Research map / Image generation / Autoregressive and latent generation
KPLM-STA: Physically-Accurate Shadow Synthesis for Human Relighting via Keypoint-Based Light Modeling
Xinhui Yin, Qifei Li, Yilin Guo, Hongxia Xie, Xiaoli Zhang
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 STA to ControlNet base (ablation) | ControlNet base · DESOBAv2 (BOS test) | GRMSE: 14.317 -> 13.172 | Improved |
| add KPLM to ControlNet base (ablation) | ControlNet base · DESOBAv2 (BOS test) | GRMSE: 14.317 -> 13.213 | Improved |
| add KPLM on top of STA (ablation) | KPLM-STA · DESOBAv2 (BOS test) | GRMSE: 13.172 -> 12.620 | Improved |
| add STA on top of KPLM (ablation) | KPLM-STA · DESOBAv2 (BOS test) | LRMSE: 55.744 -> 54.243 | Improved |
| use KPLM-STA instead of GPSDiffusion | KPLM-STA · DESOBAv2 (BOS test) | GRMSE: 5.896 -> 4.486 | Improved |
| use KPLM-STA instead of GPSDiffusion | KPLM-STA · DESOBAv2 (BOS-free test) | GBer: 0.197 -> 0.198 | No clear effect |
| use KPLM-STA instead of GPSDiffusion | KPLM-STA · DESOBA (BOS test) | GRMSE: 3.613 -> 3.012 | Improved |
| add shadow synthesis after IC-Light relighting | KPLM-STA · IC-Light relit, different scene | SSIM: 0.166 -> 0.235 | Improved |
Closest papers by meaning
- Illumination-Consistent Human-Scene Reconstruction from Monocular Video (CVPR 2026 · similarity 0.88)
- Geometry-Aware Single-Image Full-Body Human Relighting (ECCV 2022 · similarity 0.88)
- Shadow Generation Using Diffusion Model with Geometry Prior (CVPR 2025 · similarity 0.86)
- RNG: Relightable Neural Gaussians (CVPR 2025 · similarity 0.86)
- Relightable and Animatable Neural Avatars from Videos (AAAI 2024 · similarity 0.86)
- SynthLight: Portrait Relighting with Diffusion Model by Learning to Re-render Synthetic Faces (CVPR 2025 · similarity 0.85)
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.
