HuangJunJie2017/BEVDet

Health v2 · maintenance only

Adoption not part of the score

1801 stars · 311 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

BEVDet is a Python research codebase implementing the BEVDet series of bird's-eye-view (BEV) 3D object detection models for autonomous driving, including camera-only, temporal (4D), stereo, and LiDAR-camera fusion (DAL) variants. It provides training/evaluation configs, pretrained checkpoints, and published nuScenes benchmarks with mAP/NDS scores and latency/FPS measurements.

Use cases

• train a camera-based 3D object detection model on nuScenes
• build bird's-eye-view perception for autonomous driving
• benchmark multi-view camera 3D detection models on mAP, NDS, latency, and FPS
• fuse LiDAR and camera data for 3D object detection
• download pretrained BEV detection checkpoints and reproduce paper results
• experiment with temporal aggregation and stereo depth for 3D detection

When to choose

• You need a reproducible, well-cited baseline for camera-only BEV 3D object detection on nuScenes
• You are researching BEV perception architectures such as depth estimation, temporal fusion, or LiDAR-camera fusion
• You want official configs and pretrained models matching published accuracy and efficiency numbers

When to avoid

• You need production-grade, safety-certified perception for a deployed vehicle rather than a research codebase
• Your task is not 3D object detection, or your datasets are not nuScenes-compatible without significant adaptation
• You lack CUDA-capable GPUs or prefer not to work in the Python/PyTorch (MMDetection3D) ecosystem

Facets

framework · maturity active

machine-learning deep-learning computer-vision autonomous-vehicles computer-vision deep-learning machine-learning artificial-intelligence python 3d-object-detection bird-eye-view nuscenes autonomous-driving camera-based-perception lidar-camera-fusion multi-view-cameras model-zoo research-code pytorch gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="HuangJunJie2017/BEVDet")