Greater Impact with Machine Learning

Radiant MLHub is an open library for geospatial training data to advance machine learning applications on Earth Observations. It serves as a resource for a community of practice, giving data scientists benchmarks they can use to train and validate their models and improve its performance.

Radiant MLHub hosts open training datasets generated by Radiant Earth Foundation's team as well as other training data catalogs contributed by Radiant Earth’s partners. Radiant MLHub is open to anyone to access, store, register and/or share their training datasets for high-quality Earth observations. All of the training datasets are stored using a SpatioTemporal Asset Catalog (STAC) compliant catalog, and exposed through a common API. In future, model hosting and Python client applications will be added to MLHub services.

Training datasets include pairs of imagery and labels for different types of ML problems including image classification, object detection, and semantic segmentation. Labels are generated from ground reference data and/or annotation of imagery.

Datasets

.

For more details check registry.mlhub.earth

African Crops

The following crop type datasets are currently hosted on Radiant MLHub. Check each dataset's documentation on the API for more information.

Kenya

Surveyed in 2019 growing season by PlantVillage team, this dataset includes 319 plots. Crops types in this dataset are: Banana, Bean, Cabbage, Cassava, Cowpea, Fallowland, Groundnut, Maize, Millet, Sorghum, Soybean, Sugarcane, Sweet Potato, Tomato.

Tanzania

This dataset contains 392 plots provided by Great African Food Company. Data are collected in 2018 growing season and include Bush Bean, Dry Bean, Safflower, Sunflower, White Sorghum, Yellow Maize.

Uganda

This dataset contains 234 plots provided by Dalberg Data Insights. Data are collected in 2017 growing season and include Maize and Sorghum.

Chesapeake Bay Land cover

Land cover classes and associated imagery from NAIP, Landsat 8, as well as Bing building masks produced by the Chesapeake Conservancy and Microsoft AI for Earth.

BigEarthNet

Land cover image classification data across 10 European countries consistig of 590,326 Sentinel-2 image patches produced by BigEarth project.

SpaceNet

Roads and buildings training dataset across Las Vegas, Paris, Shanghai and Khartoum from SpaceNet 2 collection produced by SpaceNet LLC.

Global Land Cover [Future Data]

Radiant Earth Foundation is generating a global land cover training dataset based on annual time series of Sentinel-2 observations, which will include a representative set of image patches capturing global diversity of land cover classes. (Expected release: Summer 2020)

FAQ

.

What’s the purpose of Radiant MLHub?

Radiant MLHub is an open library for geospatial training data to advance machine learning applications on Earth Observations. It serves as a resource for a community of practice, giving data scientists benchmarks they can use to train and validate their models and improve its performance.

How do I get data?

Accessing data currently requires using the API. Go to dashboard.mlhub.earth, sign up (or sign in if you have already registered) and get your API access token. API documentation is available at docs.mlhub.earth. You can read more about accessing data in this guide.

Do I have to pay to use the data?

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

How can I find each dataset’s license?

License information are included in each dataset’s metadata under `license` key. You can see an example in this tutorial for retrieving the license.

Where do you get data?

Datasets are generated by Radiant Earth Foundation or our partners and other members of the community. Partners that have created training datasets store them either in our cloud repository or in a different cloud location but register it on Radiant MLHub, meaning that it can be accessed through our API.

Can I contribute data?

Yes! If you have a complete training dataset currently on the cloud that you want registered through Radiant MLHub, a dataset that you want to be hosted, or labels that you think can be used to create training data, please contact ml@radiant.earth.

Why should I share my data?

Open training data allows for faster and more collaborative advancement of machine learning models. Training data tends to be expensive to create, making it very hard for people without access to data to build machine learning applications. This is especially a problem when conducting research in developing countries where data access and availability is limited. By sharing training data on an open-access repository, efforts can be focused on improving machine learning models and creating new datasets where needed, rather than duplicating existing data.

Radiant MLHub on Slack

Join Radian MLHub community on Slack to be part of this collaborative open source group. If you are looking for new training data or models, or are interested in collaborating with us to develop new ones, get in touch!

Join Radiant MLHub Slack

Newsletter

Subscribe to Radiant Earth Foundation's newsletter to get the latest updates about new datasets, workshops and hackathons around machine learning and Earth observations.