Research map / 3D vision / Face and shape
An Approximate Shading Model with Detail Decomposition for Object Relighting
Zicheng Liao, Kevin Karsch, Hongyi Zhang, David Forsyth
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 ground truth shading to derive detail layers instead of default | Ours (oracle weights) · MIT Intrinsic (Lab illumination) | re-rendering MSE: 0.0586 -> 0.0228 | Improved |
| use Barron and Malik shading estimation for detail derivation instead of default (ablation) | Ours (regress) · MIT Intrinsic (Natural illumination) | re-rendering MSE: 0.0358 -> 0.0320 | Improved |
| replace Barron and Malik SIRFS shape with approximate shading model | Ours (regress) · MIT Intrinsic (Natural illumination) | re-rendering MSE: 0.0172 -> 0.0358 | Hurt |
| replace Barron and Malik SIRFS shape with approximate shading model | Ours (regress) · MIT Intrinsic (Lab illumination) | re-rendering MSE: 0.0372 -> 0.0641 | Mixed |
| use SIRFS-stereo base shape instead of SFC (ablation) | Ours (oracle) · Relighting dataset (10 objects x 10 lights) | re-rendering MSE: 0.0295 -> 0.0194 | Improved |
| use Kinect depth instead of SFC base shape (ablation) | Ours (default) · Relighting dataset (10 objects x 10 lights) | re-rendering MSE: 0.0723 -> 0.0732 | No clear effect |
| use SIRFS depth instead of SFC base shape (ablation) | Ours (oracle) · Relighting dataset (10 objects x 10 lights) | re-rendering MSE: 0.0295 -> 0.0334 | Hurt |
| add parametric shading detail layer (ablation) | Ours insertion · User study detail layers | % preferred: 61% of 764 viewed pairs | Improved |
| add geometric detail layer on top of parametric detail (ablation) | Ours insertion · User study detail layers | % preferred: 65% of 756 viewed pairs | Improved |
| insert real image fragments with approximate model instead of 3D synthetic models (Karsch et al.) | insertion · User study | % chosen as real: 34% -> 44% | Improved |
Closest papers by meaning
- RNG: Relightable Neural Gaussians (CVPR 2025 · similarity 0.88)
- MetaGS: A Meta-Learned Gaussian-Phong Model for Out-of-Distribution 3D Scene Relighting (NeurIPS 2025 · similarity 0.87)
- Neural Gaffer: Relighting Any Object via Diffusion (NeurIPS 2024 · similarity 0.87)
- Generative Multiview Relighting for 3D Reconstruction under Extreme Illumination Variation (CVPR 2025 · similarity 0.87)
- ReCap: Better Gaussian Relighting with Cross-Environment Captures (CVPR 2025 · similarity 0.87)
- IllumiNeRF: 3D Relighting Without Inverse Rendering (NeurIPS 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.
