Research map / 3D vision / 3D generation
Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation
Sherwin Bahmani, Tianchang Shen, Jiawei Ren, Jiahui Huang, Yifeng Jiang, Haithem Turki, Andrea Tagliasacchi, David B. Lindell, Zan Gojcic, Sanja Fidler, Huan Ling, Jun Gao and 1 more
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 3DGS decoder with self-distillation from video model instead of real multi-view data only (ablation) | Lyra · Lyra dataset ablation | PSNR: 19.08 -> 24.77 | Improved |
| add real multi-view data to self-distillation (ablation) | Lyra · Lyra dataset ablation | PSNR: 24.77 -> 24.74 | No clear effect |
| add depth supervision loss (ablation) | Lyra · Lyra dataset ablation | PSNR: 24.31 -> 24.77 | Improved |
| add opacity-based pruning (ablation) | Lyra · Lyra dataset ablation | PSNR: 24.55 -> 24.77 | Improved |
| add opacity-based pruning (ablation) | Lyra · render time at 704x1280 | ms: 30ms -> 18ms | Improved |
| add LPIPS loss (ablation) | Lyra · Lyra dataset ablation | LPIPS: 0.370 -> 0.224 | Improved |
| learn multi-view fusion across trajectories instead of independent fusion (ablation) | Lyra · Lyra dataset ablation | PSNR: 17.73 -> 24.77 | Improved |
| use Transformer plus Mamba-2 blocks instead of Transformer only (ablation) | Lyra · Lyra dataset ablation | PSNR: 24.58 -> 24.77 | Improved |
| operate 3DGS decoder in video latent space instead of pixel space (ablation) | Lyra · memory | out-of-memory | Improved |
| use Lyra instead of Bolt3D | Lyra · RealEstate10K | PSNR: 21.54 -> 21.79 | Improved |
| use Lyra instead of Wonderland | Lyra · DL3DV | PSNR: 16.64 -> 20.09 | 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.
- ViewCrafter — ViewCrafter: Taming Video Diffusion Models for High-fidelity Novel View Synthesis (TPAMI 2025) · better on 5 of 5 shared results
- Bolt3D — Bolt3D: Generating 3D Scenes in Seconds (ICCV 2025) · better on 3 of 3 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
- Wonderland: Navigating 3D Scenes from a Single Image (CVPR 2025 · similarity 0.89)
- LT3SD: Latent Trees for 3D Scene Diffusion (CVPR 2025 · similarity 0.89)
- Denoising Diffusion via Image-Based Rendering (ICLR 2024 · similarity 0.89)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.89)
- Scene Splatter: Momentum 3D Scene Generation from Single Image with Video Diffusion Model (CVPR 2025 · similarity 0.89)
- CAD: Photorealistic 3D Generation via Adversarial Distillation (CVPR 2024 · similarity 0.88)
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.
