Research map / 3D vision / 3D generation
MapPrior: Bird's-Eye View Map Layout Estimation with Generative Models
Xiyue Zhu, Vlas Zyrianov, Zhijian Liu, Shenlong 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 generative prior stage on top of BEVFusion-L | BEVFusion-L · nuScenes val (LiDAR) | mean IoU: 48.6 -> 51.1 | Improved |
| add generative prior stage on top of BEVFusion-L | BEVFusion-L · nuScenes val (LiDAR) | MMD: 115.2 -> 35.8 | Improved |
| add generative prior stage on top of BEVFusion-C | BEVFusion-C · nuScenes val (camera) | mean IoU: 56.6 -> 56.7 | No clear effect |
| add generative prior stage on top of BEVFusion-C | BEVFusion-C · nuScenes val (camera) | ECE: 0.038 -> 0.026 | Improved |
| add generative prior stage on top of camera+LiDAR BEVFusion | BEVFusion · nuScenes val (camera+LiDAR) | MMD: 21.6 -> 28.0 | Hurt |
| add output L2 loss to transformer training (ablation) | MapPrior-L · nuScenes val (LiDAR) | mIoU / MMD: mIoU 45.6 -> 49.0, MMD 33.3 -> 45.6 | Mixed |
| feed BEV features to the transformer in addition to guidance code (ablation) | MapPrior-L · nuScenes val (LiDAR) | mIoU: 49.0 -> 51.1, MMD 45.6 -> 35.8 | Improved |
| one-step generation instead of autoregressive sampling (LiDAR) (ablation) | MapPrior-L · nuScenes val (LiDAR) | FPS / mIoU: 0.57 -> 4.88 FPS, mIoU 51.1 -> 50.0, MMD 35.8 -> 50.2 | Mixed |
| one-step generation instead of autoregressive sampling (camera) (ablation) | MapPrior-C · nuScenes val (camera) | FPS: 0.60 -> 4.26 FPS, mIoU 56.7 -> 56.7 | Improved |
Closest papers by meaning
- VQ-Map: Bird's-Eye-View Map Layout Estimation in Tokenized Discrete Space via Vector Quantization (NeurIPS 2024 · similarity 0.89)
- A Black-Box Evaluation Framework for Semantic Robustness in Bird’s Eye View Detection (AAAI 2025 · similarity 0.88)
- Map It Anywhere: Empowering BEV Map Prediction using Large-scale Public Datasets (NeurIPS 2024 · similarity 0.87)
- Improving Bird's Eye View Semantic Segmentation by Task Decomposition (CVPR 2024 · similarity 0.87)
- Toward Real-world BEV Perception: Depth Uncertainty Estimation via Gaussian Splatting (CVPR 2025 · similarity 0.87)
- MamBEV: Enabling State Space Models to Learn Birds-Eye-View Representations (ICLR 2025 · 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.
