Research map / 3D vision / Gaussian splatting
A3GS: Arbitrary Artistic Style into Arbitrary 3D Gaussian Splatting
Zhiyuan Fang, Rengan Xie, Xuancheng Jin, Qi Ye, Wei Chen, Wenting Zheng, Rui Wang, Yuchi Huo
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 |
|---|---|---|---|
| replace the 3D-GCN autoencoder layers with an MLP (ablation) | A3GS · qualitative stylization | visual quality | Hurt |
| replace the 3D-GCN layers with a simple GCN without learnable kernel shape (ablation) | A3GS · qualitative stylization | content preservation | Hurt |
| apply the AdaIN stylizer at each layer instead of once in the latent space (ablation) | A3GS · qualitative stylization | fine detail preservation | Hurt |
| add background masking to the style loss (ablation) | A3GS · qualitative stylization | color accuracy | Improved |
| add the autoencoder (latent-space stylization) rather than stylizing colors directly (ablation) | A3GS · qualitative stylization | style features | Improved |
| use feed-forward A3GS instead of per-scene optimization (StyleSplat) | 3DGS stylization · TNT dataset | style transfer time: around 300 seconds -> 10 seconds | Improved |
| use A3GS instead of StyleGaussian for scene scale | 3DGS stylization · TNT dataset | max Gaussian primitives: 0.3 million -> 15 million per batch | Improved |
| use A3GS instead of StyleRF | 3D stylization · short-range consistency | LPIPS: 0.092 -> 0.033 | Improved |
| use A3GS instead of StyleSplat | 3D stylization · long-range consistency | LPIPS: 0.052 -> 0.061 | Hurt |
| use A3GS instead of StyleGaussian | 3D stylization · short-range consistency | RMSE: 0.033 -> 0.029 | Improved |
| use A3GS instead of StyleGaussian | 3D stylization · user study | Content Preservation: 3.3 -> 4.5 | Improved |
| use A3GS instead of StyleSplat | 3D stylization · user study | Style Consistency: 3.9 -> 4.1 | Improved |
Closest papers by meaning
- Multi-StyleGS: Stylized Gaussian Splatting with Multiple Styles (AAAI 2025 · similarity 0.90)
- ReGS: Reference-based Controllable Scene Stylization with Gaussian Splatting (NeurIPS 2024 · similarity 0.89)
- Styl3R: Instant 3D Stylized Reconstruction for Arbitrary Scenes and Styles (NeurIPS 2025 · similarity 0.89)
- SGSST: Scaling Gaussian Splatting Style Transfer (CVPR 2025 · similarity 0.88)
- Morpheus: Text-Driven 3D Gaussian Splat Shape and Color Stylization (CVPR 2025 · similarity 0.88)
- CLIPGaussian: Universal and Multimodal Style Transfer Based on Gaussian Splatting (NeurIPS 2025 · similarity 0.87)
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.
