Research map / 3D vision / 3D generation
4D-VGGT: A SpatioTemporal Foundation Model for Dynamic Scene Geometry Estimation
Haonan Wang, Hanyu Zhou, Haoyue Liu, Luxin Yan
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 Adaptive Visual Grid (ablation) | 4D-VGGT without CVGF/CTLF · camera pose ablation | ATE: 0.045 -> 0.030 | Improved |
| add Cross-View Global Fusion (with AVG) (ablation) | 4D-VGGT · camera pose ablation | ATE: 0.030 -> 0.017 | Improved |
| add Cross-View Global Fusion (with AVG) (ablation) | 4D-VGGT · point map ablation | Acc.: 0.215 -> 0.085 | Improved |
| add Cross-Time Local Fusion (with AVG) (ablation) | 4D-VGGT · point map ablation | Acc.: 0.215 -> 0.091 | Improved |
| add Cross-Time Local Fusion (with AVG) (ablation) | 4D-VGGT · point tracking ablation | L-12: 24.76 -> 36.12 | Improved |
| add Cross-Time Local Fusion (with AVG) (ablation) | 4D-VGGT · camera pose ablation | ATE: 0.030 -> 0.030 | No clear effect |
| use two-stage training instead of single-stage (ablation) | 4D-VGGT · point map ablation | Acc.: 0.093 -> 0.065 | Improved |
| add random sampling strategy (ablation) | 4D-VGGT two-stage · tracking ablation | L-12: 33.49 -> 36.12 | Improved |
| add spatial attention mask (ablation) | 4D-VGGT · camera pose ablation | ATE: 0.024 -> 0.017 | Improved |
| add temporal attention mask (ablation) | 4D-VGGT · tracking ablation | L-12: 29.53 -> 36.12 | Improved |
| feed temporal features in addition to spatial features to camera head (ablation) | 4D-VGGT · camera pose | ATE: 0.039 -> 0.045 | Hurt |
| feed spatial features in addition to temporal features to tracking head (ablation) | 4D-VGGT · point tracking | L-12: 33.39 -> 33.87 | No clear effect |
Closest papers by meaning
- IGGT4D: Streaming 4D Instance-Grounded Geometry Transformer (NeurIPS 2026 · similarity 0.89)
- VGGT: Visual Geometry Grounded Transformer (CVPR 2025 · similarity 0.89)
- Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting (ICLR 2024 · similarity 0.88)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.88)
- TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention (NeurIPS 2026 · similarity 0.87)
- Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding (NeurIPS 2025 · similarity 0.87)
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.
