Research map / 3D vision / NeRF and radiance fields
City-on-Web: Real-time Neural Rendering of Large-scale Scenes on the Web
Kaiwen Song, Xiaoyi Zeng, Chenqu Ren, Juyong Zhang
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 depth-sorted alpha blending across blocks (ablation) | City-on-Web · Campus (one part) | LPIPS: 0.345 -> 0.190 | Improved |
| train with the same block-based rendering pipeline as at render time (ablation) | City-on-Web · Campus (one part) | LPIPS: 0.281 -> 0.190 | Improved |
| add global deferred MLP regularization (ablation) | City-on-Web · Campus (one part) | SSIM: 0.712 -> 0.741 | Improved |
| split into 4 blocks instead of a single higher-resolution MERF (1024^3) (ablation) | MERF · Campus (one part) | SSIM / DISK: SSIM 0.710 -> 0.741; DISK 147 -> 114 | Improved |
| increase blocks from 4 to 16 (ablation) | City-on-Web · Campus (one part) | PSNR / VRAM / FPS: PSNR 24.82 -> 25.13; VRAM 526.6 -> 2040.7; FPS 46 -> 34 | Mixed |
| generate LODs by finetuning from the finest LOD instead of retraining from scratch (ablation) | City-on-Web LOD · Campus | training time: 12hours -> 4hours (LPIPS 0.224 -> 0.204) | Improved |
| re-optimize after downsampling for LOD generation (ablation) | City-on-Web LOD · Campus | PSNR: 22.43 -> 24.20 | Improved |
| use coarser LOD levels (LOD1 to LOD3) | City-on-Web · Campus whole scene | VRAM / PSNR: VRAM 3970.2 -> 132.1; PSNR 24.73 -> 23.72 | Mixed |
| volume rendering on blocks versus mesh rasterization baselines | City-on-Web · Campus (one part) | FPS: 46 vs 223 (BakedSDF) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Mega-NeRF — Mega-NERF: Scalable Construction of Large-Scale NeRFs for Virtual Fly-Throughs (CVPR 2022) · better on 1 of 2 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
- Block-NeRF: Scalable Large Scene Neural View Synthesis (CVPR 2022 · similarity 0.87)
- Real-Time Neural Rasterization for Large Scenes (ICCV 2023 · similarity 0.86)
- Multi-Level Neural Scene Graphs for Dynamic Urban Environments (CVPR 2024 · similarity 0.85)
- VideoRF: Rendering Dynamic Radiance Fields as 2D Feature Video Streams (CVPR 2024 · similarity 0.84)
- Dynamic 3D Gaussian Fields for Urban Areas (NeurIPS 2024 · similarity 0.84)
- Ced-NeRF: A Compact and Efficient Method for Dynamic Neural Radiance Fields (AAAI 2024 · similarity 0.83)
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.
