Research map / 3D vision / 3D generation
VideoRFSplat: Direct Scene-Level Text-to-3D Gaussian Splatting Generation with Flexible Pose and Multi-View Joint Modeling
Hyojun Go, Byeongjun Park, Hyelin Nam, Byung-Hoon Kim, Hyungjin Chung, Changick Kim
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 asynchronous sampling (delta=0.2) (ablation) | VideoRFSplat vanilla sampling · T3Bench | BRISQUE: 34.9 -> 32.2 | Improved |
| add asynchronous sampling (delta=0.2) (ablation) | VideoRFSplat vanilla sampling · T3Bench | NIQE: 5.32 -> 4.20 | Improved |
| add modified CFG to asynchronous sampling (delta=0.2) (ablation) | VideoRFSplat · T3Bench | BRISQUE: 34.1 -> 32.2 | Improved |
| increase delta from 0.2 to 0.5 with async sampling (ablation) | VideoRFSplat · T3Bench | BRISQUE: 32.2 -> 34.4 | Hurt |
| add asynchronous sampling with 200 steps (ablation) | VideoRFSplat · T3Bench | BRISQUE: 33.8 -> 31.8 | Improved |
| add timestep division in training (Eq. 4) (ablation) | VideoRFSplat vanilla sampling · T3Bench | BRISQUE: 35.3 -> 34.9 | No clear effect |
| use dual-stream architecture instead of channel concatenation (ablation) | joint pose-image generation · T3Bench | NIQE: 7.83 -> 6.19 | Improved |
| use dual-stream architecture instead of channel concatenation (ablation) | joint pose-image generation · T3Bench | CLIPScore: 29.8 -> 32.5 | Improved |
| use VideoRFSplat without SDS++ instead of SplatFlow | text-to-3DGS · MVImgNet val | FID-10K: 34.85 -> 30.33 | Improved |
| add SDS++ refinement to SplatFlow | SplatFlow · MVImgNet val | FID-10K: 34.85 -> 35.46 | Hurt |
| add SDS++ refinement to Director3D | Director3D · DL3DV val | FID-2.4K: 88.44 -> 95.88 | Hurt |
| use VideoRFSplat instead of CameraCtrl | camera-conditioned generation · RealEstate10K | FID-8K: 62.97 -> 43.07 | Improved |
Later papers that compared with this method
1 later accepted paper reports VideoRFSplat as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Text-to-3D by Stitching a Multi-view Reconstruction Network to a Video Generator (ICLR 2026 · better on 5 of 5 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Director3D — Director3D: Real-world Camera Trajectory and 3D Scene Generation from Text (NeurIPS 2024) · better on 5 of 5 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
- SplatFlow: Multi-View Rectified Flow Model for 3D Gaussian Splatting Synthesis (CVPR 2025 · similarity 0.91)
- DiffSplat: Repurposing Image Diffusion Models for Scalable Gaussian Splat Generation (ICLR 2025 · similarity 0.89)
- No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images (ICLR 2025 · similarity 0.89)
- MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024 · similarity 0.89)
- SelfSplat: Pose-Free and 3D Prior-Free Generalizable 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
- KeyGS: A Keyframe-Centric Gaussian Splatting Method for Monocular Image Sequences (AAAI 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.
