Research map / 3D vision / 3D generation
LumiTokens: 3D Relighting via Token-Space Lighting Transformation
Yiwen Chen, Matheus Gadelha, Huaizu Jiang
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 |
|---|---|---|---|
| unfreeze the encoder during relighting training (ablation) | LumiTokens · Novel-view relighting | PSNR: 25.06 -> 26.12 | Improved |
| add lighting invariance loss to unfrozen encoder (ablation) | LumiTokens · Novel-view relighting | PSNR: 26.12 -> 27.05 | Improved |
| add explicit albedo prediction head regularization (ablation) | LumiTokens · Novel-view relighting | PSNR: 26.12 -> 27.09 | Improved |
| replace explicit albedo supervision with invariance loss (ablation) | LumiTokens · Novel-view relighting | SSIM: 0.934 -> 0.952 | Improved |
| replace explicit albedo supervision with invariance loss (ablation) | LumiTokens · Novel-view relighting | LPIPS: 0.057 -> 0.045 | Improved |
| add multi-step chain training for the editor, edit in token space instead of pixel space (ablation) | LumiTokens · Progressive relighting, 10 steps | fidelity under repeated edits | Improved |
| use LumiTokens instead of Neural Gaffer | LumiTokens · Multi-view relighting (ILR) | PSNR: 28.40 -> 30.48 | Improved |
| use LumiTokens instead of Neural Gaffer | LumiTokens · Multi-view relighting (ILR) | LPIPS: 0.030 -> 0.045 | Hurt |
| use LumiTokens instead of DiffusionRenderer under scene-level rescaling | LumiTokens · Multi-view relighting (SLR) | LPIPS: 0.086 -> 0.086 | No clear effect |
| use LumiTokens (8 views) instead of TensoIR (50 views) | LumiTokens · Novel-view relighting (Synthetic) | PSNR: 26.17 -> 27.76 | Improved |
| use LumiTokens instead of LightSwitch | LumiTokens · Novel-view relighting (Stanford-ORB) | PSNR: 32.02 -> 31.01 | Hurt |
| use LumiTokens instead of Neural Gaffer for runtime | LumiTokens · Runtime end-to-end | seconds: 16.85 -> 3.79 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LightSwitch — LightSwitch: Multi-view Relighting with Material-guided Diffusion (ICCV 2025) · better on 6 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- UniLumos: Fast and Unified Image and Video Relighting with Physics-Plausible Feedback (NeurIPS 2025 · similarity 0.91)
- UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting (NeurIPS 2025 · similarity 0.91)
- LumiNet: Latent Intrinsics Meets Diffusion Models for Indoor Scene Relighting (CVPR 2025 · similarity 0.90)
- SyncLight: Single-Edit Multi-View Relighting (NeurIPS 2026 · similarity 0.90)
- IllumiNeRF: 3D Relighting Without Inverse Rendering (NeurIPS 2024 · similarity 0.89)
- ROGR: Relightable 3D Objects using Generative Relighting (NeurIPS 2025 · similarity 0.88)
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.
