Research map / 3D vision / 3D generation
Neural Lighting Simulation for Urban Scenes
Ava Pun, Gary Sun, Jingkang Wang, Yun Chen, Ze Yang, Sivabalan Manivasagam, Wei-Chiu Ma, Raquel Urtasun
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 |
|---|---|---|---|
| augment BEVFormer training with LightSim relit images | BEVFormer · BEVFormer detection, unseen lighting | mAP: 32.1 -> 36.6 | Improved |
| augment with Color Transfer simulated images instead of LightSim | BEVFormer · BEVFormer detection | mAP: 32.1 -> 35.1 | Improved |
| augment with color augmentation | BEVFormer · BEVFormer detection | mAP: 32.1 -> 33.8 | Improved |
| augment with Self-OSR simulated images | BEVFormer · BEVFormer detection | mAP: 32.1 -> 30.3 | Hurt |
| augment with EPE simulated images | BEVFormer · BEVFormer detection | mAP: 32.1 -> 32.5 | No clear effect |
| use LightSim relighting instead of Color Transfer | Color Transfer · PandaSet 15 sequences | FID: 85.4 -> 87.1 | No clear effect |
| use LightSim relighting instead of EPE | EPE · PandaSet 15 sequences | FID: 93.0 -> 87.1 | Improved |
| use LightSim relighting instead of NeRF-OSR | NeRF-OSR · PandaSet 15 sequences | FID: 143.9 -> 87.1 | Improved |
| remove precomputed rendering buffers from neural deferred rendering (ablation) | LightSim · PandaSet | qualitative | Improved |
| remove edge-based content-preserving loss (ablation) | LightSim · PandaSet | qualitative | Improved |
| remove sim-real training (ablation) | LightSim · PandaSet | qualitative | Improved |
| vary the data pairs column from 200 to 800 versus 400 in Table A4 (ablation) | LightSim · PandaSet | FID: 67.1 -> 57.3 -> 55.4 | Mixed |
Closest papers by meaning
- Do Image Editing Models Understand Lighting? (NeurIPS 2026 · similarity 0.86)
- SyncLight: Single-Edit Multi-View Relighting (NeurIPS 2026 · similarity 0.85)
- UniRelight: Learning Joint Decomposition and Synthesis for Video Relighting (NeurIPS 2025 · similarity 0.85)
- ROGR: Relightable 3D Objects using Generative Relighting (NeurIPS 2025 · similarity 0.85)
- A Pre-convolved Representation for Plug-and-Play Neural Illumination Fields (AAAI 2024 · similarity 0.85)
- LightIt: Illumination Modeling and Control for Diffusion Models (CVPR 2024 · 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.
