Research map / 3D vision / 3D generation
RIGS: Radar-Informed Gaussian Splatting for Uncertainty-Aware 3D Occupancy and Motion Prediction
Zeyu Han, Junzhe Wu, Fang Zhang, Maani Ghaffari Jadidi, Jianqiang Wang
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 4D radar tensor (radar-assisted init + tri-modal fusion) to camera-only (ablation) | RIGS · K-Radar (51.2 m) | FG IoU: 10.19 -> 11.20 | Improved |
| add Doppler-supervised velocity learning (ablation) | RIGS · K-Radar (51.2 m) | FG IoU: 11.20 -> 12.36 | Improved |
| add evidential BKI refinement with ellipsoid kernel and RCS filtering (ablation) | RIGS · K-Radar (25.6 m) | FG IoU: +2.12% | Improved |
| use sparsified 4D radar tensor instead of radar points (ablation) | RIGS · K-Radar (51.2 m) | FG IoU: 12.61 -> 14.09 | Improved |
| add temporal displacement-guided Doppler de-aliasing to Doppler supervision (ablation) | RIGS · K-Radar Vdyn | MAE (m/s): 3.48 -> 2.37 | Improved |
| add Doppler supervision to displacement-only velocity (ablation) | RIGS · K-Radar Vdyn | MAE (m/s): 2.62 -> 2.37 | Improved |
| directly adapt camera-LiDAR fusion (SDGOcc, OccFusion) to radar input | SDGOcc/OccFusion · K-Radar | mIoU: 19.52 vs 21.50 (12.8 m, SDGOcc C+R vs OccFormer) | No clear effect |
| use radar-informed RIGS instead of camera-only OccFormer | RIGS · K-Radar (51.2 m) | FG IoU: 11.44 -> 14.09 | Improved |
| use RIGS instead of OccFormer at short range | RIGS · K-Radar (12.8 m) | mIoU: 21.50 vs 21.21 | Hurt |
Closest papers by meaning
- RadarOcc: Robust 3D Occupancy Prediction with 4D Imaging Radar (NeurIPS 2024 · similarity 0.89)
- ODG: Occupancy Prediction Using Dual Gaussians (NeurIPS 2025 · similarity 0.88)
- VoxelSplat: Dynamic Gaussian Splatting as an Effective Loss for Occupancy and Flow Prediction (CVPR 2025 · similarity 0.87)
- Selection, Not Fusion: Radar-Modulated State Space Models for Radar-Camera Depth Estimation (NeurIPS 2026 · similarity 0.86)
- GaussianFormer-2: Probabilistic Gaussian Superposition for Efficient 3D Occupancy Prediction (CVPR 2025 · similarity 0.86)
- DrivingGaussian: Composite Gaussian Splatting for Surrounding Dynamic Autonomous Driving Scenes (CVPR 2024 · 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.
