spla-tam/SplaTAM

Health v2 · maintenance only

Adoption not part of the score

2186 stars · 252 forks observed · 2026-08-28

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

SplaTAM is a dense RGB-D SLAM system that uses 3D Gaussian splatting for high-fidelity scene reconstruction and precise camera tracking from a single unposed RGB-D camera. It is the research code accompanying a CVPR 2024 paper, implemented in Python with PyTorch and CUDA.

Use cases

• run dense RGB-D SLAM with 3D Gaussian splatting
• track camera pose from unposed RGB-D video
• reconstruct high-fidelity 3D maps from RGB-D sequences
• benchmark SLAM methods on ScanNet, Replica, and TUM datasets
• render novel views from a reconstructed Gaussian map
• reconstruct scenes from iPhone RGB-D captures

When to choose

• you need state-of-the-art dense SLAM with explicit 3D Gaussian maps
• you want high-fidelity reconstruction and novel-view synthesis from RGB-D input
• you are doing robotics or AR research requiring accurate camera tracking
• you want a reproducible CVPR 2024 SLAM baseline for benchmarking

When to avoid

• you need real-time SLAM on resource-constrained hardware
• you only have monocular RGB without depth
• you need a production-ready, maintained SLAM product rather than research code
• you lack a CUDA-capable GPU

Facets

library · maturity active

computer-vision machine-learning simulation graphics computer-vision robotics autonomous-vehicles graphics python slam gaussian-splatting rgb-d camera-tracking 3d-reconstruction cvpr-2024 research-code linux gpu docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="spla-tam/SplaTAM")