Research map / 3D vision / NeRF and radiance fields
S-NeRF: Neural Radiance Fields for Street Views
Ziyang Xie, Junge Zhang, Wenye Li, Feihu Zhang, Li 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 supervision with learned confidence (vs. RGB-only) (ablation) | S-NeRF background scenes · nuScenes background (2 sequences) | PSNR: 24.41 -> 25.01 | Improved |
| Add learnable depth confidence (ablation) | S-NeRF background scenes · nuScenes background (2 sequences) | PSNR: 24.30 -> 25.01 | Improved |
| Add smoothness loss (ablation) | S-NeRF background scenes · nuScenes background (2 sequences) | PSNR: 25.06 -> 25.01 (slight drop) | Hurt |
| Add depth confidence (ablation) | S-NeRF moving vehicles · 4 foreground cars | PSNR: 18.09 -> 19.64 | Improved |
| Add smoothness loss (ablation) | S-NeRF moving vehicles · 4 foreground cars | LPIPS: 0.159 -> 0.136 | Improved |
| Replace Mip-NeRF with S-NeRF (camera pose processing, scene parameterization, depth supervision with confidence) | large-scale nuScenes background scenes · nuScenes 4 sequences | PSNR: 18.22 -> 26.21 | Improved |
| Use S-NeRF instead of GeoSim mesh method for moving vehicles | moving vehicle rendering · nuScenes/Waymo foreground cars | PSNR: 12.24 -> 18.00 | Improved |
| Use depth-confidence-based depth supervision with dense completed depth vs. sparse-only depth (ablation) | S-NeRF background scenes · nuScenes background (2 sequences) | PSNR: 23.65 (sparse) -> 24.05 (dense w/ confidence) | Improved |
| Use a lower-quality traditional depth completion (Ku et al.) instead of NLSPN (ablation) | S-NeRF background scenes · nuScenes background (2 sequences) | PSNR: 24.41 -> 24.40 | No clear effect |
| Remove reprojection confidence (ablation) | S-NeRF background scenes · nuScenes background (2 sequences) | PSNR: 24.05 -> 23.88 | Hurt |
| Use only SSIM confidence instead of RGB+SSIM for vehicles (ablation) | S-NeRF foreground vehicles · foreground vehicles | LPIPS: 0.136 -> 0.149 (SSIM-only) | Hurt |
| Replace Mip-NeRF360 with S-NeRF on Waymo background | Waymo background scenes · Waymo 2 sequences | PSNR: 22.10 -> 23.60 | Improved |
Later papers that compared with this method
12 later accepted papers report S-NeRF as a baseline in their results tables; 12 of them report a better number on most of the results they share with it.
- DrivingGaussian++: Toward Realistic Reconstruction and Editable Simulation for Surrounding Dynamic Driving Scenes (TPAMI 2026 · better on 3 of 3 shared results)
- S-NeRF++: Autonomous Driving Simulation via Neural Reconstruction and Generation (TPAMI 2025 · better on 3 of 3 shared results)
- Multi-StyleGS: Stylized Gaussian Splatting with Multiple Styles (AAAI 2025 · better on 2 of 2 shared results)
- Geometry Transfer for Stylizing Radiance Fields (CVPR 2024 · better on 7 of 7 shared results)
- NeRF Analogies: Example-Based Visual Attribute Transfer for NeRFs (CVPR 2024 · better on 7 of 7 shared results)
- Editable Scene Simulation for Autonomous Driving via Collaborative LLM-Agents (CVPR 2024 · better on 3 of 3 shared results)
- LidaRF: Delving into Lidar for Neural Radiance Field on Street Scenes (CVPR 2024 · better on 3 of 3 shared results)
- NeuRAD: Neural Rendering for Autonomous Driving (CVPR 2024 · better on 2 of 3 shared results)
4 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- GeoSim — GeoSim: Realistic Video Simulation via Geometry-Aware Composition for Self-Driving (CVPR 2021) · better on 3 of 3 shared results
- Mip-NeRF360 — Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields (CVPR 2022) · better on 2 of 2 shared results
- Mip-NeRF — Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields (ICCV 2021) · better on 1 of 1 shared result
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
- LidaRF: Delving into Lidar for Neural Radiance Field on Street Scenes (CVPR 2024 · similarity 0.91)
- NeuRAD: Neural Rendering for Autonomous Driving (CVPR 2024 · similarity 0.91)
- UC-NERF: Neural Radiance Field for Under-Calibrated Multi-View Cameras in Autonomous Driving (ICLR 2024 · similarity 0.90)
- Effective Multi-sensor Conditioning for Street-view Novel-view Synthesis (NeurIPS 2026 · similarity 0.90)
- Multi-Level Neural Scene Graphs for Dynamic Urban Environments (CVPR 2024 · similarity 0.89)
- NeRF On-the-go: Exploiting Uncertainty for Distractor-free NeRFs in the Wild (CVPR 2024 · similarity 0.89)
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.
