Research map / Video / Video generation
A Light Stage on Every Desk
Soumyadip Sengupta, Brian Curless, Ira Kemelmacher-Shlizerman, 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 |
|---|---|---|---|
| add perceptual loss (ablation) | monitor-lighting relighting network · 3 captured sequences, Protocol 1 | LPIPS: 0.1384 -> 0.1311 | Improved |
| add cycle consistency loss (ablation) | monitor-lighting relighting network · 3 captured sequences, Protocol 1 | LPIPS: 0.1493 -> 0.1311 | Improved |
| add adversarial PatchGAN loss (ablation) | monitor-lighting relighting network · 3 captured sequences, Protocol 1 | LPIPS: 0.1523 -> 0.1311 | Improved |
| add source monitor lighting as network input (ablation) | monitor-lighting relighting network · 3 captured sequences, Protocol 1 | LPIPS: 0.1374 -> 0.1311 | Improved |
| use proposed lighting-modulated architecture and losses instead of Sun+ | relighting network · Protocol 1, 19 sequences | LPIPS: 0.1684 -> 0.1537 | Improved |
| use proposed lighting-modulated architecture and losses instead of Sun+ | relighting network · Protocol 2, 8 sequences | PSNR: 23.14 -> 26.36 | Improved |
| use proposed lighting-modulated architecture and losses instead of Sun+ | relighting network · Protocol 2, 8 sequences | RMSE: 0.0064 -> 0.0034 | Improved |
| replace RSun losses (with reconstruction loss) by misalignment-tolerant losses (Sun+) | Sun et al. architecture · Protocol 1, 19 sequences | LPIPS: 0.2718 -> 0.1684 | Improved |
| use proposed method instead of RSun on perfectly aligned mannequin data | relighting network · Mannequin sequence (no misalignment) | Mean absolute error (%): 3.14% -> 3.35% | No clear effect |
Closest papers by meaning
- SynthLight: Portrait Relighting with Diffusion Model by Learning to Re-render Synthetic Faces (CVPR 2025 · similarity 0.84)
- Lux Post Facto: Learning Portrait Performance Relighting with Conditional Video Diffusion and a Hybrid Dataset (CVPR 2025 · similarity 0.84)
- Comprehensive Relighting: Generalizable and Consistent Monocular Human Relighting and Harmonization (CVPR 2025 · similarity 0.83)
- Real-time 3D-aware Portrait Video Relighting (CVPR 2024 · similarity 0.83)
- SwitchLight: Co-design of Physics-driven Architecture and Pre-training Framework for Human Portrait Relighting (CVPR 2024 · similarity 0.83)
- LightIt: Illumination Modeling and Control for Diffusion Models (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.
