Research map / Vision-language and multimodal / Multimodal LLMs
Vision-Centric BEV Perception: A Survey
Yuexin Ma, Tai Wang, Xuyang Bai, Huitong Yang, Yuenan Hou, Yaming Wang, Yu Qiao, Ruigang Yang, Xinge Zhu
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 temporal fusion to BEVFormer (ablation) | BEVFormer · nuScenes 3D detection | mAP / NDS / mAVE: BEVFormer-S 0.435 mAP, 0.495 NDS, 0.842 mAVE -> BEVFormer 0.481 mAP, 0.569 NDS, 0.378 mAVE | Improved |
| add temporal fusion to PolarFormer (ablation) | PolarFormer · nuScenes 3D detection | mAP / mAVE: PolarFormer-S 0.455 mAP, 0.870 mAVE -> PolarFormer 0.493 mAP, 0.440 mAVE | Improved |
| add temporal modeling to BEVDet (BEVDet4D) | BEVDet · nuScenes 3D detection (depth-based) | NDS / mAVE: 0.482 -> 0.552 NDS; 0.979 -> 0.289 mAVE | Improved |
| jointly train detection and segmentation heads on BEVFormer (ablation) | BEVFormer · nuScenes detection and segmentation | Drivable IoU: 80.1 -> 77.5 | Hurt |
| jointly train detection and segmentation heads on BEVFormer (ablation) | BEVFormer · nuScenes detection and segmentation | Car IoU: 44.8 -> 46.8 | Improved |
| jointly train detection and segmentation heads on BEVFormer (ablation) | BEVFormer · nuScenes detection | NDS / mAP: 0.517 -> 0.520 NDS; 0.416 -> 0.412 mAP | No clear effect |
| jointly train detection and segmentation on M2BEV (ablation) | M2BEV · nuScenes detection and segmentation | NDS / mAP / Drivable IoU: 0.470 -> 0.454 NDS; 0.417 -> 0.408 mAP; 77.2 -> 75.9 Drivable | Hurt |
| jointly train detection and segmentation on PETRv2 (ablation) | PETRv2 · nuScenes segmentation | Drivable IoU / Lane IoU: 80.5 -> 79.1 Drivable; 47.4 -> 44.3 Lane | Hurt |
| jointly train detection and segmentation on Ego3RT (ablation) | Ego3RT · nuScenes segmentation | Drivable IoU: 79.6 -> 74.6 | Hurt |
| jointly train detection and segmentation on PolarFormer (ablation) | PolarFormer · nuScenes segmentation | Drivable / Lane IoU: 81.0 -> 82.6 Drivable; 42.2 -> 44.5 Lane | Improved |
| replace MLP-based view transformation with transformer-based | PV-to-BEV methods · nuScenes BEV segmentation | Drivable IoU: 70.7 -> 82.6 | Improved |
| use polar instead of Cartesian parameterization | PolarFormer · nuScenes 3D detection | mAP / NDS: BEVFormer 0.481 mAP, 0.569 NDS -> PolarFormer 0.493 mAP, 0.572 NDS | Improved |
Closest papers by meaning
- Toward Real-world BEV Perception: Depth Uncertainty Estimation via Gaussian Splatting (CVPR 2025 · similarity 0.88)
- Improving Bird's Eye View Semantic Segmentation by Task Decomposition (CVPR 2024 · similarity 0.87)
- MamBEV: Enabling State Space Models to Learn Birds-Eye-View Representations (ICLR 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)
- DiffBEV: Conditional Diffusion Model for Bird’s Eye View Perception (AAAI 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.
