Research map / Detection and segmentation / 3D detection and driving
Seeing in Double: Dual-Granularity BEV Segmentation via Mamba-Driven Alignment and Polar-Decoupled Experts
Jiaxin Cai, Rui Lin, Jingze Su, Qi Li, Wenjie Yang, Yuanlong Yu, Wenxi Liu
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 dual-resolution heterogeneous network (ablation) | BEV baseline · nuScenes ablation (visibility filtering, 224x448) | IoU: 42.0 -> 43.9 | Improved |
| add MRADE height estimation (ablation) | BEV baseline · nuScenes ablation (visibility filtering, 224x448) | IoU: 42.0 -> 43.1 | Improved |
| add MCRA alignment (ablation) | full model · nuScenes ablation (visibility filtering, 224x448) | IoU: 45.3 -> 46.1 | Improved |
| add ABSS without MCRA (ablation) | dual-res + MRADE · nuScenes ablation (visibility filtering, 224x448) | IoU: 44.5 -> 45.3 | Improved |
| shared-state Mamba alignment instead of cross-attention (ablation) | MCRA · nuScenes ablation (visibility filtering, 224x448) | IoU: 45.5 -> 46.1 | Improved |
| shared instead of independent Mamba state spaces (ablation) | MCRA · nuScenes ablation (visibility filtering, 224x448) | IoU: 45.8 -> 46.1 | Improved |
| add permanent expert to MRADE (ablation) | MRADE · nuScenes ablation (visibility filtering, 224x448) | IoU: 45.5 -> 46.1 | Improved |
| decouple radial and angular experts (ablation) | MRADE · nuScenes ablation (visibility filtering, 224x448) | IoU: 45.6 -> 46.1 | Improved |
| MRADE instead of fixed height bins (ablation) | height estimation · nuScenes ablation (visibility filtering, 224x448) | IoU: 44.7 -> 46.1 | Improved |
| full method vs PointBeV | Ours · nuScenes vehicle, no visibility filtering, 224x480 | IoU: 38.7 -> 42.4 | Improved |
| full method vs PointBeV on Lyft L5 short range | Ours · Lyft L5 short range | IoU: 72.6 -> 73.3 | Improved |
| full method vs PointBeV runtime | Ours · nuScenes efficiency | FPS: 15 -> 14 FPS; memory 1.21 -> 1.58 G | Mixed |
Closest papers by meaning
- Improving Bird's Eye View Semantic Segmentation by Task Decomposition (CVPR 2024 · similarity 0.89)
- PC-BEV: An Efficient Polar-Cartesian BEV Fusion Framework for LiDAR Semantic Segmentation (AAAI 2025 · similarity 0.87)
- A Black-Box Evaluation Framework for Semantic Robustness in Bird’s Eye View Detection (AAAI 2025 · similarity 0.87)
- M-BEV: Masked BEV Perception for Robust Autonomous Driving (AAAI 2024 · similarity 0.86)
- MamBEV: Enabling State Space Models to Learn Birds-Eye-View Representations (ICLR 2025 · similarity 0.86)
- GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection (AAAI 2025 · 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.
