MIT-SPARK/VGGT-SLAM

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1129 stars · 126 forks observed · 2026-09-22

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

VGGT-SLAM is a dense RGB SLAM system that performs real-time feed-forward 3D scene reconstruction, optimizing on the SL(4) manifold using the VGGT feed-forward model. It supports incremental map construction with visualization and optional open-set 3D object detection.

Use cases

• reconstruct 3D scenes from a folder of RGB images
• run dense visual SLAM in real time
• build incremental 3D maps with loop closure
• detect open-set 3D objects from text queries
• evaluate dense SLAM on benchmark datasets

When to choose

• you need dense 3D reconstruction from plain RGB images without depth sensors
• you want real-time feed-forward SLAM with loop closure
• you need 3D object detection integrated into mapping

When to avoid

• you need LiDAR or depth-sensor-based SLAM
• you lack a GPU or cannot install large vision models
• you need a lightweight sparse visual odometry solution

Facets

library · maturity active

computer-vision machine-learning graphics simulation computer-vision robotics deep-learning python slam 3d-reconstruction vggt dense-mapping simultaneous-localization-and-mapping scene-reconstruction open-set-detection linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="MIT-SPARK/VGGT-SLAM")