Research map / 3D vision / 3D generation
MOSAIC: Generating Consistent, Privacy-Preserving Scenes from Multiple Depth Views in Multi-Room Environments
Zhixuan Liu, Haokun Zhu, Rui Chen, Jonathan Francis, Soonmin Hwang, Ji Zhang, Jean Oh
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 multi-channel inference-time optimization (vs naive averaging) (ablation) | MOSAIC · Indoor multi-room (ablation) | KID: 0.06715 -> 0.03391 | Improved |
| add multi-channel inference-time optimization (ablation) | MOSAIC · Indoor multi-room (ablation) | Warped PSNR: 15.74 -> 25.30 | Improved |
| add pixel-space refinement loss L_Pixel (ablation) | MOSAIC · Indoor multi-room (ablation) | KID: 0.04273 -> 0.03391 | Improved |
| add pixel-space refinement loss L_Pixel (ablation) | MOSAIC · Indoor multi-room (ablation) | Warped PSNR: 23.02 -> 25.30 | Improved |
| add pixel-space refinement loss L_Pixel (ablation) | MOSAIC · Indoor multi-room (ablation) | CLIPConsistency: 31.26 -> 30.85 | Hurt |
| use region-coverage key frame selection instead of uniform sampling (ablation) | MOSAIC · Indoor multi-room (ablation) | CLIPScore: 0.6746 -> 0.7166 | Improved |
| use region-coverage key frame selection instead of uniform sampling (ablation) | MOSAIC · Indoor multi-room (ablation) | KID: 0.03920 -> 0.03391 | Improved |
| increase overlapping depth views from one to four (ablation) | MOSAIC · Indoor multi-room (number of views) | NMSE of generated depth: 0.018630 -> 0.011269 | Improved |
| increase overlapping depth views from one to four (ablation) | MOSAIC · Indoor multi-room (number of views) | KID: 0.06123 -> 0.07833 | Hurt |
| increase overlapping depth views from one to four (ablation) | MOSAIC · Indoor multi-room (number of views) | CLIPScore: 0.6874 -> 0.7100 | Improved |
| replace Warp-Inpaint with MOSAIC | Warp-Inpaint · Indoor multi-room (geometric consistency) | Warped PSNR: 22.00 -> 25.30 | Improved |
| replace MVDiffusion with MOSAIC | MVDiffusion · Indoor multi-room (geometric consistency) | Warped PSNR: 13.58 -> 25.30 | 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.
- SceneCraft — SceneCraft: An LLM Agent for Synthesizing 3D Scenes as Blender Code (ICML 2024) · better on 4 of 4 shared results
- SceneTex — SceneTex: High-Quality Texture Synthesis for Indoor Scenes via Diffusion Priors (CVPR 2024) · better on 2 of 2 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
- MvFFN: Multi-view Floor-Plan Feed-Forward Network for Unposed Wide-Baseline Panorama Layout Reconstruction (NeurIPS 2026 · similarity 0.85)
- MOVIS: Enhancing Multi-Object Novel View Synthesis for Indoor Scenes (CVPR 2025 · similarity 0.84)
- DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesis (CVPR 2024 · similarity 0.83)
- CM-EVS: Sparse Panoramic RGB-D-Pose Data for Complete Scene Coverage (NeurIPS 2026 · similarity 0.83)
- DeBaRA: Denoising-Based 3D Room Arrangement Generation (NeurIPS 2024 · similarity 0.83)
- DiffInDScene: Diffusion-based High-Quality 3D Indoor Scene Generation (CVPR 2024 · similarity 0.83)
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.
