 0.7-0.8 without specific pattern. It is a very narrow band. In such cases, the "learning" task for the prune ratio may be a trivial task since the prune ratio is easy to approach. Just set a 0.75 value may be good enough for most of the experiments in Fig.4 & 6. Second, even though we assume the task is non-trivial, the experiment cannot prove the proposed strategy is efficient enough. For example, in the Bicycle-Redsplat subfigures of Fig.6, we cannot tell the prune ratio ~0.6(LP-3DGS) is better than 0.7 or 0.8. Most of the experiments are with the same problem: no convincing metric to examine the efficiency of the proposed method. Actually, I think it is the problem of this task as I said in the first paragraph. In other words, if all the "answers" fall in around 0.6-0.8, and we cannot find the concrete value which is significantly better than others, the results are too vague to support the claims. Therefore for now I will keep my score.

Authorsrebuttal2024-08-13

Dear reviewer, we hope our rebuttal has effectively addressed your comments and concerns. As the deadline for the author-reviewer discussion is tomorrow, we kindly request your prompt feedback on our response and ask you to reconsider the review score. Thank you very much for your time and consideration.

Authorsrebuttal2024-08-13

last day of rebuttal response

Dear reviewer CWFJ, today is the last day of reviewer-author discussion period. We hope you have checked our rebuttal and found it could address your previous concerns. We would be happy to provide clarification of our work if you have any further comment. thanks for your review.

Program Chairsdecision2024-09-25

Decision

Accept (poster)

© 2026 NYSGPT2525 LLC