Research map / Image generation / Editing and style transfer
Learning Latent Proxies for Controllable Single-Image Relighting
Haoze Zheng, Zihao Wang, Xianfeng Wu, Yajing Bai, Yexin Liu, Yun Li, Xiaogang Xu, Harry Yang
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 the few-shot latent proxy (ablation) | LightCtrl · ScaLight (temperature shift) | PSNR: 28.1 -> 30.2 | Improved |
| add the few-shot latent proxy (ablation) | LightCtrl · ScaLight (light energy) | PSNR: 18.0 -> 27.1 | Improved |
| add the few-shot latent proxy (ablation) | LightCtrl · ScaLight (light position) | SSIM: 0.931 vs 0.929 | No clear effect |
| add the lighting-aware mask (ablation) | LightCtrl · ScaLight (light position) | PSNR: 20.5 -> 25.6 | Improved |
| add the lighting-aware mask (ablation) | LightCtrl · ScaLight (light energy) | PSNR: 23.2 -> 27.1 | Improved |
| add DPO post-training (ablation) | LightCtrl · ScaLight (temperature shift) | PSNR: 22.8 -> 30.2 | Improved |
| add DPO post-training (ablation) | LightCtrl · ScaLight (light energy) | PSNR: 17.5 -> 27.1 | Improved |
| increase PBR-supervised samples from 1K to 12K (ablation) | LightCtrl · ScaLight (PBR supervision scale) | PSNR: 12.8 -> 28.6 | Improved |
| increase PBR samples from 9K to 12K (ablation) | LightCtrl · ScaLight (PBR supervision scale) | PSNR: 23.5 -> 28.6 | Improved |
| LightCtrl versus LumiNet on scene-level relighting | LightCtrl · MIIW (scene-level) | SSIM: 0.904 -> 0.655 | Hurt |
| LightCtrl versus LumiNet on scene-level relighting | LightCtrl · MIIW (scene-level) | PSNR: 17.20 -> 18.30 | Improved |
| LightCtrl versus LBM on light energy | LightCtrl · ScaLight (light energy) | PSNR: 25.3 -> 27.1 | Improved |
Closest papers by meaning
- UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting (NeurIPS 2025 · similarity 0.91)
- Neural Gaffer: Relighting Any Object via Diffusion (NeurIPS 2024 · similarity 0.91)
- UniLumos: Fast and Unified Image and Video Relighting with Physics-Plausible Feedback (NeurIPS 2025 · similarity 0.90)
- ReCap: Better Gaussian Relighting with Cross-Environment Captures (CVPR 2025 · similarity 0.90)
- Neural LightRig: Unlocking Accurate Object Normal and Material Estimation with Multi-Light Diffusion (CVPR 2025 · similarity 0.90)
- MetaGS: A Meta-Learned Gaussian-Phong Model for Out-of-Distribution 3D Scene Relighting (NeurIPS 2025 · similarity 0.89)
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.
