Research map / 3D vision / 3D generation
LaVR: Scene Latent Conditioned Generative Video Trajectory Re-Rendering using Large 4D Reconstruction Models
Mingyang Xie, Numair Khan, Tianfu Wang, Naina Dhingra, Seonghyeon Nam, Haitao Yang, Zhuo Hui, Christopher Metzler, Andrea Vedaldi, Hamed Pirsiavash, Lei Luo
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 CUT3R latent conditioning (with source pose) (ablation) | ReCamMaster-style video diffusion model · Cycle consistency | Cycle PSNR: 17.90 -> 20.74 | Improved |
| add CUT3R latent conditioning (ablation) | LaVR · Multi-view consistency | Multi-View: 6.832 -> 17.11 | Improved |
| add CUT3R latent conditioning (ablation) | LaVR · Target pose reconstruction | Abs(t) error: 19.70 -> 14.39 | Improved |
| add source camera pose conditioning on top of CUT3R latents (ablation) | LaVR · Cycle consistency | Cycle PSNR: 20.70 -> 20.74 | No clear effect |
| add source camera pose conditioning on top of CUT3R latents (ablation) | LaVR · Multi-view consistency | Multi-View: 16.08 -> 17.11 | Improved |
| add source camera pose conditioning on top of CUT3R latents (ablation) | LaVR · Target pose reconstruction | Abs(t) error: 16.93 -> 14.39 | Improved |
| add source pose only, without CUT3R latents (ablation) | ReCamMaster · Cycle consistency | Cycle PSNR: 17.75 -> 17.90 | No clear effect |
| condition on implicit CUT3R latents instead of re-rendered point clouds | video re-rendering · Cycle consistency (DL3DV) | Cycle PSNR: 14.84 -> 20.74 | Improved |
| condition on CUT3R latents vs unconditioned ReCamMaster | video re-rendering · VBench dynamic videos | Multi-view consistency: 5.975 -> 17.11 | Improved |
| CUT3R latent conditioning vs ReCamMaster on background consistency | video re-rendering · VBench dynamic videos | Background consistency: 92.76 -> 92.83 | No clear effect |
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.
- Gen3C — GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025) · better on 4 of 4 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
- Restage4D: Reanimating Deformable 3D Reconstruction from a Single Video (NeurIPS 2025 · similarity 0.88)
- GenFusion: Closing the Loop between Reconstruction and Generation via Videos (CVPR 2025 · similarity 0.88)
- DreamScene4D: Dynamic Multi-Object Scene Generation from Monocular Videos (NeurIPS 2024 · similarity 0.87)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.87)
- Robust Multi-Object 4D Generation for In-the-wild Videos (CVPR 2025 · similarity 0.86)
- Controlling Space and Time with Diffusion Models (ICLR 2025 · similarity 0.86)
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.
