Microsoft News and Stories reposted this
Our thoughts are with those affected by Hurricane Beryl, the earliest recorded Category 5 Atlantic Basin hurricane. Microsoft is working with Planet and humanitarian organizations to provide a preliminary analysis of the damage sustained by affected communities. We have run our damage assessment AI models on the satellite images we received from Planet and have maps of the affected buildings. If your organization would benefit from the underlying data in this report—please contact me. We are committed to sharing this information in support of the response efforts. It’s important to note that there is uncertainty in building footprints, and that building labeling efforts are ongoing by humanitarian organizations. Because of these differences, we trained a change detection model which looks at both the pre- and post-event imagery and assumed that all change was caused by the hurricane. The output of this model is in the red overlay on the visualizer. We then intersected the change layer with two building footprint layers: 1) an incomplete set of OpenStreetMap building footprints; and 2) a building footprint layer we generated based on just the pre-event imagery. We have used both OpenStreetMap and our own building segmentation model to estimate damage, with a range of damage between the two models. We were able to view 1937 buildings in the study area of the island of Carricou. Out of these, we estimate that at least 888 buildings were damaged: 777 buildings (40.11%) with damage fraction between 80 and 100% 102 buildings (5.27%) with damage fraction between 60% and 80% 75 buildings (3.87%) with damage fraction between 40% and 60% 70 buildings (3.61%) with damage fraction between 20% and 40% 913 buildings (47.13%) with damage fraction between 0% and 20% Our AI model bases its assessments on satellite imagery, which, while powerful, covers only a portion of the affected area and has limitations. The data and accompanying maps should serve as a preliminary guide and require on-the-ground verification for a thorough evaluation. The AI for Good Lab is committed to aiding recovery efforts and sharing valuable information that can accelerate the restoration of impacted communities. We appreciate the opportunity to partner with other organizations to help those in need. https://lnkd.in/gHEJaH4r Team: Caleb Robinson, Anthony Ortiz, Andrew Hassanali, Andrew Zolli, Amy Michaels, Kevin White , Santiago Salcido
Thanks: I am deeply concerned about the situation and greatly appreciate your efforts. During the tsunami disaster in South and Southeast Asia (1984-85), we witnessed the devastation and did our best to assist those severely affected. If I can be of any help to you and your team, please do not hesitate to let me know. I would be more than happy to extend any support necessary to overcome this situation.
Working with you and Planet, we’ve been able to ensure this valuable assessment data directly supports response efforts by the Government of Grenada. Thank you! Liz Malara Andrew Zolli and Kate Dooley
Paul T.
1w
Thanks for sharing💯🤠
This is a real world example of where AI capability play a crucial part in assessing & help with providing aid where needed.
Lori Sloan
2w
Wow, this is great work indeed
well done
Founder @ 210 Digital Marketing | Pioneering Healthcare Adtech & Analytics l AI Patent Holder
2dCongrats to the AI for Good Lab team. Your innovative efforts in using AI and satellite imagery are transformative.