Research map / 3D vision / 3D generation
Worldsheet: Wrapping the World in a 3D Sheet for View Synthesis From a Single Image
Ronghang Hu, Nikhila Ravi, Alexander C. Berg, Deepak Pathak
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 inpainting network after mesh rendering (ablation) | Worldsheet · Matterport visible regions | PSNR: 25.42 -> 26.00 | Improved |
| raise mesh Laplacian regularization weight to 10^-3 (ablation) | Worldsheet · RealEstate10K | PSNR: 26.24 -> 24.83 | Hurt |
| set mesh regularization weight to zero (ablation) | Worldsheet · RealEstate10K | PSNR: 26.24 -> 25.78 | Hurt |
| lower mesh regularization weight to 10^-5 (ablation) | Worldsheet · RealEstate10K | PSNR: 26.24 -> 26.18 | No clear effect |
| increase mesh resolution from 5x5 to 33x33 (ablation) | Worldsheet · RealEstate10K | PSNR: 24.39 -> 26.24 | Improved |
| increase mesh resolution from 33x33 to 65x65 (ablation) | Worldsheet · RealEstate10K | PSNR: 26.24 -> 26.74 | Improved |
| restrict grid offset to zero at 65x65 mesh, using only depth (ablation) | Worldsheet · RealEstate10K | PSNR: -0.13 PSNR | Hurt |
| add depth supervision during training (ablation) | Worldsheet · Matterport | PSNR (all pixels): 24.67 -> 24.75 | No clear effect |
| add depth supervision during training (ablation) | Worldsheet · Replica | PSNR: 23.51 -> 22.78 | Hurt |
| stack multiple mesh layers instead of one sheet (ablation) | Worldsheet · RealEstate10K | PSNR: 26.74 -> 26.62 | No clear effect |
| replace SynSin U-Net backbone with ResNet-50 (ablation) | SynSin · RealEstate10K (averaged) | PSNR: 20.47 -> 19.37 | Hurt |
| add geometry-based mesh rendering versus point-cloud features (Worldsheet vs SynSin) | Worldsheet · Matterport | PSNR: 20.91 -> 24.67 | Improved |
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.
- SynSin — SynSin: End-to-End View Synthesis From a Single Image (CVPR 2020) · better on 12 of 12 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
- One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi-View Generation and 3D Diffusion (CVPR 2024 · similarity 0.87)
- WorldGrow: Generating Infinite 3D World (AAAI 2026 · similarity 0.87)
- WorldFlow3D: Flowing Through 3D Distributions for Unbounded World Generation (ECCV 2026 · similarity 0.86)
- Differentiable Blocks World: Qualitative 3D Decomposition by Rendering Primitives (NeurIPS 2023 · similarity 0.86)
- WorldStereo: Bridging Camera-Guided Video Generation and Scene Reconstruction via 3D Geometric Memories (CVPR 2026 · similarity 0.86)
- Pano3DComposer: Feed-Forward Compositional 3D Scene Generation from Single Panoramic Image (CVPR 2026 · 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.
