Research map / 3D vision / Depth and stereo
Syn4D: A Multiview Synthetic 4D Dataset
Zeren Jiang, Yushi Lan, Yihang Luo, Yufan Deng, Zihang Lai, Edgar Sucar, Christian Rupprecht, Iro Laina, Diane Larlus, Chuanxia Zheng, Andrea Vedaldi
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 |
|---|---|---|---|
| train the geometry-aware NVS model on Syn4D instead of Kubric (ablation) | ReCamMaster-based multiview diffusion · Syn4D NVS benchmark | FVD: 631 -> 452 | Improved |
| train the geometry-aware NVS model on Syn4D instead of Kubric (ablation) | ReCamMaster-based multiview diffusion · Syn4D NVS benchmark | FVD-P: 229 -> 139 | Improved |
| train the geometry-aware NVS model on Syn4D instead of Kubric (ablation) | ReCamMaster-based multiview diffusion · Syn4D NVS benchmark | ATE: 0.064 -> 0.070 | No clear effect |
| train the geometry-aware NVS model on Syn4D instead of Kubric (ablation) | ReCamMaster-based multiview diffusion · ReCamMaster benchmark (zero-shot) | FVD-V-P: 1307 -> 397.4 | Improved |
| match clip count (2048) and frames (49) between Kubric and Syn4D training (ablation) | Ours (Syn4D) vs Ours (Kubric) · ReCamMaster benchmark (matched budget) | FVD-V: 1964 -> 1566 | Improved |
| condition on ground-truth point map instead of DA3-estimated point map; Syn4D vs Kubric training (ablation) | geometry-aware NVS model · Syn4D NVS benchmark | COR: 0.236 -> 0.124 | Improved |
| co-train 4RC with Syn4D | 4RC · Syn4D Soviet dense tracking | APD: 58.35 -> 74.46 | Improved |
| co-train 4RC with Syn4D | 4RC · Syn4D Warehouse dense tracking | EPE: 0.3302 -> 0.1915 | Improved |
| co-train 4RC with Syn4D | 4RC · ADT sparse 3D tracking | APD: 87.82 -> 89.44 | Improved |
| co-train 4RC with Syn4D | 4RC · Sintel camera pose | ATE: 0.144 -> 0.076 | Improved |
| co-train 4RC with Syn4D | 4RC · TUM-dynamics camera pose | ATE: 0.010 -> 0.012 | Hurt |
| co-train 4RC with Syn4D | 4RC · Sintel video depth (scale alignment) | delta<1.25: 62.2 -> 74.6 | Improved |
Closest papers by meaning
- CAT4D: Create Anything in 4D with Multi-View Video Diffusion Models (CVPR 2025 · similarity 0.91)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.89)
- Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding (NeurIPS 2025 · similarity 0.89)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.89)
- IGGT4D: Streaming 4D Instance-Grounded Geometry Transformer (NeurIPS 2026 · similarity 0.89)
- Uni4D: Unifying Visual Foundation Models for 4D Modeling from a Single Video (CVPR 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.
