Installation¶
Requirements¶
- Python 3.11+
- uv package manager
Install¶
Clone the repository with submodules and sync dependencies with uv:
If you already cloned without --recurse-submodules, initialize the submodule manually:
Install rd as a shell command:
uv tool install -e . # base install
uv tool install -e ".[zed]" # + ZED cameras
uv tool install -e ".[zed,tri-stereo]" # + TRI Stereo depth (ONNX)
uv tool install -e ".[zed,tri-stereo,tri-stereo-trt-cu12]" # + TensorRT (CUDA 12)
uv tool install -e ".[zed,tri-stereo,tri-stereo-trt-cu13]" # + TensorRT (CUDA 13)
rd --help
Note
Re-run uv tool install --reinstall -e ".[<extras>]" whenever you add
or change extras, or after pulling updates from the repository.
Hardware SDKs¶
ZED cameras - install the ZED SDK, then run the helper script to download the matching Python wheel:
This downloads pyzed-*.whl into packages/ and updates pyproject.toml to
reference it. uv sync --extra zed then installs it into the project environment.
Intel RealSense - pyrealsense2 is included as a core dependency and will be
installed by uv sync. No additional SDK installation is required for most
distributions.
Optional depth backends¶
- TRI Stereo Depth — TRI's learned stereo depth model tailored for robot manipulation scenes. Supports
c32andc64variants with ONNX and TensorRT backends. - Fast Foundation Stereo — foundation model stereo depth; higher quality at object boundaries and thin structures.