Research map / 3D vision / Gaussian splatting
One-Shot Refiner: Boosting Feed-forward Novel View Synthesis via One-Step Diffusion
Yitong Dong, Qi Zhang, Minchao Jiang, Zhiqiang Wu, Qingnan Fan, Ying Feng, Huaqi Zhang, Hujun Bao, Guofeng Zhang
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 frequency-domain branch (DD-DPM) over CNN-only detail module (ablation) | detail perception module · DL3DV | FID: 62.60 -> 55.32 | Improved |
| add DD-DPM over CNN-only detail module (ablation) | detail perception module · DL3DV | MUSIQ: 64.95 -> 63.47 | Hurt |
| add CNN detail module to base (ablation) | base feed-forward model · DL3DV | PSNR: 20.08 -> 22.10 | Improved |
| add Gaussian features with a 2-layer CNN refiner (ablation) | DD-DPM model · DL3DV | FID: 55.32 -> 70.51 | Hurt |
| add Gaussian features with a 2-layer CNN refiner (ablation) | DD-DPM model · DL3DV | LPIPS: 0.22 -> 0.27 | Hurt |
| add standalone one-step SD refiner (ablation) | reconstruction model · DL3DV | FID: 70.51 -> 48.58 | Improved |
| add standalone one-step SD refiner (ablation) | reconstruction model · DL3DV | PSNR: 22.22 -> 22.17 | No clear effect |
| make SD refiner feature-guided with joint training (ablation) | SD refiner · DL3DV | FID: 48.58 -> 40.46 | Improved |
| make SD refiner feature-guided with joint training (ablation) | SD refiner · DL3DV | PSNR: 22.17 -> 22.67 | Improved |
| use feature-guided SD instead of MARINER enhancement | enhancement module · DL3DV | FID: 63.04 -> 40.46 | Improved |
| use feature-guided SD instead of Difix3D+ SD | enhancement module · DL3DV | FID: 50.30 -> 40.46 | Improved |
| upsample to 1024x1024 with Tile-VAE without retraining | One-Shot Refiner · DL3DV 1024x1024 | FID: 108.74 -> 41.74 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FLARE — FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Views (CVPR 2025) · better on 6 of 6 shared results
- LatentSplat — latentSplat: Autoencoding Variational Gaussians for Fast Generalizable 3D Reconstruction (ECCV 2024) · better on 1 of 1 shared result
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
- Generative Sparse-View Gaussian Splatting (CVPR 2025 · similarity 0.89)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.89)
- Novel View Synthesis from A Few Glimpses via Test-Time Natural Video Completion (NeurIPS 2025 · similarity 0.89)
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.89)
- 3DGS-Enhancer: Enhancing Unbounded 3D Gaussian Splatting with View-consistent 2D Diffusion Priors (NeurIPS 2024 · similarity 0.88)
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · similarity 0.88)
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.
