Machine Learning

Accessing and Downloading Training Data on the Radiant MLHub API

Community Voices

Data Challenge Winners: Identifying African crop types using satellite imagery

In May, we announced the winners of the Radiant Earth Computer Vision for Crop Detection from Satellite Imagery data challenge, which took place in February and March 2020. A total of 440 data scientists signed up for the challenge, representing a wide range of educational backgrounds, institutions, and geographies. While five winners were selected, in this Q&A, we sat down with Karim Amer, the First Place Overall Winner of the Data challenge, and the First Place African Citizen winner, Femi Sotonwa. Our goal is to learn more about the people behind the top scores.

Machine Learning, News

BigEarthNet Benchmark Archive Now Available on Radiant MLHub

BigEarthNet is a new large-scale benchmark geospatial training data consisting of multi-label land cover classes in ten European countries. 

Radiant Earth Foundation, the leader in enabling access to geospatial training data, is pleased to announce the availability of the BigEarthNet large-scale benchmark archive through Radiant MLHub, the world’s first open library dedicated to Earth observation (EO) training data.

News, Standards

STAC 1.0-beta.1 Released!

The SpatioTemporal Asset Catalog community is incredibly proud to announce the release of STAC 1.0! If you want to get technical it’s 1.0-beta.1, which means that everything is not yet completely locked in. And it’s just the core specs, as we’ve split off the STAC API into its own repository, and its 1.0-beta.1 release will follow. But this is a huge milestone, as it symbolizes that the community has worked through every known issue and desired improvement. It is the beginning of the final stabilization steps, to ensure STAC will be a stable core that people can build on for years and even decades to come.

The reason we are calling it a ‘beta’ release is so that the specification is not so set that we can’t take additional feedback as we push to get it much more widely adopted. The goal between beta.1 and 1.0.0 is to update every piece of software that has implemented STAC, as well as upgrade all the existing STAC Catalogs to the latest, so we are sure our changes work for everyone.

Machine Learning, News

Announcing the Winners of Radiant Earth’s Competition for Crop Detection in Africa

Five Data Scientists emerged as winners of Radiant Earth Foundation’s competition, in partnership with Zindi Africa, to create a machine learning model that classifies farm fields in Kenya by crop type using time series of Sentinel-2 satellite imagery collected during the growing season.

Earth observations provide critical data for agricultural monitoring at scale, and machine learning (ML) techniques are best suited to learn from these data. Yet, building agricultural ML models poses a problem in Africa due to limited training data, as well as add-on hurdles created by the relatively small size of the farms. These difficulties prompted Radiant Earth to design a competition to crowdsource data science skills globally for the best crop detection model.

Machine Learning

A Guide for Collecting and Sharing Ground Reference Data for Machine Learning Applications

Machine learning (ML) applications for Earth observation (EO) can use currently available data that are collected via surveys for empirical research to investigate applied sciences or conduct socio-economic analysis. However, the chances of those survey data being incomplete are high. Many ML applications on EO require ground reference data, which are accurate observations of some property on the ground and can be used as a label or description of what a potential overhead image¹ represents.

It is, of course, possible to label images remotely using online platforms like OpenStreetMap…

Machine Learning

Which Data Science Event is Right for You?

As data-driven approaches become more integrated with global development missions like food security, practitioners need to stay up to date with data science methodologies and best practices. Trying to keep up with this rapidly evolving field can feel overwhelming. Luckily, there are numerous capacity-building data science events that provide a great way to stay abreast of new trends and developments.

Data science events are all structured in a similar participant-driven style, but each type has key differences and lends itself to distinct data science skillset and goals. In this article …

Machine Learning

Regional Training Data are Essential for Building Accurate Machine Learning Models

Thinking that having familiarity with Las Vegas in the United States would be enough to navigate the streets of Khartoum, Sudan sounds far-fetched. Yet, many machine learning models use Western-centric training data to predict features from satellite imagery in places as culturally and economically diverse as Bangladesh, Uganda, or Honduras.

Advances in computer vision and machine learning (ML) are improving the ability to accurately extract insights from frequent and high-resolution satellite imagery, shedding light on global development and progress …

Community Voices, Machine Learning

Olayinka Fadahunsi: Open Data Opportunities and Challenges in Africa

It is our pleasure to introduce Olayinka Fadahunsi, a Data Scientist with Stanbic IBTC Bank in Lagos, Nigeria and focuses on predictive customer models in personal and business banking. A graduate from the University of Lagos with a degree in Electrical and Electronics Engineering, he also moonlights as a Data Scientist on Zindi, Africa’s first data science competition platform that is focused on solving the continent’s most pressing problems. As one of the top data scientists on Zindi, Mr. Fadahunsi is enthusiastic to use his data science skills to solve real-world challenges …

Standards

STAC 0.9.0 Release — our final major release before 1.0-beta

I am pleased to announce that the incredible STAC community has just released version 0.9.0! This work on the release began in earnest during the 5th STAC Sprint that took place in early November. Having everyone in person enabled us to discuss all the major issues remaining, and we managed to get to decisions on all of them and got to at least draft pull requests of each. The last couple of months have been spent refining those and getting all the little details right, including two ‘release candidates’ — drafts that the community could give feedback on. You can see the full list of improvements in the changelog, and I’ll detail the highlights below.