Today, nearly 75% of the world’s population lives within 50 kilometers of the ocean. Coastal zones host critical ecosystems, infrastructure and economic assets. So, it’s of growing concern that these stretches of land are increasingly vulnerable to the dramatic effects of climate change. Maxar Intelligence is partnering with EY and Microsoft on the latest EY Open Science Data Challenge, which asks university students and early-career professionals to use artificial intelligence (AI), Maxar’s high-resolution satellite imagery, and Microsoft’s Planetary Computer’s Hub environment to help build a sustainable future for society and the planet.

The challenge is currently in Phase 1 during which participants will build machine learning models to detect damaged and undamaged buildings after tropical storms. The top entrants from Phase 1 will move on to Phase 2, which asks them to create a practical disaster response plan describing how their models will be used in practice and the value they offer to local beneficiaries for disaster response. The top finalists will receive a cash prize and a trip to the IEEE International Geoscience and Remote Sensing Symposium in Athens, Greece, for the awards ceremony in July 2024.

Satellite imagery used in this challenge

Through our Open Data Program, Maxar provided additional satellite imagery for the 2017 Hurricane Maria event, specifically in the Analysis-Ready Data (ARD) format. This format means the stack of imagery has undergone atmospheric correction, radiometric correction, orthorectification and pan-sharpening. In addition, ARD uses a patented process to align imagery collected on different dates. ARD provides smoother seamlines between image strips and improved alignment of vectors, which in turn yields greater accuracy for feature extraction and change detection. By Maxar handling these preprocessing steps, it enables users of ARD to jump straight into their image analysis workflows, unlocking insights from our satellite imagery faster.

Ask the Experts panel session

To help challenge participants get a jumpstart on their projects, we hosted an Ask the Experts panel session in February. Vincent Tompkins, a senior data scientist at Wovenware, and two experts from EY answered questions for an hour. Here’s the recording of the session:

This interactive element is best viewed on a larger screen

Our satellite imagery expert, Chris Orndorff, has three tips for using our satellite imagery in this challenge:

1. Benefits of using ARD for training AI and machine learning (ML) algorithms:

Maxar’s ARD enables greater control over certain atmospheric, topographic and land cover conditions within an image and provides consistency in multitemporal imagery inputs to AI/ML.

We previously published a blog post describing a study that demonstrates how ARD increases the accuracy of an object detection model and enables users to speed up their pixel-to-answer workflow by requiring fewer iterations to train their models. Challenge participants should see increased value for their models utilizing the Maxar ARD format.

2. Improving Maxar imagery with atmospheric compensation (ACOMP):

ACOMP is Maxar’s proprietary, fully automated, physics-based framework for mitigating the effects of the atmosphere between the satellite and the ground. In the simplest terms, it lets us see through haze, water vapor and particulates across diverse atmospheric conditions. This process normalizes reflectivity to a surface reflectance value for each pixel in an image, improving the accuracy of multitemporal data analysis.

3. Distinguishing between commercial and residential buildings in satellite imagery:

Maxar’s satellite imagery allows an end user to see details that can help determine if a building is used for commercial or residential purposes. Look for features like flat and white roofs (likely commercial) versus pitched, darker roofs (likely residential); parking lots around the building (commercial) versus buildings more closely together with singular sidewalks running from the street to the building (residential).

The deadline to register for the challenge and submit a solution for Phase 1 consideration is March 10.

Join the EY Open Science Data Challenge

Ready to try using satellite imagery, AI and cloud computing to build machine learning models to help coastal communities become more resilient to the effects of climate change?

Register to participate
Prev Post Back to Blog Next Post