Research map / Image generation / Text-to-image diffusion
Parametric Shadow Control for Portrait Generation in Text-to-Image Diffusion Models
Haoming Cai, Tsung-Wei Huang, Shiv Gehlot, Brandon Y. Feng, Sachin Shah, Guan-Ming Su, Christopher Metzler
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 |
|---|---|---|---|
| apply shadow optimization at early/middle denoising steps instead of later steps (ablation) | Shadow Director · 800 generated portraits | CVS (identity): 0.9325 -> 0.9460 | Improved |
| apply shadow optimization at early/middle denoising steps instead of last step only (ablation) | Shadow Director · 800 generated portraits | CVTA: 0.2356 -> 0.2692 | Improved |
| use UNet internal features instead of UNet output as estimator input (ablation) | Shadow Director · 800 generated portraits | CVS (identity): 0.8945 -> 0.9460 | Improved |
| estimate shadow in latent space instead of noisy RGB space (ablation) | Shadow Director · 800 generated portraits | CLIPIQA+: 0.3220 -> 0.4484 | Improved |
| use dedicated ID-Estimator embedding instead of L1 loss on latent features (ablation) | Shadow Director · 800 generated portraits | CVS (identity): 0.7685 -> 0.9460 | Improved |
| add ID-Estimator constraint to latent optimization (ablation) | Shadow Director · generated portraits (qualitative) | visual artifacts | Improved |
| use test-time latent optimization during generation instead of post-hoc editing (vs IC-Light) | portrait relighting · 800 generated portraits | CVS (identity): 0.9140 -> 0.9460 | Improved |
| choose Shadow Director over other relighting methods in user study | portrait relighting · User study | preference rate: 27.44% -> 45.69% (IC-Light vs ours) | Improved |
| increase optimization iterations (shadow strength) (ablation) | Shadow Director · generated portraits (qualitative) | shadow intensity | Improved |
Closest papers by meaning
- Shadow Generation Using Diffusion Model with Geometry Prior (CVPR 2025 · similarity 0.88)
- Detail-Preserving Latent Diffusion for Stable Shadow Removal (CVPR 2025 · similarity 0.87)
- LightIt: Illumination Modeling and Control for Diffusion Models (CVPR 2024 · similarity 0.87)
- Shadow Generation for Composite Image Using Diffusion Model (CVPR 2024 · similarity 0.87)
- SynthLight: Portrait Relighting with Diffusion Model by Learning to Re-render Synthetic Faces (CVPR 2025 · similarity 0.87)
- Diffusion Handles Enabling 3D Edits for Diffusion Models by Lifting Activations to 3D (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.
