Research map / Video / Video generation
Geometry-Aware Single-Image 4D Synthesis via Dense Trajectory Generation
Yanran Zhang, Ziyi Wang, Wenzhao Zheng, Zheng Zhu, Jie Zhou, Jiwen Lu
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 depth-guided motion normalization (ablation) | 4D-STraG · VBench Group III | Aesthetic quality: 0.4672 -> 0.4820 | Improved |
| add depth latents as conditioning (ablation) | 4D-STraG · VBench Group III | Dynamic degree: 0.8500 -> 0.9000 | Improved |
| add Motion Perception Module (ablation) | 4D-STraG · VBench Group III | Dynamic degree: 0.8500 -> 0.9000 | Improved |
| use patch-level motion features instead of global CLS token (ablation) | MPM · VBench Group III | Consistency: 0.8743 -> 0.8702 | Mixed |
| train on 60k filtered samples instead of 1k subset (ablation) | 4D-STraG · VBench Group III | Consistency: 0.8567 -> 0.8702 | Improved |
| filter training data instead of using unfiltered 60k samples (ablation) | 4D-STraG · VBench Group III | Aesthetic quality: 0.4654 -> 0.4820 | Improved |
| add random noise to training trajectories (ablation) | 4D-STraG · VBench Group III | Consistency: 0.8702 -> 0.8685 | No clear effect |
| use geometry-conditioned trajectory generation instead of Gen3C | MoGe4D · VLM 4D consistency Group III | Average VLM score: 2.22 -> 3.76 | Improved |
| MoGe4D vs 4Real on subject consistency | MoGe4D · VBench Group I | Subject consistency: 0.9329 -> 0.8752 | Hurt |
| MoGe4D vs DimensionX on motion smoothness | MoGe4D · VBench Group II | Motion smoothness: 0.9827 -> 0.9760 | Hurt |
| use direct trajectory generation instead of generate-then-track pipelines | MoGe4D · Group ii | Trajectory error: 0.197 -> 0.042 (Gen3C) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- Free4D — Free4D: Tuning-free 4D Scene Generation with Spatial-Temporal Consistency (ICCV 2025) · better on 3 of 4 shared results
- 4Real — 4Real: Towards Photorealistic 4D Scene Generation via Video Diffusion Models (NeurIPS 2024) · better on 0 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
- Motion 3-to-4: 3D Motion Reconstruction for 4D Synthesis (CVPR 2026 · similarity 0.93)
- Diffusion4D: Fast Spatial-temporal Consistent 4D generation via Video Diffusion Models (NeurIPS 2024 · similarity 0.91)
- Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding (NeurIPS 2025 · similarity 0.90)
- CAT4D: Create Anything in 4D with Multi-View Video Diffusion Models (CVPR 2025 · similarity 0.90)
- Controlling Space and Time with Diffusion Models (ICLR 2025 · similarity 0.89)
- Robust Multi-Object 4D Generation for In-the-wild Videos (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.
