Research map / Point clouds / Segmentation and detection
BEVFusion: A Simple and Robust LiDAR-Camera Fusion Framework
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 |
|---|---|---|---|
| Fuse independent camera and LiDAR BEV features instead of using the LiDAR-only detector | BEVFusion with PointPillars · nuScenes validation | mAP: 35.1 -> 53.5 | Helped |
| Fuse independent camera and LiDAR BEV features instead of using the LiDAR-only detector | BEVFusion with CenterPoint · nuScenes validation | mAP: 57.1 -> 64.2 | Helped |
| Fuse independent camera and LiDAR BEV features instead of using the LiDAR-only detector | BEVFusion with TransFusion-L · nuScenes validation | mAP: 64.9 -> 67.9 | Helped |
| Use BEVFusion rather than TransFusion | 3D detection · nuScenes test | mAP: 68.9 -> 69.2 | Helped |
| Add the BEV encoder module in place of the ResNet18 BEV encoder (ablation) | Camera stream with PointPillars · nuScenes validation | mAP: 13.9 -> 17.9 | Helped |
| Add adaptive feature alignment in the FPN (ablation) | Camera stream with PointPillars · nuScenes validation | mAP: 17.9 -> 18.0 | Helped |
| Use Dual-Swin-Tiny instead of ResNet50 as the image backbone (ablation) | Camera stream with PointPillars · nuScenes validation | mAP: 18.0 -> 22.9 | Helped |
| Add channel and spatial fusion (ablation) | BEVFusion with PointPillars · nuScenes validation | mAP: 35.1 -> 51.6 | Helped |
| Add adaptive feature selection to channel and spatial fusion (ablation) | BEVFusion with PointPillars · nuScenes validation | mAP: 51.6 -> 53.5 | Helped |
| Add channel and spatial fusion (ablation) | BEVFusion with CenterPoint · nuScenes validation | mAP: 57.1 -> 63.0 | Helped |
| Add adaptive feature selection to channel and spatial fusion (ablation) | BEVFusion with CenterPoint · nuScenes validation | mAP: 63.0 -> 64.2 | Helped |
| Add channel and spatial fusion (ablation) | BEVFusion with TransFusion-L · nuScenes validation | mAP: 64.9 -> 67.3 | Helped |
Closest papers by meaning
- DeepFusion: Lidar-Camera Deep Fusion for Multi-Modal 3D Object Detection (CVPR 2022 · similarity 0.92)
- GAFusion: Adaptive Fusing LiDAR and Camera with Multiple Guidance for 3D Object Detection (CVPR 2024 · similarity 0.90)
- BEV-MAE: Bird’s Eye View Masked Autoencoders for Point Cloud Pre-training in Autonomous Driving Scenarios (AAAI 2024 · similarity 0.88)
- SimDistill: Simulated Multi-Modal Distillation for BEV 3D Object Detection (AAAI 2024 · similarity 0.88)
- PC-BEV: An Efficient Polar-Cartesian BEV Fusion Framework for LiDAR Semantic Segmentation (AAAI 2025 · similarity 0.87)
- Resilient Sensor Fusion Under Adverse Sensor Failures via Multi-Modal Expert Fusion (CVPR 2025 · similarity 0.87)
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.