Research map / 3D vision / NeRF and radiance fields
Neural Dynamic GI: Random-Access Neural Compression for Temporal Lightmaps in Dynamic Lighting Environments
Jianhui Wu, Jian Zhou, Zhi Zhou, Zhangjin Huang, Chao Li
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 3D feature map and triplane to a 2D-only feature map (hybrid) (ablation) | NDGI · Temporal lightmap test scene | PSNR: 42.1 dB -> 44.9 dB | Improved |
| add 2D planes to a 3D-only feature map (hybrid) (ablation) | NDGI · Temporal lightmap test scene | PSNR: 36.6 dB -> 44.9 dB | Improved |
| add BC simulation (endpoints and weights) during training (ablation) | NDGI · Temporal lightmap test scene | PSNR: 37.96 dB -> 44.20 dB | Improved |
| add BC simulation during training (ablation) | NDGI · Temporal lightmap test scene | LPIPS: 0.013 -> 0.004 | Improved |
| use tri-plane instead of single 8-channel F2D_uv (ablation) | NDGI M. · Temporal lightmap test scene | PSNR / BPP: PSNR 41.32 -> 41.50, BPP 0.86 -> 0.67 | Improved |
| use tri-plane instead of single 8-channel F2D_uv (ablation) | NDGI M. · Temporal lightmap test scene | storage size (MB): 358 -> 279 | Improved |
| apply 8-bit quantization to feature maps | NDGI · Temporal lightmap dataset | BPP: 4.62 -> 2.32 | Improved |
| add BC7 compression on top of 8-bit quantization | NDGI · Temporal lightmap dataset | BPP: 2.32 -> 0.68 | Improved |
| enlarge MLP hidden size from 16 to 64 | NDGI M. · Temporal lightmap dataset | PSNR: 46.69 -> 47.42 (BPP 0.68 -> 0.86) | Improved |
| raise F3D_uvt resolution from 32x32x12x4 to 64x64x12x4 | NDGI · Temporal lightmap dataset | PSNR: 46.69 -> 48.68 (BPP 0.68 -> 1.39) | Improved |
| use NDGI instead of NTC for decoding | NDGI L. vs NTC L. · 1024x1024 lightmap decode | decode time: 0.886ms -> 0.201ms | Improved |
| use NDGI M. instead of NTC L. at similar bitrate | NDGI · Temporal lightmap dataset | PSNR: 43.61 -> 46.69 | Improved |
Closest papers by meaning
- A Pre-convolved Representation for Plug-and-Play Neural Illumination Fields (AAAI 2024 · similarity 0.86)
- RehearsalNeRF: Decoupling Intrinsic Neural Fields of Dynamic Illuminations for Scene Editing (IJCV 2026 · similarity 0.86)
- Temporal Interpolation Is All You Need for Dynamic Neural Radiance Fields (CVPR 2023 · similarity 0.85)
- Dynamic 3D Gaussian Fields for Urban Areas (NeurIPS 2024 · similarity 0.84)
- TC-Light: Temporally Coherent Generative Rendering for Realistic World Transfer (NeurIPS 2025 · similarity 0.84)
- Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting (ICLR 2024 · similarity 0.84)
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.
