Skip to content

NRDK: Neural Radar Development Kit

GitHub MIT License Supports Python 3.12+ Typed bear-ified CI GitHub issues

The Neural Radar Development Kit (NRDK) is an open-source and MIT-licensed Python library and framework for developing, training, and evaluating machine learning models on radar spectrum and multimodal sensor data.

Built around typed, high modular interfaces, the NRDK is designed to reduce the barrier of entry to learning on spectrum via out-of-the-box reference implementations for red-rover data and the I/Q-1M Dataset, while also providing an easy path towards customization and extensions for other radar and data collection systems.

Setup

The NRDK is intended to be used as a library; we recommend installing it as a submodule:

git submodule add git@github.com:RadarML/nrdk.git
uv pip install -e ./nrdk
git submodule add git@github.com:RadarML/nrdk.git
uv pip install -e ./nrdk[roverd]
[project]
dependencies = ["nrdk[roverd]"]

[tool.uv.extensions]
nrdk = { path = "./nrdk", editable = true }
git clone git@github.com:RadarML/nrdk.git
cd nrdk
uv sync --all-extras --frozen
uv run pre-commit install
Extras

The NRDK library also includes the following extras:

  • nrdk[roverd]: support for loading and processing data using the roverd format (i.e., as collected by the red-rover system).
  • nrdk[docs]: a mkdocs + mkdocs-material + mdocstrings-python documentation stack.
  • nrdk[dev]: linters, testing, pre-commit hooks, etc.

Development

After setting up nrdk for development (see above — don't forget to install the pre-commit hooks!), you can start contributing!

# Run all tests
uv run pytest --cov=src/nrdk --cov-report=html tests

# Build & serve documentation locally
uv run --extra docs mkdocs serve

Note that all code must pass our pre-commit hooks (ruff + pyright + pytest), which you can also run manually:

uv run pre-commit run --all-files

See Also

  • red-rover


    a multimodal mmWave Radar spectrum ecosystem

  • xwr


    python interface for collecting raw time signal data from TI mmWave radars

  • abstract_dataloader


    abstract interface for composable dataloaders and preprocessing pipelines

  • grt


    our latest work, GRT: Towards Foundational Models for Single-Chip Radar

  • 🎯 dart


    our prior work, DART: Implicit Doppler Tomography for Radar Novel View Synthesis

  • rover


    our previous data collection platform for radar time signal