Installation¶
Prerequisites¶
- Python 3.12+
- CUDA-capable GPU (tested on RTX 3080, RTX 5090)
- RGB-D camera (Intel RealSense D435i tested)
- SLAM system providing poses (RTAB-Map, ORB-SLAM3)
WSL2 Users
Tested on WSL2 Ubuntu 22.04 with RTAB-Map. WSL2 has USB passthrough limitations — you may need usbipd-win for camera access.
Install RTSM¶
git clone https://github.com/calabi-inc/rtsm.git
cd rtsm
# Install dependencies
pip install -r requirements.txt
# Install RTSM in editable mode
pip install -e .
Download Models¶
FastSAM Weights¶
Download from: FastSAM Releases
CLIP Weights¶
CLIP weights are auto-downloaded on first run, or you can pre-fetch:
Verify Installation¶
Next Steps¶
- Quick Start — Run your first query
- Configuration — Customize for your setup