ToniRV/NeRF-SLAM

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1266 stars · 155 forks observed · 2026-08-28

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

NeRF-SLAM is a real-time dense monocular SLAM system that combines neural radiance fields (Instant-NGP) with probabilistic volumetric fusion (Sigma-Fusion) for dense 3D scene reconstruction. It is the research code accompanying two MIT papers, built in Python with C++ components (GTSAM, instant-ngp).

Use cases

• run real-time dense monocular SLAM with neural radiance fields
• reconstruct 3D scenes from monocular camera video
• research NeRF-based SLAM systems
• fuse depth and pose estimates into a volumetric map
• evaluate SLAM on Replica dataset scenes
• build dense 3D maps for robotics

When to choose

• you need real-time dense monocular SLAM with NeRF-based rendering
• you want to reproduce or extend the NeRF-SLAM or Sigma-Fusion papers
• you have a CUDA GPU and want state-of-the-art dense 3D reconstruction from RGB video

When to avoid

• you need a production-ready SLAM solution with support and documentation
• you lack a CUDA-capable GPU or don't want to compile C++ dependencies like GTSAM and instant-ngp
• you need lightweight or CPU-only SLAM
• you want a simple point-cloud SLAM without neural rendering

Facets

library · maturity active

computer-vision machine-learning simulation graphics robotics computer-vision deep-learning autonomous-vehicles python cpp nerf slam monocular-slam volumetric-fusion instant-ngp gtsam research-code 3d-reconstruction linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="ToniRV/NeRF-SLAM")