Research map / 3D vision / NeRF and radiance fields
Efficient Depth-Guided Urban View Synthesis
sheng miao, Jiaxin Huang, Dongfeng Bai, Weichao Qiu, Liu Bingbing, Andreas Geiger, Yiyi Liao
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 geometric-prior feature (depth point cloud) to the model (ablation) | EDUS · KITTI-360 drop 50% feed-forward | PSNR: 15.71 -> 21.93 | Improved |
| add image-based 2D feature retrieval alongside 3D feature (ablation) | EDUS · KITTI-360 drop 50% feed-forward | PSNR: 18.28 -> 21.93 | Improved |
| use modulated SPADE CNN instead of 3D U-Net (ablation) | EDUS · KITTI-360 drop 50% feed-forward | PSNR: 21.49 -> 21.93 | Improved |
| add foreground/background/sky decomposition (ablation) | EDUS · KITTI-360 drop 50% feed-forward | PSNR: 20.65 -> 21.93 | Improved |
| add LiDAR supervision during training (ablation) | EDUS · KITTI-360 drop 50% feed-forward | PSNR: 21.75 -> 21.93 | Improved |
| add LiDAR supervision during training (ablation) | EDUS · KITTI-360 drop 50% feed-forward | LPIPS: 0.189 -> 0.178 | Improved |
| use EDUS geometric priors instead of feature-matching generalizable NeRFs at higher sparsity | generalizable NeRF · KITTI-360 drop 80% feed-forward | PSNR: 18.69 -> 19.63 (MuRF vs Ours) | Improved |
| initialize test-time optimization from EDUS generalizable model instead of 3DGS point cloud | 3D GS · KITTI-360 drop 90% test-time optimization | PSNR: 17.46 -> 19.16 | Improved |
| EDUS versus 3D GS at 50% drop rate | EDUS · KITTI-360 drop 50% test-time optimization | LPIPS: 0.109 -> 0.136 | Hurt |
| fine-tune only global feature volume with generalizable initialization | EDUS · KITTI-360 test-time optimization | training time: ~51min -> ~5min (MixNeRF vs Ours at 50%) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MuRF — MuRF: Multi-Baseline Radiance Fields (CVPR 2024) · better on 6 of 7 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
- Dynamic 3D Gaussian Fields for Urban Areas (NeurIPS 2024 · similarity 0.87)
- ViiNeuS: Volumetric Initialization for Implicit Neural Surface Reconstruction of Urban Scenes with Limited Image Overlap (CVPR 2025 · similarity 0.87)
- Swift-Mapping: Online Neural Implicit Dense Mapping in Urban Scenes (AAAI 2024 · similarity 0.87)
- Multi-Level Neural Scene Graphs for Dynamic Urban Environments (CVPR 2024 · similarity 0.87)
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias (ICLR 2025 · similarity 0.86)
- Effective Multi-sensor Conditioning for Street-view Novel-view Synthesis (NeurIPS 2026 · similarity 0.86)
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.
