Research map / Image generation / Editing and style transfer
TransLight: Image-Guided Customized Lighting Control with Generative Decoupling
Zongming Li, Lianghui Zhu, Haocheng Shen, Longjin Ran, Wenyu Liu, Xinggang Wang
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 stage-1 LoRA fine-tuning (ablation) | TransLight · Light FID (12000 triplets) | Light FID: 8.44 -> 6.02 | Improved |
| train on real triplets from generative decoupling instead of IC-Light-relit synthetic data (ablation) | TransLight · Light FID (12000 triplets) | Light FID: 7.72 -> 6.02 | Improved |
| add similarity-based data filtering to triplet construction (ablation) | TransLight · Light FID (12000 triplets) | Light FID: 6.49 -> 6.02 | Improved |
| add similarity filtering to triplet construction (ablation) | Generative Decoupling · 500 sampled triplets | total success rate: 58.74% -> 83.64% | Improved |
| add similarity filtering to triplet construction (ablation) | Generative Decoupling · 500 sampled triplets | content success rate: 83.45% -> 96.82% | Improved |
| add similarity filtering to triplet construction (ablation) | Generative Decoupling · 500 sampled triplets | light success rate: 66.43% -> 86.14% | Improved |
| use TransLight instead of IC-Light background-conditioned model | light effect transfer · 1000 held-out triplets | PSNR: 15.34 -> 19.58 | Improved |
| use TransLight instead of IC-Light | light effect transfer · 1000 held-out triplets | LPIPS: 0.2756 -> 0.1982 | Improved |
| use TransLight instead of IC-Light | light effect transfer · 12000 held-out triplets | Light FID: 10.05 -> 6.02 | Improved |
| use extracted light added directly to content instead of IC-Light | light effect transfer · 12000 held-out triplets | Light FID: 10.05 -> 8.40 | Improved |
Closest papers by meaning
- Do Image Editing Models Understand Lighting? (NeurIPS 2026 · similarity 0.87)
- SyncLight: Single-Edit Multi-View Relighting (NeurIPS 2026 · similarity 0.87)
- SynthLight: Portrait Relighting with Diffusion Model by Learning to Re-render Synthetic Faces (CVPR 2025 · similarity 0.87)
- Text2Relight: Creative Portrait Relighting with Text Guidance (AAAI 2025 · similarity 0.87)
- LightIt: Illumination Modeling and Control for Diffusion Models (CVPR 2024 · similarity 0.86)
- DreamLight: Towards Harmonious and Consistent Image Relighting (NeurIPS 2025 · 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.
