Research map / 3D vision / Gaussian splatting
Driving View Synthesis on Free-form Trajectories with Generative Prior
Zeyu Yang, Zijie Pan, Yuankun Yang, Xiatian Zhu, 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 unreliability mask (threshold 0.65) instead of masking all regions (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | FID: 102.69 -> 74.34 | Improved |
| add unreliability mask instead of no mask (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | FID: 76.26 -> 74.34 | Improved |
| add unreliability mask instead of no mask (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | IoU: 0.2193 -> 0.2263 | Improved |
| use iterative refinement with buffer interval 3,000 instead of one-time generation (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | FID: 89.06 -> 74.34 | Improved |
| reduce buffer interval from 3,000 to 500 (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | FID: 74.34 -> 75.33 | No clear effect |
| reduce buffer interval from 3,000 to 500 (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | training time: 1.6x -> 4.8x | Hurt |
| set refine noise level to 0.6 instead of 0.4 (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | FID: 96.21 -> 74.34 | Improved |
| set refine noise level to 0.6 instead of 1.0 (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | FID: 106.01 -> 74.34 | Improved |
| add DriveX to PVG | PVG · Waymo (2 segments, lane shift +-3m) | FID: 118.41 -> 91.85 | Improved |
| add DriveX to StreetGaussian | StreetGaussian · Waymo (2 segments, lane shift +-3m) | IoU: 0.1217 -> 0.2263 | Improved |
| add warped views as denoising start (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | FID: 105.77 -> 74.34 | Improved |
| add progressive shift length (ablation) | DriveX · Waymo (2 segments, lane shift +-3m) | FID: 75.41 -> 74.34 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- EmerNeRF — EmerNeRF: Emergent Spatial-Temporal Scene Decomposition via Self-Supervision (ICLR 2024) · better on 5 of 5 shared results
- ReconDreamer — ReconDreamer: Crafting World Models for Driving Scene Reconstruction via Online Restoration (CVPR 2025) · better on 2 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
- FreeSim: Toward Free-viewpoint Camera Simulation in Driving Scenes (CVPR 2025 · similarity 0.90)
- FreeVS: Generative View Synthesis on Free Driving Trajectory (ICLR 2025 · similarity 0.90)
- DriveScape: High-Resolution Driving Video Generation by Multi-View Feature Fusion (CVPR 2025 · similarity 0.90)
- X-Scene: Large-Scale Driving Scene Generation with High Fidelity and Flexible Controllability (NeurIPS 2025 · similarity 0.89)
- DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation (CVPR 2025 · similarity 0.88)
- Effective Multi-sensor Conditioning for Street-view Novel-view Synthesis (NeurIPS 2026 · similarity 0.87)
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.
