Research map / 3D vision / 3D generation
Baking Gaussian Splatting into Diffusion Denoiser for Fast and Scalable Single-stage Image-to-3D Generation and Reconstruction
Yuanhao Cai, He Zhang, Kai Zhang, Yixun Liang, Mengwei Ren, Fujun Luan, Qing Liu, Soo Ye Kim, Jianming Zhang, Zhifei Zhang, Yuqian Zhou, Yulun Zhang and 3 more
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 the 3D Gaussian diffusion framework to the baseline denoiser (ablation) | DiffusionGS baseline · GSO | PSNR / FID: 17.63 -> 20.57 PSNR; 118.31 -> 47.86 FID | Improved |
| add point distribution loss Lpd for warm-up (ablation) | DiffusionGS · GSO | PSNR / FID: 20.57 -> 20.94 PSNR; 47.86 -> 28.41 FID | Improved |
| add scene-object mixed training (without RPPC) (ablation) | DiffusionGS · GSO | PSNR / FID: 20.94 -> 21.73 PSNR; 28.41 -> 17.79 FID | Improved |
| add Reference-Point Plucker Coordinate (RPPC) camera conditioning (ablation) | DiffusionGS · GSO | PSNR / FID: 21.73 -> 22.07 PSNR; 17.79 -> 11.52 FID | Improved |
| add scene-object mixed training for scene reconstruction (ablation) | DiffusionGS · RealEstate10K | PSNR / FID: 0.61 dB in PSNR and 10.53 in FID | Improved |
| add RPPC camera conditioning for scene reconstruction (ablation) | DiffusionGS · RealEstate10K | PSNR / FID: 0.28 dB in PSNR and 7.09 in FID | Improved |
| replace 2D multi-view diffusion plus post-hoc Gaussian fitting (PhotoNVS) with single-stage DiffusionGS | single-view scene reconstruction · RealEstate10K | PSNR / FID: 15.31 -> 21.63 PSNR; 28.30 -> 15.87 FID | Improved |
| use DiffusionGS instead of DMV3D for object generation | single-view object generation · ABO / GSO | PSNR / FID: 2.2/1.25 dB in PSNR and 23.25/21.96 in FID | Improved |
| use DiffusionGS instead of Flash3D (which uses a depth estimator) | single-view scene reconstruction · RealEstate10K | PSNR / FID: 20.29 -> 21.63 PSNR; 35.03 -> 15.87 FID | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DMV3D — DMV3D: Denoising Multi-view Diffusion Using 3D Large Reconstruction Model (ICLR 2024) · better on 6 of 6 shared results
- GS-LRM — GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting (ECCV 2024) · better on 4 of 4 shared results
- Splatter-Image — Splatter Image: Ultra-Fast Single-View 3D Reconstruction (CVPR 2024) · better on 2 of 2 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
- Wonderland: Navigating 3D Scenes from a Single Image (CVPR 2025 · similarity 0.91)
- Taming Video Diffusion Prior with Scene-Grounding Guidance for 3D Gaussian Splatting from Sparse Inputs (CVPR 2025 · similarity 0.91)
- Denoising Diffusion via Image-Based Rendering (ICLR 2024 · similarity 0.91)
- GaussianPro: 3D Gaussian Splatting with Progressive Propagation (ICML 2024 · similarity 0.90)
- Coherent 3D Scene Diffusion From a Single RGB Image (NeurIPS 2024 · similarity 0.90)
- SplatFormer: Point Transformer for Robust 3D Gaussian Splatting (ICLR 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.
