Research map / 3D vision / Depth and stereo
DeepLight: Learning Illumination for Unconstrained Mobile Mixed Reality
Chloe LeGendre, Wan-Chun Ma, Graham Fyffe, John Flynn, Laurent Charbonnel, Jay Busch, Paul Debevec
Check 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 |
|---|---|---|---|
| use DeepLight instead of the prior indoor lighting-estimation method | mobile HDR illumination estimation · Unseen indoor (UI) | sphere L1 loss: diffuse: 0.21 ± 0.11 -> 0.13 ± 0.07; matte silver: 0.22 ± 0.06 -> 0.14 ± 0.05; mirror: 0.23 ± 0.06 -> 0.18 ± 0.03 | Improved |
| use DeepLight instead of the prior outdoor lighting-estimation method | mobile HDR illumination estimation · Unseen outdoor (UO) | sphere L1 loss: diffuse: 0.25 ± 0.12 -> 0.13 ± 0.08; matte silver: 0.25 ± 0.07 -> 0.14 ± 0.06; mirror: 0.34 ± 0.06 -> 0.17 ± 0.04 | Improved |
| use multiple BRDF reconstruction losses rather than only the mirror-sphere loss (ablation) | DeepLight · Unseen indoor and outdoor | L1 loss on diffuse and matte silver spheres: diffuse: 0.20, 0.18 -> 0.12, 0.13; matte silver: 0.16, 0.15 -> 0.13, 0.13 | Improved |
| use multiple BRDF reconstruction losses rather than only the diffuse-sphere loss (ablation) | DeepLight · Unseen indoor and outdoor | mirror sphere L1 loss: 0.28, 0.27 -> 0.17, 0.16 | Improved |
| use multiple BRDF reconstruction losses rather than only the matte-silver-sphere loss (ablation) | DeepLight · Unseen indoor and outdoor | mirror sphere L1 loss: 0.21, 0.20 -> 0.17, 0.16 | Improved |
| add adversarial loss (ablation) | DeepLight · Unseen lighting and rendered reflective objects | high-frequency illumination realism | Improved |
| add adversarial loss (ablation) | DeepLight · Unseen indoor and outdoor | mirror sphere L1 loss: 0.15 -> 0.17 (UI); 0.14 -> 0.16 (UO) | Hurt |
Later papers that compared with this method
1 later accepted paper reports DeepLight as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Lighthouse: Predicting Lighting Volumes for Spatially-Coherent Illumination (CVPR 2020 · better on 4 of 4 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
- Rendering-Aware HDR Environment Map Prediction from a Single Image (AAAI 2022 · similarity 0.88)
- UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting (NeurIPS 2025 · similarity 0.86)
- ReCap: Better Gaussian Relighting with Cross-Environment Captures (CVPR 2025 · similarity 0.86)
- IllumiNeRF: 3D Relighting Without Inverse Rendering (NeurIPS 2024 · similarity 0.86)
- A Pre-convolved Representation for Plug-and-Play Neural Illumination Fields (AAAI 2024 · similarity 0.86)
- LuxDiT: Lighting Estimation with Video Diffusion Transformer (NeurIPS 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.
