Radiant MLHub

Documentation

You can use the Radiant MLHub Python Client to search datasets on Radiant MLHub and easily download them to your local environment. You can use the MLHub API directly for programming languages other than Python.

Tutorials and Notebooks

Radiant MLHub Tutorials repository on GitHub has a series of Jupyter Notebooks for you to get started with using our API. Each notebook has comments and explanations that guides you through each step of accessing the API, searching for data, downloading data and loading data in Python, among others. There are also notebooks specific to each competition that walks you through building a baseline machine learning model.

Radiant Earth Foundation Youtube channel has a series of videos from our training programs and workshops that you can watch to learn more about applications of machine learning on Earth observations, and how to use Radiant MLHub resources for various applications.

Python Client Quick-Start Guide

  1. Get a Radiant MLHub API Key from your Settings page.
  2. Install and configure the MLHub Python Client:
    $ pip install radiant_mlhub
    $ mlhub configure
    API Key: Enter your API key here...
    Wrote profile to /Users/youruser/.mlhub/profiles
  3. Start coding! The radiant_mlhub Python Client documentation has examples and introductory material.

FAQ

Radiant MLHub is an open library for geospatial training data and models to advance machine learning applications on Earth observations. It serves as a resource for a community of practice, giving data scientists benchmarks to train and validate their models or use existing pre-trained models.

You can access all the training datasets and models on mlhub.earth under the Datasets and Models tabs. You would need to create a (free) account to download the collections though. In addition, you can get your API access token after you sign in on mlhub.earth , and use our Python Client to access and download the data.

No! Radiant MLHub is open-access, meaning that you can access and download all datasets and models for free. Most of these have a CC-BY-4.0 license; however, you should consult each dataset and model’s license before use, as in some cases, there may be specific license restrictions.

License information is included on the entry page for each dataset and model. In addition, you can retrieve the license using our Python Client. You can see an example here.

Radiant Earth Foundation generates datasets. We also host datasets generated by our partners and other community members. These are stored either in our cloud repository or in a different cloud location but registered on Radiant MLHub. Datasets are accessed through our API.

Yes! If you have a complete training dataset (either hosted on your cloud account or you want it to be hosted on Radiant’s cloud), or labels that you think can be used to create training data, as well as trained models that you would like to register on Radiant MLHub please contact ml@radiant.earth .

Open training data allows for the advancement of machine learning models and corresponding predictions. Training data are expensive to create or in-silos, making it very hard for people without access to data to build machine learning applications. This is especially a problem for researchers in developing countries that find it challenging to access machine learning ready data. By sharing training data on an open-access repository, users can focus their attention on improving machine learning models and creating new datasets where needed, rather than duplicating existing data.

Training machine learning models on geospatial data can be expensive and time-consuming. Sharing your model exposes your work and helps produce reusable workflows for community members interested in similar applications. Users can benchmark their model’s performance against yours or use it as a pre-trained model. The result of shared models is more accurate predictions trusted by the community.