Research map / 3D vision / 3D generation
MegaScenes: Scene-Level View Synthesis at Scale
Joseph Tung, Gene Chou, Ruojin Cai, Guandao Yang, Kai Zhang, Gordon Wetzstein, Bharath Hariharan, Noah Snavely
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 |
|---|---|---|---|
| finetune ZeroNVS on MegaScenes | ZeroNVS · MegaScenes test | FID: 69.097 -> 9.8382 | Improved |
| finetune Zero-1-to-3 on MegaScenes | Zero-1-to-3 · MegaScenes test | LPIPS: 0.5476 -> 0.4289 | Improved |
| finetune on MegaScenes | ZeroNVS · DTU | LPIPS: 0.6476 -> 0.4833 | Improved |
| finetune on MegaScenes | ZeroNVS · RealEstate10K | LPIPS: 0.4563 -> 0.2053 | Improved |
| add warped-image conditioning | ZeroNVS (MS) · MegaScenes test | LPIPS: 0.3857 -> 0.3444 | Improved |
| add warped-image conditioning | ZeroNVS (MS) · MegaScenes test | FID: 9.8382 -> 11.580 | Hurt |
| add extrinsic matrix conditioning to warp conditioning (ablation) | Ours · MegaScenes test | LPIPS: 0.3534 -> 0.3444 | Improved |
| finetune a diffusion model on 3D data instead of using SD-inpainting without finetuning | Ours · MegaScenes test | FID: 38.484 -> 11.580 | Improved |
| train on MegaScenes instead of MegaDepth (ablation) | ZeroNVS · MegaScenes test | qualitative | Improved |
| warp-conditioned model vs pose-only ZeroNVS (MS) | Ours · Mip-NeRF360 | FID: 60.677 -> 64.406 | Hurt |
Closest papers by meaning
- MegaSynth: Scaling Up 3D Scene Reconstruction with Synthesized Data (CVPR 2025 · similarity 0.90)
- WildCAT3D: Appearance-Aware Multi-View Diffusion in the Wild (NeurIPS 2025 · similarity 0.88)
- ZeroNVS: Zero-Shot 360-Degree View Synthesis from a Single Image (CVPR 2024 · similarity 0.88)
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.88)
- MVGenMaster: Scaling Multi-View Generation from Any Image via 3D Priors Enhanced Diffusion Model (CVPR 2025 · similarity 0.88)
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · 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.
