Research map / 3D vision / 3D generation
SynSin: End-to-End View Synthesis From a Single Image
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use a differentiable renderer with a large point footprint rather than a small footprint (ablation) | SynSin · Matterport3D | PSNR (Both): 21.14 -> 20.91 | No clear effect |
| Use soft z-buffer accumulation rather than hard z-buffering (ablation) | SynSin · Replica | PSNR: 20.70 -> 21.94 | Helped |
| Project learned features instead of RGB values (ablation) | SynSin · Matterport3D | PSNR (Both): 20.64 -> 20.91 | Helped |
| Use ground-truth depth as the projected depth (ablation) | SynSin · Matterport3D | PSNR (Both): 20.91 -> 22.65 | Helped |
| Supervise predicted depth with ground-truth depth (ablation) | SynSin · Matterport3D | PSNR (Both): 20.91 -> 21.59 | Helped |
| Replace the voxel UNet encoder/decoder with the SynSin-style ResNet setup (ablation) | Vox · Matterport3D | PSNR (Both): 18.52 -> 20.62 | Helped |
| Use SynSin's learned end-to-end geometry instead of pretrained depth prediction | Single-image view synthesis · RealEstate10K | PSNR: $21.88_{8.43}$ -> 22.31$_{4.97}$ | Helped |
| Use a point-cloud representation instead of a voxel representation | SynSin versus Vox w/ ours · Replica | PSNR: 19.77 -> 21.94 | Helped |
| Apply the model trained at 256×256 to 512×512 images | SynSin · Matterport3D | PSNR: 18.82$_{2.52}$ -> 22.06$_{6.30}$ | Helped |
| Use two input images at test time instead of one | View synthesis · RealEstate10K | PSNR: 22.31$_{4.97}$ -> $25.34_{9.48}$ | Helped |
| Use SynSin rather than ContView | View synthesis · KITTI | PSNR: 16.90 -> 16.70 | Hurt |
Closest papers by meaning
- Learning To Render Novel Views From Wide-Baseline Stereo Pairs (CVPR 2023 · similarity 0.89)
- Learning Efficient Point Cloud Generation for Dense 3D Object Reconstruction (AAAI 2018 · similarity 0.89)
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.89)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.87)
- ERUPT: Efficient Rendering with Unposed Patch Transformer (CVPR 2025 · similarity 0.87)
- 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.