Research map / 3D vision / Gaussian splatting
DefenseSplat: Enhancing the Robustness of 3D Gaussian Splatting via Frequency-Aware Filtering
Yiran Qiao, Yiren Lu, Yunlai Zhou, Rui Yang, Linlin Hou, Yu Yin, Jing Ma
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add ReLU scale regularization loss (ablation) | DefenseSplat · Mip-NeRF 360 (attacked, eps 16/255) | PSNR: 27.32 -> 27.47 | Improved |
| add ReLU scale regularization loss (ablation) | DefenseSplat · Mip-NeRF 360 (attacked, eps 16/255) | GPU memory (MB): 12567 -> 11400 | Improved |
| add ReLU scale regularization loss (ablation) | DefenseSplat · Tanks-and-Temples (8 scenes, attacked, eps 16/255) | PSNR: 26.29 -> 26.42 | Improved |
| add ReLU scale regularization loss (ablation) | DefenseSplat · LLFF (attacked, eps 16/255) | PSNR: 27.67 -> 27.63 | No clear effect |
| add ReLU scale loss under strong attack (ablation) | DefenseSplat · Tanks-and-Temples (attacked, eps 64/255) | PSNR: 19.27 -> 19.38 | Improved |
| apply DefenseSplat to attacked inputs instead of plain 3DGS | 3DGS · Mip-NeRF 360 (attacked, eps 16/255) | PSNR: 25.18 -> 27.32 | Improved |
| apply DefenseSplat to attacked inputs instead of plain 3DGS | 3DGS · Mip-NeRF 360 (attacked, eps 16/255) | GPU memory (MB): 20965 -> 12567 | Improved |
| increase attack strength to 64/255 with DefenseSplat | DefenseSplat · Tanks-and-Temples | PSNR: 24.89 -> 19.27 | Hurt |
| apply DefenseSplat to clean input | 3DGS · Mip-NeRF 360 (clean input) | PSNR: 29.06 -> 28.58 | Hurt |
| apply DefenseSplat to clean input | 3DGS · LLFF (clean input) | PSNR: 31.29 -> 28.39 | Hurt |
| apply DefenseSplat to clean input | 3DGS · Mip-NeRF 360 (clean input) | GPU memory (MB): 20965 -> 11012 | Improved |
| add fine-grained cross-view consistency filter (ablation) | DefenseSplat + ReLU · Mip-NeRF 360 (basic vs fine-grained filter) | PSNR: 27.47 -> 27.49 | No clear effect |
Closest papers by meaning
- RaPTGS: Render-Agnostic Post-Training Compression of 3D Gaussian Splatting (ECCV 2026 · similarity 0.92)
- D$^2$GS: Depth-and-Density Guided Gaussian Splatting for Stable and Accurate Sparse-View Reconstruction (ICLR 2026 · similarity 0.91)
- RogSplat: Robust Gaussian Splatting via Generative Priors (ICCV 2025 · similarity 0.91)
- From Chaos to Clarity: 3DGS in the Dark (NeurIPS 2024 · similarity 0.90)
- SplatFormer: Point Transformer for Robust 3D Gaussian Splatting (ICLR 2025 · similarity 0.90)
- Frequency-Aware Density Control via Reparameterization for High-Quality Rendering of 3D Gaussian Splatting (AAAI 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
