Research map / 3D vision / Gaussian splatting
Generative Gaussian Splatting: Generating 3D Scenes with Video Diffusion Priors
Katja Schwarz, Norman Müller, Peter Kontschieder
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 3D representation to the pose-conditional video diffusion model (ablation) | GGS vs Ours-No3D · RealEstate10K and ScanNet++ (3D scenes) | FID: ~20% | Improved |
| add novel-view reconstruction loss L_nv (ablation) | GGS · RealEstate10K (two-view conditional) | PSNR and TSED | Improved |
| predict splats in feature space with v-prediction instead of latent space with x0-prediction (ablation) | GGS · RealEstate10K (two-view conditional) | image quality and consistency | Improved |
| approximate depth with a Gaussian distribution instead of discrete depth buckets (ablation) | GGS with depth supervision · RealEstate10K (two-view conditional) | all metrics | Improved |
| use convolutional instead of transformer 2D upsampler in 3D decoder (ablation) | GGS 3D decoder · RealEstate10K (two-view conditional) | image quality and consistency | Improved |
| replace 2D decoder with 3D decoder (ablation) | GGS · RealEstate10K (two-view conditional) | 3D consistency vs fidelity | Mixed |
| train the 3D decoder on a frozen diffusion model without 3D representation (Ours-No3D) instead of GGS (ablation) | 3D decoder · RealEstate10K (two-view conditional) | quality and consistency | Hurt |
| add depth supervision (ablation) | GGS · ScanNet++ and RealEstate10K | consistency | Improved |
| add 3D representation to the model (ablation) | GGS vs Ours-No3D · ScanNet++ (single view) | TSED | Improved |
Closest papers by meaning
- Generative Sparse-View Gaussian Splatting (CVPR 2025 · similarity 0.94)
- FewViewGS: Gaussian Splatting with Few View Matching and Multi-stage Training (NeurIPS 2024 · similarity 0.93)
- GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025 · similarity 0.93)
- 3DGS-Enhancer: Enhancing Unbounded 3D Gaussian Splatting with View-consistent 2D Diffusion Priors (NeurIPS 2024 · similarity 0.93)
- Taming Video Diffusion Prior with Scene-Grounding Guidance for 3D Gaussian Splatting from Sparse Inputs (CVPR 2025 · similarity 0.92)
- DiffSplat: Repurposing Image Diffusion Models for Scalable Gaussian Splat Generation (ICLR 2025 · similarity 0.92)
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.
