open-mmlab/mmdetection3d

Health v2 · maintenance only

Adoption not part of the score

6518 stars · 1786 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

MMDetection3D is OpenMMLab's next-generation platform for general 3D object detection, built on PyTorch. It provides a modular toolbox with a model zoo covering LiDAR-based, camera-based, and multi-modal 3D detection methods on standard benchmarks like nuScenes, KITTI, and Waymo.

Use cases

• train 3d object detection models on point clouds
• run lidar-based detection on nuScenes or KITTI
• evaluate 3d detection models on a benchmark
• detect objects from multi-view camera images
• fine-tune a pretrained 3d detection model
• benchmark point cloud detection methods
• build a custom 3d detection pipeline for autonomous driving

When to choose

• you need a modular, config-driven framework for 3D detection research
• you want pretrained models and a large model zoo for LiDAR or camera-based detection
• you work with standard autonomous driving datasets like nuScenes, KITTI, or Waymo
• you want consistency with the OpenMMLab ecosystem (mmcv, mmdetection)

When to avoid

• you need simple inference in production with minimal dependencies
• you work outside 3D detection, e.g. 2D detection or segmentation only
• you prefer lightweight PyTorch codebases you can modify freely
• you need Windows support or non-CUDA hardware

Facets

framework · maturity active

machine-learning deep-learning computer-vision computer-vision autonomous-vehicles deep-learning machine-learning python 3d-object-detection point-cloud lidar pytorch model-zoo mmlab gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="open-mmlab/mmdetection3d")