NASA Harvest is seeking innovative solutions for our GRAVITY Challenge to tackle the complex task of mapping crop information to satellite imagery. The GRAVITY Challenge is a global technology innovation program for corporates, entrepreneurs and universities to design and build solutions to real industry, social and environmental problems using space data and capabilities. Now in its third year, GRAVITY fosters a community of over 300 space tech innovators to connect world-changing problems with the people who can solve them.
The critical goal of the NASA Harvest GRAVITY Challenge is to generate a consistent labeled dataset from street-level images required to develop high quality products from satellite images using machine learning. If you are interested in satellite technology applications for bolstering global food security, teams of 3-10 individuals can apply via the online application form by April 27th, 2021.
The NASA Harvest Gravity Challenge is coordinated by Robert Huppertz (Harvest Machine Learning Engineer), Hannah Kerner (Harvest Machine Learning and AI Lead), and Catherine Nakalembe (Harvest Africa Lead). NASA Harvest faces the challenge of limited access to training data on crop-types required to train machine learning algorithms.To tackle this challenge, field teams are collecting street-level images under the Lacuna Fund grant (Helmets Labeling Crops) of crop-fields to derive information on crops. The primary goal is to automate extracting information from these images and create high quality labels.
NASA Harvest researchers are currently developing machine learning solutions to map where crops are grown around the world in support of food security applications and informed agricultural decision-making. Regions that are most at risk of food insecurity often lack the necessary data to address threats to the food system. By monitoring crop conditions and food production around the world from space, we can improve the cost, timeliness and accessibility of this information - and help decision-makers better mitigate potential shortfalls across the agrifood system. Development of these types of algorithms requires ground-truth labeled data to train models and evaluate their performance. Such ground-truth datasets typically require visiting fields and recording observations in person, which is expensive and difficult to do at scale. This is compounded by the challenge of limited timely and publicly-available data.
If you win the NASA Harvest Challenge, you will continue into the Scale Phase where you will work with us to form a commercial arrangement with the intent to roll out a pilot of your solution. Some benefits of applying to the GRAVITY challenge include: