Research map / 3D vision / Depth and stereo
People as Scene Probes
Yifan Wang, Brian L. Curless, Steven M. Seitz
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 |
|---|---|---|---|
| use x-y mask shadow network (bias and gain maps) instead of pix2pix direct synthesis | shadow synthesis · Human study 1 (test images) | human preference: 58.4% vs 41.6% | Improved |
| use x-y mask shadow network instead of pix2pix | shadow synthesis · Human study 2 (inserted objects) | human preference: 70.0% | Improved |
| use x-y mask shadow network instead of pix2pix | shadow synthesis · Human study 1, scene 2 | human preference: 50.0 | No clear effect |
| use x-y mask shadow network instead of pix2pix | shadow synthesis · cloudy-day insertion | shadow quality | No clear effect |
| add x-y coordinate channels to the mask input (ablation) | shadow network · test set shadow synthesis | shadow warping and occlusion | Improved |
| generate bias and gain maps instead of the final image directly (ablation) | shadow network · shadow synthesis | artifacts | Improved |
| predict relative height from a reference instead of absolute height (ablation) | ground plane estimation · height prediction | height error: 13.28% -> 3% | Improved |
| use occlusion buffer from people motion instead of MiDaS depth | occlusion handling · occlusion in composites | qualitative occlusion correctness | Improved |
Closest papers by meaning
- Illumination-Consistent Human-Scene Reconstruction from Monocular Video (CVPR 2026 · similarity 0.88)
- What You Can Reconstruct From a Shadow (CVPR 2023 · similarity 0.87)
- Shadow-Enlightened Image Outpainting (CVPR 2024 · similarity 0.84)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.84)
- UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting (NeurIPS 2025 · similarity 0.83)
- Guess The Unseen: Dynamic 3D Scene Reconstruction from Partial 2D Glimpses (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.
