Research map / Video / Video generation
Gaussian Variation Field Diffusion for High-fidelity Video-to-4D Synthesis
Bowen Zhang, Sicheng Xu, Chuxin Wang, Jiaolong Yang, Feng Zhao, Dong Chen, Baining Guo
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 mesh-guided loss (ablation) | GVFDiffusion VAE · VAE ablation | PSNR: 23.25 -> 26.17 | Improved |
| add mesh-guided loss (ablation) | GVFDiffusion VAE · VAE ablation | LPIPS: 0.0678 -> 0.0544 | Improved |
| use motion-aware displacement query via mesh-guided interpolation (ablation) | GVFDiffusion VAE · VAE ablation | PSNR: 26.17 -> 28.58 | Improved |
| add color and opacity variations to variation fields (ablation) | GVFDiffusion VAE · VAE ablation | PSNR: 28.58 -> 29.28 | Improved |
| add color and opacity variations (ablation) | GVFDiffusion VAE · VAE ablation | LPIPS: 0.0478 -> 0.0439 | Improved |
| add positional embeddings from canonical GS positions (ablation) | GVFDiffusion diffusion model · Video-to-4D test set | PSNR: 17.86 -> 18.47 | Improved |
| add positional embeddings from canonical GS positions (ablation) | GVFDiffusion diffusion model · Video-to-4D test set | FVD: 547.20 -> 476.83 | Improved |
| jointly finetune canonical GS decoder D_GS (ablation) | GVFDiffusion VAE · VAE ablation | PSNR: 28.80 -> 29.28 | Improved |
| set K=8 neighbors rather than K=16 (ablation) | mesh-guided interpolation · VAE ablation | PSNR: 28.38 -> 29.28 | Improved |
| set K=8 neighbors rather than K=1 (ablation) | mesh-guided interpolation · VAE ablation | PSNR: 28.22 -> 29.28 | Improved |
| set beta=7.0 rather than 10.0 (ablation) | mesh-guided interpolation · VAE ablation | PSNR: 28.55 -> 29.28 | Improved |
| use GVFDiffusion instead of L4GM | video-to-4D generation · Video-to-4D test set | FVD: 529.10 -> 476.83 | 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.
- Consistent4D — Consistent4D: Consistent 360° Dynamic Object Generation from Monocular Video (ICLR 2024) · better on 2 of 2 shared results
- STAG4D — STAG4D: Spatial-Temporal Anchored Generative 4D Gaussians (ECCV 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
- Align Your Gaussians: Text-to-4D with Dynamic 3D Gaussians and Composed Diffusion Models (CVPR 2024 · similarity 0.93)
- Diffusion4D: Fast Spatial-temporal Consistent 4D generation via Video Diffusion Models (NeurIPS 2024 · similarity 0.89)
- Generative Sparse-View Gaussian Splatting (CVPR 2025 · similarity 0.89)
- Wonderland: Navigating 3D Scenes from a Single Image (CVPR 2025 · similarity 0.89)
- DiffSplat: Repurposing Image Diffusion Models for Scalable Gaussian Splat Generation (ICLR 2025 · similarity 0.89)
- L4GM: Large 4D Gaussian Reconstruction Model (NeurIPS 2024 · 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.
