Radiant Earth Foundation Awarded Cooperative Agreement from NASA to Expand Radiant MLHub, the Geospatial Training Data Repository
Radiant Earth Foundation is pleased to announce receiving a 3-year cooperative agreement from NASA Advancing Collaborative Connections for Earth System Science (ACCESS) program to expand Radiant MLHub services, and support the growing demand for benchmarks in the geospatial machine learning community.
Radiant MLHub is the world’s first open-access cloud-based geospatial training data library designed to accelerate the adoption of ML to help solve global development challenges such as climate change and food insecurity. As an interoperable repository for sharing training data, Radiant MLHub allows anyone to discover and access high-quality ML-ready training data. It also invites individuals and organizations to register or share their geospatial training datasets to maximize its utility and facilitate the reproducibility of research results.
This new award will support the development of Radiant MLHub’s library to host ML models. The Radiant team will also develop a Python client to enhance the use of Radiant MLHub’s API for both training datasets and models. Finally, the existing training data catalog will be expanded by generating the first multi-mission global land cover training dataset based on Landsat-8, Sentinel-2, and Sentinel-1. Radiant Earth recently released the first version of LandCoverNet, a land cover classification benchmark training dataset based on Sentinel-2 observations for Africa, and this award will support extending that globally.
“This project is designed to standardize cataloging and access to geospatial training data and models so developers and researchers can easily reproduce results of previous studies and develop new ones,” said Hamed Alemohammad, Chief Data Scientist at Radiant Earth. “Essentially, we are building a ML commons that will facilitate browsing and usage, as well as publishing training datasets and models. By publishing model accuracy metrics, our library will also enable benchmarking models’ performance.”
Radiant Earth Foundation was founded in 2016 and works to empower organizations and individuals with open ML and EO training data, standards, and tools to address the world’s most critical international development challenges. Radiant Earth’s goal is to make Radiant MLHub the primary repository for geospatial training data and models to advance satellite imagery analysis using ML techniques.
Visit us on Twitter, LinkedIn, Facebook, Instagram, Medium, and GitHub.
