Research map / 3D vision / Gaussian splatting
DecoupleGS: Interactive 3D Gaussian Splatting for End-to-End Autonomous Driving Testing
Siying Li, Ying Ni, Haotian Shi, Jie Sun, Jian Sun
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 asset compression (pruning + VQ) (ablation) | DecoupleGS · Medium-density scenarios | FPS: 8.5 -> 35.4 | Improved |
| add map-guided geometric registration (ablation) | DecoupleGS · Medium-density scenarios | Trajectory ADE: 0.48 -> 0.05 | Improved |
| add proxy-based relighting (ablation) | DecoupleGS · Medium-density scenarios | Peak Angular Error: 48.5 -> 6.8 | Improved |
| add relighting (cost in throughput) (ablation) | DecoupleGS · Medium-density scenarios | FPS: 38.5 -> 35.4 | Hurt |
| use DecoupleGS compression vs Vanilla 3DGS for large traffic | DecoupleGS · Asset compression, size L | FPS: 12.1 -> 38.5 | Improved |
| use DecoupleGS compression vs Vanilla 3DGS | DecoupleGS · Asset compression, size S | PSNR-Veh: 28.52 -> 28.10 | Hurt |
| use DecoupleGS vs LightGaussian | DecoupleGS · Asset compression, size S | PSNR-Veh: 26.21 -> 28.10 | Improved |
| use DecoupleGS vs HUGSIM as simulator | DecoupleGS · Closed-loop simulator comparison | Driving Score: 0.765 -> 0.884 | Improved |
| render planner input with DecoupleGS vs HUGSIM (open-loop) | UniAD · Open-loop sim-to-real | mADE (m): 0.94 -> 0.82 | Improved |
| increase scenario difficulty from Easy to Extreme | UniAD · Closed-loop E2E testing | Driving Score: 0.725 -> 0.195 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- LightGaussian — LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS (NeurIPS 2024) · better on 2 of 4 shared results
- HUGSIM — HUGSIM: A Real-Time, Photo-Realistic and Closed-Loop Simulator for Autonomous Driving (TPAMI 2025) · better on 3 of 3 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
- SplatAD: Real-Time Lidar and Camera Rendering with 3D Gaussian Splatting for Autonomous Driving (CVPR 2025 · similarity 0.89)
- Unifying Appearance Codes and Bilateral Grids for Driving Scene Gaussian Splatting (NeurIPS 2025 · similarity 0.88)
- DrivingGaussian: Composite Gaussian Splatting for Surrounding Dynamic Autonomous Driving Scenes (CVPR 2024 · similarity 0.88)
- DeSiRe-GS: 4D Street Gaussians for Static-Dynamic Decomposition and Surface Reconstruction for Urban Driving Scenes (CVPR 2025 · similarity 0.88)
- FRUC: Feedforward Dynamic Scene Reconstruction from Uncalibrated Collaborative Driving Views (NeurIPS 2026 · similarity 0.87)
- VR-Drive: Viewpoint-Robust End-to-End Driving with Feed-Forward 3D Gaussian Splatting (NeurIPS 2025 · 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.
