Research map / 3D vision / Gaussian splatting
Geometry-Aware Style Transfer in 3D Gaussian Splatting
Min Hyeok Bang, Jun Hyeong Kim, Seung-Wook Kim, Se-Ho Lee
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 |
|---|---|---|---|
| use decoupled alternating color/geometry optimization instead of joint-step (ablation) | geometry-aware 3DGS style transfer · 72 scene-style combinations | SIFID: 1.4220 -> 1.1736 | Improved |
| add geometry updates to color-only optimization (ablation) | geometry-aware 3DGS style transfer · 72 scene-style combinations | SIFID: 1.5392 -> 1.1736 | Improved |
| add color updates to geometry-only optimization (ablation) | geometry-aware 3DGS style transfer · 72 scene-style combinations | SIFID: 2.5496 -> 1.1736 | Improved |
| decoupled instead of joint-step (ablation) | geometry-aware 3DGS style transfer · long-range consistency | LPIPS: 0.122 -> 0.114 | Improved |
| decoupled instead of only geometry (ablation) | geometry-aware 3DGS style transfer · long-range consistency | LPIPS: 0.113 -> 0.114 | No clear effect |
| increase geometry update steps from 0 to 90 of 100 per cycle (ablation) | geometry-aware 3DGS style transfer · 72 scene-style combinations | SIFID: 1.5244 -> 1.1736 | Improved |
| use all 100 geometry updates instead of 90 (ablation) | geometry-aware 3DGS style transfer · 72 scene-style combinations | SIFID: 1.1736 -> 1.1885 | Hurt |
| add GCFM loss (lambda_GC 0.0 to 2.0) (ablation) | geometry-aware 3DGS style transfer · 72 scene-style combinations | SIFID: 1.5718 -> 1.1736 | Improved |
| vary lambda_GC from 1.0 to 5.0 (ablation) | geometry-aware 3DGS style transfer · 72 scene-style combinations | SIFID: 1.1683 -> 1.1841 | No clear effect |
| add contrastive objective to GCFM (ablation) | geometry-aware 3DGS style transfer · depth map structure | qualitative | Improved |
| use multi-modal color/depth/edge features instead of single modality (ablation) | geometry-aware 3DGS style transfer · depth map structure | qualitative | Improved |
| compare against StylizedGS on consistency | Ours vs StylizedGS · long-range consistency | RMSE: 0.062 -> 0.084 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- StylizedGS — StylizedGS: Controllable Stylization for 3D Gaussian Splatting (TPAMI 2025) · better on 1 of 5 shared results
- SGSST — SGSST: Scaling Gaussian Splatting Style Transfer (CVPR 2025) · better on 4 of 4 shared results
- CLIPGaussian — CLIPGaussian: Universal and Multimodal Style Transfer Based on Gaussian Splatting (NeurIPS 2025) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Morpheus: Text-Driven 3D Gaussian Splat Shape and Color Stylization (CVPR 2025 · similarity 0.92)
- Multi-StyleGS: Stylized Gaussian Splatting with Multiple Styles (AAAI 2025 · similarity 0.91)
- SGSST: Scaling Gaussian Splatting Style Transfer (CVPR 2025 · similarity 0.90)
- ReGS: Reference-based Controllable Scene Stylization with Gaussian Splatting (NeurIPS 2024 · similarity 0.90)
- CLIPGaussian: Universal and Multimodal Style Transfer Based on Gaussian Splatting (NeurIPS 2025 · similarity 0.89)
- Textured Gaussians for Enhanced 3D Scene Appearance Modeling (CVPR 2025 · similarity 0.89)
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.
