Skip to content
/ nrtk Public

The Natural Robustness Toolkit (NRTK) is an open source toolkit for generating operationally-realistic perturbations to evaluate the natural robustness of computer vision algorithms.

License

Notifications You must be signed in to change notification settings

Kitware/nrtk

Repository files navigation

Natural Robustness Toolkit (NRTK)

The nrtk package is an open source toolkit for evaluating the natural robustness of computer vision algorithms to various perturbations, including sensor-specific changes to camera focal length, aperture diameter, etc. Functionality is provided through Strategy and Adapter patterns to allow for modular integration into systems and applications.

We have also created the nrtk-jatic package to support AI T&E use-cases and workflows, through interoperability with the maite library and integration with other JATIC tools. Users seeking to use NRTK to perturb MAITE-wrapped datasets or evaluate MAITE-wrapped models should start with the nrtk-jatic package.

Installation

The following steps assume the source tree has been acquired locally.

Install the current version via pip:

pip install nrtk

Alternatively, you can also use Poetry:

poetry install --sync --with dev-linting,dev-testing,dev-docs

See here for more installation documentation.

Getting Started

We provide a number of examples based on Jupyter notebooks in the ./examples/ directory to show usage of the nrtk package in a number of different contexts.

Contributions are welcome! See the CONTRIBUTING.md file for details.

Documentation

Documentation snapshots for releases as well as the latest master are hosted on ReadTheDocs.

The sphinx-based documentation may also be built locally for the most up-to-date reference:

# Install dependencies
poetry install --sync --with dev-linting,dev-testing,dev-docs
# Navigate to the documentation root.
cd docs
# Build the docs.
poetry run make html
# Open in your favorite browser!
firefox _build/html/index.html

Developer tools

pre-commit hooks pre-commit hooks are used to ensure that any code meets all linting and formatting guidelines required. After installing, this will always run before committing to ensure that any commits are following the standards, but you can also manually run the check without committing. If you want to commit despite there being errors, you can add --no-verify to your commit command.

Installing pre-commit hooks:

# Ensure that all dependencies are installed
poetry install --sync --with dev-linting,dev-testing,dev-docs
# Initialize pre-commit for the repository
poetry run pre-commit install
# Run pre-commit check on all files
poetry run pre-commit run --all-files

Contributing

License

Apache 2.0

Contacts

Principal Investigator: Brian Hu (Kitware) @brian.hu

Product Owner: Austin Whitesell (MITRE) @awhitesell

Scrum Master / Tech Lead: Brandon RichardWebster (Kitware) @b.richardwebster

Deputy Tech Lead: Emily Veenhuis (Kitware) @emily.veenhuis

About

The Natural Robustness Toolkit (NRTK) is an open source toolkit for generating operationally-realistic perturbations to evaluate the natural robustness of computer vision algorithms.

Resources

License

Stars

Watchers

Forks

Packages

No packages published