NVlabs/FoundationPose

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

3516 stars · 525 forks observed · 2026-08-28

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

FoundationPose is NVIDIA's unified foundation model for 6D object pose estimation and tracking of novel objects, supporting both model-based (CAD model) and model-free (reference images) setups without fine-tuning. It ranked first on the BOP leaderboard for model-based novel object pose estimation and is the official CVPR 2024 Highlight implementation.

Use cases

• estimate 6D pose of a novel object from RGBD images without retraining
• track unknown objects in video using only a few reference images
• enable robot grasping by localizing objects in 3D
• run pose estimation for AR applications
• benchmark pose estimation on the BOP leaderboard
• perform model-free 6DoF object tracking and 3D reconstruction

When to choose

• you need zero-shot pose estimation for objects without task-specific training
• you have RGBD input and either a CAD model or a handful of reference images
• you want state-of-the-art accuracy on novel object pose benchmarks
• you are building robotic manipulation or AR pipelines needing object localization

When to avoid

• you only have monocular RGB without depth
• you need real-time inference on embedded hardware without the Isaac ROS TensorRT variant
• you need a lightweight model for CPU-only deployment
• your objects are textureless planar markers better handled by fiducial systems like ArUco

Facets

library · maturity active

computer-vision machine-learning deep-learning computer-vision robotics artificial-intelligence autonomous-vehicles python 6d-pose-estimation object-tracking rgb-d foundation-model cvpr-2024 nvidia robotics-perception ar linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NVlabs/FoundationPose")