Research map / 3D vision / 3D generation
Learning Ego-Centric BEV Representations from a Perspective-Privileged View: Cross-View Supervision for Online HD Map Construction
Daniel Lengerer, Mathias Pechinger, Klaus Bogenberger, Carsten Markgraf
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 cross-view supervision from aerial teacher features | StreamMapNet · AID4AD nuScenes, 60x30 m | mAP: 34.1 -> 38.0 | Improved |
| add cross-view supervision from aerial teacher features | StreamMapNet · AID4AD nuScenes, 100x50 m | mAP: 22.4 -> 32.3 | Improved |
| add MapTRv2-style auxiliary BEV segmentation loss instead of aerial supervision | StreamMapNet · AID4AD nuScenes, 100x50 m | mAP: 22.4 -> 23.3 | No clear effect |
| add feature normalization to cross-view supervision (ablation) | CVS · AID4AD nuScenes, 100x50 m | mAP: 28.9 -> 31.5 | Improved |
| add affine adapter on top of normalization (ablation) | CVS · AID4AD nuScenes, 100x50 m | mAP: 31.5 -> 32.3 | Improved |
| use stronger aerial teacher (ResUNet++, teacher 47.3 -> 52.2 mAP) (ablation) | CVS · AID4AD nuScenes, 100x50 m | mAP: 32.3 -> 30.5 | Hurt |
| use UNet++ teacher (teacher 55.0 mAP) (ablation) | CVS · AID4AD nuScenes, 100x50 m | mAP: 32.3 -> 32.2 | No clear effect |
| increase BEVFormer student encoder from 1 to 2 layers (ablation) | CVS · AID4AD nuScenes, 100x50 m | mAP | No clear effect |
| vary BEV loss weight lambda_bev above zero (ablation) | CVS · AID4AD nuScenes, both ROIs | mAP | No clear effect |
| add CVS on rainy scenes | StreamMapNet · AID4AD nuScenes, 100x50 m, rainy scenes | mAP: 18.6 -> 29.7 | Improved |
Closest papers by meaning
- LetsMap: Unsupervised Representation Learning for Label-Efficient Semantic BEV Mapping (ECCV 2024 · similarity 0.88)
- Map It Anywhere: Empowering BEV Map Prediction using Large-scale Public Datasets (NeurIPS 2024 · similarity 0.87)
- VQ-Map: Bird's-Eye-View Map Layout Estimation in Tokenized Discrete Space via Vector Quantization (NeurIPS 2024 · similarity 0.87)
- Improving Bird's Eye View Semantic Segmentation by Task Decomposition (CVPR 2024 · similarity 0.86)
- Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis (NeurIPS 2026 · similarity 0.86)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (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.
