Research map / 3D vision / Depth and stereo
BEV@DC: Bird's-Eye View Assisted Training for Depth Completion
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 |
|---|---|---|---|
| train with auxiliary LiDAR BEV completion and 3D completion guidance, then discard the LiDAR branch at inference | camera-based depth completion · KITTI depth completion benchmark | RMSE (mm): 709.12 -> 697.44 | Helped |
| add BEV completion without feature alignment (ablation) | camera stream baseline · KITTI depth completion validation set | RMSE: 762.21 -> 757.66 | Helped |
| add feature alignment between camera and LiDAR branches (ablation) | BEV completion model · KITTI depth completion validation set | RMSE: 757.66 -> 736.57 | Helped |
| add PV-SPN for fine-grained 3D completion (ablation) | BEV completion model with feature alignment · KITTI depth completion validation set | RMSE: 736.57 -> 719.62 | Helped |
| use 1 or 2 PV-SPN propagation iterations instead of the selected 3 iterations (ablation) | PV-SPN · KITTI depth completion validation set | RMSE: 726.25 or 721.51 | Hurt |
| use 4 neighbors rather than 16 in PV-SPN (ablation) | PV-SPN · KITTI depth completion validation set | RMSE | Hurt |
| increase PV-SPN iterations beyond 3 (ablation) | PV-SPN · KITTI depth completion validation set | RMSE | No clear effect |
| discard the auxiliary LiDAR branch at inference | BEV@DC · KITTI depth completion validation set | inference speed: 4.20 FPS -> 7.87 FPS | Helped |
Closest papers by meaning
- BEV-MAE: Bird’s Eye View Masked Autoencoders for Point Cloud Pre-training in Autonomous Driving Scenarios (AAAI 2024 · similarity 0.90)
- Distilling Monocular Foundation Model for Fine-grained Depth Completion (CVPR 2025 · similarity 0.88)
- PC-BEV: An Efficient Polar-Cartesian BEV Fusion Framework for LiDAR Semantic Segmentation (AAAI 2025 · similarity 0.87)
- Tri-Perspective View Decomposition for Geometry-Aware Depth Completion (CVPR 2024 · similarity 0.87)
- Flexible Depth Completion for Sparse and Varying Point Densities (CVPR 2024 · similarity 0.87)
- GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection (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.