Research map / Detection and segmentation / 3D detection and driving
BEVHeight: A Robust Framework for Vision-Based Roadside 3D Object Detection
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 |
|---|---|---|---|
| replace depth-based projection with height prediction and height-based projection | BEVDepth · DAIR-V2X-I val | AP3D: 2.19%, 5.87% and 4.61% | Helped |
| replace depth-based projection with height prediction and height-based projection | BEVDepth · Rope3D val | AP3D R40: 4.97% / 4.02% | Helped |
| replace depth-based projection with height prediction and height-based projection | BEVDepth · DAIR-V2X-I val, roll and pitch disturbed | AP3D (IoU=0.5), vehicle mid: 9.48% -> 51.77% | Helped |
| use dynamic-increasing instead of uniform height discretization (ablation) | BEVHeight · DAIR-V2X-I val | AP3D: 75.63 -> 76.24; 75.63 -> 76.61 | Helped |
| set dynamic discretization α to 2.0 instead of 1.5 (ablation) | BEVHeight · DAIR-V2X-I val | AP3D: 76.24 -> 76.61 | Mixed |
| add point-cloud supervision (ablation) | BEVDepth · DAIR-V2X-I val | AP3D (IoU=0.5), vehicle mid: 60.75 -> 60.37 | Hurt |
| add point-cloud supervision (ablation) | BEVHeight · DAIR-V2X-I val | AP3D (IoU=0.5), vehicle mid: 63.49 -> 63.61 | No clear effect |
| use height-based rather than depth-based perception | BEVHeight vs BEVDepth · nuScenes val | mAP: 0.313 -> 0.291 | Hurt |
| use height-based rather than depth-based projection | BEVDepth · DAIR-V2X-I val, ResNet-101 | latency: 68 -> 62 | Helped |
Closest papers by meaning
- BEVSpread: Spread Voxel Pooling for Bird's-Eye-View Representation in Vision-based Roadside 3D Object Detection (CVPR 2024 · similarity 0.91)
- SOGDet: Semantic-Occupancy Guided Multi-View 3D Object Detection (AAAI 2024 · similarity 0.87)
- Instance-Aware Multi-Camera 3D Object Detection with Structural Priors Mining and Self-Boosting Learning (AAAI 2024 · similarity 0.87)
- CorrBEV: Multi-View 3D Object Detection by Correlation Learning with Multi-modal Prototypes (CVPR 2025 · similarity 0.87)
- M-BEV: Masked BEV Perception for Robust Autonomous Driving (AAAI 2024 · similarity 0.86)
- BEV-MAE: Bird’s Eye View Masked Autoencoders for Point Cloud Pre-training in Autonomous Driving Scenarios (AAAI 2024 · 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.