Research map / 3D vision / Face and shape
What You Can Reconstruct From a Shadow
Ruoshi Liu, Sachit Menon, Chengzhi Mao, Dennis Park, Simon Stent, Carl Vondrick
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 |
|---|---|---|---|
| latent search over generative model instead of direct regression from shadow | shadow-to-3D reconstruction with known light and pose · ShapeNet-style categories (Car, Chair, Plane, Sofa) | volumetric IoU (All): +0.086 over regression (.553 vs .467) | Improved |
| latent search over generative model instead of regression with unknown light and pose | shadow-to-3D reconstruction with unknown light and pose · ShapeNet-style categories (Car, Chair, Plane, Sofa) | volumetric IoU (All): +0.087 over regression (.390 vs .303) | Improved |
| jointly optimizing light position and object pose with latent search | shadow-based reconstruction · unknown light and pose vs known | volumetric IoU (All): -0.163 (.553 to .390) | Hurt |
| multiple random restarts (diverse samples) of latent search (ablation) | shadow reconstruction · known light and pose | best volumetric IoU: rises monotonically from about 0.41978 at 1 sample to 0.57198 at 23 samples | Improved |
| adaptive Gaussian noise injection during latent optimization | latent search · shadow reconstruction | qualitative optimization quality | Improved |
Closest papers by meaning
- Towards Learning Neural Representations from Shadows (ECCV 2022 · similarity 0.87)
- Efficient and Differentiable Shadow Computation for Inverse Problems (ICCV 2021 · similarity 0.87)
- People as Scene Probes (ECCV 2020 · similarity 0.87)
- Cast and Attached Shadow Detection via Iterative Light and Geometry Reasoning (ECCV 2026 · similarity 0.87)
- Floating No More: Object-Ground Reconstruction from a Single Image (CVPR 2025 · similarity 0.86)
- Eclipse: Disambiguating Illumination and Materials using Unintended Shadows (CVPR 2024 · 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.
