Research map / Detection and segmentation / 3D detection and driving
BEV-SAN: Accurate BEV 3D Object Detection via Slice Attention Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 local and global height slices with attention | BEVDepth R50 · nuScenes val, without CBGS | NDS: 0.336 -> 0.366 | Helped |
| Add local and global height slices with attention | BEVDepth R50 · nuScenes val, with CBGS | NDS: 0.470 -> 0.482 | Helped |
| Add BEV-SAN | BEVDepth R50 · nuScenes night val | NDS: 0.170 -> 0.210 | Helped |
| Add BEV-SAN | BEVDepth R50 · nuScenes rainy val | NDS: 0.363 -> 0.396 | Helped |
| Add local slices without global slices (ablation) | BEVDepth R50 · nuScenes val | NDS: 0.330 -> 0.351 | Helped |
| Add global slices without local slices (ablation) | BEVDepth R50 · nuScenes val | NDS: 0.330 -> 0.343 | Helped |
| Add local slices to global slices (ablation) | BEVDepth R50 · nuScenes val | NDS: 0.343 -> 0.366 | Helped |
| Use LiDAR-guided local sampling instead of average local sampling (ablation) | BEVDepth R50 with global slices · nuScenes val | NDS: 0.359 -> 0.366 | Helped |
| Use SE fusion instead of mean fusion for local slices (ablation) | BEVDepth R50 · nuScenes val, local slices only | NDS: 0.332 -> 0.350 | Helped |
| Use SE fusion instead of mean fusion in the first stage and second-stage mean fusion (ablation) | BEVDepth R50 with local + global slices · nuScenes val | NDS: 0.359 -> 0.361 | Helped |
| Use dual-branch transformer fusion instead of SE fusion in the second stage (ablation) | BEVDepth R50 with local + global slices · nuScenes val | NDS: 0.361 -> 0.366 | Helped |
| Add BEV-SAN | BEVDepth R50 with front camera occluded · nuScenes val, front camera occluded | mAP: 0.228 -> 0.258 | Helped |
Closest papers by meaning
- GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection (AAAI 2025 · similarity 0.90)
- BEV-MAE: Bird’s Eye View Masked Autoencoders for Point Cloud Pre-training in Autonomous Driving Scenarios (AAAI 2024 · similarity 0.89)
- Instance-Aware Multi-Camera 3D Object Detection with Structural Priors Mining and Self-Boosting Learning (AAAI 2024 · similarity 0.89)
- CLIP-BEVFormer: Enhancing Multi-View Image-Based BEV Detector with Ground Truth Flow (CVPR 2024 · similarity 0.88)
- M-BEV: Masked BEV Perception for Robust Autonomous Driving (AAAI 2024 · similarity 0.88)
- BEVSpread: Spread Voxel Pooling for Bird's-Eye-View Representation in Vision-based Roadside 3D Object Detection (CVPR 2024 · similarity 0.88)
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.