Research map / 3D vision / 3D generation
RadarMP: Motion Perception for 4D mmWave Radar in Autonomous Driving
Ruiqi Cheng, Huijun Di, Jian Li, Feng Liu, Wei Liang
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 radial flow segmentation loss L_rfs (ablation) | RadarMP · K-Radar (loss ablation) | Pd (%): 62.033 -> 69.458 | Improved |
| add radial flow segmentation loss L_rfs (ablation) | RadarMP · K-Radar (loss ablation) | EPE3D (m): 0.209 -> 0.157 | Improved |
| add energy flow loss L_ef (ablation) | RadarMP · K-Radar (loss ablation) | Pd (%): 56.224 -> 69.458 | Improved |
| add energy flow loss L_ef (ablation) | RadarMP · K-Radar (loss ablation) | EPE3D (m): 0.788 -> 0.157 | Improved |
| add segmentation energy loss L_se (ablation) | RadarMP · K-Radar (loss ablation) | Pd (%): 19.846 -> 69.458 | Improved |
| add segmentation energy loss L_se (ablation) | RadarMP · K-Radar (loss ablation) | Pfa (%): 17.136 -> 1.335 | Improved |
| add PWC-Net reference points for deformable attention (ablation) | RadarMP · K-Radar scene flow | EPE3D (m): worsens EPE3D by 0.31 | Improved |
| use full pretrained PWC-Net instead of weak version (ablation) | RadarMP · K-Radar scene flow | EPE3D (m): worsens EPE3D by 0.07 | Improved |
| keep full 64 Doppler bins instead of reducing to 32/16 top-K (ablation) | RadarMP · K-Radar | Pd (%): Pd drops 3.6%/4.3% | Improved |
| keep full 64 Doppler bins instead of reducing to 32/16 top-K (ablation) | RadarMP · K-Radar | EPE3D (m): EPE3D drops 0.011m/0.018m | Improved |
| use RadarMP-P target points as input to PPN instead of OS-CFAR points | PointPWC-Net · K-Radar scene flow | EPE3D (m): 0.489 -> 0.223 | Improved |
| use RadarMP-P target points as input to CMFlow instead of Radelft points | CMFlow · K-Radar scene flow | EPE3D (m): 0.190 -> 0.168 | Improved |
Closest papers by meaning
- RadarOcc: Robust 3D Occupancy Prediction with 4D Imaging Radar (NeurIPS 2024 · similarity 0.89)
- Horizon3D: Sparse Radar-Camera Fusion for Long-Range 3D Perception in Autonomous Driving (ECCV 2026 · similarity 0.88)
- V2X-Radar: A Multi-modal Dataset with 4D Radar for Cooperative Perception (NeurIPS 2025 · similarity 0.88)
- AdaRadar: Rate Adaptive Spectral Compression for Radar-based Perception (CVPR 2026 · similarity 0.88)
- RF4D:Neural Radar Fields for Novel View Synthesis in Outdoor Dynamic Scenes (CVPR 2026 · similarity 0.87)
- Instantaneous Perception of Moving Objects in 3D (CVPR 2024 · similarity 0.85)
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.
