With one third of the world population in self isolation, COVID-19 has changed each one of our lives. Also, few days ago, there was a news about how heavy rainfall had caused the dam in Michigan to collapse and the recent Cyclone(Amphan) that added to the misery of many people during the pandemic. Also, our team members from Japan and Nepal have had the experience of major earthquake and seen the reality of panic and chaos. This made out team members think about the evacuation, but at what cost? Since, evacuating to a public place increases the chances of spreading the virus, we thought that strategizing about how to prevent such natural events from facilitating the spread of the novel coronavirus at evacuation shelters should be done as soon as possible to help save more lives around the globe.

For the safe evacuation, we designed an application 'EEP' which will guide each and everyone to the respective nearest evacuation centers depending on if they are not infected or infected from the virus or feeling unwell and not sure about their body condition. This application is designed very simple and easy to use for everyone during the emergency panic. This might not be 100%  as effective as self isolation, but it might be the best we could do and easier to take care of every citizens in the case of emergency during pandemic. With some modifications like allowing the crew members to be able to access the information of health condition through each QR code given to evacuated people, we can take immediate action to not allow the infection to spread. 

For countries like Japan which has a well-maintained database for evacuation shelters, it is easy to locate them. But, in the areas where the information on shelters are not available, it is possible to locate them by detecting buildings. We thought that we could use a model pre-trained by SpaceNet's Building Detection Dataset, which is open to the public as Open Data in AWS to detect potential shelters in the area where shelters are not pre-assigned. The pre-training model is a model introduced within Solaris, an open-source machine learning pipeline for geospatial imagery. Then a database will be constructed from the results obtained and GIS information. After developing a database, we plan to make an API that can exchange information with a database about existing shelters, shelter candidates, and evacuees.

None of our team members, had experienced using the satellite data until today. So, it was quite difficult to get used to the file format and find a useable data form thousands of data. Also, while designing the application, there were not many samples designed for emergency situation and actually not having used one of them before, our team faced some challenges to imagine the situation for the users.

Screenshots

Slides

Video

1. Solaris: pre-training model of SpaceNet (Api / python)
   • github: https://github.com/CosmiQ/solaris
   • doc: https://solaris.readthedocs.io/en/latest/index.html
2. SpaceNet
   • SpaceNet Open Data: https://registry.opendata.aws/spacenet/
   • SpaceNet 2: Building Detection v2: https://spacenet.ai/spacenet-buildings-dataset-v2/